diff options
Diffstat (limited to 'src/contrib')
65 files changed, 20938 insertions, 0 deletions
diff --git a/src/contrib/Makefile.inc b/src/contrib/Makefile.inc new file mode 100644 index 0000000..a6b6576 --- /dev/null +++ b/src/contrib/Makefile.inc @@ -0,0 +1,103 @@ +noinst_LTLIBRARIES += libcontrib.la + +libcontrib_la_CPPFLAGS = $(AM_CPPFLAGS) $(CFLAG_VISIBILITY) +libcontrib_la_LDFLAGS = $(AM_LDFLAGS) $(LDFLAG_EXCLUDE_LIBS) +if !HAVE_LMDB +libcontrib_la_LDFLAGS += $(pthread_LIBS) +endif !HAVE_LMDB + +EXTRA_DIST += \ + contrib/licenses/0BSD \ + contrib/licenses/BSD-3-Clause \ + contrib/licenses/LGPL-2.0 \ + contrib/licenses/OLDAP-2.8 \ + contrib/lmdb/LICENSE \ + contrib/openbsd/LICENSE \ + contrib/ucw/LICENSE \ + contrib/dnstap/dnstap.proto + +libcontrib_la_SOURCES = \ + contrib/asan.h \ + contrib/base32hex.c \ + contrib/base32hex.h \ + contrib/base64.c \ + contrib/base64.h \ + contrib/ctype.h \ + contrib/dynarray.h \ + contrib/files.c \ + contrib/files.h \ + contrib/getline.c \ + contrib/getline.h \ + contrib/macros.h \ + contrib/mempattern.c \ + contrib/mempattern.h \ + contrib/net.c \ + contrib/net.h \ + contrib/qp-trie/trie.c \ + contrib/qp-trie/trie.h \ + contrib/sockaddr.c \ + contrib/sockaddr.h \ + contrib/string.c \ + contrib/string.h \ + contrib/strtonum.h \ + contrib/time.c \ + contrib/time.h \ + contrib/tolower.h \ + contrib/trim.h \ + contrib/wire_ctx.h \ + contrib/openbsd/siphash.c \ + contrib/openbsd/siphash.h \ + contrib/openbsd/strlcat.c \ + contrib/openbsd/strlcat.h \ + contrib/openbsd/strlcpy.c \ + contrib/openbsd/strlcpy.h \ + contrib/ucw/array-sort.h \ + contrib/ucw/binsearch.h \ + contrib/ucw/heap.c \ + contrib/ucw/heap.h \ + contrib/ucw/lists.c \ + contrib/ucw/lists.h \ + contrib/ucw/mempool.c \ + contrib/ucw/mempool.h + +if !HAVE_LMDB +libcontrib_la_SOURCES += \ + contrib/lmdb/lmdb.h \ + contrib/lmdb/mdb.c \ + contrib/lmdb/midl.c \ + contrib/lmdb/midl.h +endif !HAVE_LMDB + +if HAVE_LIBDNSTAP +noinst_LTLIBRARIES += libdnstap.la + +libdnstap_la_CPPFLAGS = $(AM_CPPFLAGS) $(DNSTAP_CFLAGS) +libdnstap_la_LDFLAGS = $(AM_LDFLAGS) $(DNSTAP_LIBS) + +SUFFIXES = .proto .pb-c.c .pb-c.h + +.proto.pb-c.c: + $(AM_V_GEN)@PROTOC_C@ --c_out=. -I$(srcdir) $< + +.proto.pb-c.h: + $(AM_V_GEN)@PROTOC_C@ --c_out=. -I$(srcdir) $< + +libdnstap_la_SOURCES = \ + contrib/dnstap/convert.c \ + contrib/dnstap/convert.h \ + contrib/dnstap/dnstap.c \ + contrib/dnstap/dnstap.h \ + contrib/dnstap/message.c \ + contrib/dnstap/message.h \ + contrib/dnstap/reader.c \ + contrib/dnstap/reader.h \ + contrib/dnstap/writer.c \ + contrib/dnstap/writer.h + +nodist_libdnstap_la_SOURCES = \ + contrib/dnstap/dnstap.pb-c.c \ + contrib/dnstap/dnstap.pb-c.h + +BUILT_SOURCES += $(nodist_libdnstap_la_SOURCES) +CLEANFILES += $(nodist_libdnstap_la_SOURCES) +endif HAVE_LIBDNSTAP diff --git a/src/contrib/asan.h b/src/contrib/asan.h new file mode 100644 index 0000000..112f81e --- /dev/null +++ b/src/contrib/asan.h @@ -0,0 +1,37 @@ +/* Copyright (C) 2015 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. +*/ + +#pragma once + +/* + * see sanitizer/asan_interface.h in compiler-rt (LLVM) + */ +#ifndef __has_feature + #define __has_feature(feature) 0 +#endif +#if __has_feature(address_sanitizer) || defined(__SANITIZE_ADDRESS__) + void __asan_poison_memory_region(void const volatile *addr, size_t size); + void __asan_unpoison_memory_region(void const volatile *addr, size_t size); + #define ASAN_POISON_MEMORY_REGION(addr, size) \ + __asan_poison_memory_region((addr), (size)) + #define ASAN_UNPOISON_MEMORY_REGION(addr, size) \ + __asan_unpoison_memory_region((addr), (size)) +#else + #define ASAN_POISON_MEMORY_REGION(addr, size) \ + ((void)(addr), (void)(size)) + #define ASAN_UNPOISON_MEMORY_REGION(addr, size) \ + ((void)(addr), (void)(size)) +#endif diff --git a/src/contrib/base32hex.c b/src/contrib/base32hex.c new file mode 100644 index 0000000..56fb427 --- /dev/null +++ b/src/contrib/base32hex.c @@ -0,0 +1,353 @@ +/* Copyright (C) 2017 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ + +#include "contrib/base32hex.h" +#include "libknot/errcode.h" + +#include <stdlib.h> +#include <stdint.h> + +/*! \brief Maximal length of binary input to Base32hex encoding. */ +#define MAX_BIN_DATA_LEN ((INT32_MAX / 8) * 5) + +/*! \brief Base32hex padding character. */ +static const uint8_t base32hex_pad = '='; +/*! \brief Base32hex alphabet. */ +static const uint8_t base32hex_enc[] = "0123456789abcdefghijklmnopqrstuv"; + +/*! \brief Indicates bad Base32hex character. */ +#define KO 255 +/*! \brief Indicates Base32hex padding character. */ +#define PD 32 + +/*! \brief Transformation and validation table for decoding Base32hex. */ +static const uint8_t base32hex_dec[256] = { + [ 0] = KO, [ 43] = KO, ['V'] = 31, [129] = KO, [172] = KO, [215] = KO, + [ 1] = KO, [ 44] = KO, ['W'] = KO, [130] = KO, [173] = KO, [216] = KO, + [ 2] = KO, [ 45] = KO, ['X'] = KO, [131] = KO, [174] = KO, [217] = KO, + [ 3] = KO, [ 46] = KO, ['Y'] = KO, [132] = KO, [175] = KO, [218] = KO, + [ 4] = KO, [ 47] = KO, ['Z'] = KO, [133] = KO, [176] = KO, [219] = KO, + [ 5] = KO, ['0'] = 0, [ 91] = KO, [134] = KO, [177] = KO, [220] = KO, + [ 6] = KO, ['1'] = 1, [ 92] = KO, [135] = KO, [178] = KO, [221] = KO, + [ 7] = KO, ['2'] = 2, [ 93] = KO, [136] = KO, [179] = KO, [222] = KO, + [ 8] = KO, ['3'] = 3, [ 94] = KO, [137] = KO, [180] = KO, [223] = KO, + [ 9] = KO, ['4'] = 4, [ 95] = KO, [138] = KO, [181] = KO, [224] = KO, + [ 10] = KO, ['5'] = 5, [ 96] = KO, [139] = KO, [182] = KO, [225] = KO, + [ 11] = KO, ['6'] = 6, ['a'] = 10, [140] = KO, [183] = KO, [226] = KO, + [ 12] = KO, ['7'] = 7, ['b'] = 11, [141] = KO, [184] = KO, [227] = KO, + [ 13] = KO, ['8'] = 8, ['c'] = 12, [142] = KO, [185] = KO, [228] = KO, + [ 14] = KO, ['9'] = 9, ['d'] = 13, [143] = KO, [186] = KO, [229] = KO, + [ 15] = KO, [ 58] = KO, ['e'] = 14, [144] = KO, [187] = KO, [230] = KO, + [ 16] = KO, [ 59] = KO, ['f'] = 15, [145] = KO, [188] = KO, [231] = KO, + [ 17] = KO, [ 60] = KO, ['g'] = 16, [146] = KO, [189] = KO, [232] = KO, + [ 18] = KO, ['='] = PD, ['h'] = 17, [147] = KO, [190] = KO, [233] = KO, + [ 19] = KO, [ 62] = KO, ['i'] = 18, [148] = KO, [191] = KO, [234] = KO, + [ 20] = KO, [ 63] = KO, ['j'] = 19, [149] = KO, [192] = KO, [235] = KO, + [ 21] = KO, [ 64] = KO, ['k'] = 20, [150] = KO, [193] = KO, [236] = KO, + [ 22] = KO, ['A'] = 10, ['l'] = 21, [151] = KO, [194] = KO, [237] = KO, + [ 23] = KO, ['B'] = 11, ['m'] = 22, [152] = KO, [195] = KO, [238] = KO, + [ 24] = KO, ['C'] = 12, ['n'] = 23, [153] = KO, [196] = KO, [239] = KO, + [ 25] = KO, ['D'] = 13, ['o'] = 24, [154] = KO, [197] = KO, [240] = KO, + [ 26] = KO, ['E'] = 14, ['p'] = 25, [155] = KO, [198] = KO, [241] = KO, + [ 27] = KO, ['F'] = 15, ['q'] = 26, [156] = KO, [199] = KO, [242] = KO, + [ 28] = KO, ['G'] = 16, ['r'] = 27, [157] = KO, [200] = KO, [243] = KO, + [ 29] = KO, ['H'] = 17, ['s'] = 28, [158] = KO, [201] = KO, [244] = KO, + [ 30] = KO, ['I'] = 18, ['t'] = 29, [159] = KO, [202] = KO, [245] = KO, + [ 31] = KO, ['J'] = 19, ['u'] = 30, [160] = KO, [203] = KO, [246] = KO, + [ 32] = KO, ['K'] = 20, ['v'] = 31, [161] = KO, [204] = KO, [247] = KO, + [ 33] = KO, ['L'] = 21, ['w'] = KO, [162] = KO, [205] = KO, [248] = KO, + [ 34] = KO, ['M'] = 22, ['x'] = KO, [163] = KO, [206] = KO, [249] = KO, + [ 35] = KO, ['N'] = 23, ['y'] = KO, [164] = KO, [207] = KO, [250] = KO, + [ 36] = KO, ['O'] = 24, ['z'] = KO, [165] = KO, [208] = KO, [251] = KO, + [ 37] = KO, ['P'] = 25, [123] = KO, [166] = KO, [209] = KO, [252] = KO, + [ 38] = KO, ['Q'] = 26, [124] = KO, [167] = KO, [210] = KO, [253] = KO, + [ 39] = KO, ['R'] = 27, [125] = KO, [168] = KO, [211] = KO, [254] = KO, + [ 40] = KO, ['S'] = 28, [126] = KO, [169] = KO, [212] = KO, [255] = KO, + [ 41] = KO, ['T'] = 29, [127] = KO, [170] = KO, [213] = KO, + [ 42] = KO, ['U'] = 30, [128] = KO, [171] = KO, [214] = KO, +}; + +int32_t base32hex_encode(const uint8_t *in, + const uint32_t in_len, + uint8_t *out, + const uint32_t out_len) +{ + // Checking inputs. + if (in == NULL || out == NULL) { + return KNOT_EINVAL; + } + if (in_len > MAX_BIN_DATA_LEN || out_len < ((in_len + 4) / 5) * 8) { + return KNOT_ERANGE; + } + + uint8_t rest_len = in_len % 5; + const uint8_t *stop = in + in_len - rest_len; + uint8_t *text = out; + + // Encoding loop takes 5 bytes and creates 8 characters. + while (in < stop) { + text[0] = base32hex_enc[in[0] >> 3]; + text[1] = base32hex_enc[(in[0] & 0x07) << 2 | in[1] >> 6]; + text[2] = base32hex_enc[(in[1] & 0x3E) >> 1]; + text[3] = base32hex_enc[(in[1] & 0x01) << 4 | in[2] >> 4]; + text[4] = base32hex_enc[(in[2] & 0x0F) << 1 | in[3] >> 7]; + text[5] = base32hex_enc[(in[3] & 0x7C) >> 2]; + text[6] = base32hex_enc[(in[3] & 0x03) << 3 | in[4] >> 5]; + text[7] = base32hex_enc[in[4] & 0x1F]; + text += 8; + in += 5; + } + + // Processing of padding, if any. + switch (rest_len) { + case 4: + text[0] = base32hex_enc[in[0] >> 3]; + text[1] = base32hex_enc[(in[0] & 0x07) << 2 | in[1] >> 6]; + text[2] = base32hex_enc[(in[1] & 0x3E) >> 1]; + text[3] = base32hex_enc[(in[1] & 0x01) << 4 | in[2] >> 4]; + text[4] = base32hex_enc[(in[2] & 0x0F) << 1 | in[3] >> 7]; + text[5] = base32hex_enc[(in[3] & 0x7C) >> 2]; + text[6] = base32hex_enc[(in[3] & 0x03) << 3]; + text[7] = base32hex_pad; + text += 8; + break; + case 3: + text[0] = base32hex_enc[in[0] >> 3]; + text[1] = base32hex_enc[(in[0] & 0x07) << 2 | in[1] >> 6]; + text[2] = base32hex_enc[(in[1] & 0x3E) >> 1]; + text[3] = base32hex_enc[(in[1] & 0x01) << 4 | in[2] >> 4]; + text[4] = base32hex_enc[(in[2] & 0x0F) << 1]; + text[5] = base32hex_pad; + text[6] = base32hex_pad; + text[7] = base32hex_pad; + text += 8; + break; + case 2: + text[0] = base32hex_enc[in[0] >> 3]; + text[1] = base32hex_enc[(in[0] & 0x07) << 2 | in[1] >> 6]; + text[2] = base32hex_enc[(in[1] & 0x3E) >> 1]; + text[3] = base32hex_enc[(in[1] & 0x01) << 4]; + text[4] = base32hex_pad; + text[5] = base32hex_pad; + text[6] = base32hex_pad; + text[7] = base32hex_pad; + text += 8; + break; + case 1: + text[0] = base32hex_enc[in[0] >> 3]; + text[1] = base32hex_enc[(in[0] & 0x07) << 2]; + text[2] = base32hex_pad; + text[3] = base32hex_pad; + text[4] = base32hex_pad; + text[5] = base32hex_pad; + text[6] = base32hex_pad; + text[7] = base32hex_pad; + text += 8; + break; + } + + return (text - out); +} + +int32_t base32hex_encode_alloc(const uint8_t *in, + const uint32_t in_len, + uint8_t **out) +{ + // Checking inputs. + if (out == NULL) { + return KNOT_EINVAL; + } + if (in_len > MAX_BIN_DATA_LEN) { + return KNOT_ERANGE; + } + + // Compute output buffer length. + uint32_t out_len = ((in_len + 4) / 5) * 8; + + // Allocate output buffer. + *out = malloc(out_len); + if (*out == NULL) { + return KNOT_ENOMEM; + } + + // Encode data. + int32_t ret = base32hex_encode(in, in_len, *out, out_len); + if (ret < 0) { + free(*out); + *out = NULL; + } + + return ret; +} + +int32_t base32hex_decode(const uint8_t *in, + const uint32_t in_len, + uint8_t *out, + const uint32_t out_len) +{ + // Checking inputs. + if (in == NULL || out == NULL) { + return KNOT_EINVAL; + } + if (in_len > INT32_MAX || out_len < ((in_len + 7) / 8) * 5) { + return KNOT_ERANGE; + } + if ((in_len % 8) != 0) { + return KNOT_BASE32HEX_ESIZE; + } + + const uint8_t *stop = in + in_len; + uint8_t *bin = out; + uint8_t pad_len = 0; + uint8_t c1, c2, c3, c4, c5, c6, c7, c8; + + // Decoding loop takes 8 characters and creates 5 bytes. + while (in < stop) { + // Filling and transforming 8 Base32hex chars. + c1 = base32hex_dec[in[0]]; + c2 = base32hex_dec[in[1]]; + c3 = base32hex_dec[in[2]]; + c4 = base32hex_dec[in[3]]; + c5 = base32hex_dec[in[4]]; + c6 = base32hex_dec[in[5]]; + c7 = base32hex_dec[in[6]]; + c8 = base32hex_dec[in[7]]; + + // Check 8. char if is bad or padding. + if (c8 >= PD) { + if (c8 == PD && pad_len == 0) { + pad_len = 1; + } else { + return KNOT_BASE32HEX_ECHAR; + } + } + + // Check 7. char if is bad or padding (if so, 6. must be too). + if (c7 >= PD) { + if (c7 == PD && c6 == PD && pad_len == 1) { + pad_len = 3; + } else { + return KNOT_BASE32HEX_ECHAR; + } + } + + // Check 6. char if is bad or padding. + if (c6 >= PD) { + if (!(c6 == PD && pad_len == 3)) { + return KNOT_BASE32HEX_ECHAR; + } + } + + // Check 5. char if is bad or padding. + if (c5 >= PD) { + if (c5 == PD && pad_len == 3) { + pad_len = 4; + } else { + return KNOT_BASE32HEX_ECHAR; + } + } + + // Check 4. char if is bad or padding (if so, 3. must be too). + if (c4 >= PD) { + if (c4 == PD && c3 == PD && pad_len == 4) { + pad_len = 6; + } else { + return KNOT_BASE32HEX_ECHAR; + } + } + + // Check 3. char if is bad or padding. + if (c3 >= PD) { + if (!(c3 == PD && pad_len == 6)) { + return KNOT_BASE32HEX_ECHAR; + } + } + + // 1. and 2. chars must not be padding. + if (c2 >= PD || c1 >= PD) { + return KNOT_BASE32HEX_ECHAR; + } + + // Computing of output data based on padding length. + switch (pad_len) { + case 0: + bin[4] = (c7 << 5) + c8; + // FALLTHROUGH + case 1: + bin[3] = (c5 << 7) + (c6 << 2) + (c7 >> 3); + // FALLTHROUGH + case 3: + bin[2] = (c4 << 4) + (c5 >> 1); + // FALLTHROUGH + case 4: + bin[1] = (c2 << 6) + (c3 << 1) + (c4 >> 4); + // FALLTHROUGH + case 6: + bin[0] = (c1 << 3) + (c2 >> 2); + } + + // Update output end. + switch (pad_len) { + case 0: + bin += 5; + break; + case 1: + bin += 4; + break; + case 3: + bin += 3; + break; + case 4: + bin += 2; + break; + case 6: + bin += 1; + break; + } + + in += 8; + } + + return (bin - out); +} + +int32_t base32hex_decode_alloc(const uint8_t *in, + const uint32_t in_len, + uint8_t **out) +{ + // Checking inputs. + if (out == NULL) { + return KNOT_EINVAL; + } + + // Compute output buffer length. + uint32_t out_len = ((in_len + 7) / 8) * 5; + + // Allocate output buffer. + *out = malloc(out_len); + if (*out == NULL) { + return KNOT_ENOMEM; + } + + // Decode data. + int32_t ret = base32hex_decode(in, in_len, *out, out_len); + if (ret < 0) { + free(*out); + *out = NULL; + } + + return ret; +} diff --git a/src/contrib/base32hex.h b/src/contrib/base32hex.h new file mode 100644 index 0000000..027acd8 --- /dev/null +++ b/src/contrib/base32hex.h @@ -0,0 +1,103 @@ +/* Copyright (C) 2018 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ +/*! + * \brief Base32hex implementation (RFC 4648). + * + * \note Input Base32hex string can contain a-v characters. These characters + * are considered as A-V equivalent. Lower-case variant is used for encoding! + */ + +#pragma once + +#include <stdint.h> + +/*! + * \brief Encodes binary data using Base32hex. + * + * \note Output data buffer contains Base32hex text string which isn't + * terminated with '\0'! + * + * \param in Input binary data. + * \param in_len Length of input data. + * \param out Output data buffer. + * \param out_len Size of output buffer. + * + * \retval >=0 length of output string. + * \retval KNOT_E* if error. + */ +int32_t base32hex_encode(const uint8_t *in, + const uint32_t in_len, + uint8_t *out, + const uint32_t out_len); + +/*! + * \brief Encodes binary data using Base32hex and output stores to own buffer. + * + * \note Output data buffer contains Base32hex text string which isn't + * terminated with '\0'! + * + * \note Output buffer should be deallocated after use. + * + * \param in Input binary data. + * \param in_len Length of input data. + * \param out Output data buffer. + * + * \retval >=0 length of output string. + * \retval KNOT_E* if error. + */ +int32_t base32hex_encode_alloc(const uint8_t *in, + const uint32_t in_len, + uint8_t **out); + +/*! + * \brief Decodes text data using Base32hex. + * + * \note Input data needn't be terminated with '\0'. + * + * \note Input data must be continuous Base32hex string! + * + * \param in Input text data. + * \param in_len Length of input string. + * \param out Output data buffer. + * \param out_len Size of output buffer. + * + * \retval >=0 length of output data. + * \retval KNOT_E* if error. + */ +int32_t base32hex_decode(const uint8_t *in, + const uint32_t in_len, + uint8_t *out, + const uint32_t out_len); + +/*! + * \brief Decodes text data using Base32hex and output stores to own buffer. + * + * \note Input data needn't be terminated with '\0'. + * + * \note Input data must be continuous Base32hex string! + * + * \note Output buffer should be deallocated after use. + * + * \param in Input text data. + * \param in_len Length of input string. + * \param out Output data buffer. + * + * \retval >=0 length of output data. + * \retval KNOT_E* if error. + */ +int32_t base32hex_decode_alloc(const uint8_t *in, + const uint32_t in_len, + uint8_t **out); diff --git a/src/contrib/base64.c b/src/contrib/base64.c new file mode 100644 index 0000000..a0d83a4 --- /dev/null +++ b/src/contrib/base64.c @@ -0,0 +1,272 @@ +/* Copyright (C) 2017 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ + +#include "contrib/base64.h" +#include "libknot/errcode.h" + +#include <stdlib.h> +#include <stdint.h> + +/*! \brief Maximal length of binary input to Base64 encoding. */ +#define MAX_BIN_DATA_LEN ((INT32_MAX / 4) * 3) + +/*! \brief Base64 padding character. */ +static const uint8_t base64_pad = '='; +/*! \brief Base64 alphabet. */ +static const uint8_t base64_enc[] = + "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/"; + +/*! \brief Indicates bad Base64 character. */ +#define KO 255 +/*! \brief Indicates Base64 padding character. */ +#define PD 64 + +/*! \brief Transformation and validation table for decoding Base64. */ +static const uint8_t base64_dec[256] = { + [ 0] = KO, ['+'] = 62, ['V'] = 21, [129] = KO, [172] = KO, [215] = KO, + [ 1] = KO, [ 44] = KO, ['W'] = 22, [130] = KO, [173] = KO, [216] = KO, + [ 2] = KO, [ 45] = KO, ['X'] = 23, [131] = KO, [174] = KO, [217] = KO, + [ 3] = KO, [ 46] = KO, ['Y'] = 24, [132] = KO, [175] = KO, [218] = KO, + [ 4] = KO, ['/'] = 63, ['Z'] = 25, [133] = KO, [176] = KO, [219] = KO, + [ 5] = KO, ['0'] = 52, [ 91] = KO, [134] = KO, [177] = KO, [220] = KO, + [ 6] = KO, ['1'] = 53, [ 92] = KO, [135] = KO, [178] = KO, [221] = KO, + [ 7] = KO, ['2'] = 54, [ 93] = KO, [136] = KO, [179] = KO, [222] = KO, + [ 8] = KO, ['3'] = 55, [ 94] = KO, [137] = KO, [180] = KO, [223] = KO, + [ 9] = KO, ['4'] = 56, [ 95] = KO, [138] = KO, [181] = KO, [224] = KO, + [ 10] = KO, ['5'] = 57, [ 96] = KO, [139] = KO, [182] = KO, [225] = KO, + [ 11] = KO, ['6'] = 58, ['a'] = 26, [140] = KO, [183] = KO, [226] = KO, + [ 12] = KO, ['7'] = 59, ['b'] = 27, [141] = KO, [184] = KO, [227] = KO, + [ 13] = KO, ['8'] = 60, ['c'] = 28, [142] = KO, [185] = KO, [228] = KO, + [ 14] = KO, ['9'] = 61, ['d'] = 29, [143] = KO, [186] = KO, [229] = KO, + [ 15] = KO, [ 58] = KO, ['e'] = 30, [144] = KO, [187] = KO, [230] = KO, + [ 16] = KO, [ 59] = KO, ['f'] = 31, [145] = KO, [188] = KO, [231] = KO, + [ 17] = KO, [ 60] = KO, ['g'] = 32, [146] = KO, [189] = KO, [232] = KO, + [ 18] = KO, ['='] = PD, ['h'] = 33, [147] = KO, [190] = KO, [233] = KO, + [ 19] = KO, [ 62] = KO, ['i'] = 34, [148] = KO, [191] = KO, [234] = KO, + [ 20] = KO, [ 63] = KO, ['j'] = 35, [149] = KO, [192] = KO, [235] = KO, + [ 21] = KO, [ 64] = KO, ['k'] = 36, [150] = KO, [193] = KO, [236] = KO, + [ 22] = KO, ['A'] = 0, ['l'] = 37, [151] = KO, [194] = KO, [237] = KO, + [ 23] = KO, ['B'] = 1, ['m'] = 38, [152] = KO, [195] = KO, [238] = KO, + [ 24] = KO, ['C'] = 2, ['n'] = 39, [153] = KO, [196] = KO, [239] = KO, + [ 25] = KO, ['D'] = 3, ['o'] = 40, [154] = KO, [197] = KO, [240] = KO, + [ 26] = KO, ['E'] = 4, ['p'] = 41, [155] = KO, [198] = KO, [241] = KO, + [ 27] = KO, ['F'] = 5, ['q'] = 42, [156] = KO, [199] = KO, [242] = KO, + [ 28] = KO, ['G'] = 6, ['r'] = 43, [157] = KO, [200] = KO, [243] = KO, + [ 29] = KO, ['H'] = 7, ['s'] = 44, [158] = KO, [201] = KO, [244] = KO, + [ 30] = KO, ['I'] = 8, ['t'] = 45, [159] = KO, [202] = KO, [245] = KO, + [ 31] = KO, ['J'] = 9, ['u'] = 46, [160] = KO, [203] = KO, [246] = KO, + [ 32] = KO, ['K'] = 10, ['v'] = 47, [161] = KO, [204] = KO, [247] = KO, + [ 33] = KO, ['L'] = 11, ['w'] = 48, [162] = KO, [205] = KO, [248] = KO, + [ 34] = KO, ['M'] = 12, ['x'] = 49, [163] = KO, [206] = KO, [249] = KO, + [ 35] = KO, ['N'] = 13, ['y'] = 50, [164] = KO, [207] = KO, [250] = KO, + [ 36] = KO, ['O'] = 14, ['z'] = 51, [165] = KO, [208] = KO, [251] = KO, + [ 37] = KO, ['P'] = 15, [123] = KO, [166] = KO, [209] = KO, [252] = KO, + [ 38] = KO, ['Q'] = 16, [124] = KO, [167] = KO, [210] = KO, [253] = KO, + [ 39] = KO, ['R'] = 17, [125] = KO, [168] = KO, [211] = KO, [254] = KO, + [ 40] = KO, ['S'] = 18, [126] = KO, [169] = KO, [212] = KO, [255] = KO, + [ 41] = KO, ['T'] = 19, [127] = KO, [170] = KO, [213] = KO, + [ 42] = KO, ['U'] = 20, [128] = KO, [171] = KO, [214] = KO, +}; + +int32_t base64_encode(const uint8_t *in, + const uint32_t in_len, + uint8_t *out, + const uint32_t out_len) +{ + // Checking inputs. + if (in == NULL || out == NULL) { + return KNOT_EINVAL; + } + if (in_len > MAX_BIN_DATA_LEN || out_len < ((in_len + 2) / 3) * 4) { + return KNOT_ERANGE; + } + + uint8_t rest_len = in_len % 3; + const uint8_t *stop = in + in_len - rest_len; + uint8_t *text = out; + + // Encoding loop takes 3 bytes and creates 4 characters. + while (in < stop) { + text[0] = base64_enc[in[0] >> 2]; + text[1] = base64_enc[(in[0] & 0x03) << 4 | in[1] >> 4]; + text[2] = base64_enc[(in[1] & 0x0F) << 2 | in[2] >> 6]; + text[3] = base64_enc[in[2] & 0x3F]; + text += 4; + in += 3; + } + + // Processing of padding, if any. + switch (rest_len) { + case 2: + text[0] = base64_enc[in[0] >> 2]; + text[1] = base64_enc[(in[0] & 0x03) << 4 | in[1] >> 4]; + text[2] = base64_enc[(in[1] & 0x0F) << 2]; + text[3] = base64_pad; + text += 4; + break; + case 1: + text[0] = base64_enc[in[0] >> 2]; + text[1] = base64_enc[(in[0] & 0x03) << 4]; + text[2] = base64_pad; + text[3] = base64_pad; + text += 4; + break; + } + + return (text - out); +} + +int32_t base64_encode_alloc(const uint8_t *in, + const uint32_t in_len, + uint8_t **out) +{ + // Checking inputs. + if (out == NULL) { + return KNOT_EINVAL; + } + if (in_len > MAX_BIN_DATA_LEN) { + return KNOT_ERANGE; + } + + // Compute output buffer length. + uint32_t out_len = ((in_len + 2) / 3) * 4; + + // Allocate output buffer. + *out = malloc(out_len); + if (*out == NULL) { + return KNOT_ENOMEM; + } + + // Encode data. + int32_t ret = base64_encode(in, in_len, *out, out_len); + if (ret < 0) { + free(*out); + *out = NULL; + } + + return ret; +} + +int32_t base64_decode(const uint8_t *in, + const uint32_t in_len, + uint8_t *out, + const uint32_t out_len) +{ + // Checking inputs. + if (in == NULL || out == NULL) { + return KNOT_EINVAL; + } + if (in_len > INT32_MAX || out_len < ((in_len + 3) / 4) * 3) { + return KNOT_ERANGE; + } + if ((in_len % 4) != 0) { + return KNOT_BASE64_ESIZE; + } + + const uint8_t *stop = in + in_len; + uint8_t *bin = out; + uint8_t pad_len = 0; + uint8_t c1, c2, c3, c4; + + // Decoding loop takes 4 characters and creates 3 bytes. + while (in < stop) { + // Filling and transforming 4 Base64 chars. + c1 = base64_dec[in[0]]; + c2 = base64_dec[in[1]]; + c3 = base64_dec[in[2]]; + c4 = base64_dec[in[3]]; + + // Check 4. char if is bad or padding. + if (c4 >= PD) { + if (c4 == PD && pad_len == 0) { + pad_len = 1; + } else { + return KNOT_BASE64_ECHAR; + } + } + + // Check 3. char if is bad or padding. + if (c3 >= PD) { + if (c3 == PD && pad_len == 1) { + pad_len = 2; + } else { + return KNOT_BASE64_ECHAR; + } + } + + // Check 1. and 2. chars if are not padding. + if (c2 >= PD || c1 >= PD) { + return KNOT_BASE64_ECHAR; + } + + // Computing of output data based on padding length. + switch (pad_len) { + case 0: + bin[2] = (c3 << 6) + c4; + // FALLTHROUGH + case 1: + bin[1] = (c2 << 4) + (c3 >> 2); + // FALLTHROUGH + case 2: + bin[0] = (c1 << 2) + (c2 >> 4); + } + + // Update output end. + switch (pad_len) { + case 0: + bin += 3; + break; + case 1: + bin += 2; + break; + case 2: + bin += 1; + break; + } + + in += 4; + } + + return (bin - out); +} + +int32_t base64_decode_alloc(const uint8_t *in, + const uint32_t in_len, + uint8_t **out) +{ + // Checking inputs. + if (out == NULL) { + return KNOT_EINVAL; + } + + // Compute output buffer length. + uint32_t out_len = ((in_len + 3) / 4) * 3; + + // Allocate output buffer. + *out = malloc(out_len); + if (*out == NULL) { + return KNOT_ENOMEM; + } + + // Decode data. + int32_t ret = base64_decode(in, in_len, *out, out_len); + if (ret < 0) { + free(*out); + *out = NULL; + } + + return ret; +} diff --git a/src/contrib/base64.h b/src/contrib/base64.h new file mode 100644 index 0000000..611574e --- /dev/null +++ b/src/contrib/base64.h @@ -0,0 +1,102 @@ +/* Copyright (C) 2018 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ +/*! + * \brief Base64 implementation (RFC 4648). + */ + +#pragma once + +#include <stdint.h> + +/*! + * \brief Encodes binary data using Base64. + * + * \note Output data buffer contains Base64 text string which isn't + * terminated with '\0'! + * + * \param in Input binary data. + * \param in_len Length of input data. + * \param out Output data buffer. + * \param out_len Size of output buffer. + * + * \retval >=0 length of output string. + * \retval KNOT_E* if error. + */ +int32_t base64_encode(const uint8_t *in, + const uint32_t in_len, + uint8_t *out, + const uint32_t out_len); + +/*! + * \brief Encodes binary data using Base64 and output stores to own buffer. + * + * \note Output data buffer contains Base64 text string which isn't + * terminated with '\0'! + * + * \note Output buffer should be deallocated after use. + * + * \param in Input binary data. + * \param in_len Length of input data. + * \param out Output data buffer. + * + * \retval >=0 length of output string. + * \retval KNOT_E* if error. + */ +int32_t base64_encode_alloc(const uint8_t *in, + const uint32_t in_len, + uint8_t **out); + +/*! + * \brief Decodes text data using Base64. + * + * \note Input data needn't be terminated with '\0'. + * + * \note Input data must be continuous Base64 string! + * + * \param in Input text data. + * \param in_len Length of input string. + * \param out Output data buffer. + * \param out_len Size of output buffer. + * + * \retval >=0 length of output data. + * \retval KNOT_E* if error. + */ +int32_t base64_decode(const uint8_t *in, + const uint32_t in_len, + uint8_t *out, + const uint32_t out_len); + +/*! + * \brief Decodes text data using Base64 and output stores to own buffer. + * + * \note Input data needn't be terminated with '\0'. + * + * \note Input data must be continuous Base64 string! + * + * \note Output buffer should be deallocated after use. + * + * \param in Input text data. + * \param in_len Length of input string. + * \param out Output data buffer. + * + * \retval >=0 length of output data. + * \retval KNOT_E* if error. + */ +int32_t base64_decode_alloc(const uint8_t *in, + const uint32_t in_len, + uint8_t **out); + +/*! @} */ diff --git a/src/contrib/ctype.h b/src/contrib/ctype.h new file mode 100644 index 0000000..93b85db --- /dev/null +++ b/src/contrib/ctype.h @@ -0,0 +1,192 @@ +/* Copyright (C) 2017 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ +/*! + * \brief Locale-independent ctype functions. + */ + +#pragma once + +#include <ctype.h> +#include <stdbool.h> +#include <stdint.h> + +enum { + CT_DIGIT = 1 << 0, + CT_UPPER = 1 << 1, + CT_LOWER = 1 << 2, + CT_XDIGT = 1 << 3, + CT_PUNCT = 1 << 4, + CT_PRINT = 1 << 5, + CT_SPACE = 1 << 6, +}; + +static const uint8_t char_mask[256] = { + // 0 - 8 + ['\t'] = CT_SPACE, + ['\n'] = CT_SPACE, + ['\v'] = CT_SPACE, + ['\f'] = CT_SPACE, + ['\r'] = CT_SPACE, + // 14 - 31 + [' '] = CT_PRINT | CT_SPACE, + + ['!'] = CT_PRINT | CT_PUNCT, + ['"'] = CT_PRINT | CT_PUNCT, + ['#'] = CT_PRINT | CT_PUNCT, + ['$'] = CT_PRINT | CT_PUNCT, + ['%'] = CT_PRINT | CT_PUNCT, + ['&'] = CT_PRINT | CT_PUNCT, + ['\''] = CT_PRINT | CT_PUNCT, + ['('] = CT_PRINT | CT_PUNCT, + [')'] = CT_PRINT | CT_PUNCT, + ['*'] = CT_PRINT | CT_PUNCT, + ['+'] = CT_PRINT | CT_PUNCT, + [','] = CT_PRINT | CT_PUNCT, + ['-'] = CT_PRINT | CT_PUNCT, + ['.'] = CT_PRINT | CT_PUNCT, + ['/'] = CT_PRINT | CT_PUNCT, + + ['0'] = CT_PRINT | CT_DIGIT | CT_XDIGT, + ['1'] = CT_PRINT | CT_DIGIT | CT_XDIGT, + ['2'] = CT_PRINT | CT_DIGIT | CT_XDIGT, + ['3'] = CT_PRINT | CT_DIGIT | CT_XDIGT, + ['4'] = CT_PRINT | CT_DIGIT | CT_XDIGT, + ['5'] = CT_PRINT | CT_DIGIT | CT_XDIGT, + ['6'] = CT_PRINT | CT_DIGIT | CT_XDIGT, + ['7'] = CT_PRINT | CT_DIGIT | CT_XDIGT, + ['8'] = CT_PRINT | CT_DIGIT | CT_XDIGT, + ['9'] = CT_PRINT | CT_DIGIT | CT_XDIGT, + + [':'] = CT_PRINT | CT_PUNCT, + [';'] = CT_PRINT | CT_PUNCT, + ['<'] = CT_PRINT | CT_PUNCT, + ['='] = CT_PRINT | CT_PUNCT, + ['>'] = CT_PRINT | CT_PUNCT, + ['?'] = CT_PRINT | CT_PUNCT, + ['@'] = CT_PRINT | CT_PUNCT, + + ['A'] = CT_PRINT | CT_UPPER | CT_XDIGT, + ['B'] = CT_PRINT | CT_UPPER | CT_XDIGT, + ['C'] = CT_PRINT | CT_UPPER | CT_XDIGT, + ['D'] = CT_PRINT | CT_UPPER | CT_XDIGT, + ['E'] = CT_PRINT | CT_UPPER | CT_XDIGT, + ['F'] = CT_PRINT | CT_UPPER | CT_XDIGT, + ['G'] = CT_PRINT | CT_UPPER, + ['H'] = CT_PRINT | CT_UPPER, + ['I'] = CT_PRINT | CT_UPPER, + ['J'] = CT_PRINT | CT_UPPER, + ['K'] = CT_PRINT | CT_UPPER, + ['L'] = CT_PRINT | CT_UPPER, + ['M'] = CT_PRINT | CT_UPPER, + ['N'] = CT_PRINT | CT_UPPER, + ['O'] = CT_PRINT | CT_UPPER, + ['P'] = CT_PRINT | CT_UPPER, + ['Q'] = CT_PRINT | CT_UPPER, + ['R'] = CT_PRINT | CT_UPPER, + ['S'] = CT_PRINT | CT_UPPER, + ['T'] = CT_PRINT | CT_UPPER, + ['U'] = CT_PRINT | CT_UPPER, + ['V'] = CT_PRINT | CT_UPPER, + ['W'] = CT_PRINT | CT_UPPER, + ['X'] = CT_PRINT | CT_UPPER, + ['Y'] = CT_PRINT | CT_UPPER, + ['Z'] = CT_PRINT | CT_UPPER, + + ['['] = CT_PRINT | CT_PUNCT, + ['\\'] = CT_PRINT | CT_PUNCT, + [']'] = CT_PRINT | CT_PUNCT, + ['^'] = CT_PRINT | CT_PUNCT, + ['_'] = CT_PRINT | CT_PUNCT, + ['`'] = CT_PRINT | CT_PUNCT, + + ['a'] = CT_PRINT | CT_LOWER | CT_XDIGT, + ['b'] = CT_PRINT | CT_LOWER | CT_XDIGT, + ['c'] = CT_PRINT | CT_LOWER | CT_XDIGT, + ['d'] = CT_PRINT | CT_LOWER | CT_XDIGT, + ['e'] = CT_PRINT | CT_LOWER | CT_XDIGT, + ['f'] = CT_PRINT | CT_LOWER | CT_XDIGT, + ['g'] = CT_PRINT | CT_LOWER, + ['h'] = CT_PRINT | CT_LOWER, + ['i'] = CT_PRINT | CT_LOWER, + ['j'] = CT_PRINT | CT_LOWER, + ['k'] = CT_PRINT | CT_LOWER, + ['l'] = CT_PRINT | CT_LOWER, + ['m'] = CT_PRINT | CT_LOWER, + ['n'] = CT_PRINT | CT_LOWER, + ['o'] = CT_PRINT | CT_LOWER, + ['p'] = CT_PRINT | CT_LOWER, + ['q'] = CT_PRINT | CT_LOWER, + ['r'] = CT_PRINT | CT_LOWER, + ['s'] = CT_PRINT | CT_LOWER, + ['t'] = CT_PRINT | CT_LOWER, + ['u'] = CT_PRINT | CT_LOWER, + ['v'] = CT_PRINT | CT_LOWER, + ['w'] = CT_PRINT | CT_LOWER, + ['x'] = CT_PRINT | CT_LOWER, + ['y'] = CT_PRINT | CT_LOWER, + ['z'] = CT_PRINT | CT_LOWER, + + ['{'] = CT_PRINT | CT_PUNCT, + ['|'] = CT_PRINT | CT_PUNCT, + ['}'] = CT_PRINT | CT_PUNCT, + ['~'] = CT_PRINT | CT_PUNCT, + // 127 - 255 +}; + +static inline bool is_alnum(uint8_t c) +{ + return char_mask[c] & (CT_DIGIT | CT_UPPER | CT_LOWER); +} + +static inline bool is_alpha(uint8_t c) +{ + return char_mask[c] & (CT_UPPER | CT_LOWER); +} + +static inline bool is_digit(uint8_t c) +{ + return char_mask[c] & CT_DIGIT; +} + +static inline bool is_xdigit(uint8_t c) +{ + return char_mask[c] & CT_XDIGT; +} + +static inline bool is_lower(uint8_t c) +{ + return char_mask[c] & CT_LOWER; +} + +static inline bool is_upper(uint8_t c) +{ + return char_mask[c] & CT_UPPER; +} + +static inline bool is_print(uint8_t c) +{ + return char_mask[c] & CT_PRINT; +} + +static inline bool is_punct(uint8_t c) +{ + return char_mask[c] & CT_PUNCT; +} + +static inline bool is_space(uint8_t c) +{ + return char_mask[c] & CT_SPACE; +} diff --git a/src/contrib/dnstap/convert.c b/src/contrib/dnstap/convert.c new file mode 100644 index 0000000..f59bacd --- /dev/null +++ b/src/contrib/dnstap/convert.c @@ -0,0 +1,142 @@ +/* Copyright (C) 2016 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. +*/ + +#include <sys/socket.h> +#include <sys/types.h> +#include <netinet/in.h> +#include <netinet/ip.h> + +#include "contrib/dnstap/convert.h" +#include "contrib/dnstap/dnstap.pb-c.h" + +/*! + * \brief Translation between real and Dnstap value. + */ +typedef struct mapping { + int real; + int dnstap; +} mapping_t; + +/*! + * \brief Mapping for network family. + */ +static const mapping_t SOCKET_FAMILY_MAPPING[] = { + { AF_INET, DNSTAP__SOCKET_FAMILY__INET }, + { AF_INET6, DNSTAP__SOCKET_FAMILY__INET6 }, + { 0 } +}; + +/*! + * \brief Mapping from network protocol. + */ +static const mapping_t SOCKET_PROTOCOL_MAPPING[] = { + { IPPROTO_UDP, DNSTAP__SOCKET_PROTOCOL__UDP }, + { IPPROTO_TCP, DNSTAP__SOCKET_PROTOCOL__TCP }, + { 0 } +}; + +/*! + * \brief Get Dnstap value for a given real value. + */ +static int encode(const mapping_t *mapping, int real) +{ + for (const mapping_t *m = mapping; m->real != 0; m += 1) { + if (m->real == real) { + return m->dnstap; + } + } + + return 0; +} + +/*! + * \brief Get real value for a given Dnstap value. + */ +static int decode(const mapping_t *mapping, int dnstap) +{ + for (const mapping_t *m = mapping; m->real != 0; m += 1) { + if (m->dnstap == dnstap) { + return m->real; + } + } + + return 0; +} + +/* -- public API ----------------------------------------------------------- */ + +Dnstap__SocketFamily dt_family_encode(int family) +{ + return encode(SOCKET_FAMILY_MAPPING, family); +} + +int dt_family_decode(Dnstap__SocketFamily dnstap_family) +{ + return decode(SOCKET_FAMILY_MAPPING, dnstap_family); +} + +Dnstap__SocketProtocol dt_protocol_encode(int protocol) +{ + return encode(SOCKET_PROTOCOL_MAPPING, protocol); +} + +int dt_protocol_decode(Dnstap__SocketProtocol dnstap_protocol) +{ + return decode(SOCKET_PROTOCOL_MAPPING, dnstap_protocol); +} + +bool dt_message_type_is_query(Dnstap__Message__Type type) +{ + switch (type) { + case DNSTAP__MESSAGE__TYPE__AUTH_QUERY: + case DNSTAP__MESSAGE__TYPE__CLIENT_QUERY: + case DNSTAP__MESSAGE__TYPE__FORWARDER_QUERY: + case DNSTAP__MESSAGE__TYPE__RESOLVER_QUERY: + case DNSTAP__MESSAGE__TYPE__STUB_QUERY: + case DNSTAP__MESSAGE__TYPE__TOOL_QUERY: + return true; + default: + return false; + } +} + +bool dt_message_type_is_response(Dnstap__Message__Type type) +{ + switch (type) { + case DNSTAP__MESSAGE__TYPE__AUTH_RESPONSE: + case DNSTAP__MESSAGE__TYPE__CLIENT_RESPONSE: + case DNSTAP__MESSAGE__TYPE__FORWARDER_RESPONSE: + case DNSTAP__MESSAGE__TYPE__RESOLVER_RESPONSE: + case DNSTAP__MESSAGE__TYPE__STUB_RESPONSE: + case DNSTAP__MESSAGE__TYPE__TOOL_RESPONSE: + return true; + default: + return false; + } +} + +bool dt_message_role_is_initiator(Dnstap__Message__Type type) +{ + switch (type) { + case DNSTAP__MESSAGE__TYPE__AUTH_QUERY: + case DNSTAP__MESSAGE__TYPE__AUTH_RESPONSE: + case DNSTAP__MESSAGE__TYPE__CLIENT_QUERY: + case DNSTAP__MESSAGE__TYPE__CLIENT_RESPONSE: + return false; + default: + return true; + } +} diff --git a/src/contrib/dnstap/convert.h b/src/contrib/dnstap/convert.h new file mode 100644 index 0000000..bf00ffb --- /dev/null +++ b/src/contrib/dnstap/convert.h @@ -0,0 +1,59 @@ +/* Copyright (C) 2018 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ +/*! + * \brief Dnstap identifiers conversions. + */ + +#pragma once + +#include <stdbool.h> + +#include "contrib/dnstap/dnstap.pb-c.h" + +/*! + * \brief Get Dnstap socket family from the real one. + */ +Dnstap__SocketFamily dt_family_encode(int family); + +/*! + * \brief Get real socket family from the Dnstap one. + */ +int dt_family_decode(Dnstap__SocketFamily dnstap_family); + +/*! + * \brief Get Dnstap protocol from a real one. + */ +Dnstap__SocketProtocol dt_protocol_encode(int protocol); + +/*! + * \brief Get real protocol from the Dnstap one. + */ +int dt_protocol_decode(Dnstap__SocketProtocol dnstap_protocol); + +/*! + * Check if a message type is any type of a query. + */ +bool dt_message_type_is_query(Dnstap__Message__Type type); + +/*! + * Check if a message type is any type of a response. + */ +bool dt_message_type_is_response(Dnstap__Message__Type type); + +/*! + * Check if a message role is any type of an initiator. + */ +bool dt_message_role_is_initiator(Dnstap__Message__Type type); diff --git a/src/contrib/dnstap/dnstap.c b/src/contrib/dnstap/dnstap.c new file mode 100644 index 0000000..9e24eac --- /dev/null +++ b/src/contrib/dnstap/dnstap.c @@ -0,0 +1,40 @@ +/* Copyright (C) 2014 Farsight Security, Inc. <software@farsightsecurity.com> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ + +#include <stdint.h> +#include <stdlib.h> + +#include "contrib/dnstap/dnstap.pb-c.h" + +#define DNSTAP_INITIAL_BUF_SIZE 256 + +uint8_t* dt_pack(const Dnstap__Dnstap *d, uint8_t **buf, size_t *sz) +{ + ProtobufCBufferSimple sbuf = { { NULL } }; + + sbuf.base.append = protobuf_c_buffer_simple_append; + sbuf.len = 0; + sbuf.alloced = DNSTAP_INITIAL_BUF_SIZE; + sbuf.data = malloc(sbuf.alloced); + if (sbuf.data == NULL) { + return NULL; + } + sbuf.must_free_data = 1; + + *sz = dnstap__dnstap__pack_to_buffer(d, (ProtobufCBuffer *) &sbuf); + *buf = sbuf.data; + return *buf; +} diff --git a/src/contrib/dnstap/dnstap.h b/src/contrib/dnstap/dnstap.h new file mode 100644 index 0000000..41e5f65 --- /dev/null +++ b/src/contrib/dnstap/dnstap.h @@ -0,0 +1,46 @@ +/* Copyright (C) 2014 Farsight Security, Inc. <software@farsightsecurity.com> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ +/*! + * \author Robert Edmonds <edmonds@fsi.io> + * + * \brief Public interface for dnstap. + */ + +#pragma once + +#include <stddef.h> +#include <stdint.h> + +#include "contrib/dnstap/dnstap.pb-c.h" + +/*! \brief Frame Streams "Content Type" value for dnstap. */ +#define DNSTAP_CONTENT_TYPE "protobuf:dnstap.Dnstap" + +/*! + * \brief Serializes a filled out dnstap protobuf struct. Dynamically allocates + * storage for the serialized frame. + * + * \note This function returns a copy of its parameter return value 'buf' to + * make error checking slightly easier. + * + * \param d dnstap protobuf struct. + * \param[out] buf Serialized frame. + * \param[out] sz Size in bytes of the serialized frame. + * + * \return Serialized frame. + * \retval NULL if error. + */ +uint8_t* dt_pack(const Dnstap__Dnstap *d, uint8_t **buf, size_t *sz); diff --git a/src/contrib/dnstap/dnstap.proto b/src/contrib/dnstap/dnstap.proto new file mode 100644 index 0000000..ea5c77a --- /dev/null +++ b/src/contrib/dnstap/dnstap.proto @@ -0,0 +1,270 @@ +// dnstap: flexible, structured event replication format for DNS software +// +// This file contains the protobuf schemas for the "dnstap" structured event +// replication format for DNS software. + +// Written in 2013-2014 by Farsight Security, Inc. +// +// To the extent possible under law, the author(s) have dedicated all +// copyright and related and neighboring rights to this file to the public +// domain worldwide. This file is distributed without any warranty. +// +// You should have received a copy of the CC0 Public Domain Dedication along +// with this file. If not, see: +// +// <http://creativecommons.org/publicdomain/zero/1.0/>. + +syntax = "proto2"; + +package dnstap; + +// "Dnstap": this is the top-level dnstap type, which is a "union" type that +// contains other kinds of dnstap payloads, although currently only one type +// of dnstap payload is defined. +// See: https://developers.google.com/protocol-buffers/docs/techniques#union +message Dnstap { + // DNS server identity. + // If enabled, this is the identity string of the DNS server which generated + // this message. Typically this would be the same string as returned by an + // "NSID" (RFC 5001) query. + optional bytes identity = 1; + + // DNS server version. + // If enabled, this is the version string of the DNS server which generated + // this message. Typically this would be the same string as returned by a + // "version.bind" query. + optional bytes version = 2; + + // Extra data for this payload. + // This field can be used for adding an arbitrary byte-string annotation to + // the payload. No encoding or interpretation is applied or enforced. + optional bytes extra = 3; + + // Identifies which field below is filled in. + enum Type { + MESSAGE = 1; + } + required Type type = 15; + + // One of the following will be filled in. + optional Message message = 14; +} + +// SocketFamily: the network protocol family of a socket. This specifies how +// to interpret "network address" fields. +enum SocketFamily { + INET = 1; // IPv4 (RFC 791) + INET6 = 2; // IPv6 (RFC 2460) +} + +// SocketProtocol: the transport protocol of a socket. This specifies how to +// interpret "transport port" fields. +enum SocketProtocol { + UDP = 1; // User Datagram Protocol (RFC 768) + TCP = 2; // Transmission Control Protocol (RFC 793) +} + +// Message: a wire-format (RFC 1035 section 4) DNS message and associated +// metadata. Applications generating "Message" payloads should follow +// certain requirements based on the MessageType, see below. +message Message { + + // There are eight types of "Message" defined that correspond to the + // four arrows in the following diagram, slightly modified from RFC 1035 + // section 2: + + // +---------+ +----------+ +--------+ + // | | query | | query | | + // | Stub |-SQ--------CQ->| Recursive|-RQ----AQ->| Auth. | + // | Resolver| | Server | | Name | + // | |<-SR--------CR-| |<-RR----AR-| Server | + // +---------+ response | | response | | + // +----------+ +--------+ + + // Each arrow has two Type values each, one for each "end" of each arrow, + // because these are considered to be distinct events. Each end of each + // arrow on the diagram above has been marked with a two-letter Type + // mnemonic. Clockwise from upper left, these mnemonic values are: + // + // SQ: STUB_QUERY + // CQ: CLIENT_QUERY + // RQ: RESOLVER_QUERY + // AQ: AUTH_QUERY + // AR: AUTH_RESPONSE + // RR: RESOLVER_RESPONSE + // CR: CLIENT_RESPONSE + // SR: STUB_RESPONSE + + // Two additional types of "Message" have been defined for the + // "forwarding" case where an upstream DNS server is responsible for + // further recursion. These are not shown on the diagram above, but have + // the following mnemonic values: + + // FQ: FORWARDER_QUERY + // FR: FORWARDER_RESPONSE + + // The "Message" Type values are defined below. + + enum Type { + // AUTH_QUERY is a DNS query message received from a resolver by an + // authoritative name server, from the perspective of the authoritative + // name server. + AUTH_QUERY = 1; + + // AUTH_RESPONSE is a DNS response message sent from an authoritative + // name server to a resolver, from the perspective of the authoritative + // name server. + AUTH_RESPONSE = 2; + + // RESOLVER_QUERY is a DNS query message sent from a resolver to an + // authoritative name server, from the perspective of the resolver. + // Resolvers typically clear the RD (recursion desired) bit when + // sending queries. + RESOLVER_QUERY = 3; + + // RESOLVER_RESPONSE is a DNS response message received from an + // authoritative name server by a resolver, from the perspective of + // the resolver. + RESOLVER_RESPONSE = 4; + + // CLIENT_QUERY is a DNS query message sent from a client to a DNS + // server which is expected to perform further recursion, from the + // perspective of the DNS server. The client may be a stub resolver or + // forwarder or some other type of software which typically sets the RD + // (recursion desired) bit when querying the DNS server. The DNS server + // may be a simple forwarding proxy or it may be a full recursive + // resolver. + CLIENT_QUERY = 5; + + // CLIENT_RESPONSE is a DNS response message sent from a DNS server to + // a client, from the perspective of the DNS server. The DNS server + // typically sets the RA (recursion available) bit when responding. + CLIENT_RESPONSE = 6; + + // FORWARDER_QUERY is a DNS query message sent from a downstream DNS + // server to an upstream DNS server which is expected to perform + // further recursion, from the perspective of the downstream DNS + // server. + FORWARDER_QUERY = 7; + + // FORWARDER_RESPONSE is a DNS response message sent from an upstream + // DNS server performing recursion to a downstream DNS server, from the + // perspective of the downstream DNS server. + FORWARDER_RESPONSE = 8; + + // STUB_QUERY is a DNS query message sent from a stub resolver to a DNS + // server, from the perspective of the stub resolver. + STUB_QUERY = 9; + + // STUB_RESPONSE is a DNS response message sent from a DNS server to a + // stub resolver, from the perspective of the stub resolver. + STUB_RESPONSE = 10; + + // TOOL_QUERY is a DNS query message sent from a DNS software tool to a + // DNS server, from the perspective of the tool. + TOOL_QUERY = 11; + + // TOOL_RESPONSE is a DNS response message received by a DNS software + // tool from a DNS server, from the perspective of the tool. + TOOL_RESPONSE = 12; + } + + // One of the Type values described above. + required Type type = 1; + + // One of the SocketFamily values described above. + optional SocketFamily socket_family = 2; + + // One of the SocketProtocol values described above. + optional SocketProtocol socket_protocol = 3; + + // The network address of the message initiator. + // For SocketFamily INET, this field is 4 octets (IPv4 address). + // For SocketFamily INET6, this field is 16 octets (IPv6 address). + optional bytes query_address = 4; + + // The network address of the message responder. + // For SocketFamily INET, this field is 4 octets (IPv4 address). + // For SocketFamily INET6, this field is 16 octets (IPv6 address). + optional bytes response_address = 5; + + // The transport port of the message initiator. + // This is a 16-bit UDP or TCP port number, depending on SocketProtocol. + optional uint32 query_port = 6; + + // The transport port of the message responder. + // This is a 16-bit UDP or TCP port number, depending on SocketProtocol. + optional uint32 response_port = 7; + + // The time at which the DNS query message was sent or received, depending + // on whether this is an AUTH_QUERY, RESOLVER_QUERY, or CLIENT_QUERY. + // This is the number of seconds since the UNIX epoch. + optional uint64 query_time_sec = 8; + + // The time at which the DNS query message was sent or received. + // This is the seconds fraction, expressed as a count of nanoseconds. + optional fixed32 query_time_nsec = 9; + + // The initiator's original wire-format DNS query message, verbatim. + optional bytes query_message = 10; + + // The "zone" or "bailiwick" pertaining to the DNS query message. + // This is a wire-format DNS domain name. + optional bytes query_zone = 11; + + // The time at which the DNS response message was sent or received, + // depending on whether this is an AUTH_RESPONSE, RESOLVER_RESPONSE, or + // CLIENT_RESPONSE. + // This is the number of seconds since the UNIX epoch. + optional uint64 response_time_sec = 12; + + // The time at which the DNS response message was sent or received. + // This is the seconds fraction, expressed as a count of nanoseconds. + optional fixed32 response_time_nsec = 13; + + // The responder's original wire-format DNS response message, verbatim. + optional bytes response_message = 14; +} + +// All fields except for 'type' in the Message schema are optional. +// It is recommended that at least the following fields be filled in for +// particular types of Messages. + +// AUTH_QUERY: +// socket_family, socket_protocol +// query_address, query_port +// query_message +// query_time_sec, query_time_nsec + +// AUTH_RESPONSE: +// socket_family, socket_protocol +// query_address, query_port +// query_time_sec, query_time_nsec +// response_message +// response_time_sec, response_time_nsec + +// RESOLVER_QUERY: +// socket_family, socket_protocol +// query_message +// query_time_sec, query_time_nsec +// query_zone +// response_address, response_port + +// RESOLVER_RESPONSE: +// socket_family, socket_protocol +// query_time_sec, query_time_nsec +// query_zone +// response_address, response_port +// response_message +// response_time_sec, response_time_nsec + +// CLIENT_QUERY: +// socket_family, socket_protocol +// query_message +// query_time_sec, query_time_nsec + +// CLIENT_RESPONSE: +// socket_family, socket_protocol +// query_time_sec, query_time_nsec +// response_message +// response_time_sec, response_time_nsec diff --git a/src/contrib/dnstap/message.c b/src/contrib/dnstap/message.c new file mode 100644 index 0000000..0b41e9d --- /dev/null +++ b/src/contrib/dnstap/message.c @@ -0,0 +1,130 @@ +/* Copyright (C) 2017 Farsight Security, Inc. <software@farsightsecurity.com> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ + +#include <netinet/in.h> +#include <stdbool.h> +#include <stdint.h> +#include <stdlib.h> +#include <string.h> + +#include "libknot/errcode.h" + +#include "contrib/dnstap/convert.h" +#include "contrib/dnstap/message.h" + +static void set_address(const struct sockaddr *sockaddr, + ProtobufCBinaryData *addr, + protobuf_c_boolean *has_addr, + uint32_t *port, + protobuf_c_boolean *has_port) +{ + if (sockaddr == NULL) { + *has_addr = 0; + *has_port = 0; + return; + } + + *has_addr = 1; + *has_port = 1; + + if (sockaddr->sa_family == AF_INET) { + const struct sockaddr_in *sai; + sai = (const struct sockaddr_in *)sockaddr; + addr->len = sizeof(sai->sin_addr); + addr->data = (uint8_t *)&sai->sin_addr.s_addr; + *port = ntohs(sai->sin_port); + } else if (sockaddr->sa_family == AF_INET6) { + const struct sockaddr_in6 *sai6; + sai6 = (const struct sockaddr_in6 *)sockaddr; + addr->len = sizeof(sai6->sin6_addr); + addr->data = (uint8_t *)&sai6->sin6_addr.s6_addr; + *port = ntohs(sai6->sin6_port); + } +} + +static int get_family(const struct sockaddr *query_sa, + const struct sockaddr *response_sa) +{ + const struct sockaddr *source = query_sa ? query_sa : response_sa; + if (source == NULL) { + return 0; + } + + return dt_family_encode(source->sa_family); +} + +int dt_message_fill(Dnstap__Message *m, + const Dnstap__Message__Type type, + const struct sockaddr *query_sa, + const struct sockaddr *response_sa, + const int protocol, + const void *wire, + const size_t len_wire, + const struct timespec *mtime) +{ + if (m == NULL) { + return KNOT_EINVAL; + } + + memset(m, 0, sizeof(*m)); + + m->base.descriptor = &dnstap__message__descriptor; + + // Message.type + m->type = type; + + // Message.socket_family + m->socket_family = get_family(query_sa, response_sa); + m->has_socket_family = m->socket_family != 0; + + // Message.socket_protocol + m->socket_protocol = dt_protocol_encode(protocol); + m->has_socket_protocol = m->socket_protocol != 0; + + // Message addresses + set_address(query_sa, &m->query_address, &m->has_query_address, + &m->query_port, &m->has_query_port); + set_address(response_sa, &m->response_address, &m->has_response_address, + &m->response_port, &m->has_response_port); + + if (dt_message_type_is_query(type)) { + // Message.query_message + m->query_message.len = len_wire; + m->query_message.data = (uint8_t *)wire; + m->has_query_message = 1; + // Message.query_time_sec, Message.query_time_nsec + if (mtime != NULL) { + m->query_time_sec = mtime->tv_sec; + m->query_time_nsec = mtime->tv_nsec; + m->has_query_time_sec = 1; + m->has_query_time_nsec = 1; + } + } else if (dt_message_type_is_response(type)) { + // Message.response_message + m->response_message.len = len_wire; + m->response_message.data = (uint8_t *)wire; + m->has_response_message = 1; + // Message.response_time_sec, Message.response_time_nsec + if (mtime != NULL) { + m->response_time_sec = mtime->tv_sec; + m->response_time_nsec = mtime->tv_nsec; + m->has_response_time_sec = 1; + m->has_response_time_nsec = 1; + } + } + + return KNOT_EOK; +} diff --git a/src/contrib/dnstap/message.h b/src/contrib/dnstap/message.h new file mode 100644 index 0000000..11a967d --- /dev/null +++ b/src/contrib/dnstap/message.h @@ -0,0 +1,62 @@ +/* Copyright (C) 2017 Farsight Security, Inc. <software@farsightsecurity.com> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ +/*! + * \author Robert Edmonds <edmonds@fsi.io> + * + * \brief Dnstap message interface. + */ + +#pragma once + +#include <sys/socket.h> +#include <sys/time.h> +#include <stddef.h> + +#include "contrib/dnstap/dnstap.pb-c.h" + +/*! + * \brief Fill a Dnstap__Message structure with the given parameters. + * + * \param[out] m + * Dnstap__Message structure to fill. Will be zeroed first. + * \param type + * One of the DNSTAP__MESSAGE__TYPE__* values. + * \param query_sa + * sockaddr_in or sockaddr_in6 to use when filling the 'socket_family', + * 'query_address', 'query_port' fields. + * \param response_sa + * sockaddr_in or sockaddr_in6 to use when filling the 'socket_family', + * 'response_address', 'response_port' fields. + * \param protocol + * \c IPPROTO_UDP or \c IPPROTO_TCP. + * \param wire + * Wire-format query message or response message (depending on 'type'). + * \param len_wire + * Length in bytes of 'wire'. + * \param mtime + * Message time. May be NULL. + * + * \retval KNOT_EOK + * \retval KNOT_EINVAL + */ +int dt_message_fill(Dnstap__Message *m, + const Dnstap__Message__Type type, + const struct sockaddr *query_sa, + const struct sockaddr *response_sa, + const int protocol, + const void *wire, + const size_t len_wire, + const struct timespec *mtime); diff --git a/src/contrib/dnstap/reader.c b/src/contrib/dnstap/reader.c new file mode 100644 index 0000000..593c6ec --- /dev/null +++ b/src/contrib/dnstap/reader.c @@ -0,0 +1,105 @@ +/* Copyright (C) 2014 Farsight Security, Inc. <software@farsightsecurity.com> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ + +#include <stdint.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> + +#include "libknot/errcode.h" + +#include "contrib/macros.h" +#include "contrib/dnstap/dnstap.h" +#include "contrib/dnstap/reader.h" + +dt_reader_t* dt_reader_create(const char *file_path) +{ + struct fstrm_file_options *fopt = NULL; + struct fstrm_reader_options *ropt = NULL; + dt_reader_t *reader = NULL; + fstrm_res res; + + reader = calloc(1, sizeof(dt_reader_t)); + if (reader == NULL) { + goto fail; + } + + // Open reader. + fopt = fstrm_file_options_init(); + fstrm_file_options_set_file_path(fopt, file_path); + ropt = fstrm_reader_options_init(); + fstrm_reader_options_add_content_type(ropt, + (const uint8_t *) DNSTAP_CONTENT_TYPE, + strlen(DNSTAP_CONTENT_TYPE)); + reader->fr = fstrm_file_reader_init(fopt, ropt); + fstrm_file_options_destroy(&fopt); + fstrm_reader_options_destroy(&ropt); + if (reader->fr == NULL) { + goto fail; + } + res = fstrm_reader_open(reader->fr); + if (res != fstrm_res_success) { + goto fail; + } + + return reader; +fail: + dt_reader_free(reader); + return NULL; +} + +void dt_reader_free(dt_reader_t *reader) +{ + if (reader == NULL) { + return; + } + + fstrm_reader_destroy(&reader->fr); + free(reader); +} + +int dt_reader_read(dt_reader_t *reader, Dnstap__Dnstap **d) +{ + fstrm_res res; + const uint8_t *data = NULL; + size_t len = 0; + + res = fstrm_reader_read(reader->fr, &data, &len); + if (res == fstrm_res_success) { + *d = dnstap__dnstap__unpack(NULL, len, data); + if (*d == NULL) { + return KNOT_ENOMEM; + } + } else if (res == fstrm_res_failure) { + return KNOT_ERROR; + } else if (res == fstrm_res_stop) { + return KNOT_EOF; + } + + return KNOT_EOK; +} + +void dt_reader_free_frame(dt_reader_t *reader, Dnstap__Dnstap **frame_ptr) +{ + if (!*frame_ptr) { + return; + } + + UNUSED(reader); + + dnstap__dnstap__free_unpacked(*frame_ptr, NULL); + *frame_ptr = NULL; +} diff --git a/src/contrib/dnstap/reader.h b/src/contrib/dnstap/reader.h new file mode 100644 index 0000000..c8cda26 --- /dev/null +++ b/src/contrib/dnstap/reader.h @@ -0,0 +1,72 @@ +/* Copyright (C) 2017 Farsight Security, Inc. <software@farsightsecurity.com> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ +/*! + * \brief Dnstap file reader. + */ + +#pragma once + +#include <fstrm.h> +#include <protobuf-c/protobuf-c.h> + +#include "contrib/dnstap/dnstap.pb-c.h" + +/*! \brief Structure for dnstap file reader. */ +typedef struct { + /*!< Input reader. */ + struct fstrm_reader *fr; +} dt_reader_t; + +/*! + * \brief Creates dnstap file reader structure. + * + * \param file_path Name of file to read input from. + * + * \retval reader if success. + * \retval NULL if error. + */ +dt_reader_t* dt_reader_create(const char *file_path); + +/*! + * \brief Close dnstap file reader. + * + * \param reader dnstap file reader structure. + */ +void dt_reader_free(dt_reader_t *reader); + +/*! + * \brief Read a dnstap protobuf from a dnstap file reader. + * + * Caller must deallocate the returned protobuf with the + * dnstap__dnstap__free_unpacked() function. + * + * \param[in] reader dnstap file reader structure. + * \param[out] d Unpacked dnstap protobuf. + * + * \retval KNOT_EOK + * \retval KNOT_ERROR + * \retval KNOT_EOF + * \retval KNOT_ENOMEM + */ +int dt_reader_read(dt_reader_t *reader, Dnstap__Dnstap **d); + +/*! + * \brief free the frame allocated by dt_read_data. + * + * \param reader Dnstap reader context. + * \param d The frame to be freed. + */ +void dt_reader_free_frame(dt_reader_t *reader, Dnstap__Dnstap **d); diff --git a/src/contrib/dnstap/writer.c b/src/contrib/dnstap/writer.c new file mode 100644 index 0000000..08f4519 --- /dev/null +++ b/src/contrib/dnstap/writer.c @@ -0,0 +1,120 @@ +/* Copyright (C) 2014 Farsight Security, Inc. <software@farsightsecurity.com> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ + +#include <stdint.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> + +#include "libknot/errcode.h" + +#include "contrib/dnstap/dnstap.h" +#include "contrib/dnstap/writer.h" + +dt_writer_t* dt_writer_create(const char *file_path, const char *version) +{ + struct fstrm_file_options *fopt = NULL; + struct fstrm_writer_options *wopt = NULL; + dt_writer_t *writer = NULL; + fstrm_res res; + + writer = calloc(1, sizeof(dt_writer_t)); + if (writer == NULL) { + goto fail; + } + + // Set "version". + if (version != NULL) { + writer->len_version = strlen(version); + writer->version = strdup(version); + if (!writer->version) { + goto fail; + } + } + + // Open writer. + fopt = fstrm_file_options_init(); + fstrm_file_options_set_file_path(fopt, file_path); + wopt = fstrm_writer_options_init(); + fstrm_writer_options_add_content_type(wopt, + (const uint8_t *) DNSTAP_CONTENT_TYPE, + strlen(DNSTAP_CONTENT_TYPE)); + writer->fw = fstrm_file_writer_init(fopt, wopt); + fstrm_file_options_destroy(&fopt); + fstrm_writer_options_destroy(&wopt); + if (writer->fw == NULL) { + goto fail; + } + + res = fstrm_writer_open(writer->fw); + if (res != fstrm_res_success) { + goto fail; + } + + return writer; +fail: + dt_writer_free(writer); + return NULL; +} + +void dt_writer_free(dt_writer_t *writer) +{ + if (writer == NULL) { + return; + } + + fstrm_writer_destroy(&writer->fw); + free(writer->version); + free(writer); +} + +int dt_writer_write(dt_writer_t *writer, const ProtobufCMessage *msg) +{ + Dnstap__Dnstap dnstap = DNSTAP__DNSTAP__INIT; + size_t len; + uint8_t *data; + + if (writer->fw == NULL) { + return KNOT_EOK; + } + + // Only handle dnstap/Message. + assert(msg->descriptor == &dnstap__message__descriptor); + + // Fill out 'dnstap'. + if (writer->version) { + dnstap.version.data = writer->version; + dnstap.version.len = writer->len_version; + dnstap.has_version = 1; + } + dnstap.type = DNSTAP__DNSTAP__TYPE__MESSAGE; + dnstap.message = (Dnstap__Message *)msg; + + // Serialize the dnstap frame. + if (!dt_pack(&dnstap, &data, &len)) { + return KNOT_ENOMEM; + } + + // Write the dnstap frame to the output stream. + if (fstrm_writer_write(writer->fw, data, len) != fstrm_res_success) { + return KNOT_ERROR; + } + + // Cleanup. + free(data); + + return KNOT_EOK; +} diff --git a/src/contrib/dnstap/writer.h b/src/contrib/dnstap/writer.h new file mode 100644 index 0000000..4fe7326 --- /dev/null +++ b/src/contrib/dnstap/writer.h @@ -0,0 +1,70 @@ +/* Copyright (C) 2014 Farsight Security, Inc. <software@farsightsecurity.com> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ +/*! + * \author Robert Edmonds <edmonds@fsi.io> + * + * \brief Dnstap file writer. + */ + +#pragma once + +#include <fstrm.h> +#include <protobuf-c/protobuf-c.h> + +/*! \brief Structure for dnstap file writer. */ +typedef struct { + /*!< Output writer. */ + struct fstrm_writer *fw; + + /*!< dnstap "version" field. */ + void *version; + + /*!< length of dnstap "version" field. */ + size_t len_version; +} dt_writer_t; + +/*! + * \brief Creates dnstap file writer structure. + * + * \param file_path Name of file to write output to. + * \param version Version string of software. May be NULL. + * + * \retval writer if success. + * \retval NULL if error. + */ +dt_writer_t* dt_writer_create(const char *file_path, const char *version); + +/*! + * \brief Finish writing dnstap file writer and free resources. + * + * \param writer dnstap file writer structure. + */ +void dt_writer_free(dt_writer_t *writer); + +/*! + * \brief Write a protobuf to the dnstap file writer. + * + * Supported protobuf types for the 'msg' parameter: + * \c Dnstap__Message + * + * \param writer dnstap file writer structure. + * \param msg dnstap protobuf. Must be a supported type. + * + * \retval KNOT_EOK + * \retval KNOT_EINVAL + * \retval KNOT_ENOMEM + */ +int dt_writer_write(dt_writer_t *writer, const ProtobufCMessage *msg); diff --git a/src/contrib/dynarray.h b/src/contrib/dynarray.h new file mode 100644 index 0000000..79eab31 --- /dev/null +++ b/src/contrib/dynarray.h @@ -0,0 +1,123 @@ +/* Copyright (C) 2018 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ +/*! + * \brief Simple write-once allocation-optimal dynamic array. + * + * Include it into your .c file + * + * prefix - identifier prefix, e.g. ptr -> struct ptr_dynarray, ptr_dynarray_add(), ... + * ntype - data type to be stored. Let it be a number, pointer or small struct + * initial_capacity - how many data items will be allocated on stac and copied with assignment + * + * prefix_dynarray_add() - add a data item + * prefix_dynarray_fix() - call EVERYTIME the array is copied from some already invalid stack + * prefix_dynarray_free() - call EVERYTIME you dismiss all copies of the array + * + */ + +#include <stdlib.h> +#include <assert.h> + +#pragma once + +#define DYNARRAY_VISIBILITY_STATIC static +#define DYNARRAY_VISIBILITY_PUBLIC +#define DYNARRAY_VISIBILITY_LIBRARY __public__ + +#define dynarray_declare(prefix, ntype, visibility, initial_capacity) \ + typedef struct prefix ## _dynarray { \ + ssize_t capacity; \ + ssize_t size; \ + ntype *(*arr)(struct prefix ## _dynarray *dynarray); \ + ntype init[initial_capacity]; \ + ntype *_arr; \ + } prefix ## _dynarray_t; \ + \ + visibility ntype *prefix ## _dynarray_arr(prefix ## _dynarray_t *dynarray); \ + visibility void prefix ## _dynarray_add(prefix ## _dynarray_t *dynarray, \ + ntype const *to_add); \ + visibility void prefix ## _dynarray_free(prefix ## _dynarray_t *dynarray); + +#define dynarray_foreach(prefix, ntype, ptr, array) \ + for (ntype *ptr = prefix ## _dynarray_arr(&(array)); \ + ptr < prefix ## _dynarray_arr(&(array)) + (array).size; ptr++) + +#define dynarray_define(prefix, ntype, visibility) \ + \ + static void prefix ## _dynarray_free__(struct prefix ## _dynarray *dynarray) \ + { \ + if (dynarray->capacity > sizeof(dynarray->init) / sizeof(*dynarray->init)) { \ + free(dynarray->_arr); \ + } \ + } \ + \ + __attribute__((unused)) \ + visibility ntype *prefix ## _dynarray_arr(struct prefix ## _dynarray *dynarray) \ + { \ + assert(dynarray->size <= dynarray->capacity); \ + return (dynarray->capacity <= sizeof(dynarray->init) / sizeof(*dynarray->init) ? \ + dynarray->init : dynarray->_arr); \ + } \ + \ + static ntype *prefix ## _dynarray_arr_init__(struct prefix ## _dynarray *dynarray) \ + { \ + assert(dynarray->capacity == sizeof(dynarray->init) / sizeof(*dynarray->init)); \ + return dynarray->init; \ + } \ + \ + static ntype *prefix ## _dynarray_arr_arr__(struct prefix ## _dynarray *dynarray) \ + { \ + assert(dynarray->capacity > sizeof(dynarray->init) / sizeof(*dynarray->init)); \ + return dynarray->_arr; \ + } \ + \ + __attribute__((unused)) \ + visibility void prefix ## _dynarray_add(struct prefix ## _dynarray *dynarray, \ + ntype const *to_add) \ + { \ + if (dynarray->capacity < 0) { \ + return; \ + } \ + if (dynarray->capacity == 0) { \ + dynarray->capacity = sizeof(dynarray->init) / sizeof(*dynarray->init); \ + dynarray->arr = prefix ## _dynarray_arr_init__; \ + } \ + if (dynarray->size >= dynarray->capacity) { \ + ssize_t new_capacity = dynarray->capacity * 2 + 1; \ + ntype *new_arr = calloc(new_capacity, sizeof(ntype)); \ + if (new_arr == NULL) { \ + prefix ## _dynarray_free__(dynarray); \ + dynarray->capacity = dynarray->size = -1; \ + return; \ + } \ + if (dynarray->capacity > 0) { \ + memcpy(new_arr, prefix ## _dynarray_arr(dynarray), \ + dynarray->capacity * sizeof(ntype)); \ + } \ + prefix ## _dynarray_free__(dynarray); \ + dynarray->_arr = new_arr; \ + dynarray->capacity = new_capacity; \ + dynarray->arr = prefix ## _dynarray_arr_arr__; \ + } \ + prefix ## _dynarray_arr(dynarray)[dynarray->size++] = *to_add; \ + } \ + \ + __attribute__((unused)) \ + visibility void prefix ## _dynarray_free(struct prefix ## _dynarray *dynarray) \ + { \ + prefix ## _dynarray_free__(dynarray); \ + memset(dynarray, 0, sizeof(*dynarray)); \ + } diff --git a/src/contrib/files.c b/src/contrib/files.c new file mode 100644 index 0000000..3181386 --- /dev/null +++ b/src/contrib/files.c @@ -0,0 +1,134 @@ +/* Copyright (C) 2016 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. +*/ + +#include <assert.h> +#include <dirent.h> +#include <fcntl.h> +#include <ftw.h> +#include <stdlib.h> +#include <string.h> +#include <sys/stat.h> +#include <unistd.h> + +#include "contrib/files.h" +#include "contrib/string.h" +#include "libknot/errcode.h" + +static int remove_file(const char *path, const struct stat *stat, int type, struct FTW *ftw) +{ + (void)stat; + (void)ftw; + if (type == FTW_DP) { + return rmdir(path); + } else { + return unlink(path); + } +} + +bool remove_path(const char *path) +{ + return (0 == nftw(path, remove_file, 1, FTW_DEPTH | FTW_PHYS)); +} + +int make_dir(const char *path, mode_t mode, bool ignore_existing) +{ + if (mkdir(path, mode) == 0) { + return KNOT_EOK; + } + + if (!ignore_existing || errno != EEXIST) { + return knot_map_errno(); + } + + assert(errno == EEXIST); + + struct stat st = { 0 }; + if (stat(path, &st) != 0) { + return knot_map_errno(); + } + + if (!S_ISDIR(st.st_mode)) { + return knot_map_errno_code(ENOTDIR); + } + + return KNOT_EOK; +} + +int make_path(const char *path, mode_t mode) +{ + if (path == NULL) { + return KNOT_EINVAL; + } + + char *dir = strdup(path); + if (dir == NULL) { + return KNOT_ENOMEM; + } + + for (char *p = strchr(dir + 1, '/'); p != NULL; p = strchr(p + 1, '/')) { + *p = '\0'; + if (mkdir(dir, mode) == -1 && errno != EEXIST) { + free(dir); + return knot_map_errno(); + } + *p = '/'; + } + + free(dir); + + return KNOT_EOK; +} + +int open_tmp_file(const char *path, char **tmp_name, FILE **file, mode_t mode) +{ + int ret; + + *tmp_name = sprintf_alloc("%s.XXXXXX", path); + if (*tmp_name == NULL) { + ret = KNOT_ENOMEM; + goto open_tmp_failed; + } + + int fd = mkstemp(*tmp_name); + if (fd < 0) { + ret = knot_map_errno(); + goto open_tmp_failed; + } + + if (fchmod(fd, mode) != 0) { + ret = knot_map_errno(); + close(fd); + unlink(*tmp_name); + goto open_tmp_failed; + } + + *file = fdopen(fd, "w"); + if (*file == NULL) { + ret = knot_map_errno(); + close(fd); + unlink(*tmp_name); + goto open_tmp_failed; + } + + return KNOT_EOK; +open_tmp_failed: + free(*tmp_name); + *tmp_name = NULL; + *file = NULL; + + assert(ret != KNOT_EOK); + return ret; +} diff --git a/src/contrib/files.h b/src/contrib/files.h new file mode 100644 index 0000000..c4c681f --- /dev/null +++ b/src/contrib/files.h @@ -0,0 +1,43 @@ +/* Copyright (C) 2016 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. +*/ + +#pragma once + +#include <stdbool.h> +#include <stdio.h> +#include <sys/types.h> + +/*! + * \brief Delete file or directory (recursive). + * + * \return true on success, false when one or more files failed to be removed. + */ +bool remove_path(const char *path); + +/*! + * Equivalent to mkdir(2), can succeed if the directory already exists. + */ +int make_dir(const char *path, mode_t mode, bool ignore_existing); + +/*! + * Makes a directory part of the path with all parent directories if not exist. + */ +int make_path(const char *path, mode_t mode); + +/*! + * Creates and opens for writing a temporary file based on given path. + */ +int open_tmp_file(const char *path, char **tmp_name, FILE **file, mode_t mode); diff --git a/src/contrib/getline.c b/src/contrib/getline.c new file mode 100644 index 0000000..074c511 --- /dev/null +++ b/src/contrib/getline.c @@ -0,0 +1,60 @@ +/* Copyright (C) 2011 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. +*/ + +// FreeBSD POSIX2008 getline +#ifndef _WITH_GETLINE +#define _WITH_GETLINE +#endif + +#include "contrib/getline.h" + +#include <stdio.h> // getline or fgetln +#include <stdlib.h> // free +#include <string.h> // memcpy + +ssize_t knot_getline(char **lineptr, size_t *n, FILE *stream) +{ +#ifdef HAVE_GETLINE + return getline(lineptr, n, stream); +#else +#ifdef HAVE_FGETLN + size_t length = 0; + char *buffer = fgetln(stream, &length); + if (buffer == NULL) { + return -1; + } + + /* NOTE: Function fgetln doesn't return terminated string! + * Output buffer from the fgetln can't be freed. + */ + + // If the output buffer is not specified or is small, extend it. + if (*lineptr == NULL || *n <= length) { + char *tmp = realloc(*lineptr, length + 1); + if (tmp == NULL) { + return -1; + } + *lineptr = tmp; + *n = length + 1; + } + + memcpy(*lineptr, buffer, length); + (*lineptr)[length] = '\0'; + + return length; +#endif +#endif +} diff --git a/src/contrib/getline.h b/src/contrib/getline.h new file mode 100644 index 0000000..11c7fe3 --- /dev/null +++ b/src/contrib/getline.h @@ -0,0 +1,38 @@ +/* Copyright (C) 2018 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ +/*! + * \brief Multiplatform getline wrapper. + */ + +#pragma once + +#include <stdio.h> +#include <sys/types.h> + +/*! + * \brief Reads a line from a stream. + * + * This function has the same semantics as POSIX.1-2008 getline(). + * If necessary, the output buffer will be allocated/reallocated. + * + * \param lineptr Output buffer. + * \param n Output buffer size. + * \param stream Input stream. + * + * \retval Number of characters read, including new line delimiter, + * not including terminating. -1 on error or EOF. + */ +ssize_t knot_getline(char **lineptr, size_t *n, FILE *stream); diff --git a/src/contrib/licenses/0BSD b/src/contrib/licenses/0BSD new file mode 100644 index 0000000..56c5528 --- /dev/null +++ b/src/contrib/licenses/0BSD @@ -0,0 +1,12 @@ +Permission to use, copy, modify, and distribute this software for any +purpose with or without fee is hereby granted, provided that the above +copyright notice and this permission notice appear in all copies. + +THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES +WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF +MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR +ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES +WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN +ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF +OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. + diff --git a/src/contrib/licenses/BSD-3-Clause b/src/contrib/licenses/BSD-3-Clause new file mode 100644 index 0000000..8041f21 --- /dev/null +++ b/src/contrib/licenses/BSD-3-Clause @@ -0,0 +1,21 @@ +Redistribution and use in source and binary forms, with or without modification, +are permitted provided that the following conditions are met: +1. Redistributions of source code must retain the above copyright notice, this + list of conditions and the following disclaimer. +2. Redistributions in binary form must reproduce the above copyright notice, + this list of conditions and the following disclaimer in the documentation + and/or other materials provided with the distribution. +3. Neither the name of the copyright holder nor the names of its contributors + may be used to endorse or promote products derived from this software without + specific prior written permission. + +THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND +ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED +WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. +IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, +INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, +BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, +DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF +LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE +OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED +OF THE POSSIBILITY OF SUCH DAMAGE. diff --git a/src/contrib/licenses/LGPL-2.0 b/src/contrib/licenses/LGPL-2.0 new file mode 100644 index 0000000..d159169 --- /dev/null +++ b/src/contrib/licenses/LGPL-2.0 @@ -0,0 +1,339 @@ + GNU GENERAL PUBLIC LICENSE + Version 2, June 1991 + + Copyright (C) 1989, 1991 Free Software Foundation, Inc., + 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA + Everyone is permitted to copy and distribute verbatim copies + of this license document, but changing it is not allowed. + + Preamble + + The licenses for most software are designed to take away your +freedom to share and change it. By contrast, the GNU General Public +License is intended to guarantee your freedom to share and change free +software--to make sure the software is free for all its users. This +General Public License applies to most of the Free Software +Foundation's software and to any other program whose authors commit to +using it. (Some other Free Software Foundation software is covered by +the GNU Lesser General Public License instead.) You can apply it to +your programs, too. + + When we speak of free software, we are referring to freedom, not +price. Our General Public Licenses are designed to make sure that you +have the freedom to distribute copies of free software (and charge for +this service if you wish), that you receive source code or can get it +if you want it, that you can change the software or use pieces of it +in new free programs; and that you know you can do these things. + + To protect your rights, we need to make restrictions that forbid +anyone to deny you these rights or to ask you to surrender the rights. +These restrictions translate to certain responsibilities for you if you +distribute copies of the software, or if you modify it. + + For example, if you distribute copies of such a program, whether +gratis or for a fee, you must give the recipients all the rights that +you have. You must make sure that they, too, receive or can get the +source code. And you must show them these terms so they know their +rights. + + We protect your rights with two steps: (1) copyright the software, and +(2) offer you this license which gives you legal permission to copy, +distribute and/or modify the software. + + Also, for each author's protection and ours, we want to make certain +that everyone understands that there is no warranty for this free +software. If the software is modified by someone else and passed on, we +want its recipients to know that what they have is not the original, so +that any problems introduced by others will not reflect on the original +authors' reputations. + + Finally, any free program is threatened constantly by software +patents. We wish to avoid the danger that redistributors of a free +program will individually obtain patent licenses, in effect making the +program proprietary. To prevent this, we have made it clear that any +patent must be licensed for everyone's free use or not licensed at all. + + The precise terms and conditions for copying, distribution and +modification follow. + + GNU GENERAL PUBLIC LICENSE + TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION + + 0. This License applies to any program or other work which contains +a notice placed by the copyright holder saying it may be distributed +under the terms of this General Public License. The "Program", below, +refers to any such program or work, and a "work based on the Program" +means either the Program or any derivative work under copyright law: +that is to say, a work containing the Program or a portion of it, +either verbatim or with modifications and/or translated into another +language. (Hereinafter, translation is included without limitation in +the term "modification".) Each licensee is addressed as "you". + +Activities other than copying, distribution and modification are not +covered by this License; they are outside its scope. The act of +running the Program is not restricted, and the output from the Program +is covered only if its contents constitute a work based on the +Program (independent of having been made by running the Program). +Whether that is true depends on what the Program does. + + 1. You may copy and distribute verbatim copies of the Program's +source code as you receive it, in any medium, provided that you +conspicuously and appropriately publish on each copy an appropriate +copyright notice and disclaimer of warranty; keep intact all the +notices that refer to this License and to the absence of any warranty; +and give any other recipients of the Program a copy of this License +along with the Program. + +You may charge a fee for the physical act of transferring a copy, and +you may at your option offer warranty protection in exchange for a fee. + + 2. You may modify your copy or copies of the Program or any portion +of it, thus forming a work based on the Program, and copy and +distribute such modifications or work under the terms of Section 1 +above, provided that you also meet all of these conditions: + + a) You must cause the modified files to carry prominent notices + stating that you changed the files and the date of any change. + + b) You must cause any work that you distribute or publish, that in + whole or in part contains or is derived from the Program or any + part thereof, to be licensed as a whole at no charge to all third + parties under the terms of this License. + + c) If the modified program normally reads commands interactively + when run, you must cause it, when started running for such + interactive use in the most ordinary way, to print or display an + announcement including an appropriate copyright notice and a + notice that there is no warranty (or else, saying that you provide + a warranty) and that users may redistribute the program under + these conditions, and telling the user how to view a copy of this + License. (Exception: if the Program itself is interactive but + does not normally print such an announcement, your work based on + the Program is not required to print an announcement.) + +These requirements apply to the modified work as a whole. If +identifiable sections of that work are not derived from the Program, +and can be reasonably considered independent and separate works in +themselves, then this License, and its terms, do not apply to those +sections when you distribute them as separate works. But when you +distribute the same sections as part of a whole which is a work based +on the Program, the distribution of the whole must be on the terms of +this License, whose permissions for other licensees extend to the +entire whole, and thus to each and every part regardless of who wrote it. + +Thus, it is not the intent of this section to claim rights or contest +your rights to work written entirely by you; rather, the intent is to +exercise the right to control the distribution of derivative or +collective works based on the Program. + +In addition, mere aggregation of another work not based on the Program +with the Program (or with a work based on the Program) on a volume of +a storage or distribution medium does not bring the other work under +the scope of this License. + + 3. You may copy and distribute the Program (or a work based on it, +under Section 2) in object code or executable form under the terms of +Sections 1 and 2 above provided that you also do one of the following: + + a) Accompany it with the complete corresponding machine-readable + source code, which must be distributed under the terms of Sections + 1 and 2 above on a medium customarily used for software interchange; or, + + b) Accompany it with a written offer, valid for at least three + years, to give any third party, for a charge no more than your + cost of physically performing source distribution, a complete + machine-readable copy of the corresponding source code, to be + distributed under the terms of Sections 1 and 2 above on a medium + customarily used for software interchange; or, + + c) Accompany it with the information you received as to the offer + to distribute corresponding source code. (This alternative is + allowed only for noncommercial distribution and only if you + received the program in object code or executable form with such + an offer, in accord with Subsection b above.) + +The source code for a work means the preferred form of the work for +making modifications to it. For an executable work, complete source +code means all the source code for all modules it contains, plus any +associated interface definition files, plus the scripts used to +control compilation and installation of the executable. However, as a +special exception, the source code distributed need not include +anything that is normally distributed (in either source or binary +form) with the major components (compiler, kernel, and so on) of the +operating system on which the executable runs, unless that component +itself accompanies the executable. + +If distribution of executable or object code is made by offering +access to copy from a designated place, then offering equivalent +access to copy the source code from the same place counts as +distribution of the source code, even though third parties are not +compelled to copy the source along with the object code. + + 4. You may not copy, modify, sublicense, or distribute the Program +except as expressly provided under this License. Any attempt +otherwise to copy, modify, sublicense or distribute the Program is +void, and will automatically terminate your rights under this License. +However, parties who have received copies, or rights, from you under +this License will not have their licenses terminated so long as such +parties remain in full compliance. + + 5. You are not required to accept this License, since you have not +signed it. However, nothing else grants you permission to modify or +distribute the Program or its derivative works. These actions are +prohibited by law if you do not accept this License. Therefore, by +modifying or distributing the Program (or any work based on the +Program), you indicate your acceptance of this License to do so, and +all its terms and conditions for copying, distributing or modifying +the Program or works based on it. + + 6. Each time you redistribute the Program (or any work based on the +Program), the recipient automatically receives a license from the +original licensor to copy, distribute or modify the Program subject to +these terms and conditions. You may not impose any further +restrictions on the recipients' exercise of the rights granted herein. +You are not responsible for enforcing compliance by third parties to +this License. + + 7. If, as a consequence of a court judgment or allegation of patent +infringement or for any other reason (not limited to patent issues), +conditions are imposed on you (whether by court order, agreement or +otherwise) that contradict the conditions of this License, they do not +excuse you from the conditions of this License. If you cannot +distribute so as to satisfy simultaneously your obligations under this +License and any other pertinent obligations, then as a consequence you +may not distribute the Program at all. For example, if a patent +license would not permit royalty-free redistribution of the Program by +all those who receive copies directly or indirectly through you, then +the only way you could satisfy both it and this License would be to +refrain entirely from distribution of the Program. + +If any portion of this section is held invalid or unenforceable under +any particular circumstance, the balance of the section is intended to +apply and the section as a whole is intended to apply in other +circumstances. + +It is not the purpose of this section to induce you to infringe any +patents or other property right claims or to contest validity of any +such claims; this section has the sole purpose of protecting the +integrity of the free software distribution system, which is +implemented by public license practices. Many people have made +generous contributions to the wide range of software distributed +through that system in reliance on consistent application of that +system; it is up to the author/donor to decide if he or she is willing +to distribute software through any other system and a licensee cannot +impose that choice. + +This section is intended to make thoroughly clear what is believed to +be a consequence of the rest of this License. + + 8. If the distribution and/or use of the Program is restricted in +certain countries either by patents or by copyrighted interfaces, the +original copyright holder who places the Program under this License +may add an explicit geographical distribution limitation excluding +those countries, so that distribution is permitted only in or among +countries not thus excluded. In such case, this License incorporates +the limitation as if written in the body of this License. + + 9. The Free Software Foundation may publish revised and/or new versions +of the General Public License from time to time. Such new versions will +be similar in spirit to the present version, but may differ in detail to +address new problems or concerns. + +Each version is given a distinguishing version number. If the Program +specifies a version number of this License which applies to it and "any +later version", you have the option of following the terms and conditions +either of that version or of any later version published by the Free +Software Foundation. If the Program does not specify a version number of +this License, you may choose any version ever published by the Free Software +Foundation. + + 10. If you wish to incorporate parts of the Program into other free +programs whose distribution conditions are different, write to the author +to ask for permission. For software which is copyrighted by the Free +Software Foundation, write to the Free Software Foundation; we sometimes +make exceptions for this. Our decision will be guided by the two goals +of preserving the free status of all derivatives of our free software and +of promoting the sharing and reuse of software generally. + + NO WARRANTY + + 11. BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY +FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN +OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES +PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED +OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF +MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS +TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU. SHOULD THE +PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING, +REPAIR OR CORRECTION. + + 12. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING +WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR +REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, +INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING +OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED +TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY +YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER +PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE +POSSIBILITY OF SUCH DAMAGES. + + END OF TERMS AND CONDITIONS + + How to Apply These Terms to Your New Programs + + If you develop a new program, and you want it to be of the greatest +possible use to the public, the best way to achieve this is to make it +free software which everyone can redistribute and change under these terms. + + To do so, attach the following notices to the program. It is safest +to attach them to the start of each source file to most effectively +convey the exclusion of warranty; and each file should have at least +the "copyright" line and a pointer to where the full notice is found. + + <one line to give the program's name and a brief idea of what it does.> + Copyright (C) <year> <name of author> + + This program is free software; you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 2 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License along + with this program; if not, write to the Free Software Foundation, Inc., + 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. + +Also add information on how to contact you by electronic and paper mail. + +If the program is interactive, make it output a short notice like this +when it starts in an interactive mode: + + Gnomovision version 69, Copyright (C) year name of author + Gnomovision comes with ABSOLUTELY NO WARRANTY; for details type `show w'. + This is free software, and you are welcome to redistribute it + under certain conditions; type `show c' for details. + +The hypothetical commands `show w' and `show c' should show the appropriate +parts of the General Public License. Of course, the commands you use may +be called something other than `show w' and `show c'; they could even be +mouse-clicks or menu items--whatever suits your program. + +You should also get your employer (if you work as a programmer) or your +school, if any, to sign a "copyright disclaimer" for the program, if +necessary. Here is a sample; alter the names: + + Yoyodyne, Inc., hereby disclaims all copyright interest in the program + `Gnomovision' (which makes passes at compilers) written by James Hacker. + + <signature of Ty Coon>, 1 April 1989 + Ty Coon, President of Vice + +This General Public License does not permit incorporating your program into +proprietary programs. If your program is a subroutine library, you may +consider it more useful to permit linking proprietary applications with the +library. If this is what you want to do, use the GNU Lesser General +Public License instead of this License. diff --git a/src/contrib/licenses/OLDAP-2.8 b/src/contrib/licenses/OLDAP-2.8 new file mode 100644 index 0000000..05ad757 --- /dev/null +++ b/src/contrib/licenses/OLDAP-2.8 @@ -0,0 +1,47 @@ +The OpenLDAP Public License + Version 2.8, 17 August 2003 + +Redistribution and use of this software and associated documentation +("Software"), with or without modification, are permitted provided +that the following conditions are met: + +1. Redistributions in source form must retain copyright statements + and notices, + +2. Redistributions in binary form must reproduce applicable copyright + statements and notices, this list of conditions, and the following + disclaimer in the documentation and/or other materials provided + with the distribution, and + +3. Redistributions must contain a verbatim copy of this document. + +The OpenLDAP Foundation may revise this license from time to time. +Each revision is distinguished by a version number. You may use +this Software under terms of this license revision or under the +terms of any subsequent revision of the license. + +THIS SOFTWARE IS PROVIDED BY THE OPENLDAP FOUNDATION AND ITS +CONTRIBUTORS ``AS IS'' AND ANY EXPRESSED OR IMPLIED WARRANTIES, +INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY +AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT +SHALL THE OPENLDAP FOUNDATION, ITS CONTRIBUTORS, OR THE AUTHOR(S) +OR OWNER(S) OF THE SOFTWARE BE LIABLE FOR ANY DIRECT, INDIRECT, +INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, +BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; +LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER +CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT +LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN +ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE +POSSIBILITY OF SUCH DAMAGE. + +The names of the authors and copyright holders must not be used in +advertising or otherwise to promote the sale, use or other dealing +in this Software without specific, written prior permission. Title +to copyright in this Software shall at all times remain with copyright +holders. + +OpenLDAP is a registered trademark of the OpenLDAP Foundation. + +Copyright 1999-2003 The OpenLDAP Foundation, Redwood City, +California, USA. All Rights Reserved. Permission to copy and +distribute verbatim copies of this document is granted. diff --git a/src/contrib/lmdb/LICENSE b/src/contrib/lmdb/LICENSE new file mode 100644 index 0000000..66210f0 --- /dev/null +++ b/src/contrib/lmdb/LICENSE @@ -0,0 +1 @@ +../licenses/OLDAP-2.8
\ No newline at end of file diff --git a/src/contrib/lmdb/lmdb.h b/src/contrib/lmdb/lmdb.h new file mode 100644 index 0000000..32a278e --- /dev/null +++ b/src/contrib/lmdb/lmdb.h @@ -0,0 +1,1604 @@ +/** @file lmdb.h + * @brief Lightning memory-mapped database library + * + * @mainpage Lightning Memory-Mapped Database Manager (LMDB) + * + * @section intro_sec Introduction + * LMDB is a Btree-based database management library modeled loosely on the + * BerkeleyDB API, but much simplified. The entire database is exposed + * in a memory map, and all data fetches return data directly + * from the mapped memory, so no malloc's or memcpy's occur during + * data fetches. As such, the library is extremely simple because it + * requires no page caching layer of its own, and it is extremely high + * performance and memory-efficient. It is also fully transactional with + * full ACID semantics, and when the memory map is read-only, the + * database integrity cannot be corrupted by stray pointer writes from + * application code. + * + * The library is fully thread-aware and supports concurrent read/write + * access from multiple processes and threads. Data pages use a copy-on- + * write strategy so no active data pages are ever overwritten, which + * also provides resistance to corruption and eliminates the need of any + * special recovery procedures after a system crash. Writes are fully + * serialized; only one write transaction may be active at a time, which + * guarantees that writers can never deadlock. The database structure is + * multi-versioned so readers run with no locks; writers cannot block + * readers, and readers don't block writers. + * + * Unlike other well-known database mechanisms which use either write-ahead + * transaction logs or append-only data writes, LMDB requires no maintenance + * during operation. Both write-ahead loggers and append-only databases + * require periodic checkpointing and/or compaction of their log or database + * files otherwise they grow without bound. LMDB tracks free pages within + * the database and re-uses them for new write operations, so the database + * size does not grow without bound in normal use. + * + * The memory map can be used as a read-only or read-write map. It is + * read-only by default as this provides total immunity to corruption. + * Using read-write mode offers much higher write performance, but adds + * the possibility for stray application writes thru pointers to silently + * corrupt the database. Of course if your application code is known to + * be bug-free (...) then this is not an issue. + * + * If this is your first time using a transactional embedded key/value + * store, you may find the \ref starting page to be helpful. + * + * @section caveats_sec Caveats + * Troubleshooting the lock file, plus semaphores on BSD systems: + * + * - A broken lockfile can cause sync issues. + * Stale reader transactions left behind by an aborted program + * cause further writes to grow the database quickly, and + * stale locks can block further operation. + * + * Fix: Check for stale readers periodically, using the + * #mdb_reader_check function or the \ref mdb_stat_1 "mdb_stat" tool. + * Stale writers will be cleared automatically on some systems: + * - Windows - automatic + * - Linux, systems using POSIX mutexes with Robust option - automatic + * - not on BSD, systems using POSIX semaphores. + * Otherwise just make all programs using the database close it; + * the lockfile is always reset on first open of the environment. + * + * - On BSD systems or others configured with MDB_USE_POSIX_SEM, + * startup can fail due to semaphores owned by another userid. + * + * Fix: Open and close the database as the user which owns the + * semaphores (likely last user) or as root, while no other + * process is using the database. + * + * Restrictions/caveats (in addition to those listed for some functions): + * + * - Only the database owner should normally use the database on + * BSD systems or when otherwise configured with MDB_USE_POSIX_SEM. + * Multiple users can cause startup to fail later, as noted above. + * + * - There is normally no pure read-only mode, since readers need write + * access to locks and lock file. Exceptions: On read-only filesystems + * or with the #MDB_NOLOCK flag described under #mdb_env_open(). + * + * - An LMDB configuration will often reserve considerable \b unused + * memory address space and maybe file size for future growth. + * This does not use actual memory or disk space, but users may need + * to understand the difference so they won't be scared off. + * + * - By default, in versions before 0.9.10, unused portions of the data + * file might receive garbage data from memory freed by other code. + * (This does not happen when using the #MDB_WRITEMAP flag.) As of + * 0.9.10 the default behavior is to initialize such memory before + * writing to the data file. Since there may be a slight performance + * cost due to this initialization, applications may disable it using + * the #MDB_NOMEMINIT flag. Applications handling sensitive data + * which must not be written should not use this flag. This flag is + * irrelevant when using #MDB_WRITEMAP. + * + * - A thread can only use one transaction at a time, plus any child + * transactions. Each transaction belongs to one thread. See below. + * The #MDB_NOTLS flag changes this for read-only transactions. + * + * - Use an MDB_env* in the process which opened it, not after fork(). + * + * - Do not have open an LMDB database twice in the same process at + * the same time. Not even from a plain open() call - close()ing it + * breaks fcntl() advisory locking. (It is OK to reopen it after + * fork() - exec*(), since the lockfile has FD_CLOEXEC set.) + * + * - Avoid long-lived transactions. Read transactions prevent + * reuse of pages freed by newer write transactions, thus the + * database can grow quickly. Write transactions prevent + * other write transactions, since writes are serialized. + * + * - Avoid suspending a process with active transactions. These + * would then be "long-lived" as above. Also read transactions + * suspended when writers commit could sometimes see wrong data. + * + * ...when several processes can use a database concurrently: + * + * - Avoid aborting a process with an active transaction. + * The transaction becomes "long-lived" as above until a check + * for stale readers is performed or the lockfile is reset, + * since the process may not remove it from the lockfile. + * + * This does not apply to write transactions if the system clears + * stale writers, see above. + * + * - If you do that anyway, do a periodic check for stale readers. Or + * close the environment once in a while, so the lockfile can get reset. + * + * - Do not use LMDB databases on remote filesystems, even between + * processes on the same host. This breaks flock() on some OSes, + * possibly memory map sync, and certainly sync between programs + * on different hosts. + * + * - Opening a database can fail if another process is opening or + * closing it at exactly the same time. + * + * @author Howard Chu, Symas Corporation. + * + * @copyright Copyright 2011-2018 Howard Chu, Symas Corp. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted only as authorized by the OpenLDAP + * Public License. + * + * A copy of this license is available in the file LICENSE in the + * top-level directory of the distribution or, alternatively, at + * <http://www.OpenLDAP.org/license.html>. + * + * @par Derived From: + * This code is derived from btree.c written by Martin Hedenfalk. + * + * Copyright (c) 2009, 2010 Martin Hedenfalk <martin@bzero.se> + * + * Permission to use, copy, modify, and distribute this software for any + * purpose with or without fee is hereby granted, provided that the above + * copyright notice and this permission notice appear in all copies. + * + * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES + * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF + * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR + * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES + * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN + * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF + * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. + */ +#ifndef _LMDB_H_ +#define _LMDB_H_ + +#include <sys/types.h> + +#ifdef __cplusplus +extern "C" { +#endif + +/** Unix permissions for creating files, or dummy definition for Windows */ +#ifdef _MSC_VER +typedef int mdb_mode_t; +#else +typedef mode_t mdb_mode_t; +#endif + +/** An abstraction for a file handle. + * On POSIX systems file handles are small integers. On Windows + * they're opaque pointers. + */ +#ifdef _WIN32 +typedef void *mdb_filehandle_t; +#else +typedef int mdb_filehandle_t; +#endif + +/** @defgroup mdb LMDB API + * @{ + * @brief OpenLDAP Lightning Memory-Mapped Database Manager + */ +/** @defgroup Version Version Macros + * @{ + */ +/** Library major version */ +#define MDB_VERSION_MAJOR 0 +/** Library minor version */ +#define MDB_VERSION_MINOR 9 +/** Library patch version */ +#define MDB_VERSION_PATCH 22 + +/** Combine args a,b,c into a single integer for easy version comparisons */ +#define MDB_VERINT(a,b,c) (((a) << 24) | ((b) << 16) | (c)) + +/** The full library version as a single integer */ +#define MDB_VERSION_FULL \ + MDB_VERINT(MDB_VERSION_MAJOR,MDB_VERSION_MINOR,MDB_VERSION_PATCH) + +/** The release date of this library version */ +#define MDB_VERSION_DATE "March 21, 2018" + +/** A stringifier for the version info */ +#define MDB_VERSTR(a,b,c,d) "LMDB " #a "." #b "." #c ": (" d ")" + +/** A helper for the stringifier macro */ +#define MDB_VERFOO(a,b,c,d) MDB_VERSTR(a,b,c,d) + +/** The full library version as a C string */ +#define MDB_VERSION_STRING \ + MDB_VERFOO(MDB_VERSION_MAJOR,MDB_VERSION_MINOR,MDB_VERSION_PATCH,MDB_VERSION_DATE) +/** @} */ + +/** @brief Opaque structure for a database environment. + * + * A DB environment supports multiple databases, all residing in the same + * shared-memory map. + */ +typedef struct MDB_env MDB_env; + +/** @brief Opaque structure for a transaction handle. + * + * All database operations require a transaction handle. Transactions may be + * read-only or read-write. + */ +typedef struct MDB_txn MDB_txn; + +/** @brief A handle for an individual database in the DB environment. */ +typedef unsigned int MDB_dbi; + +/** @brief Opaque structure for navigating through a database */ +typedef struct MDB_cursor MDB_cursor; + +/** @brief Generic structure used for passing keys and data in and out + * of the database. + * + * Values returned from the database are valid only until a subsequent + * update operation, or the end of the transaction. Do not modify or + * free them, they commonly point into the database itself. + * + * Key sizes must be between 1 and #mdb_env_get_maxkeysize() inclusive. + * The same applies to data sizes in databases with the #MDB_DUPSORT flag. + * Other data items can in theory be from 0 to 0xffffffff bytes long. + */ +typedef struct MDB_val { + size_t mv_size; /**< size of the data item */ + void *mv_data; /**< address of the data item */ +} MDB_val; + +/** @brief A callback function used to compare two keys in a database */ +typedef int (MDB_cmp_func)(const MDB_val *a, const MDB_val *b); + +/** @brief A callback function used to relocate a position-dependent data item + * in a fixed-address database. + * + * The \b newptr gives the item's desired address in + * the memory map, and \b oldptr gives its previous address. The item's actual + * data resides at the address in \b item. This callback is expected to walk + * through the fields of the record in \b item and modify any + * values based at the \b oldptr address to be relative to the \b newptr address. + * @param[in,out] item The item that is to be relocated. + * @param[in] oldptr The previous address. + * @param[in] newptr The new address to relocate to. + * @param[in] relctx An application-provided context, set by #mdb_set_relctx(). + * @todo This feature is currently unimplemented. + */ +typedef void (MDB_rel_func)(MDB_val *item, void *oldptr, void *newptr, void *relctx); + +/** @defgroup mdb_env Environment Flags + * @{ + */ + /** mmap at a fixed address (experimental) */ +#define MDB_FIXEDMAP 0x01 + /** no environment directory */ +#define MDB_NOSUBDIR 0x4000 + /** don't fsync after commit */ +#define MDB_NOSYNC 0x10000 + /** read only */ +#define MDB_RDONLY 0x20000 + /** don't fsync metapage after commit */ +#define MDB_NOMETASYNC 0x40000 + /** use writable mmap */ +#define MDB_WRITEMAP 0x80000 + /** use asynchronous msync when #MDB_WRITEMAP is used */ +#define MDB_MAPASYNC 0x100000 + /** tie reader locktable slots to #MDB_txn objects instead of to threads */ +#define MDB_NOTLS 0x200000 + /** don't do any locking, caller must manage their own locks */ +#define MDB_NOLOCK 0x400000 + /** don't do readahead (no effect on Windows) */ +#define MDB_NORDAHEAD 0x800000 + /** don't initialize malloc'd memory before writing to datafile */ +#define MDB_NOMEMINIT 0x1000000 +/** @} */ + +/** @defgroup mdb_dbi_open Database Flags + * @{ + */ + /** use reverse string keys */ +#define MDB_REVERSEKEY 0x02 + /** use sorted duplicates */ +#define MDB_DUPSORT 0x04 + /** numeric keys in native byte order: either unsigned int or size_t. + * The keys must all be of the same size. */ +#define MDB_INTEGERKEY 0x08 + /** with #MDB_DUPSORT, sorted dup items have fixed size */ +#define MDB_DUPFIXED 0x10 + /** with #MDB_DUPSORT, dups are #MDB_INTEGERKEY-style integers */ +#define MDB_INTEGERDUP 0x20 + /** with #MDB_DUPSORT, use reverse string dups */ +#define MDB_REVERSEDUP 0x40 + /** create DB if not already existing */ +#define MDB_CREATE 0x40000 +/** @} */ + +/** @defgroup mdb_put Write Flags + * @{ + */ +/** For put: Don't write if the key already exists. */ +#define MDB_NOOVERWRITE 0x10 +/** Only for #MDB_DUPSORT<br> + * For put: don't write if the key and data pair already exist.<br> + * For mdb_cursor_del: remove all duplicate data items. + */ +#define MDB_NODUPDATA 0x20 +/** For mdb_cursor_put: overwrite the current key/data pair */ +#define MDB_CURRENT 0x40 +/** For put: Just reserve space for data, don't copy it. Return a + * pointer to the reserved space. + */ +#define MDB_RESERVE 0x10000 +/** Data is being appended, don't split full pages. */ +#define MDB_APPEND 0x20000 +/** Duplicate data is being appended, don't split full pages. */ +#define MDB_APPENDDUP 0x40000 +/** Store multiple data items in one call. Only for #MDB_DUPFIXED. */ +#define MDB_MULTIPLE 0x80000 +/* @} */ + +/** @defgroup mdb_copy Copy Flags + * @{ + */ +/** Compacting copy: Omit free space from copy, and renumber all + * pages sequentially. + */ +#define MDB_CP_COMPACT 0x01 +/* @} */ + +/** @brief Cursor Get operations. + * + * This is the set of all operations for retrieving data + * using a cursor. + */ +typedef enum MDB_cursor_op { + MDB_FIRST, /**< Position at first key/data item */ + MDB_FIRST_DUP, /**< Position at first data item of current key. + Only for #MDB_DUPSORT */ + MDB_GET_BOTH, /**< Position at key/data pair. Only for #MDB_DUPSORT */ + MDB_GET_BOTH_RANGE, /**< position at key, nearest data. Only for #MDB_DUPSORT */ + MDB_GET_CURRENT, /**< Return key/data at current cursor position */ + MDB_GET_MULTIPLE, /**< Return key and up to a page of duplicate data items + from current cursor position. Move cursor to prepare + for #MDB_NEXT_MULTIPLE. Only for #MDB_DUPFIXED */ + MDB_LAST, /**< Position at last key/data item */ + MDB_LAST_DUP, /**< Position at last data item of current key. + Only for #MDB_DUPSORT */ + MDB_NEXT, /**< Position at next data item */ + MDB_NEXT_DUP, /**< Position at next data item of current key. + Only for #MDB_DUPSORT */ + MDB_NEXT_MULTIPLE, /**< Return key and up to a page of duplicate data items + from next cursor position. Move cursor to prepare + for #MDB_NEXT_MULTIPLE. Only for #MDB_DUPFIXED */ + MDB_NEXT_NODUP, /**< Position at first data item of next key */ + MDB_PREV, /**< Position at previous data item */ + MDB_PREV_DUP, /**< Position at previous data item of current key. + Only for #MDB_DUPSORT */ + MDB_PREV_NODUP, /**< Position at last data item of previous key */ + MDB_SET, /**< Position at specified key */ + MDB_SET_KEY, /**< Position at specified key, return key + data */ + MDB_SET_RANGE, /**< Position at first key greater than or equal to specified key. */ + MDB_PREV_MULTIPLE /**< Position at previous page and return key and up to + a page of duplicate data items. Only for #MDB_DUPFIXED */ +} MDB_cursor_op; + +/** @defgroup errors Return Codes + * + * BerkeleyDB uses -30800 to -30999, we'll go under them + * @{ + */ + /** Successful result */ +#define MDB_SUCCESS 0 + /** key/data pair already exists */ +#define MDB_KEYEXIST (-30799) + /** key/data pair not found (EOF) */ +#define MDB_NOTFOUND (-30798) + /** Requested page not found - this usually indicates corruption */ +#define MDB_PAGE_NOTFOUND (-30797) + /** Located page was wrong type */ +#define MDB_CORRUPTED (-30796) + /** Update of meta page failed or environment had fatal error */ +#define MDB_PANIC (-30795) + /** Environment version mismatch */ +#define MDB_VERSION_MISMATCH (-30794) + /** File is not a valid LMDB file */ +#define MDB_INVALID (-30793) + /** Environment mapsize reached */ +#define MDB_MAP_FULL (-30792) + /** Environment maxdbs reached */ +#define MDB_DBS_FULL (-30791) + /** Environment maxreaders reached */ +#define MDB_READERS_FULL (-30790) + /** Too many TLS keys in use - Windows only */ +#define MDB_TLS_FULL (-30789) + /** Txn has too many dirty pages */ +#define MDB_TXN_FULL (-30788) + /** Cursor stack too deep - internal error */ +#define MDB_CURSOR_FULL (-30787) + /** Page has not enough space - internal error */ +#define MDB_PAGE_FULL (-30786) + /** Database contents grew beyond environment mapsize */ +#define MDB_MAP_RESIZED (-30785) + /** Operation and DB incompatible, or DB type changed. This can mean: + * <ul> + * <li>The operation expects an #MDB_DUPSORT / #MDB_DUPFIXED database. + * <li>Opening a named DB when the unnamed DB has #MDB_DUPSORT / #MDB_INTEGERKEY. + * <li>Accessing a data record as a database, or vice versa. + * <li>The database was dropped and recreated with different flags. + * </ul> + */ +#define MDB_INCOMPATIBLE (-30784) + /** Invalid reuse of reader locktable slot */ +#define MDB_BAD_RSLOT (-30783) + /** Transaction must abort, has a child, or is invalid */ +#define MDB_BAD_TXN (-30782) + /** Unsupported size of key/DB name/data, or wrong DUPFIXED size */ +#define MDB_BAD_VALSIZE (-30781) + /** The specified DBI was changed unexpectedly */ +#define MDB_BAD_DBI (-30780) + /** The last defined error code */ +#define MDB_LAST_ERRCODE MDB_BAD_DBI +/** @} */ + +/** @brief Statistics for a database in the environment */ +typedef struct MDB_stat { + unsigned int ms_psize; /**< Size of a database page. + This is currently the same for all databases. */ + unsigned int ms_depth; /**< Depth (height) of the B-tree */ + size_t ms_branch_pages; /**< Number of internal (non-leaf) pages */ + size_t ms_leaf_pages; /**< Number of leaf pages */ + size_t ms_overflow_pages; /**< Number of overflow pages */ + size_t ms_entries; /**< Number of data items */ +} MDB_stat; + +/** @brief Information about the environment */ +typedef struct MDB_envinfo { + void *me_mapaddr; /**< Address of map, if fixed */ + size_t me_mapsize; /**< Size of the data memory map */ + size_t me_last_pgno; /**< ID of the last used page */ + size_t me_last_txnid; /**< ID of the last committed transaction */ + unsigned int me_maxreaders; /**< max reader slots in the environment */ + unsigned int me_numreaders; /**< max reader slots used in the environment */ +} MDB_envinfo; + + /** @brief Return the LMDB library version information. + * + * @param[out] major if non-NULL, the library major version number is copied here + * @param[out] minor if non-NULL, the library minor version number is copied here + * @param[out] patch if non-NULL, the library patch version number is copied here + * @retval "version string" The library version as a string + */ +char *mdb_version(int *major, int *minor, int *patch); + + /** @brief Return a string describing a given error code. + * + * This function is a superset of the ANSI C X3.159-1989 (ANSI C) strerror(3) + * function. If the error code is greater than or equal to 0, then the string + * returned by the system function strerror(3) is returned. If the error code + * is less than 0, an error string corresponding to the LMDB library error is + * returned. See @ref errors for a list of LMDB-specific error codes. + * @param[in] err The error code + * @retval "error message" The description of the error + */ +char *mdb_strerror(int err); + + /** @brief Create an LMDB environment handle. + * + * This function allocates memory for a #MDB_env structure. To release + * the allocated memory and discard the handle, call #mdb_env_close(). + * Before the handle may be used, it must be opened using #mdb_env_open(). + * Various other options may also need to be set before opening the handle, + * e.g. #mdb_env_set_mapsize(), #mdb_env_set_maxreaders(), #mdb_env_set_maxdbs(), + * depending on usage requirements. + * @param[out] env The address where the new handle will be stored + * @return A non-zero error value on failure and 0 on success. + */ +int mdb_env_create(MDB_env **env); + + /** @brief Open an environment handle. + * + * If this function fails, #mdb_env_close() must be called to discard the #MDB_env handle. + * @param[in] env An environment handle returned by #mdb_env_create() + * @param[in] path The directory in which the database files reside. This + * directory must already exist and be writable. + * @param[in] flags Special options for this environment. This parameter + * must be set to 0 or by bitwise OR'ing together one or more of the + * values described here. + * Flags set by mdb_env_set_flags() are also used. + * <ul> + * <li>#MDB_FIXEDMAP + * use a fixed address for the mmap region. This flag must be specified + * when creating the environment, and is stored persistently in the environment. + * If successful, the memory map will always reside at the same virtual address + * and pointers used to reference data items in the database will be constant + * across multiple invocations. This option may not always work, depending on + * how the operating system has allocated memory to shared libraries and other uses. + * The feature is highly experimental. + * <li>#MDB_NOSUBDIR + * By default, LMDB creates its environment in a directory whose + * pathname is given in \b path, and creates its data and lock files + * under that directory. With this option, \b path is used as-is for + * the database main data file. The database lock file is the \b path + * with "-lock" appended. + * <li>#MDB_RDONLY + * Open the environment in read-only mode. No write operations will be + * allowed. LMDB will still modify the lock file - except on read-only + * filesystems, where LMDB does not use locks. + * <li>#MDB_WRITEMAP + * Use a writeable memory map unless MDB_RDONLY is set. This uses + * fewer mallocs but loses protection from application bugs + * like wild pointer writes and other bad updates into the database. + * This may be slightly faster for DBs that fit entirely in RAM, but + * is slower for DBs larger than RAM. + * Incompatible with nested transactions. + * Do not mix processes with and without MDB_WRITEMAP on the same + * environment. This can defeat durability (#mdb_env_sync etc). + * <li>#MDB_NOMETASYNC + * Flush system buffers to disk only once per transaction, omit the + * metadata flush. Defer that until the system flushes files to disk, + * or next non-MDB_RDONLY commit or #mdb_env_sync(). This optimization + * maintains database integrity, but a system crash may undo the last + * committed transaction. I.e. it preserves the ACI (atomicity, + * consistency, isolation) but not D (durability) database property. + * This flag may be changed at any time using #mdb_env_set_flags(). + * <li>#MDB_NOSYNC + * Don't flush system buffers to disk when committing a transaction. + * This optimization means a system crash can corrupt the database or + * lose the last transactions if buffers are not yet flushed to disk. + * The risk is governed by how often the system flushes dirty buffers + * to disk and how often #mdb_env_sync() is called. However, if the + * filesystem preserves write order and the #MDB_WRITEMAP flag is not + * used, transactions exhibit ACI (atomicity, consistency, isolation) + * properties and only lose D (durability). I.e. database integrity + * is maintained, but a system crash may undo the final transactions. + * Note that (#MDB_NOSYNC | #MDB_WRITEMAP) leaves the system with no + * hint for when to write transactions to disk, unless #mdb_env_sync() + * is called. (#MDB_MAPASYNC | #MDB_WRITEMAP) may be preferable. + * This flag may be changed at any time using #mdb_env_set_flags(). + * <li>#MDB_MAPASYNC + * When using #MDB_WRITEMAP, use asynchronous flushes to disk. + * As with #MDB_NOSYNC, a system crash can then corrupt the + * database or lose the last transactions. Calling #mdb_env_sync() + * ensures on-disk database integrity until next commit. + * This flag may be changed at any time using #mdb_env_set_flags(). + * <li>#MDB_NOTLS + * Don't use Thread-Local Storage. Tie reader locktable slots to + * #MDB_txn objects instead of to threads. I.e. #mdb_txn_reset() keeps + * the slot reseved for the #MDB_txn object. A thread may use parallel + * read-only transactions. A read-only transaction may span threads if + * the user synchronizes its use. Applications that multiplex many + * user threads over individual OS threads need this option. Such an + * application must also serialize the write transactions in an OS + * thread, since LMDB's write locking is unaware of the user threads. + * <li>#MDB_NOLOCK + * Don't do any locking. If concurrent access is anticipated, the + * caller must manage all concurrency itself. For proper operation + * the caller must enforce single-writer semantics, and must ensure + * that no readers are using old transactions while a writer is + * active. The simplest approach is to use an exclusive lock so that + * no readers may be active at all when a writer begins. + * <li>#MDB_NORDAHEAD + * Turn off readahead. Most operating systems perform readahead on + * read requests by default. This option turns it off if the OS + * supports it. Turning it off may help random read performance + * when the DB is larger than RAM and system RAM is full. + * The option is not implemented on Windows. + * <li>#MDB_NOMEMINIT + * Don't initialize malloc'd memory before writing to unused spaces + * in the data file. By default, memory for pages written to the data + * file is obtained using malloc. While these pages may be reused in + * subsequent transactions, freshly malloc'd pages will be initialized + * to zeroes before use. This avoids persisting leftover data from other + * code (that used the heap and subsequently freed the memory) into the + * data file. Note that many other system libraries may allocate + * and free memory from the heap for arbitrary uses. E.g., stdio may + * use the heap for file I/O buffers. This initialization step has a + * modest performance cost so some applications may want to disable + * it using this flag. This option can be a problem for applications + * which handle sensitive data like passwords, and it makes memory + * checkers like Valgrind noisy. This flag is not needed with #MDB_WRITEMAP, + * which writes directly to the mmap instead of using malloc for pages. The + * initialization is also skipped if #MDB_RESERVE is used; the + * caller is expected to overwrite all of the memory that was + * reserved in that case. + * This flag may be changed at any time using #mdb_env_set_flags(). + * </ul> + * @param[in] mode The UNIX permissions to set on created files and semaphores. + * This parameter is ignored on Windows. + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>#MDB_VERSION_MISMATCH - the version of the LMDB library doesn't match the + * version that created the database environment. + * <li>#MDB_INVALID - the environment file headers are corrupted. + * <li>ENOENT - the directory specified by the path parameter doesn't exist. + * <li>EACCES - the user didn't have permission to access the environment files. + * <li>EAGAIN - the environment was locked by another process. + * </ul> + */ +int mdb_env_open(MDB_env *env, const char *path, unsigned int flags, mdb_mode_t mode); + + /** @brief Copy an LMDB environment to the specified path. + * + * This function may be used to make a backup of an existing environment. + * No lockfile is created, since it gets recreated at need. + * @note This call can trigger significant file size growth if run in + * parallel with write transactions, because it employs a read-only + * transaction. See long-lived transactions under @ref caveats_sec. + * @param[in] env An environment handle returned by #mdb_env_create(). It + * must have already been opened successfully. + * @param[in] path The directory in which the copy will reside. This + * directory must already exist and be writable but must otherwise be + * empty. + * @return A non-zero error value on failure and 0 on success. + */ +int mdb_env_copy(MDB_env *env, const char *path); + + /** @brief Copy an LMDB environment to the specified file descriptor. + * + * This function may be used to make a backup of an existing environment. + * No lockfile is created, since it gets recreated at need. + * @note This call can trigger significant file size growth if run in + * parallel with write transactions, because it employs a read-only + * transaction. See long-lived transactions under @ref caveats_sec. + * @param[in] env An environment handle returned by #mdb_env_create(). It + * must have already been opened successfully. + * @param[in] fd The filedescriptor to write the copy to. It must + * have already been opened for Write access. + * @return A non-zero error value on failure and 0 on success. + */ +int mdb_env_copyfd(MDB_env *env, mdb_filehandle_t fd); + + /** @brief Copy an LMDB environment to the specified path, with options. + * + * This function may be used to make a backup of an existing environment. + * No lockfile is created, since it gets recreated at need. + * @note This call can trigger significant file size growth if run in + * parallel with write transactions, because it employs a read-only + * transaction. See long-lived transactions under @ref caveats_sec. + * @param[in] env An environment handle returned by #mdb_env_create(). It + * must have already been opened successfully. + * @param[in] path The directory in which the copy will reside. This + * directory must already exist and be writable but must otherwise be + * empty. + * @param[in] flags Special options for this operation. This parameter + * must be set to 0 or by bitwise OR'ing together one or more of the + * values described here. + * <ul> + * <li>#MDB_CP_COMPACT - Perform compaction while copying: omit free + * pages and sequentially renumber all pages in output. This option + * consumes more CPU and runs more slowly than the default. + * Currently it fails if the environment has suffered a page leak. + * </ul> + * @return A non-zero error value on failure and 0 on success. + */ +int mdb_env_copy2(MDB_env *env, const char *path, unsigned int flags); + + /** @brief Copy an LMDB environment to the specified file descriptor, + * with options. + * + * This function may be used to make a backup of an existing environment. + * No lockfile is created, since it gets recreated at need. See + * #mdb_env_copy2() for further details. + * @note This call can trigger significant file size growth if run in + * parallel with write transactions, because it employs a read-only + * transaction. See long-lived transactions under @ref caveats_sec. + * @param[in] env An environment handle returned by #mdb_env_create(). It + * must have already been opened successfully. + * @param[in] fd The filedescriptor to write the copy to. It must + * have already been opened for Write access. + * @param[in] flags Special options for this operation. + * See #mdb_env_copy2() for options. + * @return A non-zero error value on failure and 0 on success. + */ +int mdb_env_copyfd2(MDB_env *env, mdb_filehandle_t fd, unsigned int flags); + + /** @brief Return statistics about the LMDB environment. + * + * @param[in] env An environment handle returned by #mdb_env_create() + * @param[out] stat The address of an #MDB_stat structure + * where the statistics will be copied + */ +int mdb_env_stat(MDB_env *env, MDB_stat *stat); + + /** @brief Return information about the LMDB environment. + * + * @param[in] env An environment handle returned by #mdb_env_create() + * @param[out] stat The address of an #MDB_envinfo structure + * where the information will be copied + */ +int mdb_env_info(MDB_env *env, MDB_envinfo *stat); + + /** @brief Flush the data buffers to disk. + * + * Data is always written to disk when #mdb_txn_commit() is called, + * but the operating system may keep it buffered. LMDB always flushes + * the OS buffers upon commit as well, unless the environment was + * opened with #MDB_NOSYNC or in part #MDB_NOMETASYNC. This call is + * not valid if the environment was opened with #MDB_RDONLY. + * @param[in] env An environment handle returned by #mdb_env_create() + * @param[in] force If non-zero, force a synchronous flush. Otherwise + * if the environment has the #MDB_NOSYNC flag set the flushes + * will be omitted, and with #MDB_MAPASYNC they will be asynchronous. + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>EACCES - the environment is read-only. + * <li>EINVAL - an invalid parameter was specified. + * <li>EIO - an error occurred during synchronization. + * </ul> + */ +int mdb_env_sync(MDB_env *env, int force); + + /** @brief Close the environment and release the memory map. + * + * Only a single thread may call this function. All transactions, databases, + * and cursors must already be closed before calling this function. Attempts to + * use any such handles after calling this function will cause a SIGSEGV. + * The environment handle will be freed and must not be used again after this call. + * @param[in] env An environment handle returned by #mdb_env_create() + */ +void mdb_env_close(MDB_env *env); + + /** @brief Set environment flags. + * + * This may be used to set some flags in addition to those from + * #mdb_env_open(), or to unset these flags. If several threads + * change the flags at the same time, the result is undefined. + * @param[in] env An environment handle returned by #mdb_env_create() + * @param[in] flags The flags to change, bitwise OR'ed together + * @param[in] onoff A non-zero value sets the flags, zero clears them. + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>EINVAL - an invalid parameter was specified. + * </ul> + */ +int mdb_env_set_flags(MDB_env *env, unsigned int flags, int onoff); + + /** @brief Get environment flags. + * + * @param[in] env An environment handle returned by #mdb_env_create() + * @param[out] flags The address of an integer to store the flags + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>EINVAL - an invalid parameter was specified. + * </ul> + */ +int mdb_env_get_flags(MDB_env *env, unsigned int *flags); + + /** @brief Return the path that was used in #mdb_env_open(). + * + * @param[in] env An environment handle returned by #mdb_env_create() + * @param[out] path Address of a string pointer to contain the path. This + * is the actual string in the environment, not a copy. It should not be + * altered in any way. + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>EINVAL - an invalid parameter was specified. + * </ul> + */ +int mdb_env_get_path(MDB_env *env, const char **path); + + /** @brief Return the filedescriptor for the given environment. + * + * This function may be called after fork(), so the descriptor can be + * closed before exec*(). Other LMDB file descriptors have FD_CLOEXEC. + * (Until LMDB 0.9.18, only the lockfile had that.) + * + * @param[in] env An environment handle returned by #mdb_env_create() + * @param[out] fd Address of a mdb_filehandle_t to contain the descriptor. + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>EINVAL - an invalid parameter was specified. + * </ul> + */ +int mdb_env_get_fd(MDB_env *env, mdb_filehandle_t *fd); + + /** @brief Set the size of the memory map to use for this environment. + * + * The size should be a multiple of the OS page size. The default is + * 10485760 bytes. The size of the memory map is also the maximum size + * of the database. The value should be chosen as large as possible, + * to accommodate future growth of the database. + * This function should be called after #mdb_env_create() and before #mdb_env_open(). + * It may be called at later times if no transactions are active in + * this process. Note that the library does not check for this condition, + * the caller must ensure it explicitly. + * + * The new size takes effect immediately for the current process but + * will not be persisted to any others until a write transaction has been + * committed by the current process. Also, only mapsize increases are + * persisted into the environment. + * + * If the mapsize is increased by another process, and data has grown + * beyond the range of the current mapsize, #mdb_txn_begin() will + * return #MDB_MAP_RESIZED. This function may be called with a size + * of zero to adopt the new size. + * + * Any attempt to set a size smaller than the space already consumed + * by the environment will be silently changed to the current size of the used space. + * @param[in] env An environment handle returned by #mdb_env_create() + * @param[in] size The size in bytes + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>EINVAL - an invalid parameter was specified, or the environment has + * an active write transaction. + * </ul> + */ +int mdb_env_set_mapsize(MDB_env *env, size_t size); + + /** @brief Set the maximum number of threads/reader slots for the environment. + * + * This defines the number of slots in the lock table that is used to track readers in the + * the environment. The default is 126. + * Starting a read-only transaction normally ties a lock table slot to the + * current thread until the environment closes or the thread exits. If + * MDB_NOTLS is in use, #mdb_txn_begin() instead ties the slot to the + * MDB_txn object until it or the #MDB_env object is destroyed. + * This function may only be called after #mdb_env_create() and before #mdb_env_open(). + * @param[in] env An environment handle returned by #mdb_env_create() + * @param[in] readers The maximum number of reader lock table slots + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>EINVAL - an invalid parameter was specified, or the environment is already open. + * </ul> + */ +int mdb_env_set_maxreaders(MDB_env *env, unsigned int readers); + + /** @brief Get the maximum number of threads/reader slots for the environment. + * + * @param[in] env An environment handle returned by #mdb_env_create() + * @param[out] readers Address of an integer to store the number of readers + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>EINVAL - an invalid parameter was specified. + * </ul> + */ +int mdb_env_get_maxreaders(MDB_env *env, unsigned int *readers); + + /** @brief Set the maximum number of named databases for the environment. + * + * This function is only needed if multiple databases will be used in the + * environment. Simpler applications that use the environment as a single + * unnamed database can ignore this option. + * This function may only be called after #mdb_env_create() and before #mdb_env_open(). + * + * Currently a moderate number of slots are cheap but a huge number gets + * expensive: 7-120 words per transaction, and every #mdb_dbi_open() + * does a linear search of the opened slots. + * @param[in] env An environment handle returned by #mdb_env_create() + * @param[in] dbs The maximum number of databases + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>EINVAL - an invalid parameter was specified, or the environment is already open. + * </ul> + */ +int mdb_env_set_maxdbs(MDB_env *env, MDB_dbi dbs); + + /** @brief Get the maximum size of keys and #MDB_DUPSORT data we can write. + * + * Depends on the compile-time constant #MDB_MAXKEYSIZE. Default 511. + * See @ref MDB_val. + * @param[in] env An environment handle returned by #mdb_env_create() + * @return The maximum size of a key we can write + */ +int mdb_env_get_maxkeysize(MDB_env *env); + + /** @brief Set application information associated with the #MDB_env. + * + * @param[in] env An environment handle returned by #mdb_env_create() + * @param[in] ctx An arbitrary pointer for whatever the application needs. + * @return A non-zero error value on failure and 0 on success. + */ +int mdb_env_set_userctx(MDB_env *env, void *ctx); + + /** @brief Get the application information associated with the #MDB_env. + * + * @param[in] env An environment handle returned by #mdb_env_create() + * @return The pointer set by #mdb_env_set_userctx(). + */ +void *mdb_env_get_userctx(MDB_env *env); + + /** @brief A callback function for most LMDB assert() failures, + * called before printing the message and aborting. + * + * @param[in] env An environment handle returned by #mdb_env_create(). + * @param[in] msg The assertion message, not including newline. + */ +typedef void MDB_assert_func(MDB_env *env, const char *msg); + + /** Set or reset the assert() callback of the environment. + * Disabled if liblmdb is buillt with NDEBUG. + * @note This hack should become obsolete as lmdb's error handling matures. + * @param[in] env An environment handle returned by #mdb_env_create(). + * @param[in] func An #MDB_assert_func function, or 0. + * @return A non-zero error value on failure and 0 on success. + */ +int mdb_env_set_assert(MDB_env *env, MDB_assert_func *func); + + /** @brief Create a transaction for use with the environment. + * + * The transaction handle may be discarded using #mdb_txn_abort() or #mdb_txn_commit(). + * @note A transaction and its cursors must only be used by a single + * thread, and a thread may only have a single transaction at a time. + * If #MDB_NOTLS is in use, this does not apply to read-only transactions. + * @note Cursors may not span transactions. + * @param[in] env An environment handle returned by #mdb_env_create() + * @param[in] parent If this parameter is non-NULL, the new transaction + * will be a nested transaction, with the transaction indicated by \b parent + * as its parent. Transactions may be nested to any level. A parent + * transaction and its cursors may not issue any other operations than + * mdb_txn_commit and mdb_txn_abort while it has active child transactions. + * @param[in] flags Special options for this transaction. This parameter + * must be set to 0 or by bitwise OR'ing together one or more of the + * values described here. + * <ul> + * <li>#MDB_RDONLY + * This transaction will not perform any write operations. + * </ul> + * @param[out] txn Address where the new #MDB_txn handle will be stored + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>#MDB_PANIC - a fatal error occurred earlier and the environment + * must be shut down. + * <li>#MDB_MAP_RESIZED - another process wrote data beyond this MDB_env's + * mapsize and this environment's map must be resized as well. + * See #mdb_env_set_mapsize(). + * <li>#MDB_READERS_FULL - a read-only transaction was requested and + * the reader lock table is full. See #mdb_env_set_maxreaders(). + * <li>ENOMEM - out of memory. + * </ul> + */ +int mdb_txn_begin(MDB_env *env, MDB_txn *parent, unsigned int flags, MDB_txn **txn); + + /** @brief Returns the transaction's #MDB_env + * + * @param[in] txn A transaction handle returned by #mdb_txn_begin() + */ +MDB_env *mdb_txn_env(MDB_txn *txn); + + /** @brief Return the transaction's ID. + * + * This returns the identifier associated with this transaction. For a + * read-only transaction, this corresponds to the snapshot being read; + * concurrent readers will frequently have the same transaction ID. + * + * @param[in] txn A transaction handle returned by #mdb_txn_begin() + * @return A transaction ID, valid if input is an active transaction. + */ +size_t mdb_txn_id(MDB_txn *txn); + + /** @brief Commit all the operations of a transaction into the database. + * + * The transaction handle is freed. It and its cursors must not be used + * again after this call, except with #mdb_cursor_renew(). + * @note Earlier documentation incorrectly said all cursors would be freed. + * Only write-transactions free cursors. + * @param[in] txn A transaction handle returned by #mdb_txn_begin() + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>EINVAL - an invalid parameter was specified. + * <li>ENOSPC - no more disk space. + * <li>EIO - a low-level I/O error occurred while writing. + * <li>ENOMEM - out of memory. + * </ul> + */ +int mdb_txn_commit(MDB_txn *txn); + + /** @brief Abandon all the operations of the transaction instead of saving them. + * + * The transaction handle is freed. It and its cursors must not be used + * again after this call, except with #mdb_cursor_renew(). + * @note Earlier documentation incorrectly said all cursors would be freed. + * Only write-transactions free cursors. + * @param[in] txn A transaction handle returned by #mdb_txn_begin() + */ +void mdb_txn_abort(MDB_txn *txn); + + /** @brief Reset a read-only transaction. + * + * Abort the transaction like #mdb_txn_abort(), but keep the transaction + * handle. #mdb_txn_renew() may reuse the handle. This saves allocation + * overhead if the process will start a new read-only transaction soon, + * and also locking overhead if #MDB_NOTLS is in use. The reader table + * lock is released, but the table slot stays tied to its thread or + * #MDB_txn. Use mdb_txn_abort() to discard a reset handle, and to free + * its lock table slot if MDB_NOTLS is in use. + * Cursors opened within the transaction must not be used + * again after this call, except with #mdb_cursor_renew(). + * Reader locks generally don't interfere with writers, but they keep old + * versions of database pages allocated. Thus they prevent the old pages + * from being reused when writers commit new data, and so under heavy load + * the database size may grow much more rapidly than otherwise. + * @param[in] txn A transaction handle returned by #mdb_txn_begin() + */ +void mdb_txn_reset(MDB_txn *txn); + + /** @brief Renew a read-only transaction. + * + * This acquires a new reader lock for a transaction handle that had been + * released by #mdb_txn_reset(). It must be called before a reset transaction + * may be used again. + * @param[in] txn A transaction handle returned by #mdb_txn_begin() + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>#MDB_PANIC - a fatal error occurred earlier and the environment + * must be shut down. + * <li>EINVAL - an invalid parameter was specified. + * </ul> + */ +int mdb_txn_renew(MDB_txn *txn); + +/** Compat with version <= 0.9.4, avoid clash with libmdb from MDB Tools project */ +#define mdb_open(txn,name,flags,dbi) mdb_dbi_open(txn,name,flags,dbi) +/** Compat with version <= 0.9.4, avoid clash with libmdb from MDB Tools project */ +#define mdb_close(env,dbi) mdb_dbi_close(env,dbi) + + /** @brief Open a database in the environment. + * + * A database handle denotes the name and parameters of a database, + * independently of whether such a database exists. + * The database handle may be discarded by calling #mdb_dbi_close(). + * The old database handle is returned if the database was already open. + * The handle may only be closed once. + * + * The database handle will be private to the current transaction until + * the transaction is successfully committed. If the transaction is + * aborted the handle will be closed automatically. + * After a successful commit the handle will reside in the shared + * environment, and may be used by other transactions. + * + * This function must not be called from multiple concurrent + * transactions in the same process. A transaction that uses + * this function must finish (either commit or abort) before + * any other transaction in the process may use this function. + * + * To use named databases (with name != NULL), #mdb_env_set_maxdbs() + * must be called before opening the environment. Database names are + * keys in the unnamed database, and may be read but not written. + * + * @param[in] txn A transaction handle returned by #mdb_txn_begin() + * @param[in] name The name of the database to open. If only a single + * database is needed in the environment, this value may be NULL. + * @param[in] flags Special options for this database. This parameter + * must be set to 0 or by bitwise OR'ing together one or more of the + * values described here. + * <ul> + * <li>#MDB_REVERSEKEY + * Keys are strings to be compared in reverse order, from the end + * of the strings to the beginning. By default, Keys are treated as strings and + * compared from beginning to end. + * <li>#MDB_DUPSORT + * Duplicate keys may be used in the database. (Or, from another perspective, + * keys may have multiple data items, stored in sorted order.) By default + * keys must be unique and may have only a single data item. + * <li>#MDB_INTEGERKEY + * Keys are binary integers in native byte order, either unsigned int + * or size_t, and will be sorted as such. + * The keys must all be of the same size. + * <li>#MDB_DUPFIXED + * This flag may only be used in combination with #MDB_DUPSORT. This option + * tells the library that the data items for this database are all the same + * size, which allows further optimizations in storage and retrieval. When + * all data items are the same size, the #MDB_GET_MULTIPLE, #MDB_NEXT_MULTIPLE + * and #MDB_PREV_MULTIPLE cursor operations may be used to retrieve multiple + * items at once. + * <li>#MDB_INTEGERDUP + * This option specifies that duplicate data items are binary integers, + * similar to #MDB_INTEGERKEY keys. + * <li>#MDB_REVERSEDUP + * This option specifies that duplicate data items should be compared as + * strings in reverse order. + * <li>#MDB_CREATE + * Create the named database if it doesn't exist. This option is not + * allowed in a read-only transaction or a read-only environment. + * </ul> + * @param[out] dbi Address where the new #MDB_dbi handle will be stored + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>#MDB_NOTFOUND - the specified database doesn't exist in the environment + * and #MDB_CREATE was not specified. + * <li>#MDB_DBS_FULL - too many databases have been opened. See #mdb_env_set_maxdbs(). + * </ul> + */ +int mdb_dbi_open(MDB_txn *txn, const char *name, unsigned int flags, MDB_dbi *dbi); + + /** @brief Retrieve statistics for a database. + * + * @param[in] txn A transaction handle returned by #mdb_txn_begin() + * @param[in] dbi A database handle returned by #mdb_dbi_open() + * @param[out] stat The address of an #MDB_stat structure + * where the statistics will be copied + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>EINVAL - an invalid parameter was specified. + * </ul> + */ +int mdb_stat(MDB_txn *txn, MDB_dbi dbi, MDB_stat *stat); + + /** @brief Retrieve the DB flags for a database handle. + * + * @param[in] txn A transaction handle returned by #mdb_txn_begin() + * @param[in] dbi A database handle returned by #mdb_dbi_open() + * @param[out] flags Address where the flags will be returned. + * @return A non-zero error value on failure and 0 on success. + */ +int mdb_dbi_flags(MDB_txn *txn, MDB_dbi dbi, unsigned int *flags); + + /** @brief Close a database handle. Normally unnecessary. Use with care: + * + * This call is not mutex protected. Handles should only be closed by + * a single thread, and only if no other threads are going to reference + * the database handle or one of its cursors any further. Do not close + * a handle if an existing transaction has modified its database. + * Doing so can cause misbehavior from database corruption to errors + * like MDB_BAD_VALSIZE (since the DB name is gone). + * + * Closing a database handle is not necessary, but lets #mdb_dbi_open() + * reuse the handle value. Usually it's better to set a bigger + * #mdb_env_set_maxdbs(), unless that value would be large. + * + * @param[in] env An environment handle returned by #mdb_env_create() + * @param[in] dbi A database handle returned by #mdb_dbi_open() + */ +void mdb_dbi_close(MDB_env *env, MDB_dbi dbi); + + /** @brief Empty or delete+close a database. + * + * See #mdb_dbi_close() for restrictions about closing the DB handle. + * @param[in] txn A transaction handle returned by #mdb_txn_begin() + * @param[in] dbi A database handle returned by #mdb_dbi_open() + * @param[in] del 0 to empty the DB, 1 to delete it from the + * environment and close the DB handle. + * @return A non-zero error value on failure and 0 on success. + */ +int mdb_drop(MDB_txn *txn, MDB_dbi dbi, int del); + + /** @brief Set a custom key comparison function for a database. + * + * The comparison function is called whenever it is necessary to compare a + * key specified by the application with a key currently stored in the database. + * If no comparison function is specified, and no special key flags were specified + * with #mdb_dbi_open(), the keys are compared lexically, with shorter keys collating + * before longer keys. + * @warning This function must be called before any data access functions are used, + * otherwise data corruption may occur. The same comparison function must be used by every + * program accessing the database, every time the database is used. + * @param[in] txn A transaction handle returned by #mdb_txn_begin() + * @param[in] dbi A database handle returned by #mdb_dbi_open() + * @param[in] cmp A #MDB_cmp_func function + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>EINVAL - an invalid parameter was specified. + * </ul> + */ +int mdb_set_compare(MDB_txn *txn, MDB_dbi dbi, MDB_cmp_func *cmp); + + /** @brief Set a custom data comparison function for a #MDB_DUPSORT database. + * + * This comparison function is called whenever it is necessary to compare a data + * item specified by the application with a data item currently stored in the database. + * This function only takes effect if the database was opened with the #MDB_DUPSORT + * flag. + * If no comparison function is specified, and no special key flags were specified + * with #mdb_dbi_open(), the data items are compared lexically, with shorter items collating + * before longer items. + * @warning This function must be called before any data access functions are used, + * otherwise data corruption may occur. The same comparison function must be used by every + * program accessing the database, every time the database is used. + * @param[in] txn A transaction handle returned by #mdb_txn_begin() + * @param[in] dbi A database handle returned by #mdb_dbi_open() + * @param[in] cmp A #MDB_cmp_func function + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>EINVAL - an invalid parameter was specified. + * </ul> + */ +int mdb_set_dupsort(MDB_txn *txn, MDB_dbi dbi, MDB_cmp_func *cmp); + + /** @brief Set a relocation function for a #MDB_FIXEDMAP database. + * + * @todo The relocation function is called whenever it is necessary to move the data + * of an item to a different position in the database (e.g. through tree + * balancing operations, shifts as a result of adds or deletes, etc.). It is + * intended to allow address/position-dependent data items to be stored in + * a database in an environment opened with the #MDB_FIXEDMAP option. + * Currently the relocation feature is unimplemented and setting + * this function has no effect. + * @param[in] txn A transaction handle returned by #mdb_txn_begin() + * @param[in] dbi A database handle returned by #mdb_dbi_open() + * @param[in] rel A #MDB_rel_func function + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>EINVAL - an invalid parameter was specified. + * </ul> + */ +int mdb_set_relfunc(MDB_txn *txn, MDB_dbi dbi, MDB_rel_func *rel); + + /** @brief Set a context pointer for a #MDB_FIXEDMAP database's relocation function. + * + * See #mdb_set_relfunc and #MDB_rel_func for more details. + * @param[in] txn A transaction handle returned by #mdb_txn_begin() + * @param[in] dbi A database handle returned by #mdb_dbi_open() + * @param[in] ctx An arbitrary pointer for whatever the application needs. + * It will be passed to the callback function set by #mdb_set_relfunc + * as its \b relctx parameter whenever the callback is invoked. + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>EINVAL - an invalid parameter was specified. + * </ul> + */ +int mdb_set_relctx(MDB_txn *txn, MDB_dbi dbi, void *ctx); + + /** @brief Get items from a database. + * + * This function retrieves key/data pairs from the database. The address + * and length of the data associated with the specified \b key are returned + * in the structure to which \b data refers. + * If the database supports duplicate keys (#MDB_DUPSORT) then the + * first data item for the key will be returned. Retrieval of other + * items requires the use of #mdb_cursor_get(). + * + * @note The memory pointed to by the returned values is owned by the + * database. The caller need not dispose of the memory, and may not + * modify it in any way. For values returned in a read-only transaction + * any modification attempts will cause a SIGSEGV. + * @note Values returned from the database are valid only until a + * subsequent update operation, or the end of the transaction. + * @param[in] txn A transaction handle returned by #mdb_txn_begin() + * @param[in] dbi A database handle returned by #mdb_dbi_open() + * @param[in] key The key to search for in the database + * @param[out] data The data corresponding to the key + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>#MDB_NOTFOUND - the key was not in the database. + * <li>EINVAL - an invalid parameter was specified. + * </ul> + */ +int mdb_get(MDB_txn *txn, MDB_dbi dbi, MDB_val *key, MDB_val *data); + + /** @brief Store items into a database. + * + * This function stores key/data pairs in the database. The default behavior + * is to enter the new key/data pair, replacing any previously existing key + * if duplicates are disallowed, or adding a duplicate data item if + * duplicates are allowed (#MDB_DUPSORT). + * @param[in] txn A transaction handle returned by #mdb_txn_begin() + * @param[in] dbi A database handle returned by #mdb_dbi_open() + * @param[in] key The key to store in the database + * @param[in,out] data The data to store + * @param[in] flags Special options for this operation. This parameter + * must be set to 0 or by bitwise OR'ing together one or more of the + * values described here. + * <ul> + * <li>#MDB_NODUPDATA - enter the new key/data pair only if it does not + * already appear in the database. This flag may only be specified + * if the database was opened with #MDB_DUPSORT. The function will + * return #MDB_KEYEXIST if the key/data pair already appears in the + * database. + * <li>#MDB_NOOVERWRITE - enter the new key/data pair only if the key + * does not already appear in the database. The function will return + * #MDB_KEYEXIST if the key already appears in the database, even if + * the database supports duplicates (#MDB_DUPSORT). The \b data + * parameter will be set to point to the existing item. + * <li>#MDB_RESERVE - reserve space for data of the given size, but + * don't copy the given data. Instead, return a pointer to the + * reserved space, which the caller can fill in later - before + * the next update operation or the transaction ends. This saves + * an extra memcpy if the data is being generated later. + * LMDB does nothing else with this memory, the caller is expected + * to modify all of the space requested. This flag must not be + * specified if the database was opened with #MDB_DUPSORT. + * <li>#MDB_APPEND - append the given key/data pair to the end of the + * database. This option allows fast bulk loading when keys are + * already known to be in the correct order. Loading unsorted keys + * with this flag will cause a #MDB_KEYEXIST error. + * <li>#MDB_APPENDDUP - as above, but for sorted dup data. + * </ul> + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>#MDB_MAP_FULL - the database is full, see #mdb_env_set_mapsize(). + * <li>#MDB_TXN_FULL - the transaction has too many dirty pages. + * <li>EACCES - an attempt was made to write in a read-only transaction. + * <li>EINVAL - an invalid parameter was specified. + * </ul> + */ +int mdb_put(MDB_txn *txn, MDB_dbi dbi, MDB_val *key, MDB_val *data, + unsigned int flags); + + /** @brief Delete items from a database. + * + * This function removes key/data pairs from the database. + * If the database does not support sorted duplicate data items + * (#MDB_DUPSORT) the data parameter is ignored. + * If the database supports sorted duplicates and the data parameter + * is NULL, all of the duplicate data items for the key will be + * deleted. Otherwise, if the data parameter is non-NULL + * only the matching data item will be deleted. + * This function will return #MDB_NOTFOUND if the specified key/data + * pair is not in the database. + * @param[in] txn A transaction handle returned by #mdb_txn_begin() + * @param[in] dbi A database handle returned by #mdb_dbi_open() + * @param[in] key The key to delete from the database + * @param[in] data The data to delete + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>EACCES - an attempt was made to write in a read-only transaction. + * <li>EINVAL - an invalid parameter was specified. + * </ul> + */ +int mdb_del(MDB_txn *txn, MDB_dbi dbi, MDB_val *key, MDB_val *data); + + /** @brief Create a cursor handle. + * + * A cursor is associated with a specific transaction and database. + * A cursor cannot be used when its database handle is closed. Nor + * when its transaction has ended, except with #mdb_cursor_renew(). + * It can be discarded with #mdb_cursor_close(). + * A cursor in a write-transaction can be closed before its transaction + * ends, and will otherwise be closed when its transaction ends. + * A cursor in a read-only transaction must be closed explicitly, before + * or after its transaction ends. It can be reused with + * #mdb_cursor_renew() before finally closing it. + * @note Earlier documentation said that cursors in every transaction + * were closed when the transaction committed or aborted. + * @param[in] txn A transaction handle returned by #mdb_txn_begin() + * @param[in] dbi A database handle returned by #mdb_dbi_open() + * @param[out] cursor Address where the new #MDB_cursor handle will be stored + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>EINVAL - an invalid parameter was specified. + * </ul> + */ +int mdb_cursor_open(MDB_txn *txn, MDB_dbi dbi, MDB_cursor **cursor); + + /** @brief Close a cursor handle. + * + * The cursor handle will be freed and must not be used again after this call. + * Its transaction must still be live if it is a write-transaction. + * @param[in] cursor A cursor handle returned by #mdb_cursor_open() + */ +void mdb_cursor_close(MDB_cursor *cursor); + + /** @brief Renew a cursor handle. + * + * A cursor is associated with a specific transaction and database. + * Cursors that are only used in read-only + * transactions may be re-used, to avoid unnecessary malloc/free overhead. + * The cursor may be associated with a new read-only transaction, and + * referencing the same database handle as it was created with. + * This may be done whether the previous transaction is live or dead. + * @param[in] txn A transaction handle returned by #mdb_txn_begin() + * @param[in] cursor A cursor handle returned by #mdb_cursor_open() + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>EINVAL - an invalid parameter was specified. + * </ul> + */ +int mdb_cursor_renew(MDB_txn *txn, MDB_cursor *cursor); + + /** @brief Return the cursor's transaction handle. + * + * @param[in] cursor A cursor handle returned by #mdb_cursor_open() + */ +MDB_txn *mdb_cursor_txn(MDB_cursor *cursor); + + /** @brief Return the cursor's database handle. + * + * @param[in] cursor A cursor handle returned by #mdb_cursor_open() + */ +MDB_dbi mdb_cursor_dbi(MDB_cursor *cursor); + + /** @brief Retrieve by cursor. + * + * This function retrieves key/data pairs from the database. The address and length + * of the key are returned in the object to which \b key refers (except for the + * case of the #MDB_SET option, in which the \b key object is unchanged), and + * the address and length of the data are returned in the object to which \b data + * refers. + * See #mdb_get() for restrictions on using the output values. + * @param[in] cursor A cursor handle returned by #mdb_cursor_open() + * @param[in,out] key The key for a retrieved item + * @param[in,out] data The data of a retrieved item + * @param[in] op A cursor operation #MDB_cursor_op + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>#MDB_NOTFOUND - no matching key found. + * <li>EINVAL - an invalid parameter was specified. + * </ul> + */ +int mdb_cursor_get(MDB_cursor *cursor, MDB_val *key, MDB_val *data, + MDB_cursor_op op); + + /** @brief Store by cursor. + * + * This function stores key/data pairs into the database. + * The cursor is positioned at the new item, or on failure usually near it. + * @note Earlier documentation incorrectly said errors would leave the + * state of the cursor unchanged. + * @param[in] cursor A cursor handle returned by #mdb_cursor_open() + * @param[in] key The key operated on. + * @param[in] data The data operated on. + * @param[in] flags Options for this operation. This parameter + * must be set to 0 or one of the values described here. + * <ul> + * <li>#MDB_CURRENT - replace the item at the current cursor position. + * The \b key parameter must still be provided, and must match it. + * If using sorted duplicates (#MDB_DUPSORT) the data item must still + * sort into the same place. This is intended to be used when the + * new data is the same size as the old. Otherwise it will simply + * perform a delete of the old record followed by an insert. + * <li>#MDB_NODUPDATA - enter the new key/data pair only if it does not + * already appear in the database. This flag may only be specified + * if the database was opened with #MDB_DUPSORT. The function will + * return #MDB_KEYEXIST if the key/data pair already appears in the + * database. + * <li>#MDB_NOOVERWRITE - enter the new key/data pair only if the key + * does not already appear in the database. The function will return + * #MDB_KEYEXIST if the key already appears in the database, even if + * the database supports duplicates (#MDB_DUPSORT). + * <li>#MDB_RESERVE - reserve space for data of the given size, but + * don't copy the given data. Instead, return a pointer to the + * reserved space, which the caller can fill in later - before + * the next update operation or the transaction ends. This saves + * an extra memcpy if the data is being generated later. This flag + * must not be specified if the database was opened with #MDB_DUPSORT. + * <li>#MDB_APPEND - append the given key/data pair to the end of the + * database. No key comparisons are performed. This option allows + * fast bulk loading when keys are already known to be in the + * correct order. Loading unsorted keys with this flag will cause + * a #MDB_KEYEXIST error. + * <li>#MDB_APPENDDUP - as above, but for sorted dup data. + * <li>#MDB_MULTIPLE - store multiple contiguous data elements in a + * single request. This flag may only be specified if the database + * was opened with #MDB_DUPFIXED. The \b data argument must be an + * array of two MDB_vals. The mv_size of the first MDB_val must be + * the size of a single data element. The mv_data of the first MDB_val + * must point to the beginning of the array of contiguous data elements. + * The mv_size of the second MDB_val must be the count of the number + * of data elements to store. On return this field will be set to + * the count of the number of elements actually written. The mv_data + * of the second MDB_val is unused. + * </ul> + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>#MDB_MAP_FULL - the database is full, see #mdb_env_set_mapsize(). + * <li>#MDB_TXN_FULL - the transaction has too many dirty pages. + * <li>EACCES - an attempt was made to write in a read-only transaction. + * <li>EINVAL - an invalid parameter was specified. + * </ul> + */ +int mdb_cursor_put(MDB_cursor *cursor, MDB_val *key, MDB_val *data, + unsigned int flags); + + /** @brief Delete current key/data pair + * + * This function deletes the key/data pair to which the cursor refers. + * @param[in] cursor A cursor handle returned by #mdb_cursor_open() + * @param[in] flags Options for this operation. This parameter + * must be set to 0 or one of the values described here. + * <ul> + * <li>#MDB_NODUPDATA - delete all of the data items for the current key. + * This flag may only be specified if the database was opened with #MDB_DUPSORT. + * </ul> + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>EACCES - an attempt was made to write in a read-only transaction. + * <li>EINVAL - an invalid parameter was specified. + * </ul> + */ +int mdb_cursor_del(MDB_cursor *cursor, unsigned int flags); + + /** @brief Return count of duplicates for current key. + * + * This call is only valid on databases that support sorted duplicate + * data items #MDB_DUPSORT. + * @param[in] cursor A cursor handle returned by #mdb_cursor_open() + * @param[out] countp Address where the count will be stored + * @return A non-zero error value on failure and 0 on success. Some possible + * errors are: + * <ul> + * <li>EINVAL - cursor is not initialized, or an invalid parameter was specified. + * </ul> + */ +int mdb_cursor_count(MDB_cursor *cursor, size_t *countp); + + /** @brief Compare two data items according to a particular database. + * + * This returns a comparison as if the two data items were keys in the + * specified database. + * @param[in] txn A transaction handle returned by #mdb_txn_begin() + * @param[in] dbi A database handle returned by #mdb_dbi_open() + * @param[in] a The first item to compare + * @param[in] b The second item to compare + * @return < 0 if a < b, 0 if a == b, > 0 if a > b + */ +int mdb_cmp(MDB_txn *txn, MDB_dbi dbi, const MDB_val *a, const MDB_val *b); + + /** @brief Compare two data items according to a particular database. + * + * This returns a comparison as if the two items were data items of + * the specified database. The database must have the #MDB_DUPSORT flag. + * @param[in] txn A transaction handle returned by #mdb_txn_begin() + * @param[in] dbi A database handle returned by #mdb_dbi_open() + * @param[in] a The first item to compare + * @param[in] b The second item to compare + * @return < 0 if a < b, 0 if a == b, > 0 if a > b + */ +int mdb_dcmp(MDB_txn *txn, MDB_dbi dbi, const MDB_val *a, const MDB_val *b); + + /** @brief A callback function used to print a message from the library. + * + * @param[in] msg The string to be printed. + * @param[in] ctx An arbitrary context pointer for the callback. + * @return < 0 on failure, >= 0 on success. + */ +typedef int (MDB_msg_func)(const char *msg, void *ctx); + + /** @brief Dump the entries in the reader lock table. + * + * @param[in] env An environment handle returned by #mdb_env_create() + * @param[in] func A #MDB_msg_func function + * @param[in] ctx Anything the message function needs + * @return < 0 on failure, >= 0 on success. + */ +int mdb_reader_list(MDB_env *env, MDB_msg_func *func, void *ctx); + + /** @brief Check for stale entries in the reader lock table. + * + * @param[in] env An environment handle returned by #mdb_env_create() + * @param[out] dead Number of stale slots that were cleared + * @return 0 on success, non-zero on failure. + */ +int mdb_reader_check(MDB_env *env, int *dead); +/** @} */ + +#ifdef __cplusplus +} +#endif +/** @page tools LMDB Command Line Tools + The following describes the command line tools that are available for LMDB. + \li \ref mdb_copy_1 + \li \ref mdb_dump_1 + \li \ref mdb_load_1 + \li \ref mdb_stat_1 +*/ + +#endif /* _LMDB_H_ */ diff --git a/src/contrib/lmdb/mdb.c b/src/contrib/lmdb/mdb.c new file mode 100644 index 0000000..d9e7c5e --- /dev/null +++ b/src/contrib/lmdb/mdb.c @@ -0,0 +1,10266 @@ +/** @file mdb.c + * @brief Lightning memory-mapped database library + * + * A Btree-based database management library modeled loosely on the + * BerkeleyDB API, but much simplified. + */ +/* + * Copyright 2011-2018 Howard Chu, Symas Corp. + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted only as authorized by the OpenLDAP + * Public License. + * + * A copy of this license is available in the file LICENSE in the + * top-level directory of the distribution or, alternatively, at + * <http://www.OpenLDAP.org/license.html>. + * + * This code is derived from btree.c written by Martin Hedenfalk. + * + * Copyright (c) 2009, 2010 Martin Hedenfalk <martin@bzero.se> + * + * Permission to use, copy, modify, and distribute this software for any + * purpose with or without fee is hereby granted, provided that the above + * copyright notice and this permission notice appear in all copies. + * + * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES + * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF + * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR + * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES + * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN + * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF + * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. + */ +#ifndef _GNU_SOURCE +#define _GNU_SOURCE 1 +#endif +#if defined(__WIN64__) +#define _FILE_OFFSET_BITS 64 +#endif +#ifdef _WIN32 +#include <malloc.h> +#include <windows.h> +#include <wchar.h> /* get wcscpy() */ + +/** getpid() returns int; MinGW defines pid_t but MinGW64 typedefs it + * as int64 which is wrong. MSVC doesn't define it at all, so just + * don't use it. + */ +#define MDB_PID_T int +#define MDB_THR_T DWORD +#include <sys/types.h> +#include <sys/stat.h> +#ifdef __GNUC__ +# include <sys/param.h> +#else +# define LITTLE_ENDIAN 1234 +# define BIG_ENDIAN 4321 +# define BYTE_ORDER LITTLE_ENDIAN +# ifndef SSIZE_MAX +# define SSIZE_MAX INT_MAX +# endif +#endif +#else +#include <sys/types.h> +#include <sys/stat.h> +#define MDB_PID_T pid_t +#define MDB_THR_T pthread_t +#include <sys/param.h> +#include <sys/uio.h> +#include <sys/mman.h> +#ifdef HAVE_SYS_FILE_H +#include <sys/file.h> +#endif +#include <fcntl.h> +#endif + +#if defined(__mips) && defined(__linux) +/* MIPS has cache coherency issues, requires explicit cache control */ +#include <asm/cachectl.h> +extern int cacheflush(char *addr, int nbytes, int cache); +#define CACHEFLUSH(addr, bytes, cache) cacheflush(addr, bytes, cache) +#else +#define CACHEFLUSH(addr, bytes, cache) +#endif + +#if defined(__linux) && !defined(MDB_FDATASYNC_WORKS) +/** fdatasync is broken on ext3/ext4fs on older kernels, see + * description in #mdb_env_open2 comments. You can safely + * define MDB_FDATASYNC_WORKS if this code will only be run + * on kernels 3.6 and newer. + */ +#define BROKEN_FDATASYNC +#endif + +#include <errno.h> +#include <limits.h> +#include <stddef.h> +#include <inttypes.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <time.h> + +#ifdef _MSC_VER +#include <io.h> +typedef SSIZE_T ssize_t; +#else +#include <unistd.h> +#endif + +#if defined(__sun) || defined(ANDROID) +/* Most platforms have posix_memalign, older may only have memalign */ +#define HAVE_MEMALIGN 1 +#include <malloc.h> +/* On Solaris, we need the POSIX sigwait function */ +#if defined (__sun) +# define _POSIX_PTHREAD_SEMANTICS 1 +#endif +#endif + +#if !(defined(BYTE_ORDER) || defined(__BYTE_ORDER)) +#include <netinet/in.h> +#include <resolv.h> /* defines BYTE_ORDER on HPUX and Solaris */ +#endif + +#if defined(__APPLE__) || defined (BSD) || defined(__FreeBSD_kernel__) +# define MDB_USE_POSIX_SEM 1 +# define MDB_FDATASYNC fsync +#elif defined(ANDROID) +# define MDB_FDATASYNC fsync +#endif + +#ifndef _WIN32 +#include <pthread.h> +#include <signal.h> +#ifdef MDB_USE_POSIX_SEM +# define MDB_USE_HASH 1 +#include <semaphore.h> +#else +#define MDB_USE_POSIX_MUTEX 1 +#endif +#endif + +#if defined(_WIN32) + defined(MDB_USE_POSIX_SEM) \ + + defined(MDB_USE_POSIX_MUTEX) != 1 +# error "Ambiguous shared-lock implementation" +#endif + +#ifdef USE_VALGRIND +#include <valgrind/memcheck.h> +#define VGMEMP_CREATE(h,r,z) VALGRIND_CREATE_MEMPOOL(h,r,z) +#define VGMEMP_ALLOC(h,a,s) VALGRIND_MEMPOOL_ALLOC(h,a,s) +#define VGMEMP_FREE(h,a) VALGRIND_MEMPOOL_FREE(h,a) +#define VGMEMP_DESTROY(h) VALGRIND_DESTROY_MEMPOOL(h) +#define VGMEMP_DEFINED(a,s) VALGRIND_MAKE_MEM_DEFINED(a,s) +#else +#define VGMEMP_CREATE(h,r,z) +#define VGMEMP_ALLOC(h,a,s) +#define VGMEMP_FREE(h,a) +#define VGMEMP_DESTROY(h) +#define VGMEMP_DEFINED(a,s) +#endif + +#ifndef BYTE_ORDER +# if (defined(_LITTLE_ENDIAN) || defined(_BIG_ENDIAN)) && !(defined(_LITTLE_ENDIAN) && defined(_BIG_ENDIAN)) +/* Solaris just defines one or the other */ +# define LITTLE_ENDIAN 1234 +# define BIG_ENDIAN 4321 +# ifdef _LITTLE_ENDIAN +# define BYTE_ORDER LITTLE_ENDIAN +# else +# define BYTE_ORDER BIG_ENDIAN +# endif +# else +# define BYTE_ORDER __BYTE_ORDER +# endif +#endif + +#ifndef LITTLE_ENDIAN +#define LITTLE_ENDIAN __LITTLE_ENDIAN +#endif +#ifndef BIG_ENDIAN +#define BIG_ENDIAN __BIG_ENDIAN +#endif + +#if defined(__i386) || defined(__x86_64) || defined(_M_IX86) +#define MISALIGNED_OK 1 +#endif + +#include "lmdb.h" +#include "midl.h" + +#if (BYTE_ORDER == LITTLE_ENDIAN) == (BYTE_ORDER == BIG_ENDIAN) +# error "Unknown or unsupported endianness (BYTE_ORDER)" +#elif (-6 & 5) || CHAR_BIT != 8 || UINT_MAX < 0xffffffff || ULONG_MAX % 0xFFFF +# error "Two's complement, reasonably sized integer types, please" +#endif + +#ifdef __GNUC__ +/** Put infrequently used env functions in separate section */ +# ifdef __APPLE__ +# define ESECT __attribute__ ((section("__TEXT,text_env"))) +# else +# define ESECT __attribute__ ((section("text_env"))) +# endif +#else +#define ESECT +#endif + +#ifdef _WIN32 +#define CALL_CONV WINAPI +#else +#define CALL_CONV +#endif + +/** @defgroup internal LMDB Internals + * @{ + */ +/** @defgroup compat Compatibility Macros + * A bunch of macros to minimize the amount of platform-specific ifdefs + * needed throughout the rest of the code. When the features this library + * needs are similar enough to POSIX to be hidden in a one-or-two line + * replacement, this macro approach is used. + * @{ + */ + + /** Features under development */ +#ifndef MDB_DEVEL +#define MDB_DEVEL 0 +#endif + + /** Wrapper around __func__, which is a C99 feature */ +#if __STDC_VERSION__ >= 199901L +# define mdb_func_ __func__ +#elif __GNUC__ >= 2 || _MSC_VER >= 1300 +# define mdb_func_ __FUNCTION__ +#else +/* If a debug message says <mdb_unknown>(), update the #if statements above */ +# define mdb_func_ "<mdb_unknown>" +#endif + +/* Internal error codes, not exposed outside liblmdb */ +#define MDB_NO_ROOT (MDB_LAST_ERRCODE + 10) +#ifdef _WIN32 +#define MDB_OWNERDEAD ((int) WAIT_ABANDONED) +#elif defined(MDB_USE_POSIX_MUTEX) && defined(EOWNERDEAD) +#define MDB_OWNERDEAD EOWNERDEAD /**< #LOCK_MUTEX0() result if dead owner */ +#endif + +#ifdef __GLIBC__ +#define GLIBC_VER ((__GLIBC__ << 16 )| __GLIBC_MINOR__) +#endif +/** Some platforms define the EOWNERDEAD error code + * even though they don't support Robust Mutexes. + * Compile with -DMDB_USE_ROBUST=0, or use some other + * mechanism like -DMDB_USE_POSIX_SEM instead of + * -DMDB_USE_POSIX_MUTEX. + * (Posix semaphores are not robust.) + */ +#ifndef MDB_USE_ROBUST +/* Android currently lacks Robust Mutex support. So does glibc < 2.4. */ +# if defined(MDB_USE_POSIX_MUTEX) && (defined(ANDROID) || \ + (defined(__GLIBC__) && GLIBC_VER < 0x020004)) +# define MDB_USE_ROBUST 0 +# else +# define MDB_USE_ROBUST 1 +# endif +#endif /* !MDB_USE_ROBUST */ + +#if defined(MDB_USE_POSIX_MUTEX) && (MDB_USE_ROBUST) +/* glibc < 2.12 only provided _np API */ +# if (defined(__GLIBC__) && GLIBC_VER < 0x02000c) || \ + (defined(PTHREAD_MUTEX_ROBUST_NP) && !defined(PTHREAD_MUTEX_ROBUST)) +# define PTHREAD_MUTEX_ROBUST PTHREAD_MUTEX_ROBUST_NP +# define pthread_mutexattr_setrobust(attr, flag) pthread_mutexattr_setrobust_np(attr, flag) +# define pthread_mutex_consistent(mutex) pthread_mutex_consistent_np(mutex) +# endif +#endif /* MDB_USE_POSIX_MUTEX && MDB_USE_ROBUST */ + +#if defined(MDB_OWNERDEAD) && (MDB_USE_ROBUST) +#define MDB_ROBUST_SUPPORTED 1 +#endif + +#ifdef _WIN32 +#define MDB_USE_HASH 1 +#define MDB_PIDLOCK 0 +#define THREAD_RET DWORD +#define pthread_t HANDLE +#define pthread_mutex_t HANDLE +#define pthread_cond_t HANDLE +typedef HANDLE mdb_mutex_t, mdb_mutexref_t; +#define pthread_key_t DWORD +#define pthread_self() GetCurrentThreadId() +#define pthread_key_create(x,y) \ + ((*(x) = TlsAlloc()) == TLS_OUT_OF_INDEXES ? ErrCode() : 0) +#define pthread_key_delete(x) TlsFree(x) +#define pthread_getspecific(x) TlsGetValue(x) +#define pthread_setspecific(x,y) (TlsSetValue(x,y) ? 0 : ErrCode()) +#define pthread_mutex_unlock(x) ReleaseMutex(*x) +#define pthread_mutex_lock(x) WaitForSingleObject(*x, INFINITE) +#define pthread_cond_signal(x) SetEvent(*x) +#define pthread_cond_wait(cond,mutex) do{SignalObjectAndWait(*mutex, *cond, INFINITE, FALSE); WaitForSingleObject(*mutex, INFINITE);}while(0) +#define THREAD_CREATE(thr,start,arg) \ + (((thr) = CreateThread(NULL, 0, start, arg, 0, NULL)) ? 0 : ErrCode()) +#define THREAD_FINISH(thr) \ + (WaitForSingleObject(thr, INFINITE) ? ErrCode() : 0) +#define LOCK_MUTEX0(mutex) WaitForSingleObject(mutex, INFINITE) +#define UNLOCK_MUTEX(mutex) ReleaseMutex(mutex) +#define mdb_mutex_consistent(mutex) 0 +#define getpid() GetCurrentProcessId() +#define MDB_FDATASYNC(fd) (!FlushFileBuffers(fd)) +#define MDB_MSYNC(addr,len,flags) (!FlushViewOfFile(addr,len)) +#define ErrCode() GetLastError() +#define GET_PAGESIZE(x) {SYSTEM_INFO si; GetSystemInfo(&si); (x) = si.dwPageSize;} +#define close(fd) (CloseHandle(fd) ? 0 : -1) +#define munmap(ptr,len) UnmapViewOfFile(ptr) +#ifdef PROCESS_QUERY_LIMITED_INFORMATION +#define MDB_PROCESS_QUERY_LIMITED_INFORMATION PROCESS_QUERY_LIMITED_INFORMATION +#else +#define MDB_PROCESS_QUERY_LIMITED_INFORMATION 0x1000 +#endif +#define Z "I" +#else +#define THREAD_RET void * +#define THREAD_CREATE(thr,start,arg) pthread_create(&thr,NULL,start,arg) +#define THREAD_FINISH(thr) pthread_join(thr,NULL) +#define Z "z" /**< printf format modifier for size_t */ + + /** For MDB_LOCK_FORMAT: True if readers take a pid lock in the lockfile */ +#define MDB_PIDLOCK 1 + +#ifdef MDB_USE_POSIX_SEM + +typedef sem_t *mdb_mutex_t, *mdb_mutexref_t; +#define LOCK_MUTEX0(mutex) mdb_sem_wait(mutex) +#define UNLOCK_MUTEX(mutex) sem_post(mutex) + +static int +mdb_sem_wait(sem_t *sem) +{ + int rc; + while ((rc = sem_wait(sem)) && (rc = errno) == EINTR) ; + return rc; +} + +#else /* MDB_USE_POSIX_MUTEX: */ + /** Shared mutex/semaphore as the original is stored. + * + * Not for copies. Instead it can be assigned to an #mdb_mutexref_t. + * When mdb_mutexref_t is a pointer and mdb_mutex_t is not, then it + * is array[size 1] so it can be assigned to the pointer. + */ +typedef pthread_mutex_t mdb_mutex_t[1]; + /** Reference to an #mdb_mutex_t */ +typedef pthread_mutex_t *mdb_mutexref_t; + /** Lock the reader or writer mutex. + * Returns 0 or a code to give #mdb_mutex_failed(), as in #LOCK_MUTEX(). + */ +#define LOCK_MUTEX0(mutex) pthread_mutex_lock(mutex) + /** Unlock the reader or writer mutex. + */ +#define UNLOCK_MUTEX(mutex) pthread_mutex_unlock(mutex) + /** Mark mutex-protected data as repaired, after death of previous owner. + */ +#define mdb_mutex_consistent(mutex) pthread_mutex_consistent(mutex) +#endif /* MDB_USE_POSIX_SEM */ + + /** Get the error code for the last failed system function. + */ +#define ErrCode() errno + + /** An abstraction for a file handle. + * On POSIX systems file handles are small integers. On Windows + * they're opaque pointers. + */ +#define HANDLE int + + /** A value for an invalid file handle. + * Mainly used to initialize file variables and signify that they are + * unused. + */ +#define INVALID_HANDLE_VALUE (-1) + + /** Get the size of a memory page for the system. + * This is the basic size that the platform's memory manager uses, and is + * fundamental to the use of memory-mapped files. + */ +#define GET_PAGESIZE(x) ((x) = sysconf(_SC_PAGE_SIZE)) +#endif + +#if defined(_WIN32) || defined(MDB_USE_POSIX_SEM) +#define MNAME_LEN 32 +#else +#define MNAME_LEN (sizeof(pthread_mutex_t)) +#endif + +/** @} */ + +#ifdef MDB_ROBUST_SUPPORTED + /** Lock mutex, handle any error, set rc = result. + * Return 0 on success, nonzero (not rc) on error. + */ +#define LOCK_MUTEX(rc, env, mutex) \ + (((rc) = LOCK_MUTEX0(mutex)) && \ + ((rc) = mdb_mutex_failed(env, mutex, rc))) +static int mdb_mutex_failed(MDB_env *env, mdb_mutexref_t mutex, int rc); +#else +#define LOCK_MUTEX(rc, env, mutex) ((rc) = LOCK_MUTEX0(mutex)) +#define mdb_mutex_failed(env, mutex, rc) (rc) +#endif + +#ifndef _WIN32 +/** A flag for opening a file and requesting synchronous data writes. + * This is only used when writing a meta page. It's not strictly needed; + * we could just do a normal write and then immediately perform a flush. + * But if this flag is available it saves us an extra system call. + * + * @note If O_DSYNC is undefined but exists in /usr/include, + * preferably set some compiler flag to get the definition. + */ +#ifndef MDB_DSYNC +# ifdef O_DSYNC +# define MDB_DSYNC O_DSYNC +# else +# define MDB_DSYNC O_SYNC +# endif +#endif +#endif + +/** Function for flushing the data of a file. Define this to fsync + * if fdatasync() is not supported. + */ +#ifndef MDB_FDATASYNC +# define MDB_FDATASYNC fdatasync +#endif + +#ifndef MDB_MSYNC +# define MDB_MSYNC(addr,len,flags) msync(addr,len,flags) +#endif + +#ifndef MS_SYNC +#define MS_SYNC 1 +#endif + +#ifndef MS_ASYNC +#define MS_ASYNC 0 +#endif + + /** A page number in the database. + * Note that 64 bit page numbers are overkill, since pages themselves + * already represent 12-13 bits of addressable memory, and the OS will + * always limit applications to a maximum of 63 bits of address space. + * + * @note In the #MDB_node structure, we only store 48 bits of this value, + * which thus limits us to only 60 bits of addressable data. + */ +typedef MDB_ID pgno_t; + + /** A transaction ID. + * See struct MDB_txn.mt_txnid for details. + */ +typedef MDB_ID txnid_t; + +/** @defgroup debug Debug Macros + * @{ + */ +#ifndef MDB_DEBUG + /** Enable debug output. Needs variable argument macros (a C99 feature). + * Set this to 1 for copious tracing. Set to 2 to add dumps of all IDLs + * read from and written to the database (used for free space management). + */ +#define MDB_DEBUG 0 +#endif + +#if MDB_DEBUG +static int mdb_debug; +static txnid_t mdb_debug_start; + + /** Print a debug message with printf formatting. + * Requires double parenthesis around 2 or more args. + */ +# define DPRINTF(args) ((void) ((mdb_debug) && DPRINTF0 args)) +# define DPRINTF0(fmt, ...) \ + fprintf(stderr, "%s:%d " fmt "\n", mdb_func_, __LINE__, __VA_ARGS__) +#else +# define DPRINTF(args) ((void) 0) +#endif + /** Print a debug string. + * The string is printed literally, with no format processing. + */ +#define DPUTS(arg) DPRINTF(("%s", arg)) + /** Debuging output value of a cursor DBI: Negative in a sub-cursor. */ +#define DDBI(mc) \ + (((mc)->mc_flags & C_SUB) ? -(int)(mc)->mc_dbi : (int)(mc)->mc_dbi) +/** @} */ + + /** @brief The maximum size of a database page. + * + * It is 32k or 64k, since value-PAGEBASE must fit in + * #MDB_page.%mp_upper. + * + * LMDB will use database pages < OS pages if needed. + * That causes more I/O in write transactions: The OS must + * know (read) the whole page before writing a partial page. + * + * Note that we don't currently support Huge pages. On Linux, + * regular data files cannot use Huge pages, and in general + * Huge pages aren't actually pageable. We rely on the OS + * demand-pager to read our data and page it out when memory + * pressure from other processes is high. So until OSs have + * actual paging support for Huge pages, they're not viable. + */ +#define MAX_PAGESIZE (PAGEBASE ? 0x10000 : 0x8000) + + /** The minimum number of keys required in a database page. + * Setting this to a larger value will place a smaller bound on the + * maximum size of a data item. Data items larger than this size will + * be pushed into overflow pages instead of being stored directly in + * the B-tree node. This value used to default to 4. With a page size + * of 4096 bytes that meant that any item larger than 1024 bytes would + * go into an overflow page. That also meant that on average 2-3KB of + * each overflow page was wasted space. The value cannot be lower than + * 2 because then there would no longer be a tree structure. With this + * value, items larger than 2KB will go into overflow pages, and on + * average only 1KB will be wasted. + */ +#define MDB_MINKEYS 2 + + /** A stamp that identifies a file as an LMDB file. + * There's nothing special about this value other than that it is easily + * recognizable, and it will reflect any byte order mismatches. + */ +#define MDB_MAGIC 0xBEEFC0DE + + /** The version number for a database's datafile format. */ +#define MDB_DATA_VERSION ((MDB_DEVEL) ? 999 : 1) + /** The version number for a database's lockfile format. */ +#define MDB_LOCK_VERSION 1 + + /** @brief The max size of a key we can write, or 0 for computed max. + * + * This macro should normally be left alone or set to 0. + * Note that a database with big keys or dupsort data cannot be + * reliably modified by a liblmdb which uses a smaller max. + * The default is 511 for backwards compat, or 0 when #MDB_DEVEL. + * + * Other values are allowed, for backwards compat. However: + * A value bigger than the computed max can break if you do not + * know what you are doing, and liblmdb <= 0.9.10 can break when + * modifying a DB with keys/dupsort data bigger than its max. + * + * Data items in an #MDB_DUPSORT database are also limited to + * this size, since they're actually keys of a sub-DB. Keys and + * #MDB_DUPSORT data items must fit on a node in a regular page. + */ +#ifndef MDB_MAXKEYSIZE +#define MDB_MAXKEYSIZE ((MDB_DEVEL) ? 0 : 511) +#endif + + /** The maximum size of a key we can write to the environment. */ +#if MDB_MAXKEYSIZE +#define ENV_MAXKEY(env) (MDB_MAXKEYSIZE) +#else +#define ENV_MAXKEY(env) ((env)->me_maxkey) +#endif + + /** @brief The maximum size of a data item. + * + * We only store a 32 bit value for node sizes. + */ +#define MAXDATASIZE 0xffffffffUL + +#if MDB_DEBUG + /** Key size which fits in a #DKBUF. + * @ingroup debug + */ +#define DKBUF_MAXKEYSIZE ((MDB_MAXKEYSIZE) > 0 ? (MDB_MAXKEYSIZE) : 511) + /** A key buffer. + * @ingroup debug + * This is used for printing a hex dump of a key's contents. + */ +#define DKBUF char kbuf[DKBUF_MAXKEYSIZE*2+1] + /** Display a key in hex. + * @ingroup debug + * Invoke a function to display a key in hex. + */ +#define DKEY(x) mdb_dkey(x, kbuf) +#else +#define DKBUF +#define DKEY(x) 0 +#endif + + /** An invalid page number. + * Mainly used to denote an empty tree. + */ +#define P_INVALID (~(pgno_t)0) + + /** Test if the flags \b f are set in a flag word \b w. */ +#define F_ISSET(w, f) (((w) & (f)) == (f)) + + /** Round \b n up to an even number. */ +#define EVEN(n) (((n) + 1U) & -2) /* sign-extending -2 to match n+1U */ + + /** Used for offsets within a single page. + * Since memory pages are typically 4 or 8KB in size, 12-13 bits, + * this is plenty. + */ +typedef uint16_t indx_t; + + /** Default size of memory map. + * This is certainly too small for any actual applications. Apps should always set + * the size explicitly using #mdb_env_set_mapsize(). + */ +#define DEFAULT_MAPSIZE 1048576 + +/** @defgroup readers Reader Lock Table + * Readers don't acquire any locks for their data access. Instead, they + * simply record their transaction ID in the reader table. The reader + * mutex is needed just to find an empty slot in the reader table. The + * slot's address is saved in thread-specific data so that subsequent read + * transactions started by the same thread need no further locking to proceed. + * + * If #MDB_NOTLS is set, the slot address is not saved in thread-specific data. + * + * No reader table is used if the database is on a read-only filesystem, or + * if #MDB_NOLOCK is set. + * + * Since the database uses multi-version concurrency control, readers don't + * actually need any locking. This table is used to keep track of which + * readers are using data from which old transactions, so that we'll know + * when a particular old transaction is no longer in use. Old transactions + * that have discarded any data pages can then have those pages reclaimed + * for use by a later write transaction. + * + * The lock table is constructed such that reader slots are aligned with the + * processor's cache line size. Any slot is only ever used by one thread. + * This alignment guarantees that there will be no contention or cache + * thrashing as threads update their own slot info, and also eliminates + * any need for locking when accessing a slot. + * + * A writer thread will scan every slot in the table to determine the oldest + * outstanding reader transaction. Any freed pages older than this will be + * reclaimed by the writer. The writer doesn't use any locks when scanning + * this table. This means that there's no guarantee that the writer will + * see the most up-to-date reader info, but that's not required for correct + * operation - all we need is to know the upper bound on the oldest reader, + * we don't care at all about the newest reader. So the only consequence of + * reading stale information here is that old pages might hang around a + * while longer before being reclaimed. That's actually good anyway, because + * the longer we delay reclaiming old pages, the more likely it is that a + * string of contiguous pages can be found after coalescing old pages from + * many old transactions together. + * @{ + */ + /** Number of slots in the reader table. + * This value was chosen somewhat arbitrarily. 126 readers plus a + * couple mutexes fit exactly into 8KB on my development machine. + * Applications should set the table size using #mdb_env_set_maxreaders(). + */ +#define DEFAULT_READERS 126 + + /** The size of a CPU cache line in bytes. We want our lock structures + * aligned to this size to avoid false cache line sharing in the + * lock table. + * This value works for most CPUs. For Itanium this should be 128. + */ +#ifndef CACHELINE +#define CACHELINE 64 +#endif + + /** The information we store in a single slot of the reader table. + * In addition to a transaction ID, we also record the process and + * thread ID that owns a slot, so that we can detect stale information, + * e.g. threads or processes that went away without cleaning up. + * @note We currently don't check for stale records. We simply re-init + * the table when we know that we're the only process opening the + * lock file. + */ +typedef struct MDB_rxbody { + /** Current Transaction ID when this transaction began, or (txnid_t)-1. + * Multiple readers that start at the same time will probably have the + * same ID here. Again, it's not important to exclude them from + * anything; all we need to know is which version of the DB they + * started from so we can avoid overwriting any data used in that + * particular version. + */ + volatile txnid_t mrb_txnid; + /** The process ID of the process owning this reader txn. */ + volatile MDB_PID_T mrb_pid; + /** The thread ID of the thread owning this txn. */ + volatile MDB_THR_T mrb_tid; +} MDB_rxbody; + + /** The actual reader record, with cacheline padding. */ +typedef struct MDB_reader { + union { + MDB_rxbody mrx; + /** shorthand for mrb_txnid */ +#define mr_txnid mru.mrx.mrb_txnid +#define mr_pid mru.mrx.mrb_pid +#define mr_tid mru.mrx.mrb_tid + /** cache line alignment */ + char pad[(sizeof(MDB_rxbody)+CACHELINE-1) & ~(CACHELINE-1)]; + } mru; +} MDB_reader; + + /** The header for the reader table. + * The table resides in a memory-mapped file. (This is a different file + * than is used for the main database.) + * + * For POSIX the actual mutexes reside in the shared memory of this + * mapped file. On Windows, mutexes are named objects allocated by the + * kernel; we store the mutex names in this mapped file so that other + * processes can grab them. This same approach is also used on + * MacOSX/Darwin (using named semaphores) since MacOSX doesn't support + * process-shared POSIX mutexes. For these cases where a named object + * is used, the object name is derived from a 64 bit FNV hash of the + * environment pathname. As such, naming collisions are extremely + * unlikely. If a collision occurs, the results are unpredictable. + */ +typedef struct MDB_txbody { + /** Stamp identifying this as an LMDB file. It must be set + * to #MDB_MAGIC. */ + uint32_t mtb_magic; + /** Format of this lock file. Must be set to #MDB_LOCK_FORMAT. */ + uint32_t mtb_format; +#if defined(_WIN32) || defined(MDB_USE_POSIX_SEM) + char mtb_rmname[MNAME_LEN]; +#else + /** Mutex protecting access to this table. + * This is the reader table lock used with LOCK_MUTEX(). + */ + mdb_mutex_t mtb_rmutex; +#endif + /** The ID of the last transaction committed to the database. + * This is recorded here only for convenience; the value can always + * be determined by reading the main database meta pages. + */ + volatile txnid_t mtb_txnid; + /** The number of slots that have been used in the reader table. + * This always records the maximum count, it is not decremented + * when readers release their slots. + */ + volatile unsigned mtb_numreaders; +} MDB_txbody; + + /** The actual reader table definition. */ +typedef struct MDB_txninfo { + union { + MDB_txbody mtb; +#define mti_magic mt1.mtb.mtb_magic +#define mti_format mt1.mtb.mtb_format +#define mti_rmutex mt1.mtb.mtb_rmutex +#define mti_rmname mt1.mtb.mtb_rmname +#define mti_txnid mt1.mtb.mtb_txnid +#define mti_numreaders mt1.mtb.mtb_numreaders + char pad[(sizeof(MDB_txbody)+CACHELINE-1) & ~(CACHELINE-1)]; + } mt1; + union { +#if defined(_WIN32) || defined(MDB_USE_POSIX_SEM) + char mt2_wmname[MNAME_LEN]; +#define mti_wmname mt2.mt2_wmname +#else + mdb_mutex_t mt2_wmutex; +#define mti_wmutex mt2.mt2_wmutex +#endif + char pad[(MNAME_LEN+CACHELINE-1) & ~(CACHELINE-1)]; + } mt2; + MDB_reader mti_readers[1]; +} MDB_txninfo; + + /** Lockfile format signature: version, features and field layout */ +#define MDB_LOCK_FORMAT \ + ((uint32_t) \ + ((MDB_LOCK_VERSION) \ + /* Flags which describe functionality */ \ + + (((MDB_PIDLOCK) != 0) << 16))) +/** @} */ + +/** Common header for all page types. The page type depends on #mp_flags. + * + * #P_BRANCH and #P_LEAF pages have unsorted '#MDB_node's at the end, with + * sorted #mp_ptrs[] entries referring to them. Exception: #P_LEAF2 pages + * omit mp_ptrs and pack sorted #MDB_DUPFIXED values after the page header. + * + * #P_OVERFLOW records occupy one or more contiguous pages where only the + * first has a page header. They hold the real data of #F_BIGDATA nodes. + * + * #P_SUBP sub-pages are small leaf "pages" with duplicate data. + * A node with flag #F_DUPDATA but not #F_SUBDATA contains a sub-page. + * (Duplicate data can also go in sub-databases, which use normal pages.) + * + * #P_META pages contain #MDB_meta, the start point of an LMDB snapshot. + * + * Each non-metapage up to #MDB_meta.%mm_last_pg is reachable exactly once + * in the snapshot: Either used by a database or listed in a freeDB record. + */ +typedef struct MDB_page { +#define mp_pgno mp_p.p_pgno +#define mp_next mp_p.p_next + union { + pgno_t p_pgno; /**< page number */ + struct MDB_page *p_next; /**< for in-memory list of freed pages */ + } mp_p; + uint16_t mp_pad; /**< key size if this is a LEAF2 page */ +/** @defgroup mdb_page Page Flags + * @ingroup internal + * Flags for the page headers. + * @{ + */ +#define P_BRANCH 0x01 /**< branch page */ +#define P_LEAF 0x02 /**< leaf page */ +#define P_OVERFLOW 0x04 /**< overflow page */ +#define P_META 0x08 /**< meta page */ +#define P_DIRTY 0x10 /**< dirty page, also set for #P_SUBP pages */ +#define P_LEAF2 0x20 /**< for #MDB_DUPFIXED records */ +#define P_SUBP 0x40 /**< for #MDB_DUPSORT sub-pages */ +#define P_LOOSE 0x4000 /**< page was dirtied then freed, can be reused */ +#define P_KEEP 0x8000 /**< leave this page alone during spill */ +/** @} */ + uint16_t mp_flags; /**< @ref mdb_page */ +#define mp_lower mp_pb.pb.pb_lower +#define mp_upper mp_pb.pb.pb_upper +#define mp_pages mp_pb.pb_pages + union { + struct { + indx_t pb_lower; /**< lower bound of free space */ + indx_t pb_upper; /**< upper bound of free space */ + } pb; + uint32_t pb_pages; /**< number of overflow pages */ + } mp_pb; + indx_t mp_ptrs[1]; /**< dynamic size */ +} MDB_page; + + /** Size of the page header, excluding dynamic data at the end */ +#define PAGEHDRSZ ((unsigned) offsetof(MDB_page, mp_ptrs)) + + /** Address of first usable data byte in a page, after the header */ +#define METADATA(p) ((void *)((char *)(p) + PAGEHDRSZ)) + + /** ITS#7713, change PAGEBASE to handle 65536 byte pages */ +#define PAGEBASE ((MDB_DEVEL) ? PAGEHDRSZ : 0) + + /** Number of nodes on a page */ +#define NUMKEYS(p) (((p)->mp_lower - (PAGEHDRSZ-PAGEBASE)) >> 1) + + /** The amount of space remaining in the page */ +#define SIZELEFT(p) (indx_t)((p)->mp_upper - (p)->mp_lower) + + /** The percentage of space used in the page, in tenths of a percent. */ +#define PAGEFILL(env, p) (1000L * ((env)->me_psize - PAGEHDRSZ - SIZELEFT(p)) / \ + ((env)->me_psize - PAGEHDRSZ)) + /** The minimum page fill factor, in tenths of a percent. + * Pages emptier than this are candidates for merging. + */ +#define FILL_THRESHOLD 250 + + /** Test if a page is a leaf page */ +#define IS_LEAF(p) F_ISSET((p)->mp_flags, P_LEAF) + /** Test if a page is a LEAF2 page */ +#define IS_LEAF2(p) F_ISSET((p)->mp_flags, P_LEAF2) + /** Test if a page is a branch page */ +#define IS_BRANCH(p) F_ISSET((p)->mp_flags, P_BRANCH) + /** Test if a page is an overflow page */ +#define IS_OVERFLOW(p) F_ISSET((p)->mp_flags, P_OVERFLOW) + /** Test if a page is a sub page */ +#define IS_SUBP(p) F_ISSET((p)->mp_flags, P_SUBP) + + /** The number of overflow pages needed to store the given size. */ +#define OVPAGES(size, psize) ((PAGEHDRSZ-1 + (size)) / (psize) + 1) + + /** Link in #MDB_txn.%mt_loose_pgs list. + * Kept outside the page header, which is needed when reusing the page. + */ +#define NEXT_LOOSE_PAGE(p) (*(MDB_page **)((p) + 2)) + + /** Header for a single key/data pair within a page. + * Used in pages of type #P_BRANCH and #P_LEAF without #P_LEAF2. + * We guarantee 2-byte alignment for 'MDB_node's. + * + * #mn_lo and #mn_hi are used for data size on leaf nodes, and for child + * pgno on branch nodes. On 64 bit platforms, #mn_flags is also used + * for pgno. (Branch nodes have no flags). Lo and hi are in host byte + * order in case some accesses can be optimized to 32-bit word access. + * + * Leaf node flags describe node contents. #F_BIGDATA says the node's + * data part is the page number of an overflow page with actual data. + * #F_DUPDATA and #F_SUBDATA can be combined giving duplicate data in + * a sub-page/sub-database, and named databases (just #F_SUBDATA). + */ +typedef struct MDB_node { + /** part of data size or pgno + * @{ */ +#if BYTE_ORDER == LITTLE_ENDIAN + unsigned short mn_lo, mn_hi; +#else + unsigned short mn_hi, mn_lo; +#endif + /** @} */ +/** @defgroup mdb_node Node Flags + * @ingroup internal + * Flags for node headers. + * @{ + */ +#define F_BIGDATA 0x01 /**< data put on overflow page */ +#define F_SUBDATA 0x02 /**< data is a sub-database */ +#define F_DUPDATA 0x04 /**< data has duplicates */ + +/** valid flags for #mdb_node_add() */ +#define NODE_ADD_FLAGS (F_DUPDATA|F_SUBDATA|MDB_RESERVE|MDB_APPEND) + +/** @} */ + unsigned short mn_flags; /**< @ref mdb_node */ + unsigned short mn_ksize; /**< key size */ + char mn_data[1]; /**< key and data are appended here */ +} MDB_node; + + /** Size of the node header, excluding dynamic data at the end */ +#define NODESIZE offsetof(MDB_node, mn_data) + + /** Bit position of top word in page number, for shifting mn_flags */ +#define PGNO_TOPWORD ((pgno_t)-1 > 0xffffffffu ? 32 : 0) + + /** Size of a node in a branch page with a given key. + * This is just the node header plus the key, there is no data. + */ +#define INDXSIZE(k) (NODESIZE + ((k) == NULL ? 0 : (k)->mv_size)) + + /** Size of a node in a leaf page with a given key and data. + * This is node header plus key plus data size. + */ +#define LEAFSIZE(k, d) (NODESIZE + (k)->mv_size + (d)->mv_size) + + /** Address of node \b i in page \b p */ +#define NODEPTR(p, i) ((MDB_node *)((char *)(p) + (p)->mp_ptrs[i] + PAGEBASE)) + + /** Address of the key for the node */ +#define NODEKEY(node) (void *)((node)->mn_data) + + /** Address of the data for a node */ +#define NODEDATA(node) (void *)((char *)(node)->mn_data + (node)->mn_ksize) + + /** Get the page number pointed to by a branch node */ +#define NODEPGNO(node) \ + ((node)->mn_lo | ((pgno_t) (node)->mn_hi << 16) | \ + (PGNO_TOPWORD ? ((pgno_t) (node)->mn_flags << PGNO_TOPWORD) : 0)) + /** Set the page number in a branch node */ +#define SETPGNO(node,pgno) do { \ + (node)->mn_lo = (pgno) & 0xffff; (node)->mn_hi = (pgno) >> 16; \ + if (PGNO_TOPWORD) (node)->mn_flags = (pgno) >> PGNO_TOPWORD; } while(0) + + /** Get the size of the data in a leaf node */ +#define NODEDSZ(node) ((node)->mn_lo | ((unsigned)(node)->mn_hi << 16)) + /** Set the size of the data for a leaf node */ +#define SETDSZ(node,size) do { \ + (node)->mn_lo = (size) & 0xffff; (node)->mn_hi = (size) >> 16;} while(0) + /** The size of a key in a node */ +#define NODEKSZ(node) ((node)->mn_ksize) + + /** Copy a page number from src to dst */ +#ifdef MISALIGNED_OK +#define COPY_PGNO(dst,src) dst = src +#else +#if SIZE_MAX > 4294967295UL +#define COPY_PGNO(dst,src) do { \ + unsigned short *s, *d; \ + s = (unsigned short *)&(src); \ + d = (unsigned short *)&(dst); \ + *d++ = *s++; \ + *d++ = *s++; \ + *d++ = *s++; \ + *d = *s; \ +} while (0) +#else +#define COPY_PGNO(dst,src) do { \ + unsigned short *s, *d; \ + s = (unsigned short *)&(src); \ + d = (unsigned short *)&(dst); \ + *d++ = *s++; \ + *d = *s; \ +} while (0) +#endif +#endif + /** The address of a key in a LEAF2 page. + * LEAF2 pages are used for #MDB_DUPFIXED sorted-duplicate sub-DBs. + * There are no node headers, keys are stored contiguously. + */ +#define LEAF2KEY(p, i, ks) ((char *)(p) + PAGEHDRSZ + ((i)*(ks))) + + /** Set the \b node's key into \b keyptr, if requested. */ +#define MDB_GET_KEY(node, keyptr) { if ((keyptr) != NULL) { \ + (keyptr)->mv_size = NODEKSZ(node); (keyptr)->mv_data = NODEKEY(node); } } + + /** Set the \b node's key into \b key. */ +#define MDB_GET_KEY2(node, key) { key.mv_size = NODEKSZ(node); key.mv_data = NODEKEY(node); } + + /** Information about a single database in the environment. */ +typedef struct MDB_db { + uint32_t md_pad; /**< also ksize for LEAF2 pages */ + uint16_t md_flags; /**< @ref mdb_dbi_open */ + uint16_t md_depth; /**< depth of this tree */ + pgno_t md_branch_pages; /**< number of internal pages */ + pgno_t md_leaf_pages; /**< number of leaf pages */ + pgno_t md_overflow_pages; /**< number of overflow pages */ + size_t md_entries; /**< number of data items */ + pgno_t md_root; /**< the root page of this tree */ +} MDB_db; + +#define MDB_VALID 0x8000 /**< DB handle is valid, for me_dbflags */ +#define PERSISTENT_FLAGS (0xffff & ~(MDB_VALID)) + /** #mdb_dbi_open() flags */ +#define VALID_FLAGS (MDB_REVERSEKEY|MDB_DUPSORT|MDB_INTEGERKEY|MDB_DUPFIXED|\ + MDB_INTEGERDUP|MDB_REVERSEDUP|MDB_CREATE) + + /** Handle for the DB used to track free pages. */ +#define FREE_DBI 0 + /** Handle for the default DB. */ +#define MAIN_DBI 1 + /** Number of DBs in metapage (free and main) - also hardcoded elsewhere */ +#define CORE_DBS 2 + + /** Number of meta pages - also hardcoded elsewhere */ +#define NUM_METAS 2 + + /** Meta page content. + * A meta page is the start point for accessing a database snapshot. + * Pages 0-1 are meta pages. Transaction N writes meta page #(N % 2). + */ +typedef struct MDB_meta { + /** Stamp identifying this as an LMDB file. It must be set + * to #MDB_MAGIC. */ + uint32_t mm_magic; + /** Version number of this file. Must be set to #MDB_DATA_VERSION. */ + uint32_t mm_version; + void *mm_address; /**< address for fixed mapping */ + size_t mm_mapsize; /**< size of mmap region */ + MDB_db mm_dbs[CORE_DBS]; /**< first is free space, 2nd is main db */ + /** The size of pages used in this DB */ +#define mm_psize mm_dbs[FREE_DBI].md_pad + /** Any persistent environment flags. @ref mdb_env */ +#define mm_flags mm_dbs[FREE_DBI].md_flags + /** Last used page in the datafile. + * Actually the file may be shorter if the freeDB lists the final pages. + */ + pgno_t mm_last_pg; + volatile txnid_t mm_txnid; /**< txnid that committed this page */ +} MDB_meta; + + /** Buffer for a stack-allocated meta page. + * The members define size and alignment, and silence type + * aliasing warnings. They are not used directly; that could + * mean incorrectly using several union members in parallel. + */ +typedef union MDB_metabuf { + MDB_page mb_page; + struct { + char mm_pad[PAGEHDRSZ]; + MDB_meta mm_meta; + } mb_metabuf; +} MDB_metabuf; + + /** Auxiliary DB info. + * The information here is mostly static/read-only. There is + * only a single copy of this record in the environment. + */ +typedef struct MDB_dbx { + MDB_val md_name; /**< name of the database */ + MDB_cmp_func *md_cmp; /**< function for comparing keys */ + MDB_cmp_func *md_dcmp; /**< function for comparing data items */ + MDB_rel_func *md_rel; /**< user relocate function */ + void *md_relctx; /**< user-provided context for md_rel */ +} MDB_dbx; + + /** A database transaction. + * Every operation requires a transaction handle. + */ +struct MDB_txn { + MDB_txn *mt_parent; /**< parent of a nested txn */ + /** Nested txn under this txn, set together with flag #MDB_TXN_HAS_CHILD */ + MDB_txn *mt_child; + pgno_t mt_next_pgno; /**< next unallocated page */ + /** The ID of this transaction. IDs are integers incrementing from 1. + * Only committed write transactions increment the ID. If a transaction + * aborts, the ID may be re-used by the next writer. + */ + txnid_t mt_txnid; + MDB_env *mt_env; /**< the DB environment */ + /** The list of pages that became unused during this transaction. + */ + MDB_IDL mt_free_pgs; + /** The list of loose pages that became unused and may be reused + * in this transaction, linked through #NEXT_LOOSE_PAGE(page). + */ + MDB_page *mt_loose_pgs; + /** Number of loose pages (#mt_loose_pgs) */ + int mt_loose_count; + /** The sorted list of dirty pages we temporarily wrote to disk + * because the dirty list was full. page numbers in here are + * shifted left by 1, deleted slots have the LSB set. + */ + MDB_IDL mt_spill_pgs; + union { + /** For write txns: Modified pages. Sorted when not MDB_WRITEMAP. */ + MDB_ID2L dirty_list; + /** For read txns: This thread/txn's reader table slot, or NULL. */ + MDB_reader *reader; + } mt_u; + /** Array of records for each DB known in the environment. */ + MDB_dbx *mt_dbxs; + /** Array of MDB_db records for each known DB */ + MDB_db *mt_dbs; + /** Array of sequence numbers for each DB handle */ + unsigned int *mt_dbiseqs; +/** @defgroup mt_dbflag Transaction DB Flags + * @ingroup internal + * @{ + */ +#define DB_DIRTY 0x01 /**< DB was written in this txn */ +#define DB_STALE 0x02 /**< Named-DB record is older than txnID */ +#define DB_NEW 0x04 /**< Named-DB handle opened in this txn */ +#define DB_VALID 0x08 /**< DB handle is valid, see also #MDB_VALID */ +#define DB_USRVALID 0x10 /**< As #DB_VALID, but not set for #FREE_DBI */ +#define DB_DUPDATA 0x20 /**< DB is #MDB_DUPSORT data */ +/** @} */ + /** In write txns, array of cursors for each DB */ + MDB_cursor **mt_cursors; + /** Array of flags for each DB */ + unsigned char *mt_dbflags; + /** Number of DB records in use, or 0 when the txn is finished. + * This number only ever increments until the txn finishes; we + * don't decrement it when individual DB handles are closed. + */ + MDB_dbi mt_numdbs; + +/** @defgroup mdb_txn Transaction Flags + * @ingroup internal + * @{ + */ + /** #mdb_txn_begin() flags */ +#define MDB_TXN_BEGIN_FLAGS MDB_RDONLY +#define MDB_TXN_RDONLY MDB_RDONLY /**< read-only transaction */ + /* internal txn flags */ +#define MDB_TXN_WRITEMAP MDB_WRITEMAP /**< copy of #MDB_env flag in writers */ +#define MDB_TXN_FINISHED 0x01 /**< txn is finished or never began */ +#define MDB_TXN_ERROR 0x02 /**< txn is unusable after an error */ +#define MDB_TXN_DIRTY 0x04 /**< must write, even if dirty list is empty */ +#define MDB_TXN_SPILLS 0x08 /**< txn or a parent has spilled pages */ +#define MDB_TXN_HAS_CHILD 0x10 /**< txn has an #MDB_txn.%mt_child */ + /** most operations on the txn are currently illegal */ +#define MDB_TXN_BLOCKED (MDB_TXN_FINISHED|MDB_TXN_ERROR|MDB_TXN_HAS_CHILD) +/** @} */ + unsigned int mt_flags; /**< @ref mdb_txn */ + /** #dirty_list room: Array size - \#dirty pages visible to this txn. + * Includes ancestor txns' dirty pages not hidden by other txns' + * dirty/spilled pages. Thus commit(nested txn) has room to merge + * dirty_list into mt_parent after freeing hidden mt_parent pages. + */ + unsigned int mt_dirty_room; +}; + +/** Enough space for 2^32 nodes with minimum of 2 keys per node. I.e., plenty. + * At 4 keys per node, enough for 2^64 nodes, so there's probably no need to + * raise this on a 64 bit machine. + */ +#define CURSOR_STACK 32 + +struct MDB_xcursor; + + /** Cursors are used for all DB operations. + * A cursor holds a path of (page pointer, key index) from the DB + * root to a position in the DB, plus other state. #MDB_DUPSORT + * cursors include an xcursor to the current data item. Write txns + * track their cursors and keep them up to date when data moves. + * Exception: An xcursor's pointer to a #P_SUBP page can be stale. + * (A node with #F_DUPDATA but no #F_SUBDATA contains a subpage). + */ +struct MDB_cursor { + /** Next cursor on this DB in this txn */ + MDB_cursor *mc_next; + /** Backup of the original cursor if this cursor is a shadow */ + MDB_cursor *mc_backup; + /** Context used for databases with #MDB_DUPSORT, otherwise NULL */ + struct MDB_xcursor *mc_xcursor; + /** The transaction that owns this cursor */ + MDB_txn *mc_txn; + /** The database handle this cursor operates on */ + MDB_dbi mc_dbi; + /** The database record for this cursor */ + MDB_db *mc_db; + /** The database auxiliary record for this cursor */ + MDB_dbx *mc_dbx; + /** The @ref mt_dbflag for this database */ + unsigned char *mc_dbflag; + unsigned short mc_snum; /**< number of pushed pages */ + unsigned short mc_top; /**< index of top page, normally mc_snum-1 */ +/** @defgroup mdb_cursor Cursor Flags + * @ingroup internal + * Cursor state flags. + * @{ + */ +#define C_INITIALIZED 0x01 /**< cursor has been initialized and is valid */ +#define C_EOF 0x02 /**< No more data */ +#define C_SUB 0x04 /**< Cursor is a sub-cursor */ +#define C_DEL 0x08 /**< last op was a cursor_del */ +#define C_UNTRACK 0x40 /**< Un-track cursor when closing */ +/** @} */ + unsigned int mc_flags; /**< @ref mdb_cursor */ + MDB_page *mc_pg[CURSOR_STACK]; /**< stack of pushed pages */ + indx_t mc_ki[CURSOR_STACK]; /**< stack of page indices */ +}; + + /** Context for sorted-dup records. + * We could have gone to a fully recursive design, with arbitrarily + * deep nesting of sub-databases. But for now we only handle these + * levels - main DB, optional sub-DB, sorted-duplicate DB. + */ +typedef struct MDB_xcursor { + /** A sub-cursor for traversing the Dup DB */ + MDB_cursor mx_cursor; + /** The database record for this Dup DB */ + MDB_db mx_db; + /** The auxiliary DB record for this Dup DB */ + MDB_dbx mx_dbx; + /** The @ref mt_dbflag for this Dup DB */ + unsigned char mx_dbflag; +} MDB_xcursor; + + /** Check if there is an inited xcursor */ +#define XCURSOR_INITED(mc) \ + ((mc)->mc_xcursor && ((mc)->mc_xcursor->mx_cursor.mc_flags & C_INITIALIZED)) + + /** Update the xcursor's sub-page pointer, if any, in \b mc. Needed + * when the node which contains the sub-page may have moved. Called + * with leaf page \b mp = mc->mc_pg[\b top]. + */ +#define XCURSOR_REFRESH(mc, top, mp) do { \ + MDB_page *xr_pg = (mp); \ + MDB_node *xr_node; \ + if (!XCURSOR_INITED(mc) || (mc)->mc_ki[top] >= NUMKEYS(xr_pg)) break; \ + xr_node = NODEPTR(xr_pg, (mc)->mc_ki[top]); \ + if ((xr_node->mn_flags & (F_DUPDATA|F_SUBDATA)) == F_DUPDATA) \ + (mc)->mc_xcursor->mx_cursor.mc_pg[0] = NODEDATA(xr_node); \ +} while (0) + + /** State of FreeDB old pages, stored in the MDB_env */ +typedef struct MDB_pgstate { + pgno_t *mf_pghead; /**< Reclaimed freeDB pages, or NULL before use */ + txnid_t mf_pglast; /**< ID of last used record, or 0 if !mf_pghead */ +} MDB_pgstate; + + /** The database environment. */ +struct MDB_env { + HANDLE me_fd; /**< The main data file */ + HANDLE me_lfd; /**< The lock file */ + HANDLE me_mfd; /**< For writing and syncing the meta pages */ + /** Failed to update the meta page. Probably an I/O error. */ +#define MDB_FATAL_ERROR 0x80000000U + /** Some fields are initialized. */ +#define MDB_ENV_ACTIVE 0x20000000U + /** me_txkey is set */ +#define MDB_ENV_TXKEY 0x10000000U + /** fdatasync is unreliable */ +#define MDB_FSYNCONLY 0x08000000U + uint32_t me_flags; /**< @ref mdb_env */ + unsigned int me_psize; /**< DB page size, inited from me_os_psize */ + unsigned int me_os_psize; /**< OS page size, from #GET_PAGESIZE */ + unsigned int me_maxreaders; /**< size of the reader table */ + /** Max #MDB_txninfo.%mti_numreaders of interest to #mdb_env_close() */ + volatile int me_close_readers; + MDB_dbi me_numdbs; /**< number of DBs opened */ + MDB_dbi me_maxdbs; /**< size of the DB table */ + MDB_PID_T me_pid; /**< process ID of this env */ + char *me_path; /**< path to the DB files */ + char *me_map; /**< the memory map of the data file */ + MDB_txninfo *me_txns; /**< the memory map of the lock file or NULL */ + MDB_meta *me_metas[NUM_METAS]; /**< pointers to the two meta pages */ + void *me_pbuf; /**< scratch area for DUPSORT put() */ + MDB_txn *me_txn; /**< current write transaction */ + MDB_txn *me_txn0; /**< prealloc'd write transaction */ + size_t me_mapsize; /**< size of the data memory map */ + off_t me_size; /**< current file size */ + pgno_t me_maxpg; /**< me_mapsize / me_psize */ + MDB_dbx *me_dbxs; /**< array of static DB info */ + uint16_t *me_dbflags; /**< array of flags from MDB_db.md_flags */ + unsigned int *me_dbiseqs; /**< array of dbi sequence numbers */ + pthread_key_t me_txkey; /**< thread-key for readers */ + txnid_t me_pgoldest; /**< ID of oldest reader last time we looked */ + MDB_pgstate me_pgstate; /**< state of old pages from freeDB */ +# define me_pglast me_pgstate.mf_pglast +# define me_pghead me_pgstate.mf_pghead + MDB_page *me_dpages; /**< list of malloc'd blocks for re-use */ + /** IDL of pages that became unused in a write txn */ + MDB_IDL me_free_pgs; + /** ID2L of pages written during a write txn. Length MDB_IDL_UM_SIZE. */ + MDB_ID2L me_dirty_list; + /** Max number of freelist items that can fit in a single overflow page */ + int me_maxfree_1pg; + /** Max size of a node on a page */ + unsigned int me_nodemax; +#if !(MDB_MAXKEYSIZE) + unsigned int me_maxkey; /**< max size of a key */ +#endif + int me_live_reader; /**< have liveness lock in reader table */ +#ifdef _WIN32 + int me_pidquery; /**< Used in OpenProcess */ +#endif +#ifdef MDB_USE_POSIX_MUTEX /* Posix mutexes reside in shared mem */ +# define me_rmutex me_txns->mti_rmutex /**< Shared reader lock */ +# define me_wmutex me_txns->mti_wmutex /**< Shared writer lock */ +#else + mdb_mutex_t me_rmutex; + mdb_mutex_t me_wmutex; +#endif + void *me_userctx; /**< User-settable context */ + MDB_assert_func *me_assert_func; /**< Callback for assertion failures */ +}; + + /** Nested transaction */ +typedef struct MDB_ntxn { + MDB_txn mnt_txn; /**< the transaction */ + MDB_pgstate mnt_pgstate; /**< parent transaction's saved freestate */ +} MDB_ntxn; + + /** max number of pages to commit in one writev() call */ +#define MDB_COMMIT_PAGES 64 +#if defined(IOV_MAX) && IOV_MAX < MDB_COMMIT_PAGES +#undef MDB_COMMIT_PAGES +#define MDB_COMMIT_PAGES IOV_MAX +#endif + + /** max bytes to write in one call */ +#define MAX_WRITE (0x40000000U >> (sizeof(ssize_t) == 4)) + + /** Check \b txn and \b dbi arguments to a function */ +#define TXN_DBI_EXIST(txn, dbi, validity) \ + ((txn) && (dbi)<(txn)->mt_numdbs && ((txn)->mt_dbflags[dbi] & (validity))) + + /** Check for misused \b dbi handles */ +#define TXN_DBI_CHANGED(txn, dbi) \ + ((txn)->mt_dbiseqs[dbi] != (txn)->mt_env->me_dbiseqs[dbi]) + +static int mdb_page_alloc(MDB_cursor *mc, int num, MDB_page **mp); +static int mdb_page_new(MDB_cursor *mc, uint32_t flags, int num, MDB_page **mp); +static int mdb_page_touch(MDB_cursor *mc); + +#define MDB_END_NAMES {"committed", "empty-commit", "abort", "reset", \ + "reset-tmp", "fail-begin", "fail-beginchild"} +enum { + /* mdb_txn_end operation number, for logging */ + MDB_END_COMMITTED, MDB_END_EMPTY_COMMIT, MDB_END_ABORT, MDB_END_RESET, + MDB_END_RESET_TMP, MDB_END_FAIL_BEGIN, MDB_END_FAIL_BEGINCHILD +}; +#define MDB_END_OPMASK 0x0F /**< mask for #mdb_txn_end() operation number */ +#define MDB_END_UPDATE 0x10 /**< update env state (DBIs) */ +#define MDB_END_FREE 0x20 /**< free txn unless it is #MDB_env.%me_txn0 */ +#define MDB_END_SLOT MDB_NOTLS /**< release any reader slot if #MDB_NOTLS */ +static void mdb_txn_end(MDB_txn *txn, unsigned mode); + +static int mdb_page_get(MDB_cursor *mc, pgno_t pgno, MDB_page **mp, int *lvl); +static int mdb_page_search_root(MDB_cursor *mc, + MDB_val *key, int modify); +#define MDB_PS_MODIFY 1 +#define MDB_PS_ROOTONLY 2 +#define MDB_PS_FIRST 4 +#define MDB_PS_LAST 8 +static int mdb_page_search(MDB_cursor *mc, + MDB_val *key, int flags); +static int mdb_page_merge(MDB_cursor *csrc, MDB_cursor *cdst); + +#define MDB_SPLIT_REPLACE MDB_APPENDDUP /**< newkey is not new */ +static int mdb_page_split(MDB_cursor *mc, MDB_val *newkey, MDB_val *newdata, + pgno_t newpgno, unsigned int nflags); + +static int mdb_env_read_header(MDB_env *env, MDB_meta *meta); +static MDB_meta *mdb_env_pick_meta(const MDB_env *env); +static int mdb_env_write_meta(MDB_txn *txn); +#ifdef MDB_USE_POSIX_MUTEX /* Drop unused excl arg */ +# define mdb_env_close0(env, excl) mdb_env_close1(env) +#endif +static void mdb_env_close0(MDB_env *env, int excl); + +static MDB_node *mdb_node_search(MDB_cursor *mc, MDB_val *key, int *exactp); +static int mdb_node_add(MDB_cursor *mc, indx_t indx, + MDB_val *key, MDB_val *data, pgno_t pgno, unsigned int flags); +static void mdb_node_del(MDB_cursor *mc, int ksize); +static void mdb_node_shrink(MDB_page *mp, indx_t indx); +static int mdb_node_move(MDB_cursor *csrc, MDB_cursor *cdst, int fromleft); +static int mdb_node_read(MDB_cursor *mc, MDB_node *leaf, MDB_val *data); +static size_t mdb_leaf_size(MDB_env *env, MDB_val *key, MDB_val *data); +static size_t mdb_branch_size(MDB_env *env, MDB_val *key); + +static int mdb_rebalance(MDB_cursor *mc); +static int mdb_update_key(MDB_cursor *mc, MDB_val *key); + +static void mdb_cursor_pop(MDB_cursor *mc); +static int mdb_cursor_push(MDB_cursor *mc, MDB_page *mp); + +static int mdb_cursor_del0(MDB_cursor *mc); +static int mdb_del0(MDB_txn *txn, MDB_dbi dbi, MDB_val *key, MDB_val *data, unsigned flags); +static int mdb_cursor_sibling(MDB_cursor *mc, int move_right); +static int mdb_cursor_next(MDB_cursor *mc, MDB_val *key, MDB_val *data, MDB_cursor_op op); +static int mdb_cursor_prev(MDB_cursor *mc, MDB_val *key, MDB_val *data, MDB_cursor_op op); +static int mdb_cursor_set(MDB_cursor *mc, MDB_val *key, MDB_val *data, MDB_cursor_op op, + int *exactp); +static int mdb_cursor_first(MDB_cursor *mc, MDB_val *key, MDB_val *data); +static int mdb_cursor_last(MDB_cursor *mc, MDB_val *key, MDB_val *data); + +static void mdb_cursor_init(MDB_cursor *mc, MDB_txn *txn, MDB_dbi dbi, MDB_xcursor *mx); +static void mdb_xcursor_init0(MDB_cursor *mc); +static void mdb_xcursor_init1(MDB_cursor *mc, MDB_node *node); +static void mdb_xcursor_init2(MDB_cursor *mc, MDB_xcursor *src_mx, int force); + +static int mdb_drop0(MDB_cursor *mc, int subs); +static void mdb_default_cmp(MDB_txn *txn, MDB_dbi dbi); +static int mdb_reader_check0(MDB_env *env, int rlocked, int *dead); + +/** @cond */ +static MDB_cmp_func mdb_cmp_memn, mdb_cmp_memnr, mdb_cmp_int, mdb_cmp_cint, mdb_cmp_long; +/** @endcond */ + +/** Compare two items pointing at size_t's of unknown alignment. */ +#ifdef MISALIGNED_OK +# define mdb_cmp_clong mdb_cmp_long +#else +# define mdb_cmp_clong mdb_cmp_cint +#endif + +#ifdef _WIN32 +static SECURITY_DESCRIPTOR mdb_null_sd; +static SECURITY_ATTRIBUTES mdb_all_sa; +static int mdb_sec_inited; + +struct MDB_name; +static int utf8_to_utf16(const char *src, struct MDB_name *dst, int xtra); +#endif + +/** Return the library version info. */ +char * ESECT +mdb_version(int *major, int *minor, int *patch) +{ + if (major) *major = MDB_VERSION_MAJOR; + if (minor) *minor = MDB_VERSION_MINOR; + if (patch) *patch = MDB_VERSION_PATCH; + return MDB_VERSION_STRING; +} + +/** Table of descriptions for LMDB @ref errors */ +static char *const mdb_errstr[] = { + "MDB_KEYEXIST: Key/data pair already exists", + "MDB_NOTFOUND: No matching key/data pair found", + "MDB_PAGE_NOTFOUND: Requested page not found", + "MDB_CORRUPTED: Located page was wrong type", + "MDB_PANIC: Update of meta page failed or environment had fatal error", + "MDB_VERSION_MISMATCH: Database environment version mismatch", + "MDB_INVALID: File is not an LMDB file", + "MDB_MAP_FULL: Environment mapsize limit reached", + "MDB_DBS_FULL: Environment maxdbs limit reached", + "MDB_READERS_FULL: Environment maxreaders limit reached", + "MDB_TLS_FULL: Thread-local storage keys full - too many environments open", + "MDB_TXN_FULL: Transaction has too many dirty pages - transaction too big", + "MDB_CURSOR_FULL: Internal error - cursor stack limit reached", + "MDB_PAGE_FULL: Internal error - page has no more space", + "MDB_MAP_RESIZED: Database contents grew beyond environment mapsize", + "MDB_INCOMPATIBLE: Operation and DB incompatible, or DB flags changed", + "MDB_BAD_RSLOT: Invalid reuse of reader locktable slot", + "MDB_BAD_TXN: Transaction must abort, has a child, or is invalid", + "MDB_BAD_VALSIZE: Unsupported size of key/DB name/data, or wrong DUPFIXED size", + "MDB_BAD_DBI: The specified DBI handle was closed/changed unexpectedly", +}; + +char * +mdb_strerror(int err) +{ +#ifdef _WIN32 + /** HACK: pad 4KB on stack over the buf. Return system msgs in buf. + * This works as long as no function between the call to mdb_strerror + * and the actual use of the message uses more than 4K of stack. + */ +#define MSGSIZE 1024 +#define PADSIZE 4096 + char buf[MSGSIZE+PADSIZE], *ptr = buf; +#endif + int i; + if (!err) + return ("Successful return: 0"); + + if (err >= MDB_KEYEXIST && err <= MDB_LAST_ERRCODE) { + i = err - MDB_KEYEXIST; + return mdb_errstr[i]; + } + +#ifdef _WIN32 + /* These are the C-runtime error codes we use. The comment indicates + * their numeric value, and the Win32 error they would correspond to + * if the error actually came from a Win32 API. A major mess, we should + * have used LMDB-specific error codes for everything. + */ + switch(err) { + case ENOENT: /* 2, FILE_NOT_FOUND */ + case EIO: /* 5, ACCESS_DENIED */ + case ENOMEM: /* 12, INVALID_ACCESS */ + case EACCES: /* 13, INVALID_DATA */ + case EBUSY: /* 16, CURRENT_DIRECTORY */ + case EINVAL: /* 22, BAD_COMMAND */ + case ENOSPC: /* 28, OUT_OF_PAPER */ + return strerror(err); + default: + ; + } + buf[0] = 0; + FormatMessageA(FORMAT_MESSAGE_FROM_SYSTEM | + FORMAT_MESSAGE_IGNORE_INSERTS, + NULL, err, 0, ptr, MSGSIZE, (va_list *)buf+MSGSIZE); + return ptr; +#else + return strerror(err); +#endif +} + +/** assert(3) variant in cursor context */ +#define mdb_cassert(mc, expr) mdb_assert0((mc)->mc_txn->mt_env, expr, #expr) +/** assert(3) variant in transaction context */ +#define mdb_tassert(txn, expr) mdb_assert0((txn)->mt_env, expr, #expr) +/** assert(3) variant in environment context */ +#define mdb_eassert(env, expr) mdb_assert0(env, expr, #expr) + +#ifndef NDEBUG +# define mdb_assert0(env, expr, expr_txt) ((expr) ? (void)0 : \ + mdb_assert_fail(env, expr_txt, mdb_func_, __FILE__, __LINE__)) + +static void ESECT +mdb_assert_fail(MDB_env *env, const char *expr_txt, + const char *func, const char *file, int line) +{ + char buf[400]; + sprintf(buf, "%.100s:%d: Assertion '%.200s' failed in %.40s()", + file, line, expr_txt, func); + if (env->me_assert_func) + env->me_assert_func(env, buf); + fprintf(stderr, "%s\n", buf); + abort(); +} +#else +# define mdb_assert0(env, expr, expr_txt) ((void) 0) +#endif /* NDEBUG */ + +#if MDB_DEBUG +/** Return the page number of \b mp which may be sub-page, for debug output */ +static pgno_t +mdb_dbg_pgno(MDB_page *mp) +{ + pgno_t ret; + COPY_PGNO(ret, mp->mp_pgno); + return ret; +} + +/** Display a key in hexadecimal and return the address of the result. + * @param[in] key the key to display + * @param[in] buf the buffer to write into. Should always be #DKBUF. + * @return The key in hexadecimal form. + */ +char * +mdb_dkey(MDB_val *key, char *buf) +{ + char *ptr = buf; + unsigned char *c = key->mv_data; + unsigned int i; + + if (!key) + return ""; + + if (key->mv_size > DKBUF_MAXKEYSIZE) + return "MDB_MAXKEYSIZE"; + /* may want to make this a dynamic check: if the key is mostly + * printable characters, print it as-is instead of converting to hex. + */ +#if 1 + buf[0] = '\0'; + for (i=0; i<key->mv_size; i++) + ptr += sprintf(ptr, "%02x", *c++); +#else + sprintf(buf, "%.*s", key->mv_size, key->mv_data); +#endif + return buf; +} + +static const char * +mdb_leafnode_type(MDB_node *n) +{ + static char *const tp[2][2] = {{"", ": DB"}, {": sub-page", ": sub-DB"}}; + return F_ISSET(n->mn_flags, F_BIGDATA) ? ": overflow page" : + tp[F_ISSET(n->mn_flags, F_DUPDATA)][F_ISSET(n->mn_flags, F_SUBDATA)]; +} + +/** Display all the keys in the page. */ +void +mdb_page_list(MDB_page *mp) +{ + pgno_t pgno = mdb_dbg_pgno(mp); + const char *type, *state = (mp->mp_flags & P_DIRTY) ? ", dirty" : ""; + MDB_node *node; + unsigned int i, nkeys, nsize, total = 0; + MDB_val key; + DKBUF; + + switch (mp->mp_flags & (P_BRANCH|P_LEAF|P_LEAF2|P_META|P_OVERFLOW|P_SUBP)) { + case P_BRANCH: type = "Branch page"; break; + case P_LEAF: type = "Leaf page"; break; + case P_LEAF|P_SUBP: type = "Sub-page"; break; + case P_LEAF|P_LEAF2: type = "LEAF2 page"; break; + case P_LEAF|P_LEAF2|P_SUBP: type = "LEAF2 sub-page"; break; + case P_OVERFLOW: + fprintf(stderr, "Overflow page %"Z"u pages %u%s\n", + pgno, mp->mp_pages, state); + return; + case P_META: + fprintf(stderr, "Meta-page %"Z"u txnid %"Z"u\n", + pgno, ((MDB_meta *)METADATA(mp))->mm_txnid); + return; + default: + fprintf(stderr, "Bad page %"Z"u flags 0x%X\n", pgno, mp->mp_flags); + return; + } + + nkeys = NUMKEYS(mp); + fprintf(stderr, "%s %"Z"u numkeys %d%s\n", type, pgno, nkeys, state); + + for (i=0; i<nkeys; i++) { + if (IS_LEAF2(mp)) { /* LEAF2 pages have no mp_ptrs[] or node headers */ + key.mv_size = nsize = mp->mp_pad; + key.mv_data = LEAF2KEY(mp, i, nsize); + total += nsize; + fprintf(stderr, "key %d: nsize %d, %s\n", i, nsize, DKEY(&key)); + continue; + } + node = NODEPTR(mp, i); + key.mv_size = node->mn_ksize; + key.mv_data = node->mn_data; + nsize = NODESIZE + key.mv_size; + if (IS_BRANCH(mp)) { + fprintf(stderr, "key %d: page %"Z"u, %s\n", i, NODEPGNO(node), + DKEY(&key)); + total += nsize; + } else { + if (F_ISSET(node->mn_flags, F_BIGDATA)) + nsize += sizeof(pgno_t); + else + nsize += NODEDSZ(node); + total += nsize; + nsize += sizeof(indx_t); + fprintf(stderr, "key %d: nsize %d, %s%s\n", + i, nsize, DKEY(&key), mdb_leafnode_type(node)); + } + total = EVEN(total); + } + fprintf(stderr, "Total: header %d + contents %d + unused %d\n", + IS_LEAF2(mp) ? PAGEHDRSZ : PAGEBASE + mp->mp_lower, total, SIZELEFT(mp)); +} + +void +mdb_cursor_chk(MDB_cursor *mc) +{ + unsigned int i; + MDB_node *node; + MDB_page *mp; + + if (!mc->mc_snum || !(mc->mc_flags & C_INITIALIZED)) return; + for (i=0; i<mc->mc_top; i++) { + mp = mc->mc_pg[i]; + node = NODEPTR(mp, mc->mc_ki[i]); + if (NODEPGNO(node) != mc->mc_pg[i+1]->mp_pgno) + printf("oops!\n"); + } + if (mc->mc_ki[i] >= NUMKEYS(mc->mc_pg[i])) + printf("ack!\n"); + if (XCURSOR_INITED(mc)) { + node = NODEPTR(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top]); + if (((node->mn_flags & (F_DUPDATA|F_SUBDATA)) == F_DUPDATA) && + mc->mc_xcursor->mx_cursor.mc_pg[0] != NODEDATA(node)) { + printf("blah!\n"); + } + } +} +#endif + +#if (MDB_DEBUG) > 2 +/** Count all the pages in each DB and in the freelist + * and make sure it matches the actual number of pages + * being used. + * All named DBs must be open for a correct count. + */ +static void mdb_audit(MDB_txn *txn) +{ + MDB_cursor mc; + MDB_val key, data; + MDB_ID freecount, count; + MDB_dbi i; + int rc; + + freecount = 0; + mdb_cursor_init(&mc, txn, FREE_DBI, NULL); + while ((rc = mdb_cursor_get(&mc, &key, &data, MDB_NEXT)) == 0) + freecount += *(MDB_ID *)data.mv_data; + mdb_tassert(txn, rc == MDB_NOTFOUND); + + count = 0; + for (i = 0; i<txn->mt_numdbs; i++) { + MDB_xcursor mx; + if (!(txn->mt_dbflags[i] & DB_VALID)) + continue; + mdb_cursor_init(&mc, txn, i, &mx); + if (txn->mt_dbs[i].md_root == P_INVALID) + continue; + count += txn->mt_dbs[i].md_branch_pages + + txn->mt_dbs[i].md_leaf_pages + + txn->mt_dbs[i].md_overflow_pages; + if (txn->mt_dbs[i].md_flags & MDB_DUPSORT) { + rc = mdb_page_search(&mc, NULL, MDB_PS_FIRST); + for (; rc == MDB_SUCCESS; rc = mdb_cursor_sibling(&mc, 1)) { + unsigned j; + MDB_page *mp; + mp = mc.mc_pg[mc.mc_top]; + for (j=0; j<NUMKEYS(mp); j++) { + MDB_node *leaf = NODEPTR(mp, j); + if (leaf->mn_flags & F_SUBDATA) { + MDB_db db; + memcpy(&db, NODEDATA(leaf), sizeof(db)); + count += db.md_branch_pages + db.md_leaf_pages + + db.md_overflow_pages; + } + } + } + mdb_tassert(txn, rc == MDB_NOTFOUND); + } + } + if (freecount + count + NUM_METAS != txn->mt_next_pgno) { + fprintf(stderr, "audit: %"Z"u freecount: %"Z"u count: %"Z"u total: %"Z"u next_pgno: %"Z"u\n", + txn->mt_txnid, freecount, count+NUM_METAS, + freecount+count+NUM_METAS, txn->mt_next_pgno); + } +} +#endif + +int +mdb_cmp(MDB_txn *txn, MDB_dbi dbi, const MDB_val *a, const MDB_val *b) +{ + return txn->mt_dbxs[dbi].md_cmp(a, b); +} + +int +mdb_dcmp(MDB_txn *txn, MDB_dbi dbi, const MDB_val *a, const MDB_val *b) +{ + MDB_cmp_func *dcmp = txn->mt_dbxs[dbi].md_dcmp; +#if UINT_MAX < SIZE_MAX + if (dcmp == mdb_cmp_int && a->mv_size == sizeof(size_t)) + dcmp = mdb_cmp_clong; +#endif + return dcmp(a, b); +} + +/** Allocate memory for a page. + * Re-use old malloc'd pages first for singletons, otherwise just malloc. + * Set #MDB_TXN_ERROR on failure. + */ +static MDB_page * +mdb_page_malloc(MDB_txn *txn, unsigned num) +{ + MDB_env *env = txn->mt_env; + MDB_page *ret = env->me_dpages; + size_t psize = env->me_psize, sz = psize, off; + /* For ! #MDB_NOMEMINIT, psize counts how much to init. + * For a single page alloc, we init everything after the page header. + * For multi-page, we init the final page; if the caller needed that + * many pages they will be filling in at least up to the last page. + */ + if (num == 1) { + if (ret) { + VGMEMP_ALLOC(env, ret, sz); + VGMEMP_DEFINED(ret, sizeof(ret->mp_next)); + env->me_dpages = ret->mp_next; + return ret; + } + psize -= off = PAGEHDRSZ; + } else { + sz *= num; + off = sz - psize; + } + if ((ret = malloc(sz)) != NULL) { + VGMEMP_ALLOC(env, ret, sz); + if (!(env->me_flags & MDB_NOMEMINIT)) { + memset((char *)ret + off, 0, psize); + ret->mp_pad = 0; + } + } else { + txn->mt_flags |= MDB_TXN_ERROR; + } + return ret; +} +/** Free a single page. + * Saves single pages to a list, for future reuse. + * (This is not used for multi-page overflow pages.) + */ +static void +mdb_page_free(MDB_env *env, MDB_page *mp) +{ + mp->mp_next = env->me_dpages; + VGMEMP_FREE(env, mp); + env->me_dpages = mp; +} + +/** Free a dirty page */ +static void +mdb_dpage_free(MDB_env *env, MDB_page *dp) +{ + if (!IS_OVERFLOW(dp) || dp->mp_pages == 1) { + mdb_page_free(env, dp); + } else { + /* large pages just get freed directly */ + VGMEMP_FREE(env, dp); + free(dp); + } +} + +/** Return all dirty pages to dpage list */ +static void +mdb_dlist_free(MDB_txn *txn) +{ + MDB_env *env = txn->mt_env; + MDB_ID2L dl = txn->mt_u.dirty_list; + unsigned i, n = dl[0].mid; + + for (i = 1; i <= n; i++) { + mdb_dpage_free(env, dl[i].mptr); + } + dl[0].mid = 0; +} + +/** Loosen or free a single page. + * Saves single pages to a list for future reuse + * in this same txn. It has been pulled from the freeDB + * and already resides on the dirty list, but has been + * deleted. Use these pages first before pulling again + * from the freeDB. + * + * If the page wasn't dirtied in this txn, just add it + * to this txn's free list. + */ +static int +mdb_page_loose(MDB_cursor *mc, MDB_page *mp) +{ + int loose = 0; + pgno_t pgno = mp->mp_pgno; + MDB_txn *txn = mc->mc_txn; + + if ((mp->mp_flags & P_DIRTY) && mc->mc_dbi != FREE_DBI) { + if (txn->mt_parent) { + MDB_ID2 *dl = txn->mt_u.dirty_list; + /* If txn has a parent, make sure the page is in our + * dirty list. + */ + if (dl[0].mid) { + unsigned x = mdb_mid2l_search(dl, pgno); + if (x <= dl[0].mid && dl[x].mid == pgno) { + if (mp != dl[x].mptr) { /* bad cursor? */ + mc->mc_flags &= ~(C_INITIALIZED|C_EOF); + txn->mt_flags |= MDB_TXN_ERROR; + return MDB_CORRUPTED; + } + /* ok, it's ours */ + loose = 1; + } + } + } else { + /* no parent txn, so it's just ours */ + loose = 1; + } + } + if (loose) { + DPRINTF(("loosen db %d page %"Z"u", DDBI(mc), + mp->mp_pgno)); + NEXT_LOOSE_PAGE(mp) = txn->mt_loose_pgs; + txn->mt_loose_pgs = mp; + txn->mt_loose_count++; + mp->mp_flags |= P_LOOSE; + } else { + int rc = mdb_midl_append(&txn->mt_free_pgs, pgno); + if (rc) + return rc; + } + + return MDB_SUCCESS; +} + +/** Set or clear P_KEEP in dirty, non-overflow, non-sub pages watched by txn. + * @param[in] mc A cursor handle for the current operation. + * @param[in] pflags Flags of the pages to update: + * P_DIRTY to set P_KEEP, P_DIRTY|P_KEEP to clear it. + * @param[in] all No shortcuts. Needed except after a full #mdb_page_flush(). + * @return 0 on success, non-zero on failure. + */ +static int +mdb_pages_xkeep(MDB_cursor *mc, unsigned pflags, int all) +{ + enum { Mask = P_SUBP|P_DIRTY|P_LOOSE|P_KEEP }; + MDB_txn *txn = mc->mc_txn; + MDB_cursor *m3, *m0 = mc; + MDB_xcursor *mx; + MDB_page *dp, *mp; + MDB_node *leaf; + unsigned i, j; + int rc = MDB_SUCCESS, level; + + /* Mark pages seen by cursors */ + if (mc->mc_flags & C_UNTRACK) + mc = NULL; /* will find mc in mt_cursors */ + for (i = txn->mt_numdbs;; mc = txn->mt_cursors[--i]) { + for (; mc; mc=mc->mc_next) { + if (!(mc->mc_flags & C_INITIALIZED)) + continue; + for (m3 = mc;; m3 = &mx->mx_cursor) { + mp = NULL; + for (j=0; j<m3->mc_snum; j++) { + mp = m3->mc_pg[j]; + if ((mp->mp_flags & Mask) == pflags) + mp->mp_flags ^= P_KEEP; + } + mx = m3->mc_xcursor; + /* Proceed to mx if it is at a sub-database */ + if (! (mx && (mx->mx_cursor.mc_flags & C_INITIALIZED))) + break; + if (! (mp && (mp->mp_flags & P_LEAF))) + break; + leaf = NODEPTR(mp, m3->mc_ki[j-1]); + if (!(leaf->mn_flags & F_SUBDATA)) + break; + } + } + if (i == 0) + break; + } + + if (all) { + /* Mark dirty root pages */ + for (i=0; i<txn->mt_numdbs; i++) { + if (txn->mt_dbflags[i] & DB_DIRTY) { + pgno_t pgno = txn->mt_dbs[i].md_root; + if (pgno == P_INVALID) + continue; + if ((rc = mdb_page_get(m0, pgno, &dp, &level)) != MDB_SUCCESS) + break; + if ((dp->mp_flags & Mask) == pflags && level <= 1) + dp->mp_flags ^= P_KEEP; + } + } + } + + return rc; +} + +static int mdb_page_flush(MDB_txn *txn, int keep); + +/** Spill pages from the dirty list back to disk. + * This is intended to prevent running into #MDB_TXN_FULL situations, + * but note that they may still occur in a few cases: + * 1) our estimate of the txn size could be too small. Currently this + * seems unlikely, except with a large number of #MDB_MULTIPLE items. + * 2) child txns may run out of space if their parents dirtied a + * lot of pages and never spilled them. TODO: we probably should do + * a preemptive spill during #mdb_txn_begin() of a child txn, if + * the parent's dirty_room is below a given threshold. + * + * Otherwise, if not using nested txns, it is expected that apps will + * not run into #MDB_TXN_FULL any more. The pages are flushed to disk + * the same way as for a txn commit, e.g. their P_DIRTY flag is cleared. + * If the txn never references them again, they can be left alone. + * If the txn only reads them, they can be used without any fuss. + * If the txn writes them again, they can be dirtied immediately without + * going thru all of the work of #mdb_page_touch(). Such references are + * handled by #mdb_page_unspill(). + * + * Also note, we never spill DB root pages, nor pages of active cursors, + * because we'll need these back again soon anyway. And in nested txns, + * we can't spill a page in a child txn if it was already spilled in a + * parent txn. That would alter the parent txns' data even though + * the child hasn't committed yet, and we'd have no way to undo it if + * the child aborted. + * + * @param[in] m0 cursor A cursor handle identifying the transaction and + * database for which we are checking space. + * @param[in] key For a put operation, the key being stored. + * @param[in] data For a put operation, the data being stored. + * @return 0 on success, non-zero on failure. + */ +static int +mdb_page_spill(MDB_cursor *m0, MDB_val *key, MDB_val *data) +{ + MDB_txn *txn = m0->mc_txn; + MDB_page *dp; + MDB_ID2L dl = txn->mt_u.dirty_list; + unsigned int i, j, need; + int rc; + + if (m0->mc_flags & C_SUB) + return MDB_SUCCESS; + + /* Estimate how much space this op will take */ + i = m0->mc_db->md_depth; + /* Named DBs also dirty the main DB */ + if (m0->mc_dbi >= CORE_DBS) + i += txn->mt_dbs[MAIN_DBI].md_depth; + /* For puts, roughly factor in the key+data size */ + if (key) + i += (LEAFSIZE(key, data) + txn->mt_env->me_psize) / txn->mt_env->me_psize; + i += i; /* double it for good measure */ + need = i; + + if (txn->mt_dirty_room > i) + return MDB_SUCCESS; + + if (!txn->mt_spill_pgs) { + txn->mt_spill_pgs = mdb_midl_alloc(MDB_IDL_UM_MAX); + if (!txn->mt_spill_pgs) + return ENOMEM; + } else { + /* purge deleted slots */ + MDB_IDL sl = txn->mt_spill_pgs; + unsigned int num = sl[0]; + j=0; + for (i=1; i<=num; i++) { + if (!(sl[i] & 1)) + sl[++j] = sl[i]; + } + sl[0] = j; + } + + /* Preserve pages which may soon be dirtied again */ + if ((rc = mdb_pages_xkeep(m0, P_DIRTY, 1)) != MDB_SUCCESS) + goto done; + + /* Less aggressive spill - we originally spilled the entire dirty list, + * with a few exceptions for cursor pages and DB root pages. But this + * turns out to be a lot of wasted effort because in a large txn many + * of those pages will need to be used again. So now we spill only 1/8th + * of the dirty pages. Testing revealed this to be a good tradeoff, + * better than 1/2, 1/4, or 1/10. + */ + if (need < MDB_IDL_UM_MAX / 8) + need = MDB_IDL_UM_MAX / 8; + + /* Save the page IDs of all the pages we're flushing */ + /* flush from the tail forward, this saves a lot of shifting later on. */ + for (i=dl[0].mid; i && need; i--) { + MDB_ID pn = dl[i].mid << 1; + dp = dl[i].mptr; + if (dp->mp_flags & (P_LOOSE|P_KEEP)) + continue; + /* Can't spill twice, make sure it's not already in a parent's + * spill list. + */ + if (txn->mt_parent) { + MDB_txn *tx2; + for (tx2 = txn->mt_parent; tx2; tx2 = tx2->mt_parent) { + if (tx2->mt_spill_pgs) { + j = mdb_midl_search(tx2->mt_spill_pgs, pn); + if (j <= tx2->mt_spill_pgs[0] && tx2->mt_spill_pgs[j] == pn) { + dp->mp_flags |= P_KEEP; + break; + } + } + } + if (tx2) + continue; + } + if ((rc = mdb_midl_append(&txn->mt_spill_pgs, pn))) + goto done; + need--; + } + mdb_midl_sort(txn->mt_spill_pgs); + + /* Flush the spilled part of dirty list */ + if ((rc = mdb_page_flush(txn, i)) != MDB_SUCCESS) + goto done; + + /* Reset any dirty pages we kept that page_flush didn't see */ + rc = mdb_pages_xkeep(m0, P_DIRTY|P_KEEP, i); + +done: + txn->mt_flags |= rc ? MDB_TXN_ERROR : MDB_TXN_SPILLS; + return rc; +} + +/** Find oldest txnid still referenced. Expects txn->mt_txnid > 0. */ +static txnid_t +mdb_find_oldest(MDB_txn *txn) +{ + int i; + txnid_t mr, oldest = txn->mt_txnid - 1; + if (txn->mt_env->me_txns) { + MDB_reader *r = txn->mt_env->me_txns->mti_readers; + for (i = txn->mt_env->me_txns->mti_numreaders; --i >= 0; ) { + if (r[i].mr_pid) { + mr = r[i].mr_txnid; + if (oldest > mr) + oldest = mr; + } + } + } + return oldest; +} + +/** Add a page to the txn's dirty list */ +static void +mdb_page_dirty(MDB_txn *txn, MDB_page *mp) +{ + MDB_ID2 mid; + int rc, (*insert)(MDB_ID2L, MDB_ID2 *); + + if (txn->mt_flags & MDB_TXN_WRITEMAP) { + insert = mdb_mid2l_append; + } else { + insert = mdb_mid2l_insert; + } + mid.mid = mp->mp_pgno; + mid.mptr = mp; + rc = insert(txn->mt_u.dirty_list, &mid); + mdb_tassert(txn, rc == 0); + txn->mt_dirty_room--; +} + +/** Allocate page numbers and memory for writing. Maintain me_pglast, + * me_pghead and mt_next_pgno. Set #MDB_TXN_ERROR on failure. + * + * If there are free pages available from older transactions, they + * are re-used first. Otherwise allocate a new page at mt_next_pgno. + * Do not modify the freedB, just merge freeDB records into me_pghead[] + * and move me_pglast to say which records were consumed. Only this + * function can create me_pghead and move me_pglast/mt_next_pgno. + * @param[in] mc cursor A cursor handle identifying the transaction and + * database for which we are allocating. + * @param[in] num the number of pages to allocate. + * @param[out] mp Address of the allocated page(s). Requests for multiple pages + * will always be satisfied by a single contiguous chunk of memory. + * @return 0 on success, non-zero on failure. + */ +static int +mdb_page_alloc(MDB_cursor *mc, int num, MDB_page **mp) +{ +#ifdef MDB_PARANOID /* Seems like we can ignore this now */ + /* Get at most <Max_retries> more freeDB records once me_pghead + * has enough pages. If not enough, use new pages from the map. + * If <Paranoid> and mc is updating the freeDB, only get new + * records if me_pghead is empty. Then the freelist cannot play + * catch-up with itself by growing while trying to save it. + */ + enum { Paranoid = 1, Max_retries = 500 }; +#else + enum { Paranoid = 0, Max_retries = INT_MAX /*infinite*/ }; +#endif + int rc, retry = num * 60; + MDB_txn *txn = mc->mc_txn; + MDB_env *env = txn->mt_env; + pgno_t pgno, *mop = env->me_pghead; + unsigned i, j, mop_len = mop ? mop[0] : 0, n2 = num-1; + MDB_page *np; + txnid_t oldest = 0, last; + MDB_cursor_op op; + MDB_cursor m2; + int found_old = 0; + + /* If there are any loose pages, just use them */ + if (num == 1 && txn->mt_loose_pgs) { + np = txn->mt_loose_pgs; + txn->mt_loose_pgs = NEXT_LOOSE_PAGE(np); + txn->mt_loose_count--; + DPRINTF(("db %d use loose page %"Z"u", DDBI(mc), + np->mp_pgno)); + *mp = np; + return MDB_SUCCESS; + } + + *mp = NULL; + + /* If our dirty list is already full, we can't do anything */ + if (txn->mt_dirty_room == 0) { + rc = MDB_TXN_FULL; + goto fail; + } + + for (op = MDB_FIRST;; op = MDB_NEXT) { + MDB_val key, data; + MDB_node *leaf; + pgno_t *idl; + + /* Seek a big enough contiguous page range. Prefer + * pages at the tail, just truncating the list. + */ + if (mop_len > n2) { + i = mop_len; + do { + pgno = mop[i]; + if (mop[i-n2] == pgno+n2) + goto search_done; + } while (--i > n2); + if (--retry < 0) + break; + } + + if (op == MDB_FIRST) { /* 1st iteration */ + /* Prepare to fetch more and coalesce */ + last = env->me_pglast; + oldest = env->me_pgoldest; + mdb_cursor_init(&m2, txn, FREE_DBI, NULL); + if (last) { + op = MDB_SET_RANGE; + key.mv_data = &last; /* will look up last+1 */ + key.mv_size = sizeof(last); + } + if (Paranoid && mc->mc_dbi == FREE_DBI) + retry = -1; + } + if (Paranoid && retry < 0 && mop_len) + break; + + last++; + /* Do not fetch more if the record will be too recent */ + if (oldest <= last) { + if (!found_old) { + oldest = mdb_find_oldest(txn); + env->me_pgoldest = oldest; + found_old = 1; + } + if (oldest <= last) + break; + } + rc = mdb_cursor_get(&m2, &key, NULL, op); + if (rc) { + if (rc == MDB_NOTFOUND) + break; + goto fail; + } + last = *(txnid_t*)key.mv_data; + if (oldest <= last) { + if (!found_old) { + oldest = mdb_find_oldest(txn); + env->me_pgoldest = oldest; + found_old = 1; + } + if (oldest <= last) + break; + } + np = m2.mc_pg[m2.mc_top]; + leaf = NODEPTR(np, m2.mc_ki[m2.mc_top]); + if ((rc = mdb_node_read(&m2, leaf, &data)) != MDB_SUCCESS) + goto fail; + + idl = (MDB_ID *) data.mv_data; + i = idl[0]; + if (!mop) { + if (!(env->me_pghead = mop = mdb_midl_alloc(i))) { + rc = ENOMEM; + goto fail; + } + } else { + if ((rc = mdb_midl_need(&env->me_pghead, i)) != 0) + goto fail; + mop = env->me_pghead; + } + env->me_pglast = last; +#if (MDB_DEBUG) > 1 + DPRINTF(("IDL read txn %"Z"u root %"Z"u num %u", + last, txn->mt_dbs[FREE_DBI].md_root, i)); + for (j = i; j; j--) + DPRINTF(("IDL %"Z"u", idl[j])); +#endif + /* Merge in descending sorted order */ + mdb_midl_xmerge(mop, idl); + mop_len = mop[0]; + } + + /* Use new pages from the map when nothing suitable in the freeDB */ + i = 0; + pgno = txn->mt_next_pgno; + if (pgno + num >= env->me_maxpg) { + DPUTS("DB size maxed out"); + rc = MDB_MAP_FULL; + goto fail; + } + +search_done: + if (env->me_flags & MDB_WRITEMAP) { + np = (MDB_page *)(env->me_map + env->me_psize * pgno); + } else { + if (!(np = mdb_page_malloc(txn, num))) { + rc = ENOMEM; + goto fail; + } + } + if (i) { + mop[0] = mop_len -= num; + /* Move any stragglers down */ + for (j = i-num; j < mop_len; ) + mop[++j] = mop[++i]; + } else { + txn->mt_next_pgno = pgno + num; + } + np->mp_pgno = pgno; + mdb_page_dirty(txn, np); + *mp = np; + + return MDB_SUCCESS; + +fail: + txn->mt_flags |= MDB_TXN_ERROR; + return rc; +} + +/** Copy the used portions of a non-overflow page. + * @param[in] dst page to copy into + * @param[in] src page to copy from + * @param[in] psize size of a page + */ +static void +mdb_page_copy(MDB_page *dst, MDB_page *src, unsigned int psize) +{ + enum { Align = sizeof(pgno_t) }; + indx_t upper = src->mp_upper, lower = src->mp_lower, unused = upper-lower; + + /* If page isn't full, just copy the used portion. Adjust + * alignment so memcpy may copy words instead of bytes. + */ + if ((unused &= -Align) && !IS_LEAF2(src)) { + upper = (upper + PAGEBASE) & -Align; + memcpy(dst, src, (lower + PAGEBASE + (Align-1)) & -Align); + memcpy((pgno_t *)((char *)dst+upper), (pgno_t *)((char *)src+upper), + psize - upper); + } else { + memcpy(dst, src, psize - unused); + } +} + +/** Pull a page off the txn's spill list, if present. + * If a page being referenced was spilled to disk in this txn, bring + * it back and make it dirty/writable again. + * @param[in] txn the transaction handle. + * @param[in] mp the page being referenced. It must not be dirty. + * @param[out] ret the writable page, if any. ret is unchanged if + * mp wasn't spilled. + */ +static int +mdb_page_unspill(MDB_txn *txn, MDB_page *mp, MDB_page **ret) +{ + MDB_env *env = txn->mt_env; + const MDB_txn *tx2; + unsigned x; + pgno_t pgno = mp->mp_pgno, pn = pgno << 1; + + for (tx2 = txn; tx2; tx2=tx2->mt_parent) { + if (!tx2->mt_spill_pgs) + continue; + x = mdb_midl_search(tx2->mt_spill_pgs, pn); + if (x <= tx2->mt_spill_pgs[0] && tx2->mt_spill_pgs[x] == pn) { + MDB_page *np; + int num; + if (txn->mt_dirty_room == 0) + return MDB_TXN_FULL; + if (IS_OVERFLOW(mp)) + num = mp->mp_pages; + else + num = 1; + if (env->me_flags & MDB_WRITEMAP) { + np = mp; + } else { + np = mdb_page_malloc(txn, num); + if (!np) + return ENOMEM; + if (num > 1) + memcpy(np, mp, num * env->me_psize); + else + mdb_page_copy(np, mp, env->me_psize); + } + if (tx2 == txn) { + /* If in current txn, this page is no longer spilled. + * If it happens to be the last page, truncate the spill list. + * Otherwise mark it as deleted by setting the LSB. + */ + if (x == txn->mt_spill_pgs[0]) + txn->mt_spill_pgs[0]--; + else + txn->mt_spill_pgs[x] |= 1; + } /* otherwise, if belonging to a parent txn, the + * page remains spilled until child commits + */ + + mdb_page_dirty(txn, np); + np->mp_flags |= P_DIRTY; + *ret = np; + break; + } + } + return MDB_SUCCESS; +} + +/** Touch a page: make it dirty and re-insert into tree with updated pgno. + * Set #MDB_TXN_ERROR on failure. + * @param[in] mc cursor pointing to the page to be touched + * @return 0 on success, non-zero on failure. + */ +static int +mdb_page_touch(MDB_cursor *mc) +{ + MDB_page *mp = mc->mc_pg[mc->mc_top], *np; + MDB_txn *txn = mc->mc_txn; + MDB_cursor *m2, *m3; + pgno_t pgno; + int rc; + + if (!F_ISSET(mp->mp_flags, P_DIRTY)) { + if (txn->mt_flags & MDB_TXN_SPILLS) { + np = NULL; + rc = mdb_page_unspill(txn, mp, &np); + if (rc) + goto fail; + if (np) + goto done; + } + if ((rc = mdb_midl_need(&txn->mt_free_pgs, 1)) || + (rc = mdb_page_alloc(mc, 1, &np))) + goto fail; + pgno = np->mp_pgno; + DPRINTF(("touched db %d page %"Z"u -> %"Z"u", DDBI(mc), + mp->mp_pgno, pgno)); + mdb_cassert(mc, mp->mp_pgno != pgno); + mdb_midl_xappend(txn->mt_free_pgs, mp->mp_pgno); + /* Update the parent page, if any, to point to the new page */ + if (mc->mc_top) { + MDB_page *parent = mc->mc_pg[mc->mc_top-1]; + MDB_node *node = NODEPTR(parent, mc->mc_ki[mc->mc_top-1]); + SETPGNO(node, pgno); + } else { + mc->mc_db->md_root = pgno; + } + } else if (txn->mt_parent && !IS_SUBP(mp)) { + MDB_ID2 mid, *dl = txn->mt_u.dirty_list; + pgno = mp->mp_pgno; + /* If txn has a parent, make sure the page is in our + * dirty list. + */ + if (dl[0].mid) { + unsigned x = mdb_mid2l_search(dl, pgno); + if (x <= dl[0].mid && dl[x].mid == pgno) { + if (mp != dl[x].mptr) { /* bad cursor? */ + mc->mc_flags &= ~(C_INITIALIZED|C_EOF); + txn->mt_flags |= MDB_TXN_ERROR; + return MDB_CORRUPTED; + } + return 0; + } + } + mdb_cassert(mc, dl[0].mid < MDB_IDL_UM_MAX); + /* No - copy it */ + np = mdb_page_malloc(txn, 1); + if (!np) + return ENOMEM; + mid.mid = pgno; + mid.mptr = np; + rc = mdb_mid2l_insert(dl, &mid); + mdb_cassert(mc, rc == 0); + } else { + return 0; + } + + mdb_page_copy(np, mp, txn->mt_env->me_psize); + np->mp_pgno = pgno; + np->mp_flags |= P_DIRTY; + +done: + /* Adjust cursors pointing to mp */ + mc->mc_pg[mc->mc_top] = np; + m2 = txn->mt_cursors[mc->mc_dbi]; + if (mc->mc_flags & C_SUB) { + for (; m2; m2=m2->mc_next) { + m3 = &m2->mc_xcursor->mx_cursor; + if (m3->mc_snum < mc->mc_snum) continue; + if (m3->mc_pg[mc->mc_top] == mp) + m3->mc_pg[mc->mc_top] = np; + } + } else { + for (; m2; m2=m2->mc_next) { + if (m2->mc_snum < mc->mc_snum) continue; + if (m2 == mc) continue; + if (m2->mc_pg[mc->mc_top] == mp) { + m2->mc_pg[mc->mc_top] = np; + if (IS_LEAF(np)) + XCURSOR_REFRESH(m2, mc->mc_top, np); + } + } + } + return 0; + +fail: + txn->mt_flags |= MDB_TXN_ERROR; + return rc; +} + +int +mdb_env_sync(MDB_env *env, int force) +{ + int rc = 0; + if (env->me_flags & MDB_RDONLY) + return EACCES; + if (force || !F_ISSET(env->me_flags, MDB_NOSYNC)) { + if (env->me_flags & MDB_WRITEMAP) { + int flags = ((env->me_flags & MDB_MAPASYNC) && !force) + ? MS_ASYNC : MS_SYNC; + if (MDB_MSYNC(env->me_map, env->me_mapsize, flags)) + rc = ErrCode(); +#ifdef _WIN32 + else if (flags == MS_SYNC && MDB_FDATASYNC(env->me_fd)) + rc = ErrCode(); +#endif + } else { +#ifdef BROKEN_FDATASYNC + if (env->me_flags & MDB_FSYNCONLY) { + if (fsync(env->me_fd)) + rc = ErrCode(); + } else +#endif + if (MDB_FDATASYNC(env->me_fd)) + rc = ErrCode(); + } + } + return rc; +} + +/** Back up parent txn's cursors, then grab the originals for tracking */ +static int +mdb_cursor_shadow(MDB_txn *src, MDB_txn *dst) +{ + MDB_cursor *mc, *bk; + MDB_xcursor *mx; + size_t size; + int i; + + for (i = src->mt_numdbs; --i >= 0; ) { + if ((mc = src->mt_cursors[i]) != NULL) { + size = sizeof(MDB_cursor); + if (mc->mc_xcursor) + size += sizeof(MDB_xcursor); + for (; mc; mc = bk->mc_next) { + bk = malloc(size); + if (!bk) + return ENOMEM; + *bk = *mc; + mc->mc_backup = bk; + mc->mc_db = &dst->mt_dbs[i]; + /* Kill pointers into src to reduce abuse: The + * user may not use mc until dst ends. But we need a valid + * txn pointer here for cursor fixups to keep working. + */ + mc->mc_txn = dst; + mc->mc_dbflag = &dst->mt_dbflags[i]; + if ((mx = mc->mc_xcursor) != NULL) { + *(MDB_xcursor *)(bk+1) = *mx; + mx->mx_cursor.mc_txn = dst; + } + mc->mc_next = dst->mt_cursors[i]; + dst->mt_cursors[i] = mc; + } + } + } + return MDB_SUCCESS; +} + +/** Close this write txn's cursors, give parent txn's cursors back to parent. + * @param[in] txn the transaction handle. + * @param[in] merge true to keep changes to parent cursors, false to revert. + * @return 0 on success, non-zero on failure. + */ +static void +mdb_cursors_close(MDB_txn *txn, unsigned merge) +{ + MDB_cursor **cursors = txn->mt_cursors, *mc, *next, *bk; + MDB_xcursor *mx; + int i; + + for (i = txn->mt_numdbs; --i >= 0; ) { + for (mc = cursors[i]; mc; mc = next) { + next = mc->mc_next; + if ((bk = mc->mc_backup) != NULL) { + if (merge) { + /* Commit changes to parent txn */ + mc->mc_next = bk->mc_next; + mc->mc_backup = bk->mc_backup; + mc->mc_txn = bk->mc_txn; + mc->mc_db = bk->mc_db; + mc->mc_dbflag = bk->mc_dbflag; + if ((mx = mc->mc_xcursor) != NULL) + mx->mx_cursor.mc_txn = bk->mc_txn; + } else { + /* Abort nested txn */ + *mc = *bk; + if ((mx = mc->mc_xcursor) != NULL) + *mx = *(MDB_xcursor *)(bk+1); + } + mc = bk; + } + /* Only malloced cursors are permanently tracked. */ + free(mc); + } + cursors[i] = NULL; + } +} + +#if !(MDB_PIDLOCK) /* Currently the same as defined(_WIN32) */ +enum Pidlock_op { + Pidset, Pidcheck +}; +#else +enum Pidlock_op { + Pidset = F_SETLK, Pidcheck = F_GETLK +}; +#endif + +/** Set or check a pid lock. Set returns 0 on success. + * Check returns 0 if the process is certainly dead, nonzero if it may + * be alive (the lock exists or an error happened so we do not know). + * + * On Windows Pidset is a no-op, we merely check for the existence + * of the process with the given pid. On POSIX we use a single byte + * lock on the lockfile, set at an offset equal to the pid. + */ +static int +mdb_reader_pid(MDB_env *env, enum Pidlock_op op, MDB_PID_T pid) +{ +#if !(MDB_PIDLOCK) /* Currently the same as defined(_WIN32) */ + int ret = 0; + HANDLE h; + if (op == Pidcheck) { + h = OpenProcess(env->me_pidquery, FALSE, pid); + /* No documented "no such process" code, but other program use this: */ + if (!h) + return ErrCode() != ERROR_INVALID_PARAMETER; + /* A process exists until all handles to it close. Has it exited? */ + ret = WaitForSingleObject(h, 0) != 0; + CloseHandle(h); + } + return ret; +#else + for (;;) { + int rc; + struct flock lock_info; + memset(&lock_info, 0, sizeof(lock_info)); + lock_info.l_type = F_WRLCK; + lock_info.l_whence = SEEK_SET; + lock_info.l_start = pid; + lock_info.l_len = 1; + if ((rc = fcntl(env->me_lfd, op, &lock_info)) == 0) { + if (op == F_GETLK && lock_info.l_type != F_UNLCK) + rc = -1; + } else if ((rc = ErrCode()) == EINTR) { + continue; + } + return rc; + } +#endif +} + +/** Common code for #mdb_txn_begin() and #mdb_txn_renew(). + * @param[in] txn the transaction handle to initialize + * @return 0 on success, non-zero on failure. + */ +static int +mdb_txn_renew0(MDB_txn *txn) +{ + MDB_env *env = txn->mt_env; + MDB_txninfo *ti = env->me_txns; + MDB_meta *meta; + unsigned int i, nr, flags = txn->mt_flags; + uint16_t x; + int rc, new_notls = 0; + + if ((flags &= MDB_TXN_RDONLY) != 0) { + if (!ti) { + meta = mdb_env_pick_meta(env); + txn->mt_txnid = meta->mm_txnid; + txn->mt_u.reader = NULL; + } else { + MDB_reader *r = (env->me_flags & MDB_NOTLS) ? txn->mt_u.reader : + pthread_getspecific(env->me_txkey); + if (r) { + if (r->mr_pid != env->me_pid || r->mr_txnid != (txnid_t)-1) + return MDB_BAD_RSLOT; + } else { + MDB_PID_T pid = env->me_pid; + MDB_THR_T tid = pthread_self(); + mdb_mutexref_t rmutex = env->me_rmutex; + + if (!env->me_live_reader) { + rc = mdb_reader_pid(env, Pidset, pid); + if (rc) + return rc; + env->me_live_reader = 1; + } + + if (LOCK_MUTEX(rc, env, rmutex)) + return rc; + nr = ti->mti_numreaders; + for (i=0; i<nr; i++) + if (ti->mti_readers[i].mr_pid == 0) + break; + if (i == env->me_maxreaders) { + UNLOCK_MUTEX(rmutex); + return MDB_READERS_FULL; + } + r = &ti->mti_readers[i]; + /* Claim the reader slot, carefully since other code + * uses the reader table un-mutexed: First reset the + * slot, next publish it in mti_numreaders. After + * that, it is safe for mdb_env_close() to touch it. + * When it will be closed, we can finally claim it. + */ + r->mr_pid = 0; + r->mr_txnid = (txnid_t)-1; + r->mr_tid = tid; + if (i == nr) + ti->mti_numreaders = ++nr; + env->me_close_readers = nr; + r->mr_pid = pid; + UNLOCK_MUTEX(rmutex); + + new_notls = (env->me_flags & MDB_NOTLS); + if (!new_notls && (rc=pthread_setspecific(env->me_txkey, r))) { + r->mr_pid = 0; + return rc; + } + } + do /* LY: Retry on a race, ITS#7970. */ + r->mr_txnid = ti->mti_txnid; + while(r->mr_txnid != ti->mti_txnid); + txn->mt_txnid = r->mr_txnid; + txn->mt_u.reader = r; + meta = env->me_metas[txn->mt_txnid & 1]; + } + + } else { + /* Not yet touching txn == env->me_txn0, it may be active */ + if (ti) { + if (LOCK_MUTEX(rc, env, env->me_wmutex)) + return rc; + txn->mt_txnid = ti->mti_txnid; + meta = env->me_metas[txn->mt_txnid & 1]; + } else { + meta = mdb_env_pick_meta(env); + txn->mt_txnid = meta->mm_txnid; + } + txn->mt_txnid++; +#if MDB_DEBUG + if (txn->mt_txnid == mdb_debug_start) + mdb_debug = 1; +#endif + txn->mt_child = NULL; + txn->mt_loose_pgs = NULL; + txn->mt_loose_count = 0; + txn->mt_dirty_room = MDB_IDL_UM_MAX; + txn->mt_u.dirty_list = env->me_dirty_list; + txn->mt_u.dirty_list[0].mid = 0; + txn->mt_free_pgs = env->me_free_pgs; + txn->mt_free_pgs[0] = 0; + txn->mt_spill_pgs = NULL; + env->me_txn = txn; + memcpy(txn->mt_dbiseqs, env->me_dbiseqs, env->me_maxdbs * sizeof(unsigned int)); + } + + /* Copy the DB info and flags */ + memcpy(txn->mt_dbs, meta->mm_dbs, CORE_DBS * sizeof(MDB_db)); + + /* Moved to here to avoid a data race in read TXNs */ + txn->mt_next_pgno = meta->mm_last_pg+1; + + txn->mt_flags = flags; + + /* Setup db info */ + txn->mt_numdbs = env->me_numdbs; + for (i=CORE_DBS; i<txn->mt_numdbs; i++) { + x = env->me_dbflags[i]; + txn->mt_dbs[i].md_flags = x & PERSISTENT_FLAGS; + txn->mt_dbflags[i] = (x & MDB_VALID) ? DB_VALID|DB_USRVALID|DB_STALE : 0; + } + txn->mt_dbflags[MAIN_DBI] = DB_VALID|DB_USRVALID; + txn->mt_dbflags[FREE_DBI] = DB_VALID; + + if (env->me_flags & MDB_FATAL_ERROR) { + DPUTS("environment had fatal error, must shutdown!"); + rc = MDB_PANIC; + } else if (env->me_maxpg < txn->mt_next_pgno) { + rc = MDB_MAP_RESIZED; + } else { + return MDB_SUCCESS; + } + mdb_txn_end(txn, new_notls /*0 or MDB_END_SLOT*/ | MDB_END_FAIL_BEGIN); + return rc; +} + +int +mdb_txn_renew(MDB_txn *txn) +{ + int rc; + + if (!txn || !F_ISSET(txn->mt_flags, MDB_TXN_RDONLY|MDB_TXN_FINISHED)) + return EINVAL; + + rc = mdb_txn_renew0(txn); + if (rc == MDB_SUCCESS) { + DPRINTF(("renew txn %"Z"u%c %p on mdbenv %p, root page %"Z"u", + txn->mt_txnid, (txn->mt_flags & MDB_TXN_RDONLY) ? 'r' : 'w', + (void *)txn, (void *)txn->mt_env, txn->mt_dbs[MAIN_DBI].md_root)); + } + return rc; +} + +int +mdb_txn_begin(MDB_env *env, MDB_txn *parent, unsigned int flags, MDB_txn **ret) +{ + MDB_txn *txn; + MDB_ntxn *ntxn; + int rc, size, tsize; + + flags &= MDB_TXN_BEGIN_FLAGS; + flags |= env->me_flags & MDB_WRITEMAP; + + if (env->me_flags & MDB_RDONLY & ~flags) /* write txn in RDONLY env */ + return EACCES; + + if (parent) { + /* Nested transactions: Max 1 child, write txns only, no writemap */ + flags |= parent->mt_flags; + if (flags & (MDB_RDONLY|MDB_WRITEMAP|MDB_TXN_BLOCKED)) { + return (parent->mt_flags & MDB_TXN_RDONLY) ? EINVAL : MDB_BAD_TXN; + } + /* Child txns save MDB_pgstate and use own copy of cursors */ + size = env->me_maxdbs * (sizeof(MDB_db)+sizeof(MDB_cursor *)+1); + size += tsize = sizeof(MDB_ntxn); + } else if (flags & MDB_RDONLY) { + size = env->me_maxdbs * (sizeof(MDB_db)+1); + size += tsize = sizeof(MDB_txn); + } else { + /* Reuse preallocated write txn. However, do not touch it until + * mdb_txn_renew0() succeeds, since it currently may be active. + */ + txn = env->me_txn0; + goto renew; + } + if ((txn = calloc(1, size)) == NULL) { + DPRINTF(("calloc: %s", strerror(errno))); + return ENOMEM; + } + txn->mt_dbxs = env->me_dbxs; /* static */ + txn->mt_dbs = (MDB_db *) ((char *)txn + tsize); + txn->mt_dbflags = (unsigned char *)txn + size - env->me_maxdbs; + txn->mt_flags = flags; + txn->mt_env = env; + + if (parent) { + unsigned int i; + txn->mt_cursors = (MDB_cursor **)(txn->mt_dbs + env->me_maxdbs); + txn->mt_dbiseqs = parent->mt_dbiseqs; + txn->mt_u.dirty_list = malloc(sizeof(MDB_ID2)*MDB_IDL_UM_SIZE); + if (!txn->mt_u.dirty_list || + !(txn->mt_free_pgs = mdb_midl_alloc(MDB_IDL_UM_MAX))) + { + free(txn->mt_u.dirty_list); + free(txn); + return ENOMEM; + } + txn->mt_txnid = parent->mt_txnid; + txn->mt_dirty_room = parent->mt_dirty_room; + txn->mt_u.dirty_list[0].mid = 0; + txn->mt_spill_pgs = NULL; + txn->mt_next_pgno = parent->mt_next_pgno; + parent->mt_flags |= MDB_TXN_HAS_CHILD; + parent->mt_child = txn; + txn->mt_parent = parent; + txn->mt_numdbs = parent->mt_numdbs; + memcpy(txn->mt_dbs, parent->mt_dbs, txn->mt_numdbs * sizeof(MDB_db)); + /* Copy parent's mt_dbflags, but clear DB_NEW */ + for (i=0; i<txn->mt_numdbs; i++) + txn->mt_dbflags[i] = parent->mt_dbflags[i] & ~DB_NEW; + rc = 0; + ntxn = (MDB_ntxn *)txn; + ntxn->mnt_pgstate = env->me_pgstate; /* save parent me_pghead & co */ + if (env->me_pghead) { + size = MDB_IDL_SIZEOF(env->me_pghead); + env->me_pghead = mdb_midl_alloc(env->me_pghead[0]); + if (env->me_pghead) + memcpy(env->me_pghead, ntxn->mnt_pgstate.mf_pghead, size); + else + rc = ENOMEM; + } + if (!rc) + rc = mdb_cursor_shadow(parent, txn); + if (rc) + mdb_txn_end(txn, MDB_END_FAIL_BEGINCHILD); + } else { /* MDB_RDONLY */ + txn->mt_dbiseqs = env->me_dbiseqs; +renew: + rc = mdb_txn_renew0(txn); + } + if (rc) { + if (txn != env->me_txn0) + free(txn); + } else { + txn->mt_flags |= flags; /* could not change txn=me_txn0 earlier */ + *ret = txn; + DPRINTF(("begin txn %"Z"u%c %p on mdbenv %p, root page %"Z"u", + txn->mt_txnid, (flags & MDB_RDONLY) ? 'r' : 'w', + (void *) txn, (void *) env, txn->mt_dbs[MAIN_DBI].md_root)); + } + + return rc; +} + +MDB_env * +mdb_txn_env(MDB_txn *txn) +{ + if(!txn) return NULL; + return txn->mt_env; +} + +size_t +mdb_txn_id(MDB_txn *txn) +{ + if(!txn) return 0; + return txn->mt_txnid; +} + +/** Export or close DBI handles opened in this txn. */ +static void +mdb_dbis_update(MDB_txn *txn, int keep) +{ + int i; + MDB_dbi n = txn->mt_numdbs; + MDB_env *env = txn->mt_env; + unsigned char *tdbflags = txn->mt_dbflags; + + for (i = n; --i >= CORE_DBS;) { + if (tdbflags[i] & DB_NEW) { + if (keep) { + env->me_dbflags[i] = txn->mt_dbs[i].md_flags | MDB_VALID; + } else { + char *ptr = env->me_dbxs[i].md_name.mv_data; + if (ptr) { + env->me_dbxs[i].md_name.mv_data = NULL; + env->me_dbxs[i].md_name.mv_size = 0; + env->me_dbflags[i] = 0; + env->me_dbiseqs[i]++; + free(ptr); + } + } + } + } + if (keep && env->me_numdbs < n) + env->me_numdbs = n; +} + +/** End a transaction, except successful commit of a nested transaction. + * May be called twice for readonly txns: First reset it, then abort. + * @param[in] txn the transaction handle to end + * @param[in] mode why and how to end the transaction + */ +static void +mdb_txn_end(MDB_txn *txn, unsigned mode) +{ + MDB_env *env = txn->mt_env; +#if MDB_DEBUG + static const char *const names[] = MDB_END_NAMES; +#endif + + /* Export or close DBI handles opened in this txn */ + mdb_dbis_update(txn, mode & MDB_END_UPDATE); + + DPRINTF(("%s txn %"Z"u%c %p on mdbenv %p, root page %"Z"u", + names[mode & MDB_END_OPMASK], + txn->mt_txnid, (txn->mt_flags & MDB_TXN_RDONLY) ? 'r' : 'w', + (void *) txn, (void *)env, txn->mt_dbs[MAIN_DBI].md_root)); + + if (F_ISSET(txn->mt_flags, MDB_TXN_RDONLY)) { + if (txn->mt_u.reader) { + txn->mt_u.reader->mr_txnid = (txnid_t)-1; + if (!(env->me_flags & MDB_NOTLS)) { + txn->mt_u.reader = NULL; /* txn does not own reader */ + } else if (mode & MDB_END_SLOT) { + txn->mt_u.reader->mr_pid = 0; + txn->mt_u.reader = NULL; + } /* else txn owns the slot until it does MDB_END_SLOT */ + } + txn->mt_numdbs = 0; /* prevent further DBI activity */ + txn->mt_flags |= MDB_TXN_FINISHED; + + } else if (!F_ISSET(txn->mt_flags, MDB_TXN_FINISHED)) { + pgno_t *pghead = env->me_pghead; + + if (!(mode & MDB_END_UPDATE)) /* !(already closed cursors) */ + mdb_cursors_close(txn, 0); + if (!(env->me_flags & MDB_WRITEMAP)) { + mdb_dlist_free(txn); + } + + txn->mt_numdbs = 0; + txn->mt_flags = MDB_TXN_FINISHED; + + if (!txn->mt_parent) { + mdb_midl_shrink(&txn->mt_free_pgs); + env->me_free_pgs = txn->mt_free_pgs; + /* me_pgstate: */ + env->me_pghead = NULL; + env->me_pglast = 0; + + env->me_txn = NULL; + mode = 0; /* txn == env->me_txn0, do not free() it */ + + /* The writer mutex was locked in mdb_txn_begin. */ + if (env->me_txns) + UNLOCK_MUTEX(env->me_wmutex); + } else { + txn->mt_parent->mt_child = NULL; + txn->mt_parent->mt_flags &= ~MDB_TXN_HAS_CHILD; + env->me_pgstate = ((MDB_ntxn *)txn)->mnt_pgstate; + mdb_midl_free(txn->mt_free_pgs); + mdb_midl_free(txn->mt_spill_pgs); + free(txn->mt_u.dirty_list); + } + + mdb_midl_free(pghead); + } + + if (mode & MDB_END_FREE) + free(txn); +} + +void +mdb_txn_reset(MDB_txn *txn) +{ + if (txn == NULL) + return; + + /* This call is only valid for read-only txns */ + if (!(txn->mt_flags & MDB_TXN_RDONLY)) + return; + + mdb_txn_end(txn, MDB_END_RESET); +} + +void +mdb_txn_abort(MDB_txn *txn) +{ + if (txn == NULL) + return; + + if (txn->mt_child) + mdb_txn_abort(txn->mt_child); + + mdb_txn_end(txn, MDB_END_ABORT|MDB_END_SLOT|MDB_END_FREE); +} + +/** Save the freelist as of this transaction to the freeDB. + * This changes the freelist. Keep trying until it stabilizes. + */ +static int +mdb_freelist_save(MDB_txn *txn) +{ + /* env->me_pghead[] can grow and shrink during this call. + * env->me_pglast and txn->mt_free_pgs[] can only grow. + * Page numbers cannot disappear from txn->mt_free_pgs[]. + */ + MDB_cursor mc; + MDB_env *env = txn->mt_env; + int rc, maxfree_1pg = env->me_maxfree_1pg, more = 1; + txnid_t pglast = 0, head_id = 0; + pgno_t freecnt = 0, *free_pgs, *mop; + ssize_t head_room = 0, total_room = 0, mop_len, clean_limit; + + mdb_cursor_init(&mc, txn, FREE_DBI, NULL); + + if (env->me_pghead) { + /* Make sure first page of freeDB is touched and on freelist */ + rc = mdb_page_search(&mc, NULL, MDB_PS_FIRST|MDB_PS_MODIFY); + if (rc && rc != MDB_NOTFOUND) + return rc; + } + + if (!env->me_pghead && txn->mt_loose_pgs) { + /* Put loose page numbers in mt_free_pgs, since + * we may be unable to return them to me_pghead. + */ + MDB_page *mp = txn->mt_loose_pgs; + if ((rc = mdb_midl_need(&txn->mt_free_pgs, txn->mt_loose_count)) != 0) + return rc; + for (; mp; mp = NEXT_LOOSE_PAGE(mp)) + mdb_midl_xappend(txn->mt_free_pgs, mp->mp_pgno); + txn->mt_loose_pgs = NULL; + txn->mt_loose_count = 0; + } + + /* MDB_RESERVE cancels meminit in ovpage malloc (when no WRITEMAP) */ + clean_limit = (env->me_flags & (MDB_NOMEMINIT|MDB_WRITEMAP)) + ? SSIZE_MAX : maxfree_1pg; + + for (;;) { + /* Come back here after each Put() in case freelist changed */ + MDB_val key, data; + pgno_t *pgs; + ssize_t j; + + /* If using records from freeDB which we have not yet + * deleted, delete them and any we reserved for me_pghead. + */ + while (pglast < env->me_pglast) { + rc = mdb_cursor_first(&mc, &key, NULL); + if (rc) + return rc; + pglast = head_id = *(txnid_t *)key.mv_data; + total_room = head_room = 0; + mdb_tassert(txn, pglast <= env->me_pglast); + rc = mdb_cursor_del(&mc, 0); + if (rc) + return rc; + } + + /* Save the IDL of pages freed by this txn, to a single record */ + if (freecnt < txn->mt_free_pgs[0]) { + if (!freecnt) { + /* Make sure last page of freeDB is touched and on freelist */ + rc = mdb_page_search(&mc, NULL, MDB_PS_LAST|MDB_PS_MODIFY); + if (rc && rc != MDB_NOTFOUND) + return rc; + } + free_pgs = txn->mt_free_pgs; + /* Write to last page of freeDB */ + key.mv_size = sizeof(txn->mt_txnid); + key.mv_data = &txn->mt_txnid; + do { + freecnt = free_pgs[0]; + data.mv_size = MDB_IDL_SIZEOF(free_pgs); + rc = mdb_cursor_put(&mc, &key, &data, MDB_RESERVE); + if (rc) + return rc; + /* Retry if mt_free_pgs[] grew during the Put() */ + free_pgs = txn->mt_free_pgs; + } while (freecnt < free_pgs[0]); + mdb_midl_sort(free_pgs); + memcpy(data.mv_data, free_pgs, data.mv_size); +#if (MDB_DEBUG) > 1 + { + unsigned int i = free_pgs[0]; + DPRINTF(("IDL write txn %"Z"u root %"Z"u num %u", + txn->mt_txnid, txn->mt_dbs[FREE_DBI].md_root, i)); + for (; i; i--) + DPRINTF(("IDL %"Z"u", free_pgs[i])); + } +#endif + continue; + } + + mop = env->me_pghead; + mop_len = (mop ? mop[0] : 0) + txn->mt_loose_count; + + /* Reserve records for me_pghead[]. Split it if multi-page, + * to avoid searching freeDB for a page range. Use keys in + * range [1,me_pglast]: Smaller than txnid of oldest reader. + */ + if (total_room >= mop_len) { + if (total_room == mop_len || --more < 0) + break; + } else if (head_room >= maxfree_1pg && head_id > 1) { + /* Keep current record (overflow page), add a new one */ + head_id--; + head_room = 0; + } + /* (Re)write {key = head_id, IDL length = head_room} */ + total_room -= head_room; + head_room = mop_len - total_room; + if (head_room > maxfree_1pg && head_id > 1) { + /* Overflow multi-page for part of me_pghead */ + head_room /= head_id; /* amortize page sizes */ + head_room += maxfree_1pg - head_room % (maxfree_1pg + 1); + } else if (head_room < 0) { + /* Rare case, not bothering to delete this record */ + head_room = 0; + } + key.mv_size = sizeof(head_id); + key.mv_data = &head_id; + data.mv_size = (head_room + 1) * sizeof(pgno_t); + rc = mdb_cursor_put(&mc, &key, &data, MDB_RESERVE); + if (rc) + return rc; + /* IDL is initially empty, zero out at least the length */ + pgs = (pgno_t *)data.mv_data; + j = head_room > clean_limit ? head_room : 0; + do { + pgs[j] = 0; + } while (--j >= 0); + total_room += head_room; + } + + /* Return loose page numbers to me_pghead, though usually none are + * left at this point. The pages themselves remain in dirty_list. + */ + if (txn->mt_loose_pgs) { + MDB_page *mp = txn->mt_loose_pgs; + unsigned count = txn->mt_loose_count; + MDB_IDL loose; + /* Room for loose pages + temp IDL with same */ + if ((rc = mdb_midl_need(&env->me_pghead, 2*count+1)) != 0) + return rc; + mop = env->me_pghead; + loose = mop + MDB_IDL_ALLOCLEN(mop) - count; + for (count = 0; mp; mp = NEXT_LOOSE_PAGE(mp)) + loose[ ++count ] = mp->mp_pgno; + loose[0] = count; + mdb_midl_sort(loose); + mdb_midl_xmerge(mop, loose); + txn->mt_loose_pgs = NULL; + txn->mt_loose_count = 0; + mop_len = mop[0]; + } + + /* Fill in the reserved me_pghead records */ + rc = MDB_SUCCESS; + if (mop_len) { + MDB_val key, data; + + mop += mop_len; + rc = mdb_cursor_first(&mc, &key, &data); + for (; !rc; rc = mdb_cursor_next(&mc, &key, &data, MDB_NEXT)) { + txnid_t id = *(txnid_t *)key.mv_data; + ssize_t len = (ssize_t)(data.mv_size / sizeof(MDB_ID)) - 1; + MDB_ID save; + + mdb_tassert(txn, len >= 0 && id <= env->me_pglast); + key.mv_data = &id; + if (len > mop_len) { + len = mop_len; + data.mv_size = (len + 1) * sizeof(MDB_ID); + } + data.mv_data = mop -= len; + save = mop[0]; + mop[0] = len; + rc = mdb_cursor_put(&mc, &key, &data, MDB_CURRENT); + mop[0] = save; + if (rc || !(mop_len -= len)) + break; + } + } + return rc; +} + +/** Flush (some) dirty pages to the map, after clearing their dirty flag. + * @param[in] txn the transaction that's being committed + * @param[in] keep number of initial pages in dirty_list to keep dirty. + * @return 0 on success, non-zero on failure. + */ +static int +mdb_page_flush(MDB_txn *txn, int keep) +{ + MDB_env *env = txn->mt_env; + MDB_ID2L dl = txn->mt_u.dirty_list; + unsigned psize = env->me_psize, j; + int i, pagecount = dl[0].mid, rc; + size_t size = 0, pos = 0; + pgno_t pgno = 0; + MDB_page *dp = NULL; +#ifdef _WIN32 + OVERLAPPED ov; +#else + struct iovec iov[MDB_COMMIT_PAGES]; + ssize_t wpos = 0, wsize = 0, wres; + size_t next_pos = 1; /* impossible pos, so pos != next_pos */ + int n = 0; +#endif + + j = i = keep; + + if (env->me_flags & MDB_WRITEMAP) { + /* Clear dirty flags */ + while (++i <= pagecount) { + dp = dl[i].mptr; + /* Don't flush this page yet */ + if (dp->mp_flags & (P_LOOSE|P_KEEP)) { + dp->mp_flags &= ~P_KEEP; + dl[++j] = dl[i]; + continue; + } + dp->mp_flags &= ~P_DIRTY; + } + goto done; + } + + /* Write the pages */ + for (;;) { + if (++i <= pagecount) { + dp = dl[i].mptr; + /* Don't flush this page yet */ + if (dp->mp_flags & (P_LOOSE|P_KEEP)) { + dp->mp_flags &= ~P_KEEP; + dl[i].mid = 0; + continue; + } + pgno = dl[i].mid; + /* clear dirty flag */ + dp->mp_flags &= ~P_DIRTY; + pos = pgno * psize; + size = psize; + if (IS_OVERFLOW(dp)) size *= dp->mp_pages; + } +#ifdef _WIN32 + else break; + + /* Windows actually supports scatter/gather I/O, but only on + * unbuffered file handles. Since we're relying on the OS page + * cache for all our data, that's self-defeating. So we just + * write pages one at a time. We use the ov structure to set + * the write offset, to at least save the overhead of a Seek + * system call. + */ + DPRINTF(("committing page %"Z"u", pgno)); + memset(&ov, 0, sizeof(ov)); + ov.Offset = pos & 0xffffffff; + ov.OffsetHigh = pos >> 16 >> 16; + if (!WriteFile(env->me_fd, dp, size, NULL, &ov)) { + rc = ErrCode(); + DPRINTF(("WriteFile: %d", rc)); + return rc; + } +#else + /* Write up to MDB_COMMIT_PAGES dirty pages at a time. */ + if (pos!=next_pos || n==MDB_COMMIT_PAGES || wsize+size>MAX_WRITE) { + if (n) { +retry_write: + /* Write previous page(s) */ +#ifdef MDB_USE_PWRITEV + wres = pwritev(env->me_fd, iov, n, wpos); +#else + if (n == 1) { + wres = pwrite(env->me_fd, iov[0].iov_base, wsize, wpos); + } else { +retry_seek: + if (lseek(env->me_fd, wpos, SEEK_SET) == -1) { + rc = ErrCode(); + if (rc == EINTR) + goto retry_seek; + DPRINTF(("lseek: %s", strerror(rc))); + return rc; + } + wres = writev(env->me_fd, iov, n); + } +#endif + if (wres != wsize) { + if (wres < 0) { + rc = ErrCode(); + if (rc == EINTR) + goto retry_write; + DPRINTF(("Write error: %s", strerror(rc))); + } else { + rc = EIO; /* TODO: Use which error code? */ + DPUTS("short write, filesystem full?"); + } + return rc; + } + n = 0; + } + if (i > pagecount) + break; + wpos = pos; + wsize = 0; + } + DPRINTF(("committing page %"Z"u", pgno)); + next_pos = pos + size; + iov[n].iov_len = size; + iov[n].iov_base = (char *)dp; + wsize += size; + n++; +#endif /* _WIN32 */ + } + + /* MIPS has cache coherency issues, this is a no-op everywhere else + * Note: for any size >= on-chip cache size, entire on-chip cache is + * flushed. + */ + CACHEFLUSH(env->me_map, txn->mt_next_pgno * env->me_psize, DCACHE); + + for (i = keep; ++i <= pagecount; ) { + dp = dl[i].mptr; + /* This is a page we skipped above */ + if (!dl[i].mid) { + dl[++j] = dl[i]; + dl[j].mid = dp->mp_pgno; + continue; + } + mdb_dpage_free(env, dp); + } + +done: + i--; + txn->mt_dirty_room += i - j; + dl[0].mid = j; + return MDB_SUCCESS; +} + +int +mdb_txn_commit(MDB_txn *txn) +{ + int rc; + unsigned int i, end_mode; + MDB_env *env; + + if (txn == NULL) + return EINVAL; + + /* mdb_txn_end() mode for a commit which writes nothing */ + end_mode = MDB_END_EMPTY_COMMIT|MDB_END_UPDATE|MDB_END_SLOT|MDB_END_FREE; + + if (txn->mt_child) { + rc = mdb_txn_commit(txn->mt_child); + if (rc) + goto fail; + } + + env = txn->mt_env; + + if (F_ISSET(txn->mt_flags, MDB_TXN_RDONLY)) { + goto done; + } + + if (txn->mt_flags & (MDB_TXN_FINISHED|MDB_TXN_ERROR)) { + DPUTS("txn has failed/finished, can't commit"); + if (txn->mt_parent) + txn->mt_parent->mt_flags |= MDB_TXN_ERROR; + rc = MDB_BAD_TXN; + goto fail; + } + + if (txn->mt_parent) { + MDB_txn *parent = txn->mt_parent; + MDB_page **lp; + MDB_ID2L dst, src; + MDB_IDL pspill; + unsigned x, y, len, ps_len; + + /* Append our free list to parent's */ + rc = mdb_midl_append_list(&parent->mt_free_pgs, txn->mt_free_pgs); + if (rc) + goto fail; + mdb_midl_free(txn->mt_free_pgs); + /* Failures after this must either undo the changes + * to the parent or set MDB_TXN_ERROR in the parent. + */ + + parent->mt_next_pgno = txn->mt_next_pgno; + parent->mt_flags = txn->mt_flags; + + /* Merge our cursors into parent's and close them */ + mdb_cursors_close(txn, 1); + + /* Update parent's DB table. */ + memcpy(parent->mt_dbs, txn->mt_dbs, txn->mt_numdbs * sizeof(MDB_db)); + parent->mt_numdbs = txn->mt_numdbs; + parent->mt_dbflags[FREE_DBI] = txn->mt_dbflags[FREE_DBI]; + parent->mt_dbflags[MAIN_DBI] = txn->mt_dbflags[MAIN_DBI]; + for (i=CORE_DBS; i<txn->mt_numdbs; i++) { + /* preserve parent's DB_NEW status */ + x = parent->mt_dbflags[i] & DB_NEW; + parent->mt_dbflags[i] = txn->mt_dbflags[i] | x; + } + + dst = parent->mt_u.dirty_list; + src = txn->mt_u.dirty_list; + /* Remove anything in our dirty list from parent's spill list */ + if ((pspill = parent->mt_spill_pgs) && (ps_len = pspill[0])) { + x = y = ps_len; + pspill[0] = (pgno_t)-1; + /* Mark our dirty pages as deleted in parent spill list */ + for (i=0, len=src[0].mid; ++i <= len; ) { + MDB_ID pn = src[i].mid << 1; + while (pn > pspill[x]) + x--; + if (pn == pspill[x]) { + pspill[x] = 1; + y = --x; + } + } + /* Squash deleted pagenums if we deleted any */ + for (x=y; ++x <= ps_len; ) + if (!(pspill[x] & 1)) + pspill[++y] = pspill[x]; + pspill[0] = y; + } + + /* Remove anything in our spill list from parent's dirty list */ + if (txn->mt_spill_pgs && txn->mt_spill_pgs[0]) { + for (i=1; i<=txn->mt_spill_pgs[0]; i++) { + MDB_ID pn = txn->mt_spill_pgs[i]; + if (pn & 1) + continue; /* deleted spillpg */ + pn >>= 1; + y = mdb_mid2l_search(dst, pn); + if (y <= dst[0].mid && dst[y].mid == pn) { + free(dst[y].mptr); + while (y < dst[0].mid) { + dst[y] = dst[y+1]; + y++; + } + dst[0].mid--; + } + } + } + + /* Find len = length of merging our dirty list with parent's */ + x = dst[0].mid; + dst[0].mid = 0; /* simplify loops */ + if (parent->mt_parent) { + len = x + src[0].mid; + y = mdb_mid2l_search(src, dst[x].mid + 1) - 1; + for (i = x; y && i; y--) { + pgno_t yp = src[y].mid; + while (yp < dst[i].mid) + i--; + if (yp == dst[i].mid) { + i--; + len--; + } + } + } else { /* Simplify the above for single-ancestor case */ + len = MDB_IDL_UM_MAX - txn->mt_dirty_room; + } + /* Merge our dirty list with parent's */ + y = src[0].mid; + for (i = len; y; dst[i--] = src[y--]) { + pgno_t yp = src[y].mid; + while (yp < dst[x].mid) + dst[i--] = dst[x--]; + if (yp == dst[x].mid) + free(dst[x--].mptr); + } + mdb_tassert(txn, i == x); + dst[0].mid = len; + free(txn->mt_u.dirty_list); + parent->mt_dirty_room = txn->mt_dirty_room; + if (txn->mt_spill_pgs) { + if (parent->mt_spill_pgs) { + /* TODO: Prevent failure here, so parent does not fail */ + rc = mdb_midl_append_list(&parent->mt_spill_pgs, txn->mt_spill_pgs); + if (rc) + parent->mt_flags |= MDB_TXN_ERROR; + mdb_midl_free(txn->mt_spill_pgs); + mdb_midl_sort(parent->mt_spill_pgs); + } else { + parent->mt_spill_pgs = txn->mt_spill_pgs; + } + } + + /* Append our loose page list to parent's */ + for (lp = &parent->mt_loose_pgs; *lp; lp = &NEXT_LOOSE_PAGE(*lp)) + ; + *lp = txn->mt_loose_pgs; + parent->mt_loose_count += txn->mt_loose_count; + + parent->mt_child = NULL; + mdb_midl_free(((MDB_ntxn *)txn)->mnt_pgstate.mf_pghead); + free(txn); + return rc; + } + + if (txn != env->me_txn) { + DPUTS("attempt to commit unknown transaction"); + rc = EINVAL; + goto fail; + } + + mdb_cursors_close(txn, 0); + + if (!txn->mt_u.dirty_list[0].mid && + !(txn->mt_flags & (MDB_TXN_DIRTY|MDB_TXN_SPILLS))) + goto done; + + DPRINTF(("committing txn %"Z"u %p on mdbenv %p, root page %"Z"u", + txn->mt_txnid, (void*)txn, (void*)env, txn->mt_dbs[MAIN_DBI].md_root)); + + /* Update DB root pointers */ + if (txn->mt_numdbs > CORE_DBS) { + MDB_cursor mc; + MDB_dbi i; + MDB_val data; + data.mv_size = sizeof(MDB_db); + + mdb_cursor_init(&mc, txn, MAIN_DBI, NULL); + for (i = CORE_DBS; i < txn->mt_numdbs; i++) { + if (txn->mt_dbflags[i] & DB_DIRTY) { + if (TXN_DBI_CHANGED(txn, i)) { + rc = MDB_BAD_DBI; + goto fail; + } + data.mv_data = &txn->mt_dbs[i]; + rc = mdb_cursor_put(&mc, &txn->mt_dbxs[i].md_name, &data, + F_SUBDATA); + if (rc) + goto fail; + } + } + } + + rc = mdb_freelist_save(txn); + if (rc) + goto fail; + + mdb_midl_free(env->me_pghead); + env->me_pghead = NULL; + mdb_midl_shrink(&txn->mt_free_pgs); + +#if (MDB_DEBUG) > 2 + mdb_audit(txn); +#endif + + if ((rc = mdb_page_flush(txn, 0)) || + (rc = mdb_env_sync(env, 0)) || + (rc = mdb_env_write_meta(txn))) + goto fail; + end_mode = MDB_END_COMMITTED|MDB_END_UPDATE; + +done: + mdb_txn_end(txn, end_mode); + return MDB_SUCCESS; + +fail: + mdb_txn_abort(txn); + return rc; +} + +/** Read the environment parameters of a DB environment before + * mapping it into memory. + * @param[in] env the environment handle + * @param[out] meta address of where to store the meta information + * @return 0 on success, non-zero on failure. + */ +static int ESECT +mdb_env_read_header(MDB_env *env, MDB_meta *meta) +{ + MDB_metabuf pbuf; + MDB_page *p; + MDB_meta *m; + int i, rc, off; + enum { Size = sizeof(pbuf) }; + + /* We don't know the page size yet, so use a minimum value. + * Read both meta pages so we can use the latest one. + */ + + for (i=off=0; i<NUM_METAS; i++, off += meta->mm_psize) { +#ifdef _WIN32 + DWORD len; + OVERLAPPED ov; + memset(&ov, 0, sizeof(ov)); + ov.Offset = off; + rc = ReadFile(env->me_fd, &pbuf, Size, &len, &ov) ? (int)len : -1; + if (rc == -1 && ErrCode() == ERROR_HANDLE_EOF) + rc = 0; +#else + rc = pread(env->me_fd, &pbuf, Size, off); +#endif + if (rc != Size) { + if (rc == 0 && off == 0) + return ENOENT; + rc = rc < 0 ? (int) ErrCode() : MDB_INVALID; + DPRINTF(("read: %s", mdb_strerror(rc))); + return rc; + } + + p = (MDB_page *)&pbuf; + + if (!F_ISSET(p->mp_flags, P_META)) { + DPRINTF(("page %"Z"u not a meta page", p->mp_pgno)); + return MDB_INVALID; + } + + m = METADATA(p); + if (m->mm_magic != MDB_MAGIC) { + DPUTS("meta has invalid magic"); + return MDB_INVALID; + } + + if (m->mm_version != MDB_DATA_VERSION) { + DPRINTF(("database is version %u, expected version %u", + m->mm_version, MDB_DATA_VERSION)); + return MDB_VERSION_MISMATCH; + } + + if (off == 0 || m->mm_txnid > meta->mm_txnid) + *meta = *m; + } + return 0; +} + +/** Fill in most of the zeroed #MDB_meta for an empty database environment */ +static void ESECT +mdb_env_init_meta0(MDB_env *env, MDB_meta *meta) +{ + meta->mm_magic = MDB_MAGIC; + meta->mm_version = MDB_DATA_VERSION; + meta->mm_mapsize = env->me_mapsize; + meta->mm_psize = env->me_psize; + meta->mm_last_pg = NUM_METAS-1; + meta->mm_flags = env->me_flags & 0xffff; + meta->mm_flags |= MDB_INTEGERKEY; /* this is mm_dbs[FREE_DBI].md_flags */ + meta->mm_dbs[FREE_DBI].md_root = P_INVALID; + meta->mm_dbs[MAIN_DBI].md_root = P_INVALID; +} + +/** Write the environment parameters of a freshly created DB environment. + * @param[in] env the environment handle + * @param[in] meta the #MDB_meta to write + * @return 0 on success, non-zero on failure. + */ +static int ESECT +mdb_env_init_meta(MDB_env *env, MDB_meta *meta) +{ + MDB_page *p, *q; + int rc; + unsigned int psize; +#ifdef _WIN32 + DWORD len; + OVERLAPPED ov; + memset(&ov, 0, sizeof(ov)); +#define DO_PWRITE(rc, fd, ptr, size, len, pos) do { \ + ov.Offset = pos; \ + rc = WriteFile(fd, ptr, size, &len, &ov); } while(0) +#else + int len; +#define DO_PWRITE(rc, fd, ptr, size, len, pos) do { \ + len = pwrite(fd, ptr, size, pos); \ + if (len == -1 && ErrCode() == EINTR) continue; \ + rc = (len >= 0); break; } while(1) +#endif + + DPUTS("writing new meta page"); + + psize = env->me_psize; + + p = calloc(NUM_METAS, psize); + if (!p) + return ENOMEM; + + p->mp_pgno = 0; + p->mp_flags = P_META; + *(MDB_meta *)METADATA(p) = *meta; + + q = (MDB_page *)((char *)p + psize); + q->mp_pgno = 1; + q->mp_flags = P_META; + *(MDB_meta *)METADATA(q) = *meta; + + DO_PWRITE(rc, env->me_fd, p, psize * NUM_METAS, len, 0); + if (!rc) + rc = ErrCode(); + else if ((unsigned) len == psize * NUM_METAS) + rc = MDB_SUCCESS; + else + rc = ENOSPC; + free(p); + return rc; +} + +/** Update the environment info to commit a transaction. + * @param[in] txn the transaction that's being committed + * @return 0 on success, non-zero on failure. + */ +static int +mdb_env_write_meta(MDB_txn *txn) +{ + MDB_env *env; + MDB_meta meta, metab, *mp; + unsigned flags; + size_t mapsize; + off_t off; + int rc, len, toggle; + char *ptr; + HANDLE mfd; +#ifdef _WIN32 + OVERLAPPED ov; +#else + int r2; +#endif + + toggle = txn->mt_txnid & 1; + DPRINTF(("writing meta page %d for root page %"Z"u", + toggle, txn->mt_dbs[MAIN_DBI].md_root)); + + env = txn->mt_env; + flags = env->me_flags; + mp = env->me_metas[toggle]; + mapsize = env->me_metas[toggle ^ 1]->mm_mapsize; + /* Persist any increases of mapsize config */ + if (mapsize < env->me_mapsize) + mapsize = env->me_mapsize; + + if (flags & MDB_WRITEMAP) { + mp->mm_mapsize = mapsize; + mp->mm_dbs[FREE_DBI] = txn->mt_dbs[FREE_DBI]; + mp->mm_dbs[MAIN_DBI] = txn->mt_dbs[MAIN_DBI]; + mp->mm_last_pg = txn->mt_next_pgno - 1; +#if (__GNUC__ * 100 + __GNUC_MINOR__ >= 404) && /* TODO: portability */ \ + !(defined(__i386__) || defined(__x86_64__)) + /* LY: issue a memory barrier, if not x86. ITS#7969 */ + __sync_synchronize(); +#endif + mp->mm_txnid = txn->mt_txnid; + if (!(flags & (MDB_NOMETASYNC|MDB_NOSYNC))) { + unsigned meta_size = env->me_psize; + rc = (env->me_flags & MDB_MAPASYNC) ? MS_ASYNC : MS_SYNC; + ptr = (char *)mp - PAGEHDRSZ; +#ifndef _WIN32 /* POSIX msync() requires ptr = start of OS page */ + r2 = (ptr - env->me_map) & (env->me_os_psize - 1); + ptr -= r2; + meta_size += r2; +#endif + if (MDB_MSYNC(ptr, meta_size, rc)) { + rc = ErrCode(); + goto fail; + } + } + goto done; + } + metab.mm_txnid = mp->mm_txnid; + metab.mm_last_pg = mp->mm_last_pg; + + meta.mm_mapsize = mapsize; + meta.mm_dbs[FREE_DBI] = txn->mt_dbs[FREE_DBI]; + meta.mm_dbs[MAIN_DBI] = txn->mt_dbs[MAIN_DBI]; + meta.mm_last_pg = txn->mt_next_pgno - 1; + meta.mm_txnid = txn->mt_txnid; + + off = offsetof(MDB_meta, mm_mapsize); + ptr = (char *)&meta + off; + len = sizeof(MDB_meta) - off; + off += (char *)mp - env->me_map; + + /* Write to the SYNC fd unless MDB_NOSYNC/MDB_NOMETASYNC. + * (me_mfd goes to the same file as me_fd, but writing to it + * also syncs to disk. Avoids a separate fdatasync() call.) + */ + mfd = (flags & (MDB_NOSYNC|MDB_NOMETASYNC)) ? env->me_fd : env->me_mfd; +#ifdef _WIN32 + { + memset(&ov, 0, sizeof(ov)); + ov.Offset = off; + if (!WriteFile(mfd, ptr, len, (DWORD *)&rc, &ov)) + rc = -1; + } +#else +retry_write: + rc = pwrite(mfd, ptr, len, off); +#endif + if (rc != len) { + rc = rc < 0 ? ErrCode() : EIO; +#ifndef _WIN32 + if (rc == EINTR) + goto retry_write; +#endif + DPUTS("write failed, disk error?"); + /* On a failure, the pagecache still contains the new data. + * Write some old data back, to prevent it from being used. + * Use the non-SYNC fd; we know it will fail anyway. + */ + meta.mm_last_pg = metab.mm_last_pg; + meta.mm_txnid = metab.mm_txnid; +#ifdef _WIN32 + memset(&ov, 0, sizeof(ov)); + ov.Offset = off; + WriteFile(env->me_fd, ptr, len, NULL, &ov); +#else + r2 = pwrite(env->me_fd, ptr, len, off); + (void)r2; /* Silence warnings. We don't care about pwrite's return value */ +#endif +fail: + env->me_flags |= MDB_FATAL_ERROR; + return rc; + } + /* MIPS has cache coherency issues, this is a no-op everywhere else */ + CACHEFLUSH(env->me_map + off, len, DCACHE); +done: + /* Memory ordering issues are irrelevant; since the entire writer + * is wrapped by wmutex, all of these changes will become visible + * after the wmutex is unlocked. Since the DB is multi-version, + * readers will get consistent data regardless of how fresh or + * how stale their view of these values is. + */ + if (env->me_txns) + env->me_txns->mti_txnid = txn->mt_txnid; + + return MDB_SUCCESS; +} + +/** Check both meta pages to see which one is newer. + * @param[in] env the environment handle + * @return newest #MDB_meta. + */ +static MDB_meta * +mdb_env_pick_meta(const MDB_env *env) +{ + MDB_meta *const *metas = env->me_metas; + return metas[ metas[0]->mm_txnid < metas[1]->mm_txnid ]; +} + +int ESECT +mdb_env_create(MDB_env **env) +{ + MDB_env *e; + + e = calloc(1, sizeof(MDB_env)); + if (!e) + return ENOMEM; + + e->me_maxreaders = DEFAULT_READERS; + e->me_maxdbs = e->me_numdbs = CORE_DBS; + e->me_fd = INVALID_HANDLE_VALUE; + e->me_lfd = INVALID_HANDLE_VALUE; + e->me_mfd = INVALID_HANDLE_VALUE; +#ifdef MDB_USE_POSIX_SEM + e->me_rmutex = SEM_FAILED; + e->me_wmutex = SEM_FAILED; +#endif + e->me_pid = getpid(); + GET_PAGESIZE(e->me_os_psize); + VGMEMP_CREATE(e,0,0); + *env = e; + return MDB_SUCCESS; +} + +static int ESECT +mdb_env_map(MDB_env *env, void *addr) +{ + MDB_page *p; + unsigned int flags = env->me_flags; +#ifdef _WIN32 + int rc; + HANDLE mh; + LONG sizelo, sizehi; + size_t msize; + + if (flags & MDB_RDONLY) { + /* Don't set explicit map size, use whatever exists */ + msize = 0; + sizelo = 0; + sizehi = 0; + } else { + msize = env->me_mapsize; + sizelo = msize & 0xffffffff; + sizehi = msize >> 16 >> 16; /* only needed on Win64 */ + + /* Windows won't create mappings for zero length files. + * and won't map more than the file size. + * Just set the maxsize right now. + */ + if (SetFilePointer(env->me_fd, sizelo, &sizehi, 0) != (DWORD)sizelo + || !SetEndOfFile(env->me_fd) + || SetFilePointer(env->me_fd, 0, NULL, 0) != 0) + return ErrCode(); + } + + mh = CreateFileMapping(env->me_fd, NULL, flags & MDB_WRITEMAP ? + PAGE_READWRITE : PAGE_READONLY, + sizehi, sizelo, NULL); + if (!mh) + return ErrCode(); + env->me_map = MapViewOfFileEx(mh, flags & MDB_WRITEMAP ? + FILE_MAP_WRITE : FILE_MAP_READ, + 0, 0, msize, addr); + rc = env->me_map ? 0 : ErrCode(); + CloseHandle(mh); + if (rc) + return rc; +#else + int prot = PROT_READ; + if (flags & MDB_WRITEMAP) { + prot |= PROT_WRITE; + if (ftruncate(env->me_fd, env->me_mapsize) < 0) + return ErrCode(); + } + env->me_map = mmap(addr, env->me_mapsize, prot, MAP_SHARED, + env->me_fd, 0); + if (env->me_map == MAP_FAILED) { + env->me_map = NULL; + return ErrCode(); + } + + if (flags & MDB_NORDAHEAD) { + /* Turn off readahead. It's harmful when the DB is larger than RAM. */ +#ifdef MADV_RANDOM + madvise(env->me_map, env->me_mapsize, MADV_RANDOM); +#else +#ifdef POSIX_MADV_RANDOM + posix_madvise(env->me_map, env->me_mapsize, POSIX_MADV_RANDOM); +#endif /* POSIX_MADV_RANDOM */ +#endif /* MADV_RANDOM */ + } +#endif /* _WIN32 */ + + /* Can happen because the address argument to mmap() is just a + * hint. mmap() can pick another, e.g. if the range is in use. + * The MAP_FIXED flag would prevent that, but then mmap could + * instead unmap existing pages to make room for the new map. + */ + if (addr && env->me_map != addr) + return EBUSY; /* TODO: Make a new MDB_* error code? */ + + p = (MDB_page *)env->me_map; + env->me_metas[0] = METADATA(p); + env->me_metas[1] = (MDB_meta *)((char *)env->me_metas[0] + env->me_psize); + + return MDB_SUCCESS; +} + +int ESECT +mdb_env_set_mapsize(MDB_env *env, size_t size) +{ + /* If env is already open, caller is responsible for making + * sure there are no active txns. + */ + if (env->me_map) { + int rc; + MDB_meta *meta; + void *old; + if (env->me_txn) + return EINVAL; + meta = mdb_env_pick_meta(env); + if (!size) + size = meta->mm_mapsize; + { + /* Silently round up to minimum if the size is too small */ + size_t minsize = (meta->mm_last_pg + 1) * env->me_psize; + if (size < minsize) + size = minsize; + } + munmap(env->me_map, env->me_mapsize); + env->me_mapsize = size; + old = (env->me_flags & MDB_FIXEDMAP) ? env->me_map : NULL; + rc = mdb_env_map(env, old); + if (rc) + return rc; + } + env->me_mapsize = size; + if (env->me_psize) + env->me_maxpg = env->me_mapsize / env->me_psize; + return MDB_SUCCESS; +} + +int ESECT +mdb_env_set_maxdbs(MDB_env *env, MDB_dbi dbs) +{ + if (env->me_map) + return EINVAL; + env->me_maxdbs = dbs + CORE_DBS; + return MDB_SUCCESS; +} + +int ESECT +mdb_env_set_maxreaders(MDB_env *env, unsigned int readers) +{ + if (env->me_map || readers < 1) + return EINVAL; + env->me_maxreaders = readers; + return MDB_SUCCESS; +} + +int ESECT +mdb_env_get_maxreaders(MDB_env *env, unsigned int *readers) +{ + if (!env || !readers) + return EINVAL; + *readers = env->me_maxreaders; + return MDB_SUCCESS; +} + +static int ESECT +mdb_fsize(HANDLE fd, size_t *size) +{ +#ifdef _WIN32 + LARGE_INTEGER fsize; + + if (!GetFileSizeEx(fd, &fsize)) + return ErrCode(); + + *size = fsize.QuadPart; +#else + struct stat st; + + if (fstat(fd, &st)) + return ErrCode(); + + *size = st.st_size; +#endif + return MDB_SUCCESS; +} + + +#ifdef _WIN32 +typedef wchar_t mdb_nchar_t; +# define MDB_NAME(str) L##str +# define mdb_name_cpy wcscpy +#else +/** Character type for file names: char on Unix, wchar_t on Windows */ +typedef char mdb_nchar_t; +# define MDB_NAME(str) str /**< #mdb_nchar_t[] string literal */ +# define mdb_name_cpy strcpy /**< Copy name (#mdb_nchar_t string) */ +#endif + +/** Filename - string of #mdb_nchar_t[] */ +typedef struct MDB_name { + int mn_len; /**< Length */ + int mn_alloced; /**< True if #mn_val was malloced */ + mdb_nchar_t *mn_val; /**< Contents */ +} MDB_name; + +/** Filename suffixes [datafile,lockfile][without,with MDB_NOSUBDIR] */ +static const mdb_nchar_t *const mdb_suffixes[2][2] = { + { MDB_NAME("/data.mdb"), MDB_NAME("") }, + { MDB_NAME("/lock.mdb"), MDB_NAME("-lock") } +}; + +#define MDB_SUFFLEN 9 /**< Max string length in #mdb_suffixes[] */ + +/** Set up filename + scratch area for filename suffix, for opening files. + * It should be freed with #mdb_fname_destroy(). + * On Windows, paths are converted from char *UTF-8 to wchar_t *UTF-16. + * + * @param[in] path Pathname for #mdb_env_open(). + * @param[in] envflags Whether a subdir and/or lockfile will be used. + * @param[out] fname Resulting filename, with room for a suffix if necessary. + */ +static int ESECT +mdb_fname_init(const char *path, unsigned envflags, MDB_name *fname) +{ + int no_suffix = F_ISSET(envflags, MDB_NOSUBDIR|MDB_NOLOCK); + fname->mn_alloced = 0; +#ifdef _WIN32 + return utf8_to_utf16(path, fname, no_suffix ? 0 : MDB_SUFFLEN); +#else + fname->mn_len = strlen(path); + if (no_suffix) + fname->mn_val = (char *) path; + else if ((fname->mn_val = malloc(fname->mn_len + MDB_SUFFLEN+1)) != NULL) { + fname->mn_alloced = 1; + strcpy(fname->mn_val, path); + } + else + return ENOMEM; + return MDB_SUCCESS; +#endif +} + +/** Destroy \b fname from #mdb_fname_init() */ +#define mdb_fname_destroy(fname) \ + do { if ((fname).mn_alloced) free((fname).mn_val); } while (0) + +#ifdef O_CLOEXEC /* POSIX.1-2008: Set FD_CLOEXEC atomically at open() */ +# define MDB_CLOEXEC O_CLOEXEC +#else +# define MDB_CLOEXEC 0 +#endif + +/** File type, access mode etc. for #mdb_fopen() */ +enum mdb_fopen_type { +#ifdef _WIN32 + MDB_O_RDONLY, MDB_O_RDWR, MDB_O_META, MDB_O_COPY, MDB_O_LOCKS +#else + /* A comment in mdb_fopen() explains some O_* flag choices. */ + MDB_O_RDONLY= O_RDONLY, /**< for RDONLY me_fd */ + MDB_O_RDWR = O_RDWR |O_CREAT, /**< for me_fd */ + MDB_O_META = O_WRONLY|MDB_DSYNC |MDB_CLOEXEC, /**< for me_mfd */ + MDB_O_COPY = O_WRONLY|O_CREAT|O_EXCL|MDB_CLOEXEC, /**< for #mdb_env_copy() */ + /** Bitmask for open() flags in enum #mdb_fopen_type. The other bits + * distinguish otherwise-equal MDB_O_* constants from each other. + */ + MDB_O_MASK = MDB_O_RDWR|MDB_CLOEXEC | MDB_O_RDONLY|MDB_O_META|MDB_O_COPY, + MDB_O_LOCKS = MDB_O_RDWR|MDB_CLOEXEC | ((MDB_O_MASK+1) & ~MDB_O_MASK) /**< for me_lfd */ +#endif +}; + +/** Open an LMDB file. + * @param[in] env The LMDB environment. + * @param[in,out] fname Path from from #mdb_fname_init(). A suffix is + * appended if necessary to create the filename, without changing mn_len. + * @param[in] which Determines file type, access mode, etc. + * @param[in] mode The Unix permissions for the file, if we create it. + * @param[out] res Resulting file handle. + * @return 0 on success, non-zero on failure. + */ +static int ESECT +mdb_fopen(const MDB_env *env, MDB_name *fname, + enum mdb_fopen_type which, mdb_mode_t mode, + HANDLE *res) +{ + int rc = MDB_SUCCESS; + HANDLE fd; +#ifdef _WIN32 + DWORD acc, share, disp, attrs; +#else + int flags; +#endif + + if (fname->mn_alloced) /* modifiable copy */ + mdb_name_cpy(fname->mn_val + fname->mn_len, + mdb_suffixes[which==MDB_O_LOCKS][F_ISSET(env->me_flags, MDB_NOSUBDIR)]); + + /* The directory must already exist. Usually the file need not. + * MDB_O_META requires the file because we already created it using + * MDB_O_RDWR. MDB_O_COPY must not overwrite an existing file. + * + * With MDB_O_COPY we do not want the OS to cache the writes, since + * the source data is already in the OS cache. + * + * The lockfile needs FD_CLOEXEC (close file descriptor on exec*()) + * to avoid the flock() issues noted under Caveats in lmdb.h. + * Also set it for other filehandles which the user cannot get at + * and close himself, which he may need after fork(). I.e. all but + * me_fd, which programs do use via mdb_env_get_fd(). + */ + +#ifdef _WIN32 + acc = GENERIC_READ|GENERIC_WRITE; + share = FILE_SHARE_READ|FILE_SHARE_WRITE; + disp = OPEN_ALWAYS; + attrs = FILE_ATTRIBUTE_NORMAL; + switch (which) { + case MDB_O_RDONLY: /* read-only datafile */ + acc = GENERIC_READ; + disp = OPEN_EXISTING; + break; + case MDB_O_META: /* for writing metapages */ + acc = GENERIC_WRITE; + disp = OPEN_EXISTING; + attrs = FILE_ATTRIBUTE_NORMAL|FILE_FLAG_WRITE_THROUGH; + break; + case MDB_O_COPY: /* mdb_env_copy() & co */ + acc = GENERIC_WRITE; + share = 0; + disp = CREATE_NEW; + attrs = FILE_FLAG_NO_BUFFERING|FILE_FLAG_WRITE_THROUGH; + break; + default: break; /* silence gcc -Wswitch (not all enum values handled) */ + } + fd = CreateFileW(fname->mn_val, acc, share, NULL, disp, attrs, NULL); +#else + fd = open(fname->mn_val, which & MDB_O_MASK, mode); +#endif + + if (fd == INVALID_HANDLE_VALUE) + rc = ErrCode(); +#ifndef _WIN32 + else { + if (which != MDB_O_RDONLY && which != MDB_O_RDWR) { + /* Set CLOEXEC if we could not pass it to open() */ + if (!MDB_CLOEXEC && (flags = fcntl(fd, F_GETFD)) != -1) + (void) fcntl(fd, F_SETFD, flags | FD_CLOEXEC); + } + if (which == MDB_O_COPY && env->me_psize >= env->me_os_psize) { + /* This may require buffer alignment. There is no portable + * way to ask how much, so we require OS pagesize alignment. + */ +# ifdef F_NOCACHE /* __APPLE__ */ + (void) fcntl(fd, F_NOCACHE, 1); +# elif defined O_DIRECT + /* open(...O_DIRECT...) would break on filesystems without + * O_DIRECT support (ITS#7682). Try to set it here instead. + */ + if ((flags = fcntl(fd, F_GETFL)) != -1) + (void) fcntl(fd, F_SETFL, flags | O_DIRECT); +# endif + } + } +#endif /* !_WIN32 */ + + *res = fd; + return rc; +} + + +#ifdef BROKEN_FDATASYNC +#include <sys/utsname.h> +#include <sys/vfs.h> +#endif + +/** Further setup required for opening an LMDB environment + */ +static int ESECT +mdb_env_open2(MDB_env *env) +{ + unsigned int flags = env->me_flags; + int i, newenv = 0, rc; + MDB_meta meta; + +#ifdef _WIN32 + /* See if we should use QueryLimited */ + rc = GetVersion(); + if ((rc & 0xff) > 5) + env->me_pidquery = MDB_PROCESS_QUERY_LIMITED_INFORMATION; + else + env->me_pidquery = PROCESS_QUERY_INFORMATION; +#endif /* _WIN32 */ + +#ifdef BROKEN_FDATASYNC + /* ext3/ext4 fdatasync is broken on some older Linux kernels. + * https://lkml.org/lkml/2012/9/3/83 + * Kernels after 3.6-rc6 are known good. + * https://lkml.org/lkml/2012/9/10/556 + * See if the DB is on ext3/ext4, then check for new enough kernel + * Kernels 2.6.32.60, 2.6.34.15, 3.2.30, and 3.5.4 are also known + * to be patched. + */ + { + struct statfs st; + fstatfs(env->me_fd, &st); + while (st.f_type == 0xEF53) { + struct utsname uts; + int i; + uname(&uts); + if (uts.release[0] < '3') { + if (!strncmp(uts.release, "2.6.32.", 7)) { + i = atoi(uts.release+7); + if (i >= 60) + break; /* 2.6.32.60 and newer is OK */ + } else if (!strncmp(uts.release, "2.6.34.", 7)) { + i = atoi(uts.release+7); + if (i >= 15) + break; /* 2.6.34.15 and newer is OK */ + } + } else if (uts.release[0] == '3') { + i = atoi(uts.release+2); + if (i > 5) + break; /* 3.6 and newer is OK */ + if (i == 5) { + i = atoi(uts.release+4); + if (i >= 4) + break; /* 3.5.4 and newer is OK */ + } else if (i == 2) { + i = atoi(uts.release+4); + if (i >= 30) + break; /* 3.2.30 and newer is OK */ + } + } else { /* 4.x and newer is OK */ + break; + } + env->me_flags |= MDB_FSYNCONLY; + break; + } + } +#endif + + if ((i = mdb_env_read_header(env, &meta)) != 0) { + if (i != ENOENT) + return i; + DPUTS("new mdbenv"); + newenv = 1; + env->me_psize = env->me_os_psize; + if (env->me_psize > MAX_PAGESIZE) + env->me_psize = MAX_PAGESIZE; + memset(&meta, 0, sizeof(meta)); + mdb_env_init_meta0(env, &meta); + meta.mm_mapsize = DEFAULT_MAPSIZE; + } else { + env->me_psize = meta.mm_psize; + } + + /* Was a mapsize configured? */ + if (!env->me_mapsize) { + env->me_mapsize = meta.mm_mapsize; + } + { + /* Make sure mapsize >= committed data size. Even when using + * mm_mapsize, which could be broken in old files (ITS#7789). + */ + size_t minsize = (meta.mm_last_pg + 1) * meta.mm_psize; + if (env->me_mapsize < minsize) + env->me_mapsize = minsize; + } + meta.mm_mapsize = env->me_mapsize; + + if (newenv && !(flags & MDB_FIXEDMAP)) { + /* mdb_env_map() may grow the datafile. Write the metapages + * first, so the file will be valid if initialization fails. + * Except with FIXEDMAP, since we do not yet know mm_address. + * We could fill in mm_address later, but then a different + * program might end up doing that - one with a memory layout + * and map address which does not suit the main program. + */ + rc = mdb_env_init_meta(env, &meta); + if (rc) + return rc; + newenv = 0; + } + + rc = mdb_env_map(env, (flags & MDB_FIXEDMAP) ? meta.mm_address : NULL); + if (rc) + return rc; + + if (newenv) { + if (flags & MDB_FIXEDMAP) + meta.mm_address = env->me_map; + i = mdb_env_init_meta(env, &meta); + if (i != MDB_SUCCESS) { + return i; + } + } + + env->me_maxfree_1pg = (env->me_psize - PAGEHDRSZ) / sizeof(pgno_t) - 1; + env->me_nodemax = (((env->me_psize - PAGEHDRSZ) / MDB_MINKEYS) & -2) + - sizeof(indx_t); +#if !(MDB_MAXKEYSIZE) + env->me_maxkey = env->me_nodemax - (NODESIZE + sizeof(MDB_db)); +#endif + env->me_maxpg = env->me_mapsize / env->me_psize; + +#if MDB_DEBUG + { + MDB_meta *meta = mdb_env_pick_meta(env); + MDB_db *db = &meta->mm_dbs[MAIN_DBI]; + + DPRINTF(("opened database version %u, pagesize %u", + meta->mm_version, env->me_psize)); + DPRINTF(("using meta page %d", (int) (meta->mm_txnid & 1))); + DPRINTF(("depth: %u", db->md_depth)); + DPRINTF(("entries: %"Z"u", db->md_entries)); + DPRINTF(("branch pages: %"Z"u", db->md_branch_pages)); + DPRINTF(("leaf pages: %"Z"u", db->md_leaf_pages)); + DPRINTF(("overflow pages: %"Z"u", db->md_overflow_pages)); + DPRINTF(("root: %"Z"u", db->md_root)); + } +#endif + + return MDB_SUCCESS; +} + + +/** Release a reader thread's slot in the reader lock table. + * This function is called automatically when a thread exits. + * @param[in] ptr This points to the slot in the reader lock table. + */ +static void +mdb_env_reader_dest(void *ptr) +{ + MDB_reader *reader = ptr; + +#ifndef _WIN32 + if (reader->mr_pid == getpid()) /* catch pthread_exit() in child process */ +#endif + /* We omit the mutex, so do this atomically (i.e. skip mr_txnid) */ + reader->mr_pid = 0; +} + +#ifdef _WIN32 +/** Junk for arranging thread-specific callbacks on Windows. This is + * necessarily platform and compiler-specific. Windows supports up + * to 1088 keys. Let's assume nobody opens more than 64 environments + * in a single process, for now. They can override this if needed. + */ +#ifndef MAX_TLS_KEYS +#define MAX_TLS_KEYS 64 +#endif +static pthread_key_t mdb_tls_keys[MAX_TLS_KEYS]; +static int mdb_tls_nkeys; + +static void NTAPI mdb_tls_callback(PVOID module, DWORD reason, PVOID ptr) +{ + int i; + switch(reason) { + case DLL_PROCESS_ATTACH: break; + case DLL_THREAD_ATTACH: break; + case DLL_THREAD_DETACH: + for (i=0; i<mdb_tls_nkeys; i++) { + MDB_reader *r = pthread_getspecific(mdb_tls_keys[i]); + if (r) { + mdb_env_reader_dest(r); + } + } + break; + case DLL_PROCESS_DETACH: break; + } +} +#ifdef __GNUC__ +#ifdef _WIN64 +const PIMAGE_TLS_CALLBACK mdb_tls_cbp __attribute__((section (".CRT$XLB"))) = mdb_tls_callback; +#else +PIMAGE_TLS_CALLBACK mdb_tls_cbp __attribute__((section (".CRT$XLB"))) = mdb_tls_callback; +#endif +#else +#ifdef _WIN64 +/* Force some symbol references. + * _tls_used forces the linker to create the TLS directory if not already done + * mdb_tls_cbp prevents whole-program-optimizer from dropping the symbol. + */ +#pragma comment(linker, "/INCLUDE:_tls_used") +#pragma comment(linker, "/INCLUDE:mdb_tls_cbp") +#pragma const_seg(".CRT$XLB") +extern const PIMAGE_TLS_CALLBACK mdb_tls_cbp; +const PIMAGE_TLS_CALLBACK mdb_tls_cbp = mdb_tls_callback; +#pragma const_seg() +#else /* _WIN32 */ +#pragma comment(linker, "/INCLUDE:__tls_used") +#pragma comment(linker, "/INCLUDE:_mdb_tls_cbp") +#pragma data_seg(".CRT$XLB") +PIMAGE_TLS_CALLBACK mdb_tls_cbp = mdb_tls_callback; +#pragma data_seg() +#endif /* WIN 32/64 */ +#endif /* !__GNUC__ */ +#endif + +/** Downgrade the exclusive lock on the region back to shared */ +static int ESECT +mdb_env_share_locks(MDB_env *env, int *excl) +{ + int rc = 0; + MDB_meta *meta = mdb_env_pick_meta(env); + + env->me_txns->mti_txnid = meta->mm_txnid; + +#ifdef _WIN32 + { + OVERLAPPED ov; + /* First acquire a shared lock. The Unlock will + * then release the existing exclusive lock. + */ + memset(&ov, 0, sizeof(ov)); + if (!LockFileEx(env->me_lfd, 0, 0, 1, 0, &ov)) { + rc = ErrCode(); + } else { + UnlockFile(env->me_lfd, 0, 0, 1, 0); + *excl = 0; + } + } +#else + { + struct flock lock_info; + /* The shared lock replaces the existing lock */ + memset((void *)&lock_info, 0, sizeof(lock_info)); + lock_info.l_type = F_RDLCK; + lock_info.l_whence = SEEK_SET; + lock_info.l_start = 0; + lock_info.l_len = 1; + while ((rc = fcntl(env->me_lfd, F_SETLK, &lock_info)) && + (rc = ErrCode()) == EINTR) ; + *excl = rc ? -1 : 0; /* error may mean we lost the lock */ + } +#endif + + return rc; +} + +/** Try to get exclusive lock, otherwise shared. + * Maintain *excl = -1: no/unknown lock, 0: shared, 1: exclusive. + */ +static int ESECT +mdb_env_excl_lock(MDB_env *env, int *excl) +{ + int rc = 0; +#ifdef _WIN32 + if (LockFile(env->me_lfd, 0, 0, 1, 0)) { + *excl = 1; + } else { + OVERLAPPED ov; + memset(&ov, 0, sizeof(ov)); + if (LockFileEx(env->me_lfd, 0, 0, 1, 0, &ov)) { + *excl = 0; + } else { + rc = ErrCode(); + } + } +#else + struct flock lock_info; + memset((void *)&lock_info, 0, sizeof(lock_info)); + lock_info.l_type = F_WRLCK; + lock_info.l_whence = SEEK_SET; + lock_info.l_start = 0; + lock_info.l_len = 1; + while ((rc = fcntl(env->me_lfd, F_SETLK, &lock_info)) && + (rc = ErrCode()) == EINTR) ; + if (!rc) { + *excl = 1; + } else +# ifndef MDB_USE_POSIX_MUTEX + if (*excl < 0) /* always true when MDB_USE_POSIX_MUTEX */ +# endif + { + lock_info.l_type = F_RDLCK; + while ((rc = fcntl(env->me_lfd, F_SETLKW, &lock_info)) && + (rc = ErrCode()) == EINTR) ; + if (rc == 0) + *excl = 0; + } +#endif + return rc; +} + +#ifdef MDB_USE_HASH +/* + * hash_64 - 64 bit Fowler/Noll/Vo-0 FNV-1a hash code + * + * @(#) $Revision: 5.1 $ + * @(#) $Id: hash_64a.c,v 5.1 2009/06/30 09:01:38 chongo Exp $ + * @(#) $Source: /usr/local/src/cmd/fnv/RCS/hash_64a.c,v $ + * + * http://www.isthe.com/chongo/tech/comp/fnv/index.html + * + *** + * + * Please do not copyright this code. This code is in the public domain. + * + * LANDON CURT NOLL DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, + * INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO + * EVENT SHALL LANDON CURT NOLL BE LIABLE FOR ANY SPECIAL, INDIRECT OR + * CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF + * USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR + * OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR + * PERFORMANCE OF THIS SOFTWARE. + * + * By: + * chongo <Landon Curt Noll> /\oo/\ + * http://www.isthe.com/chongo/ + * + * Share and Enjoy! :-) + */ + +typedef unsigned long long mdb_hash_t; +#define MDB_HASH_INIT ((mdb_hash_t)0xcbf29ce484222325ULL) + +/** perform a 64 bit Fowler/Noll/Vo FNV-1a hash on a buffer + * @param[in] val value to hash + * @param[in] hval initial value for hash + * @return 64 bit hash + * + * NOTE: To use the recommended 64 bit FNV-1a hash, use MDB_HASH_INIT as the + * hval arg on the first call. + */ +static mdb_hash_t +mdb_hash_val(MDB_val *val, mdb_hash_t hval) +{ + unsigned char *s = (unsigned char *)val->mv_data; /* unsigned string */ + unsigned char *end = s + val->mv_size; + /* + * FNV-1a hash each octet of the string + */ + while (s < end) { + /* xor the bottom with the current octet */ + hval ^= (mdb_hash_t)*s++; + + /* multiply by the 64 bit FNV magic prime mod 2^64 */ + hval += (hval << 1) + (hval << 4) + (hval << 5) + + (hval << 7) + (hval << 8) + (hval << 40); + } + /* return our new hash value */ + return hval; +} + +/** Hash the string and output the encoded hash. + * This uses modified RFC1924 Ascii85 encoding to accommodate systems with + * very short name limits. We don't care about the encoding being reversible, + * we just want to preserve as many bits of the input as possible in a + * small printable string. + * @param[in] str string to hash + * @param[out] encbuf an array of 11 chars to hold the hash + */ +static const char mdb_a85[]= "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz!#$%&()*+-;<=>?@^_`{|}~"; + +static void ESECT +mdb_pack85(unsigned long l, char *out) +{ + int i; + + for (i=0; i<5; i++) { + *out++ = mdb_a85[l % 85]; + l /= 85; + } +} + +static void ESECT +mdb_hash_enc(MDB_val *val, char *encbuf) +{ + mdb_hash_t h = mdb_hash_val(val, MDB_HASH_INIT); + + mdb_pack85(h, encbuf); + mdb_pack85(h>>32, encbuf+5); + encbuf[10] = '\0'; +} +#endif + +/** Open and/or initialize the lock region for the environment. + * @param[in] env The LMDB environment. + * @param[in] fname Filename + scratch area, from #mdb_fname_init(). + * @param[in] mode The Unix permissions for the file, if we create it. + * @param[in,out] excl In -1, out lock type: -1 none, 0 shared, 1 exclusive + * @return 0 on success, non-zero on failure. + */ +static int ESECT +mdb_env_setup_locks(MDB_env *env, MDB_name *fname, int mode, int *excl) +{ +#ifdef _WIN32 +# define MDB_ERRCODE_ROFS ERROR_WRITE_PROTECT +#else +# define MDB_ERRCODE_ROFS EROFS +#endif + int rc; + off_t size, rsize; + + rc = mdb_fopen(env, fname, MDB_O_LOCKS, mode, &env->me_lfd); + if (rc) { + /* Omit lockfile if read-only env on read-only filesystem */ + if (rc == MDB_ERRCODE_ROFS && (env->me_flags & MDB_RDONLY)) { + return MDB_SUCCESS; + } + goto fail; + } + + if (!(env->me_flags & MDB_NOTLS)) { + rc = pthread_key_create(&env->me_txkey, mdb_env_reader_dest); + if (rc) + goto fail; + env->me_flags |= MDB_ENV_TXKEY; +#ifdef _WIN32 + /* Windows TLS callbacks need help finding their TLS info. */ + if (mdb_tls_nkeys >= MAX_TLS_KEYS) { + rc = MDB_TLS_FULL; + goto fail; + } + mdb_tls_keys[mdb_tls_nkeys++] = env->me_txkey; +#endif + } + + /* Try to get exclusive lock. If we succeed, then + * nobody is using the lock region and we should initialize it. + */ + if ((rc = mdb_env_excl_lock(env, excl))) goto fail; + +#ifdef _WIN32 + size = GetFileSize(env->me_lfd, NULL); +#else + size = lseek(env->me_lfd, 0, SEEK_END); + if (size == -1) goto fail_errno; +#endif + rsize = (env->me_maxreaders-1) * sizeof(MDB_reader) + sizeof(MDB_txninfo); + if (size < rsize && *excl > 0) { +#ifdef _WIN32 + if (SetFilePointer(env->me_lfd, rsize, NULL, FILE_BEGIN) != (DWORD)rsize + || !SetEndOfFile(env->me_lfd)) + goto fail_errno; +#else + if (ftruncate(env->me_lfd, rsize) != 0) goto fail_errno; +#endif + } else { + rsize = size; + size = rsize - sizeof(MDB_txninfo); + env->me_maxreaders = size/sizeof(MDB_reader) + 1; + } + { +#ifdef _WIN32 + HANDLE mh; + mh = CreateFileMapping(env->me_lfd, NULL, PAGE_READWRITE, + 0, 0, NULL); + if (!mh) goto fail_errno; + env->me_txns = MapViewOfFileEx(mh, FILE_MAP_WRITE, 0, 0, rsize, NULL); + CloseHandle(mh); + if (!env->me_txns) goto fail_errno; +#else + void *m = mmap(NULL, rsize, PROT_READ|PROT_WRITE, MAP_SHARED, + env->me_lfd, 0); + if (m == MAP_FAILED) goto fail_errno; + env->me_txns = m; +#endif + } + if (*excl > 0) { +#ifdef _WIN32 + BY_HANDLE_FILE_INFORMATION stbuf; + struct { + DWORD volume; + DWORD nhigh; + DWORD nlow; + } idbuf; + MDB_val val; + char encbuf[11]; + + if (!mdb_sec_inited) { + InitializeSecurityDescriptor(&mdb_null_sd, + SECURITY_DESCRIPTOR_REVISION); + SetSecurityDescriptorDacl(&mdb_null_sd, TRUE, 0, FALSE); + mdb_all_sa.nLength = sizeof(SECURITY_ATTRIBUTES); + mdb_all_sa.bInheritHandle = FALSE; + mdb_all_sa.lpSecurityDescriptor = &mdb_null_sd; + mdb_sec_inited = 1; + } + if (!GetFileInformationByHandle(env->me_lfd, &stbuf)) goto fail_errno; + idbuf.volume = stbuf.dwVolumeSerialNumber; + idbuf.nhigh = stbuf.nFileIndexHigh; + idbuf.nlow = stbuf.nFileIndexLow; + val.mv_data = &idbuf; + val.mv_size = sizeof(idbuf); + mdb_hash_enc(&val, encbuf); + sprintf(env->me_txns->mti_rmname, "Global\\MDBr%s", encbuf); + sprintf(env->me_txns->mti_wmname, "Global\\MDBw%s", encbuf); + env->me_rmutex = CreateMutexA(&mdb_all_sa, FALSE, env->me_txns->mti_rmname); + if (!env->me_rmutex) goto fail_errno; + env->me_wmutex = CreateMutexA(&mdb_all_sa, FALSE, env->me_txns->mti_wmname); + if (!env->me_wmutex) goto fail_errno; +#elif defined(MDB_USE_POSIX_SEM) + struct stat stbuf; + struct { + dev_t dev; + ino_t ino; + } idbuf; + MDB_val val; + char encbuf[11]; + +#if defined(__NetBSD__) +#define MDB_SHORT_SEMNAMES 1 /* limited to 14 chars */ +#endif + if (fstat(env->me_lfd, &stbuf)) goto fail_errno; + idbuf.dev = stbuf.st_dev; + idbuf.ino = stbuf.st_ino; + val.mv_data = &idbuf; + val.mv_size = sizeof(idbuf); + mdb_hash_enc(&val, encbuf); +#ifdef MDB_SHORT_SEMNAMES + encbuf[9] = '\0'; /* drop name from 15 chars to 14 chars */ +#endif + sprintf(env->me_txns->mti_rmname, "/MDBr%s", encbuf); + sprintf(env->me_txns->mti_wmname, "/MDBw%s", encbuf); + /* Clean up after a previous run, if needed: Try to + * remove both semaphores before doing anything else. + */ + sem_unlink(env->me_txns->mti_rmname); + sem_unlink(env->me_txns->mti_wmname); + env->me_rmutex = sem_open(env->me_txns->mti_rmname, + O_CREAT|O_EXCL, mode, 1); + if (env->me_rmutex == SEM_FAILED) goto fail_errno; + env->me_wmutex = sem_open(env->me_txns->mti_wmname, + O_CREAT|O_EXCL, mode, 1); + if (env->me_wmutex == SEM_FAILED) goto fail_errno; +#else /* MDB_USE_POSIX_MUTEX: */ + pthread_mutexattr_t mattr; + + /* Solaris needs this before initing a robust mutex. Otherwise + * it may skip the init and return EBUSY "seems someone already + * inited" or EINVAL "it was inited differently". + */ + memset(env->me_txns->mti_rmutex, 0, sizeof(*env->me_txns->mti_rmutex)); + memset(env->me_txns->mti_wmutex, 0, sizeof(*env->me_txns->mti_wmutex)); + + if ((rc = pthread_mutexattr_init(&mattr))) + goto fail; + + rc = pthread_mutexattr_setpshared(&mattr, PTHREAD_PROCESS_SHARED); +#ifdef MDB_ROBUST_SUPPORTED + if (!rc) rc = pthread_mutexattr_setrobust(&mattr, PTHREAD_MUTEX_ROBUST); +#endif + if (!rc) rc = pthread_mutex_init(env->me_txns->mti_rmutex, &mattr); + if (!rc) rc = pthread_mutex_init(env->me_txns->mti_wmutex, &mattr); + pthread_mutexattr_destroy(&mattr); + if (rc) + goto fail; +#endif /* _WIN32 || MDB_USE_POSIX_SEM */ + + env->me_txns->mti_magic = MDB_MAGIC; + env->me_txns->mti_format = MDB_LOCK_FORMAT; + env->me_txns->mti_txnid = 0; + env->me_txns->mti_numreaders = 0; + + } else { + if (env->me_txns->mti_magic != MDB_MAGIC) { + DPUTS("lock region has invalid magic"); + rc = MDB_INVALID; + goto fail; + } + if (env->me_txns->mti_format != MDB_LOCK_FORMAT) { + DPRINTF(("lock region has format+version 0x%x, expected 0x%x", + env->me_txns->mti_format, MDB_LOCK_FORMAT)); + rc = MDB_VERSION_MISMATCH; + goto fail; + } + rc = ErrCode(); + if (rc && rc != EACCES && rc != EAGAIN) { + goto fail; + } +#ifdef _WIN32 + env->me_rmutex = OpenMutexA(SYNCHRONIZE, FALSE, env->me_txns->mti_rmname); + if (!env->me_rmutex) goto fail_errno; + env->me_wmutex = OpenMutexA(SYNCHRONIZE, FALSE, env->me_txns->mti_wmname); + if (!env->me_wmutex) goto fail_errno; +#elif defined(MDB_USE_POSIX_SEM) + env->me_rmutex = sem_open(env->me_txns->mti_rmname, 0); + if (env->me_rmutex == SEM_FAILED) goto fail_errno; + env->me_wmutex = sem_open(env->me_txns->mti_wmname, 0); + if (env->me_wmutex == SEM_FAILED) goto fail_errno; +#endif + } + return MDB_SUCCESS; + +fail_errno: + rc = ErrCode(); +fail: + return rc; +} + + /** Only a subset of the @ref mdb_env flags can be changed + * at runtime. Changing other flags requires closing the + * environment and re-opening it with the new flags. + */ +#define CHANGEABLE (MDB_NOSYNC|MDB_NOMETASYNC|MDB_MAPASYNC|MDB_NOMEMINIT) +#define CHANGELESS (MDB_FIXEDMAP|MDB_NOSUBDIR|MDB_RDONLY| \ + MDB_WRITEMAP|MDB_NOTLS|MDB_NOLOCK|MDB_NORDAHEAD) + +#if VALID_FLAGS & PERSISTENT_FLAGS & (CHANGEABLE|CHANGELESS) +# error "Persistent DB flags & env flags overlap, but both go in mm_flags" +#endif + +int ESECT +mdb_env_open(MDB_env *env, const char *path, unsigned int flags, mdb_mode_t mode) +{ + int rc, excl = -1; + MDB_name fname; + + if (env->me_fd!=INVALID_HANDLE_VALUE || (flags & ~(CHANGEABLE|CHANGELESS))) + return EINVAL; + + flags |= env->me_flags; + + rc = mdb_fname_init(path, flags, &fname); + if (rc) + return rc; + + if (flags & MDB_RDONLY) { + /* silently ignore WRITEMAP when we're only getting read access */ + flags &= ~MDB_WRITEMAP; + } else { + if (!((env->me_free_pgs = mdb_midl_alloc(MDB_IDL_UM_MAX)) && + (env->me_dirty_list = calloc(MDB_IDL_UM_SIZE, sizeof(MDB_ID2))))) + rc = ENOMEM; + } + env->me_flags = flags |= MDB_ENV_ACTIVE; + if (rc) + goto leave; + + env->me_path = strdup(path); + env->me_dbxs = calloc(env->me_maxdbs, sizeof(MDB_dbx)); + env->me_dbflags = calloc(env->me_maxdbs, sizeof(uint16_t)); + env->me_dbiseqs = calloc(env->me_maxdbs, sizeof(unsigned int)); + if (!(env->me_dbxs && env->me_path && env->me_dbflags && env->me_dbiseqs)) { + rc = ENOMEM; + goto leave; + } + env->me_dbxs[FREE_DBI].md_cmp = mdb_cmp_long; /* aligned MDB_INTEGERKEY */ + + /* For RDONLY, get lockfile after we know datafile exists */ + if (!(flags & (MDB_RDONLY|MDB_NOLOCK))) { + rc = mdb_env_setup_locks(env, &fname, mode, &excl); + if (rc) + goto leave; + } + + rc = mdb_fopen(env, &fname, + (flags & MDB_RDONLY) ? MDB_O_RDONLY : MDB_O_RDWR, + mode, &env->me_fd); + if (rc) + goto leave; + + if ((flags & (MDB_RDONLY|MDB_NOLOCK)) == MDB_RDONLY) { + rc = mdb_env_setup_locks(env, &fname, mode, &excl); + if (rc) + goto leave; + } + + if ((rc = mdb_env_open2(env)) == MDB_SUCCESS) { + if (!(flags & (MDB_RDONLY|MDB_WRITEMAP))) { + /* Synchronous fd for meta writes. Needed even with + * MDB_NOSYNC/MDB_NOMETASYNC, in case these get reset. + */ + rc = mdb_fopen(env, &fname, MDB_O_META, mode, &env->me_mfd); + if (rc) + goto leave; + } + DPRINTF(("opened dbenv %p", (void *) env)); + if (excl > 0) { + rc = mdb_env_share_locks(env, &excl); + if (rc) + goto leave; + } + if (!(flags & MDB_RDONLY)) { + MDB_txn *txn; + int tsize = sizeof(MDB_txn), size = tsize + env->me_maxdbs * + (sizeof(MDB_db)+sizeof(MDB_cursor *)+sizeof(unsigned int)+1); + if ((env->me_pbuf = calloc(1, env->me_psize)) && + (txn = calloc(1, size))) + { + txn->mt_dbs = (MDB_db *)((char *)txn + tsize); + txn->mt_cursors = (MDB_cursor **)(txn->mt_dbs + env->me_maxdbs); + txn->mt_dbiseqs = (unsigned int *)(txn->mt_cursors + env->me_maxdbs); + txn->mt_dbflags = (unsigned char *)(txn->mt_dbiseqs + env->me_maxdbs); + txn->mt_env = env; + txn->mt_dbxs = env->me_dbxs; + txn->mt_flags = MDB_TXN_FINISHED; + env->me_txn0 = txn; + } else { + rc = ENOMEM; + } + } + } + +leave: + if (rc) { + mdb_env_close0(env, excl); + } + mdb_fname_destroy(fname); + return rc; +} + +/** Destroy resources from mdb_env_open(), clear our readers & DBIs */ +static void ESECT +mdb_env_close0(MDB_env *env, int excl) +{ + int i; + + if (!(env->me_flags & MDB_ENV_ACTIVE)) + return; + + /* Doing this here since me_dbxs may not exist during mdb_env_close */ + if (env->me_dbxs) { + for (i = env->me_maxdbs; --i >= CORE_DBS; ) + free(env->me_dbxs[i].md_name.mv_data); + free(env->me_dbxs); + } + + free(env->me_pbuf); + free(env->me_dbiseqs); + free(env->me_dbflags); + free(env->me_path); + free(env->me_dirty_list); + free(env->me_txn0); + mdb_midl_free(env->me_free_pgs); + + if (env->me_flags & MDB_ENV_TXKEY) { + pthread_key_delete(env->me_txkey); +#ifdef _WIN32 + /* Delete our key from the global list */ + for (i=0; i<mdb_tls_nkeys; i++) + if (mdb_tls_keys[i] == env->me_txkey) { + mdb_tls_keys[i] = mdb_tls_keys[mdb_tls_nkeys-1]; + mdb_tls_nkeys--; + break; + } +#endif + } + + if (env->me_map) { + munmap(env->me_map, env->me_mapsize); + } + if (env->me_mfd != INVALID_HANDLE_VALUE) + (void) close(env->me_mfd); + if (env->me_fd != INVALID_HANDLE_VALUE) + (void) close(env->me_fd); + if (env->me_txns) { + MDB_PID_T pid = env->me_pid; + /* Clearing readers is done in this function because + * me_txkey with its destructor must be disabled first. + * + * We skip the the reader mutex, so we touch only + * data owned by this process (me_close_readers and + * our readers), and clear each reader atomically. + */ + for (i = env->me_close_readers; --i >= 0; ) + if (env->me_txns->mti_readers[i].mr_pid == pid) + env->me_txns->mti_readers[i].mr_pid = 0; +#ifdef _WIN32 + if (env->me_rmutex) { + CloseHandle(env->me_rmutex); + if (env->me_wmutex) CloseHandle(env->me_wmutex); + } + /* Windows automatically destroys the mutexes when + * the last handle closes. + */ +#elif defined(MDB_USE_POSIX_SEM) + if (env->me_rmutex != SEM_FAILED) { + sem_close(env->me_rmutex); + if (env->me_wmutex != SEM_FAILED) + sem_close(env->me_wmutex); + /* If we have the filelock: If we are the + * only remaining user, clean up semaphores. + */ + if (excl == 0) + mdb_env_excl_lock(env, &excl); + if (excl > 0) { + sem_unlink(env->me_txns->mti_rmname); + sem_unlink(env->me_txns->mti_wmname); + } + } +#endif + munmap((void *)env->me_txns, (env->me_maxreaders-1)*sizeof(MDB_reader)+sizeof(MDB_txninfo)); + } + if (env->me_lfd != INVALID_HANDLE_VALUE) { +#ifdef _WIN32 + if (excl >= 0) { + /* Unlock the lockfile. Windows would have unlocked it + * after closing anyway, but not necessarily at once. + */ + UnlockFile(env->me_lfd, 0, 0, 1, 0); + } +#endif + (void) close(env->me_lfd); + } + + env->me_flags &= ~(MDB_ENV_ACTIVE|MDB_ENV_TXKEY); +} + +void ESECT +mdb_env_close(MDB_env *env) +{ + MDB_page *dp; + + if (env == NULL) + return; + + VGMEMP_DESTROY(env); + while ((dp = env->me_dpages) != NULL) { + VGMEMP_DEFINED(&dp->mp_next, sizeof(dp->mp_next)); + env->me_dpages = dp->mp_next; + free(dp); + } + + mdb_env_close0(env, 0); + free(env); +} + +/** Compare two items pointing at aligned size_t's */ +static int +mdb_cmp_long(const MDB_val *a, const MDB_val *b) +{ + return (*(size_t *)a->mv_data < *(size_t *)b->mv_data) ? -1 : + *(size_t *)a->mv_data > *(size_t *)b->mv_data; +} + +/** Compare two items pointing at aligned unsigned int's. + * + * This is also set as #MDB_INTEGERDUP|#MDB_DUPFIXED's #MDB_dbx.%md_dcmp, + * but #mdb_cmp_clong() is called instead if the data type is size_t. + */ +static int +mdb_cmp_int(const MDB_val *a, const MDB_val *b) +{ + return (*(unsigned int *)a->mv_data < *(unsigned int *)b->mv_data) ? -1 : + *(unsigned int *)a->mv_data > *(unsigned int *)b->mv_data; +} + +/** Compare two items pointing at unsigned ints of unknown alignment. + * Nodes and keys are guaranteed to be 2-byte aligned. + */ +static int +mdb_cmp_cint(const MDB_val *a, const MDB_val *b) +{ +#if BYTE_ORDER == LITTLE_ENDIAN + unsigned short *u, *c; + int x; + + u = (unsigned short *) ((char *) a->mv_data + a->mv_size); + c = (unsigned short *) ((char *) b->mv_data + a->mv_size); + do { + x = *--u - *--c; + } while(!x && u > (unsigned short *)a->mv_data); + return x; +#else + unsigned short *u, *c, *end; + int x; + + end = (unsigned short *) ((char *) a->mv_data + a->mv_size); + u = (unsigned short *)a->mv_data; + c = (unsigned short *)b->mv_data; + do { + x = *u++ - *c++; + } while(!x && u < end); + return x; +#endif +} + +/** Compare two items lexically */ +static int +mdb_cmp_memn(const MDB_val *a, const MDB_val *b) +{ + int diff; + ssize_t len_diff; + unsigned int len; + + len = a->mv_size; + len_diff = (ssize_t) a->mv_size - (ssize_t) b->mv_size; + if (len_diff > 0) { + len = b->mv_size; + len_diff = 1; + } + + diff = memcmp(a->mv_data, b->mv_data, len); + return diff ? diff : len_diff<0 ? -1 : len_diff; +} + +/** Compare two items in reverse byte order */ +static int +mdb_cmp_memnr(const MDB_val *a, const MDB_val *b) +{ + const unsigned char *p1, *p2, *p1_lim; + ssize_t len_diff; + int diff; + + p1_lim = (const unsigned char *)a->mv_data; + p1 = (const unsigned char *)a->mv_data + a->mv_size; + p2 = (const unsigned char *)b->mv_data + b->mv_size; + + len_diff = (ssize_t) a->mv_size - (ssize_t) b->mv_size; + if (len_diff > 0) { + p1_lim += len_diff; + len_diff = 1; + } + + while (p1 > p1_lim) { + diff = *--p1 - *--p2; + if (diff) + return diff; + } + return len_diff<0 ? -1 : len_diff; +} + +/** Search for key within a page, using binary search. + * Returns the smallest entry larger or equal to the key. + * If exactp is non-null, stores whether the found entry was an exact match + * in *exactp (1 or 0). + * Updates the cursor index with the index of the found entry. + * If no entry larger or equal to the key is found, returns NULL. + */ +static MDB_node * +mdb_node_search(MDB_cursor *mc, MDB_val *key, int *exactp) +{ + unsigned int i = 0, nkeys; + int low, high; + int rc = 0; + MDB_page *mp = mc->mc_pg[mc->mc_top]; + MDB_node *node = NULL; + MDB_val nodekey; + MDB_cmp_func *cmp; + DKBUF; + + nkeys = NUMKEYS(mp); + + DPRINTF(("searching %u keys in %s %spage %"Z"u", + nkeys, IS_LEAF(mp) ? "leaf" : "branch", IS_SUBP(mp) ? "sub-" : "", + mdb_dbg_pgno(mp))); + + low = IS_LEAF(mp) ? 0 : 1; + high = nkeys - 1; + cmp = mc->mc_dbx->md_cmp; + + /* Branch pages have no data, so if using integer keys, + * alignment is guaranteed. Use faster mdb_cmp_int. + */ + if (cmp == mdb_cmp_cint && IS_BRANCH(mp)) { + if (NODEPTR(mp, 1)->mn_ksize == sizeof(size_t)) + cmp = mdb_cmp_long; + else + cmp = mdb_cmp_int; + } + + if (IS_LEAF2(mp)) { + nodekey.mv_size = mc->mc_db->md_pad; + node = NODEPTR(mp, 0); /* fake */ + while (low <= high) { + i = (low + high) >> 1; + nodekey.mv_data = LEAF2KEY(mp, i, nodekey.mv_size); + rc = cmp(key, &nodekey); + DPRINTF(("found leaf index %u [%s], rc = %i", + i, DKEY(&nodekey), rc)); + if (rc == 0) + break; + if (rc > 0) + low = i + 1; + else + high = i - 1; + } + } else { + while (low <= high) { + i = (low + high) >> 1; + + node = NODEPTR(mp, i); + nodekey.mv_size = NODEKSZ(node); + nodekey.mv_data = NODEKEY(node); + + rc = cmp(key, &nodekey); +#if MDB_DEBUG + if (IS_LEAF(mp)) + DPRINTF(("found leaf index %u [%s], rc = %i", + i, DKEY(&nodekey), rc)); + else + DPRINTF(("found branch index %u [%s -> %"Z"u], rc = %i", + i, DKEY(&nodekey), NODEPGNO(node), rc)); +#endif + if (rc == 0) + break; + if (rc > 0) + low = i + 1; + else + high = i - 1; + } + } + + if (rc > 0) { /* Found entry is less than the key. */ + i++; /* Skip to get the smallest entry larger than key. */ + if (!IS_LEAF2(mp)) + node = NODEPTR(mp, i); + } + if (exactp) + *exactp = (rc == 0 && nkeys > 0); + /* store the key index */ + mc->mc_ki[mc->mc_top] = i; + if (i >= nkeys) + /* There is no entry larger or equal to the key. */ + return NULL; + + /* nodeptr is fake for LEAF2 */ + return node; +} + +#if 0 +static void +mdb_cursor_adjust(MDB_cursor *mc, func) +{ + MDB_cursor *m2; + + for (m2 = mc->mc_txn->mt_cursors[mc->mc_dbi]; m2; m2=m2->mc_next) { + if (m2->mc_pg[m2->mc_top] == mc->mc_pg[mc->mc_top]) { + func(mc, m2); + } + } +} +#endif + +/** Pop a page off the top of the cursor's stack. */ +static void +mdb_cursor_pop(MDB_cursor *mc) +{ + if (mc->mc_snum) { + DPRINTF(("popping page %"Z"u off db %d cursor %p", + mc->mc_pg[mc->mc_top]->mp_pgno, DDBI(mc), (void *) mc)); + + mc->mc_snum--; + if (mc->mc_snum) { + mc->mc_top--; + } else { + mc->mc_flags &= ~C_INITIALIZED; + } + } +} + +/** Push a page onto the top of the cursor's stack. + * Set #MDB_TXN_ERROR on failure. + */ +static int +mdb_cursor_push(MDB_cursor *mc, MDB_page *mp) +{ + DPRINTF(("pushing page %"Z"u on db %d cursor %p", mp->mp_pgno, + DDBI(mc), (void *) mc)); + + if (mc->mc_snum >= CURSOR_STACK) { + mc->mc_txn->mt_flags |= MDB_TXN_ERROR; + return MDB_CURSOR_FULL; + } + + mc->mc_top = mc->mc_snum++; + mc->mc_pg[mc->mc_top] = mp; + mc->mc_ki[mc->mc_top] = 0; + + return MDB_SUCCESS; +} + +/** Find the address of the page corresponding to a given page number. + * Set #MDB_TXN_ERROR on failure. + * @param[in] mc the cursor accessing the page. + * @param[in] pgno the page number for the page to retrieve. + * @param[out] ret address of a pointer where the page's address will be stored. + * @param[out] lvl dirty_list inheritance level of found page. 1=current txn, 0=mapped page. + * @return 0 on success, non-zero on failure. + */ +static int +mdb_page_get(MDB_cursor *mc, pgno_t pgno, MDB_page **ret, int *lvl) +{ + MDB_txn *txn = mc->mc_txn; + MDB_env *env = txn->mt_env; + MDB_page *p = NULL; + int level; + + if (! (txn->mt_flags & (MDB_TXN_RDONLY|MDB_TXN_WRITEMAP))) { + MDB_txn *tx2 = txn; + level = 1; + do { + MDB_ID2L dl = tx2->mt_u.dirty_list; + unsigned x; + /* Spilled pages were dirtied in this txn and flushed + * because the dirty list got full. Bring this page + * back in from the map (but don't unspill it here, + * leave that unless page_touch happens again). + */ + if (tx2->mt_spill_pgs) { + MDB_ID pn = pgno << 1; + x = mdb_midl_search(tx2->mt_spill_pgs, pn); + if (x <= tx2->mt_spill_pgs[0] && tx2->mt_spill_pgs[x] == pn) { + p = (MDB_page *)(env->me_map + env->me_psize * pgno); + goto done; + } + } + if (dl[0].mid) { + unsigned x = mdb_mid2l_search(dl, pgno); + if (x <= dl[0].mid && dl[x].mid == pgno) { + p = dl[x].mptr; + goto done; + } + } + level++; + } while ((tx2 = tx2->mt_parent) != NULL); + } + + if (pgno < txn->mt_next_pgno) { + level = 0; + p = (MDB_page *)(env->me_map + env->me_psize * pgno); + } else { + DPRINTF(("page %"Z"u not found", pgno)); + txn->mt_flags |= MDB_TXN_ERROR; + return MDB_PAGE_NOTFOUND; + } + +done: + *ret = p; + if (lvl) + *lvl = level; + return MDB_SUCCESS; +} + +/** Finish #mdb_page_search() / #mdb_page_search_lowest(). + * The cursor is at the root page, set up the rest of it. + */ +static int +mdb_page_search_root(MDB_cursor *mc, MDB_val *key, int flags) +{ + MDB_page *mp = mc->mc_pg[mc->mc_top]; + int rc; + DKBUF; + + while (IS_BRANCH(mp)) { + MDB_node *node; + indx_t i; + + DPRINTF(("branch page %"Z"u has %u keys", mp->mp_pgno, NUMKEYS(mp))); + /* Don't assert on branch pages in the FreeDB. We can get here + * while in the process of rebalancing a FreeDB branch page; we must + * let that proceed. ITS#8336 + */ + mdb_cassert(mc, !mc->mc_dbi || NUMKEYS(mp) > 1); + DPRINTF(("found index 0 to page %"Z"u", NODEPGNO(NODEPTR(mp, 0)))); + + if (flags & (MDB_PS_FIRST|MDB_PS_LAST)) { + i = 0; + if (flags & MDB_PS_LAST) { + i = NUMKEYS(mp) - 1; + /* if already init'd, see if we're already in right place */ + if (mc->mc_flags & C_INITIALIZED) { + if (mc->mc_ki[mc->mc_top] == i) { + mc->mc_top = mc->mc_snum++; + mp = mc->mc_pg[mc->mc_top]; + goto ready; + } + } + } + } else { + int exact; + node = mdb_node_search(mc, key, &exact); + if (node == NULL) + i = NUMKEYS(mp) - 1; + else { + i = mc->mc_ki[mc->mc_top]; + if (!exact) { + mdb_cassert(mc, i > 0); + i--; + } + } + DPRINTF(("following index %u for key [%s]", i, DKEY(key))); + } + + mdb_cassert(mc, i < NUMKEYS(mp)); + node = NODEPTR(mp, i); + + if ((rc = mdb_page_get(mc, NODEPGNO(node), &mp, NULL)) != 0) + return rc; + + mc->mc_ki[mc->mc_top] = i; + if ((rc = mdb_cursor_push(mc, mp))) + return rc; + +ready: + if (flags & MDB_PS_MODIFY) { + if ((rc = mdb_page_touch(mc)) != 0) + return rc; + mp = mc->mc_pg[mc->mc_top]; + } + } + + if (!IS_LEAF(mp)) { + DPRINTF(("internal error, index points to a %02X page!?", + mp->mp_flags)); + mc->mc_txn->mt_flags |= MDB_TXN_ERROR; + return MDB_CORRUPTED; + } + + DPRINTF(("found leaf page %"Z"u for key [%s]", mp->mp_pgno, + key ? DKEY(key) : "null")); + mc->mc_flags |= C_INITIALIZED; + mc->mc_flags &= ~C_EOF; + + return MDB_SUCCESS; +} + +/** Search for the lowest key under the current branch page. + * This just bypasses a NUMKEYS check in the current page + * before calling mdb_page_search_root(), because the callers + * are all in situations where the current page is known to + * be underfilled. + */ +static int +mdb_page_search_lowest(MDB_cursor *mc) +{ + MDB_page *mp = mc->mc_pg[mc->mc_top]; + MDB_node *node = NODEPTR(mp, 0); + int rc; + + if ((rc = mdb_page_get(mc, NODEPGNO(node), &mp, NULL)) != 0) + return rc; + + mc->mc_ki[mc->mc_top] = 0; + if ((rc = mdb_cursor_push(mc, mp))) + return rc; + return mdb_page_search_root(mc, NULL, MDB_PS_FIRST); +} + +/** Search for the page a given key should be in. + * Push it and its parent pages on the cursor stack. + * @param[in,out] mc the cursor for this operation. + * @param[in] key the key to search for, or NULL for first/last page. + * @param[in] flags If MDB_PS_MODIFY is set, visited pages in the DB + * are touched (updated with new page numbers). + * If MDB_PS_FIRST or MDB_PS_LAST is set, find first or last leaf. + * This is used by #mdb_cursor_first() and #mdb_cursor_last(). + * If MDB_PS_ROOTONLY set, just fetch root node, no further lookups. + * @return 0 on success, non-zero on failure. + */ +static int +mdb_page_search(MDB_cursor *mc, MDB_val *key, int flags) +{ + int rc; + pgno_t root; + + /* Make sure the txn is still viable, then find the root from + * the txn's db table and set it as the root of the cursor's stack. + */ + if (mc->mc_txn->mt_flags & MDB_TXN_BLOCKED) { + DPUTS("transaction may not be used now"); + return MDB_BAD_TXN; + } else { + /* Make sure we're using an up-to-date root */ + if (*mc->mc_dbflag & DB_STALE) { + MDB_cursor mc2; + if (TXN_DBI_CHANGED(mc->mc_txn, mc->mc_dbi)) + return MDB_BAD_DBI; + mdb_cursor_init(&mc2, mc->mc_txn, MAIN_DBI, NULL); + rc = mdb_page_search(&mc2, &mc->mc_dbx->md_name, 0); + if (rc) + return rc; + { + MDB_val data; + int exact = 0; + uint16_t flags; + MDB_node *leaf = mdb_node_search(&mc2, + &mc->mc_dbx->md_name, &exact); + if (!exact) + return MDB_NOTFOUND; + if ((leaf->mn_flags & (F_DUPDATA|F_SUBDATA)) != F_SUBDATA) + return MDB_INCOMPATIBLE; /* not a named DB */ + rc = mdb_node_read(&mc2, leaf, &data); + if (rc) + return rc; + memcpy(&flags, ((char *) data.mv_data + offsetof(MDB_db, md_flags)), + sizeof(uint16_t)); + /* The txn may not know this DBI, or another process may + * have dropped and recreated the DB with other flags. + */ + if ((mc->mc_db->md_flags & PERSISTENT_FLAGS) != flags) + return MDB_INCOMPATIBLE; + memcpy(mc->mc_db, data.mv_data, sizeof(MDB_db)); + } + *mc->mc_dbflag &= ~DB_STALE; + } + root = mc->mc_db->md_root; + + if (root == P_INVALID) { /* Tree is empty. */ + DPUTS("tree is empty"); + return MDB_NOTFOUND; + } + } + + mdb_cassert(mc, root > 1); + if (!mc->mc_pg[0] || mc->mc_pg[0]->mp_pgno != root) + if ((rc = mdb_page_get(mc, root, &mc->mc_pg[0], NULL)) != 0) + return rc; + + mc->mc_snum = 1; + mc->mc_top = 0; + + DPRINTF(("db %d root page %"Z"u has flags 0x%X", + DDBI(mc), root, mc->mc_pg[0]->mp_flags)); + + if (flags & MDB_PS_MODIFY) { + if ((rc = mdb_page_touch(mc))) + return rc; + } + + if (flags & MDB_PS_ROOTONLY) + return MDB_SUCCESS; + + return mdb_page_search_root(mc, key, flags); +} + +static int +mdb_ovpage_free(MDB_cursor *mc, MDB_page *mp) +{ + MDB_txn *txn = mc->mc_txn; + pgno_t pg = mp->mp_pgno; + unsigned x = 0, ovpages = mp->mp_pages; + MDB_env *env = txn->mt_env; + MDB_IDL sl = txn->mt_spill_pgs; + MDB_ID pn = pg << 1; + int rc; + + DPRINTF(("free ov page %"Z"u (%d)", pg, ovpages)); + /* If the page is dirty or on the spill list we just acquired it, + * so we should give it back to our current free list, if any. + * Otherwise put it onto the list of pages we freed in this txn. + * + * Won't create me_pghead: me_pglast must be inited along with it. + * Unsupported in nested txns: They would need to hide the page + * range in ancestor txns' dirty and spilled lists. + */ + if (env->me_pghead && + !txn->mt_parent && + ((mp->mp_flags & P_DIRTY) || + (sl && (x = mdb_midl_search(sl, pn)) <= sl[0] && sl[x] == pn))) + { + unsigned i, j; + pgno_t *mop; + MDB_ID2 *dl, ix, iy; + rc = mdb_midl_need(&env->me_pghead, ovpages); + if (rc) + return rc; + if (!(mp->mp_flags & P_DIRTY)) { + /* This page is no longer spilled */ + if (x == sl[0]) + sl[0]--; + else + sl[x] |= 1; + goto release; + } + /* Remove from dirty list */ + dl = txn->mt_u.dirty_list; + x = dl[0].mid--; + for (ix = dl[x]; ix.mptr != mp; ix = iy) { + if (x > 1) { + x--; + iy = dl[x]; + dl[x] = ix; + } else { + mdb_cassert(mc, x > 1); + j = ++(dl[0].mid); + dl[j] = ix; /* Unsorted. OK when MDB_TXN_ERROR. */ + txn->mt_flags |= MDB_TXN_ERROR; + return MDB_CORRUPTED; + } + } + txn->mt_dirty_room++; + if (!(env->me_flags & MDB_WRITEMAP)) + mdb_dpage_free(env, mp); +release: + /* Insert in me_pghead */ + mop = env->me_pghead; + j = mop[0] + ovpages; + for (i = mop[0]; i && mop[i] < pg; i--) + mop[j--] = mop[i]; + while (j>i) + mop[j--] = pg++; + mop[0] += ovpages; + } else { + rc = mdb_midl_append_range(&txn->mt_free_pgs, pg, ovpages); + if (rc) + return rc; + } + mc->mc_db->md_overflow_pages -= ovpages; + return 0; +} + +/** Return the data associated with a given node. + * @param[in] mc The cursor for this operation. + * @param[in] leaf The node being read. + * @param[out] data Updated to point to the node's data. + * @return 0 on success, non-zero on failure. + */ +static int +mdb_node_read(MDB_cursor *mc, MDB_node *leaf, MDB_val *data) +{ + MDB_page *omp; /* overflow page */ + pgno_t pgno; + int rc; + + if (!F_ISSET(leaf->mn_flags, F_BIGDATA)) { + data->mv_size = NODEDSZ(leaf); + data->mv_data = NODEDATA(leaf); + return MDB_SUCCESS; + } + + /* Read overflow data. + */ + data->mv_size = NODEDSZ(leaf); + memcpy(&pgno, NODEDATA(leaf), sizeof(pgno)); + if ((rc = mdb_page_get(mc, pgno, &omp, NULL)) != 0) { + DPRINTF(("read overflow page %"Z"u failed", pgno)); + return rc; + } + data->mv_data = METADATA(omp); + + return MDB_SUCCESS; +} + +int +mdb_get(MDB_txn *txn, MDB_dbi dbi, + MDB_val *key, MDB_val *data) +{ + MDB_cursor mc; + MDB_xcursor mx; + int exact = 0; + DKBUF; + + DPRINTF(("===> get db %u key [%s]", dbi, DKEY(key))); + + if (!key || !data || !TXN_DBI_EXIST(txn, dbi, DB_USRVALID)) + return EINVAL; + + if (txn->mt_flags & MDB_TXN_BLOCKED) + return MDB_BAD_TXN; + + mdb_cursor_init(&mc, txn, dbi, &mx); + return mdb_cursor_set(&mc, key, data, MDB_SET, &exact); +} + +/** Find a sibling for a page. + * Replaces the page at the top of the cursor's stack with the + * specified sibling, if one exists. + * @param[in] mc The cursor for this operation. + * @param[in] move_right Non-zero if the right sibling is requested, + * otherwise the left sibling. + * @return 0 on success, non-zero on failure. + */ +static int +mdb_cursor_sibling(MDB_cursor *mc, int move_right) +{ + int rc; + MDB_node *indx; + MDB_page *mp; + + if (mc->mc_snum < 2) { + return MDB_NOTFOUND; /* root has no siblings */ + } + + mdb_cursor_pop(mc); + DPRINTF(("parent page is page %"Z"u, index %u", + mc->mc_pg[mc->mc_top]->mp_pgno, mc->mc_ki[mc->mc_top])); + + if (move_right ? (mc->mc_ki[mc->mc_top] + 1u >= NUMKEYS(mc->mc_pg[mc->mc_top])) + : (mc->mc_ki[mc->mc_top] == 0)) { + DPRINTF(("no more keys left, moving to %s sibling", + move_right ? "right" : "left")); + if ((rc = mdb_cursor_sibling(mc, move_right)) != MDB_SUCCESS) { + /* undo cursor_pop before returning */ + mc->mc_top++; + mc->mc_snum++; + return rc; + } + } else { + if (move_right) + mc->mc_ki[mc->mc_top]++; + else + mc->mc_ki[mc->mc_top]--; + DPRINTF(("just moving to %s index key %u", + move_right ? "right" : "left", mc->mc_ki[mc->mc_top])); + } + mdb_cassert(mc, IS_BRANCH(mc->mc_pg[mc->mc_top])); + + indx = NODEPTR(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top]); + if ((rc = mdb_page_get(mc, NODEPGNO(indx), &mp, NULL)) != 0) { + /* mc will be inconsistent if caller does mc_snum++ as above */ + mc->mc_flags &= ~(C_INITIALIZED|C_EOF); + return rc; + } + + mdb_cursor_push(mc, mp); + if (!move_right) + mc->mc_ki[mc->mc_top] = NUMKEYS(mp)-1; + + return MDB_SUCCESS; +} + +/** Move the cursor to the next data item. */ +static int +mdb_cursor_next(MDB_cursor *mc, MDB_val *key, MDB_val *data, MDB_cursor_op op) +{ + MDB_page *mp; + MDB_node *leaf; + int rc; + + if ((mc->mc_flags & C_DEL && op == MDB_NEXT_DUP)) + return MDB_NOTFOUND; + + if (!(mc->mc_flags & C_INITIALIZED)) + return mdb_cursor_first(mc, key, data); + + mp = mc->mc_pg[mc->mc_top]; + + if (mc->mc_flags & C_EOF) { + if (mc->mc_ki[mc->mc_top] >= NUMKEYS(mp)-1) + return MDB_NOTFOUND; + mc->mc_flags ^= C_EOF; + } + + if (mc->mc_db->md_flags & MDB_DUPSORT) { + leaf = NODEPTR(mp, mc->mc_ki[mc->mc_top]); + if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { + if (op == MDB_NEXT || op == MDB_NEXT_DUP) { + rc = mdb_cursor_next(&mc->mc_xcursor->mx_cursor, data, NULL, MDB_NEXT); + if (op != MDB_NEXT || rc != MDB_NOTFOUND) { + if (rc == MDB_SUCCESS) + MDB_GET_KEY(leaf, key); + return rc; + } + } + } else { + mc->mc_xcursor->mx_cursor.mc_flags &= ~(C_INITIALIZED|C_EOF); + if (op == MDB_NEXT_DUP) + return MDB_NOTFOUND; + } + } + + DPRINTF(("cursor_next: top page is %"Z"u in cursor %p", + mdb_dbg_pgno(mp), (void *) mc)); + if (mc->mc_flags & C_DEL) { + mc->mc_flags ^= C_DEL; + goto skip; + } + + if (mc->mc_ki[mc->mc_top] + 1u >= NUMKEYS(mp)) { + DPUTS("=====> move to next sibling page"); + if ((rc = mdb_cursor_sibling(mc, 1)) != MDB_SUCCESS) { + mc->mc_flags |= C_EOF; + return rc; + } + mp = mc->mc_pg[mc->mc_top]; + DPRINTF(("next page is %"Z"u, key index %u", mp->mp_pgno, mc->mc_ki[mc->mc_top])); + } else + mc->mc_ki[mc->mc_top]++; + +skip: + DPRINTF(("==> cursor points to page %"Z"u with %u keys, key index %u", + mdb_dbg_pgno(mp), NUMKEYS(mp), mc->mc_ki[mc->mc_top])); + + if (IS_LEAF2(mp)) { + key->mv_size = mc->mc_db->md_pad; + key->mv_data = LEAF2KEY(mp, mc->mc_ki[mc->mc_top], key->mv_size); + return MDB_SUCCESS; + } + + mdb_cassert(mc, IS_LEAF(mp)); + leaf = NODEPTR(mp, mc->mc_ki[mc->mc_top]); + + if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { + mdb_xcursor_init1(mc, leaf); + } + if (data) { + if ((rc = mdb_node_read(mc, leaf, data)) != MDB_SUCCESS) + return rc; + + if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { + rc = mdb_cursor_first(&mc->mc_xcursor->mx_cursor, data, NULL); + if (rc != MDB_SUCCESS) + return rc; + } + } + + MDB_GET_KEY(leaf, key); + return MDB_SUCCESS; +} + +/** Move the cursor to the previous data item. */ +static int +mdb_cursor_prev(MDB_cursor *mc, MDB_val *key, MDB_val *data, MDB_cursor_op op) +{ + MDB_page *mp; + MDB_node *leaf; + int rc; + + if (!(mc->mc_flags & C_INITIALIZED)) { + rc = mdb_cursor_last(mc, key, data); + if (rc) + return rc; + mc->mc_ki[mc->mc_top]++; + } + + mp = mc->mc_pg[mc->mc_top]; + + if (mc->mc_db->md_flags & MDB_DUPSORT) { + leaf = NODEPTR(mp, mc->mc_ki[mc->mc_top]); + if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { + if (op == MDB_PREV || op == MDB_PREV_DUP) { + rc = mdb_cursor_prev(&mc->mc_xcursor->mx_cursor, data, NULL, MDB_PREV); + if (op != MDB_PREV || rc != MDB_NOTFOUND) { + if (rc == MDB_SUCCESS) { + MDB_GET_KEY(leaf, key); + mc->mc_flags &= ~C_EOF; + } + return rc; + } + } + } else { + mc->mc_xcursor->mx_cursor.mc_flags &= ~(C_INITIALIZED|C_EOF); + if (op == MDB_PREV_DUP) + return MDB_NOTFOUND; + } + } + + DPRINTF(("cursor_prev: top page is %"Z"u in cursor %p", + mdb_dbg_pgno(mp), (void *) mc)); + + mc->mc_flags &= ~(C_EOF|C_DEL); + + if (mc->mc_ki[mc->mc_top] == 0) { + DPUTS("=====> move to prev sibling page"); + if ((rc = mdb_cursor_sibling(mc, 0)) != MDB_SUCCESS) { + return rc; + } + mp = mc->mc_pg[mc->mc_top]; + mc->mc_ki[mc->mc_top] = NUMKEYS(mp) - 1; + DPRINTF(("prev page is %"Z"u, key index %u", mp->mp_pgno, mc->mc_ki[mc->mc_top])); + } else + mc->mc_ki[mc->mc_top]--; + + DPRINTF(("==> cursor points to page %"Z"u with %u keys, key index %u", + mdb_dbg_pgno(mp), NUMKEYS(mp), mc->mc_ki[mc->mc_top])); + + if (IS_LEAF2(mp)) { + key->mv_size = mc->mc_db->md_pad; + key->mv_data = LEAF2KEY(mp, mc->mc_ki[mc->mc_top], key->mv_size); + return MDB_SUCCESS; + } + + mdb_cassert(mc, IS_LEAF(mp)); + leaf = NODEPTR(mp, mc->mc_ki[mc->mc_top]); + + if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { + mdb_xcursor_init1(mc, leaf); + } + if (data) { + if ((rc = mdb_node_read(mc, leaf, data)) != MDB_SUCCESS) + return rc; + + if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { + rc = mdb_cursor_last(&mc->mc_xcursor->mx_cursor, data, NULL); + if (rc != MDB_SUCCESS) + return rc; + } + } + + MDB_GET_KEY(leaf, key); + return MDB_SUCCESS; +} + +/** Set the cursor on a specific data item. */ +static int +mdb_cursor_set(MDB_cursor *mc, MDB_val *key, MDB_val *data, + MDB_cursor_op op, int *exactp) +{ + int rc; + MDB_page *mp; + MDB_node *leaf = NULL; + DKBUF; + + if (key->mv_size == 0) + return MDB_BAD_VALSIZE; + + if (mc->mc_xcursor) + mc->mc_xcursor->mx_cursor.mc_flags &= ~(C_INITIALIZED|C_EOF); + + /* See if we're already on the right page */ + if (mc->mc_flags & C_INITIALIZED) { + MDB_val nodekey; + + mp = mc->mc_pg[mc->mc_top]; + if (!NUMKEYS(mp)) { + mc->mc_ki[mc->mc_top] = 0; + return MDB_NOTFOUND; + } + if (mp->mp_flags & P_LEAF2) { + nodekey.mv_size = mc->mc_db->md_pad; + nodekey.mv_data = LEAF2KEY(mp, 0, nodekey.mv_size); + } else { + leaf = NODEPTR(mp, 0); + MDB_GET_KEY2(leaf, nodekey); + } + rc = mc->mc_dbx->md_cmp(key, &nodekey); + if (rc == 0) { + /* Probably happens rarely, but first node on the page + * was the one we wanted. + */ + mc->mc_ki[mc->mc_top] = 0; + if (exactp) + *exactp = 1; + goto set1; + } + if (rc > 0) { + unsigned int i; + unsigned int nkeys = NUMKEYS(mp); + if (nkeys > 1) { + if (mp->mp_flags & P_LEAF2) { + nodekey.mv_data = LEAF2KEY(mp, + nkeys-1, nodekey.mv_size); + } else { + leaf = NODEPTR(mp, nkeys-1); + MDB_GET_KEY2(leaf, nodekey); + } + rc = mc->mc_dbx->md_cmp(key, &nodekey); + if (rc == 0) { + /* last node was the one we wanted */ + mc->mc_ki[mc->mc_top] = nkeys-1; + if (exactp) + *exactp = 1; + goto set1; + } + if (rc < 0) { + if (mc->mc_ki[mc->mc_top] < NUMKEYS(mp)) { + /* This is definitely the right page, skip search_page */ + if (mp->mp_flags & P_LEAF2) { + nodekey.mv_data = LEAF2KEY(mp, + mc->mc_ki[mc->mc_top], nodekey.mv_size); + } else { + leaf = NODEPTR(mp, mc->mc_ki[mc->mc_top]); + MDB_GET_KEY2(leaf, nodekey); + } + rc = mc->mc_dbx->md_cmp(key, &nodekey); + if (rc == 0) { + /* current node was the one we wanted */ + if (exactp) + *exactp = 1; + goto set1; + } + } + rc = 0; + mc->mc_flags &= ~C_EOF; + goto set2; + } + } + /* If any parents have right-sibs, search. + * Otherwise, there's nothing further. + */ + for (i=0; i<mc->mc_top; i++) + if (mc->mc_ki[i] < + NUMKEYS(mc->mc_pg[i])-1) + break; + if (i == mc->mc_top) { + /* There are no other pages */ + mc->mc_ki[mc->mc_top] = nkeys; + return MDB_NOTFOUND; + } + } + if (!mc->mc_top) { + /* There are no other pages */ + mc->mc_ki[mc->mc_top] = 0; + if (op == MDB_SET_RANGE && !exactp) { + rc = 0; + goto set1; + } else + return MDB_NOTFOUND; + } + } else { + mc->mc_pg[0] = 0; + } + + rc = mdb_page_search(mc, key, 0); + if (rc != MDB_SUCCESS) + return rc; + + mp = mc->mc_pg[mc->mc_top]; + mdb_cassert(mc, IS_LEAF(mp)); + +set2: + leaf = mdb_node_search(mc, key, exactp); + if (exactp != NULL && !*exactp) { + /* MDB_SET specified and not an exact match. */ + return MDB_NOTFOUND; + } + + if (leaf == NULL) { + DPUTS("===> inexact leaf not found, goto sibling"); + if ((rc = mdb_cursor_sibling(mc, 1)) != MDB_SUCCESS) { + mc->mc_flags |= C_EOF; + return rc; /* no entries matched */ + } + mp = mc->mc_pg[mc->mc_top]; + mdb_cassert(mc, IS_LEAF(mp)); + leaf = NODEPTR(mp, 0); + } + +set1: + mc->mc_flags |= C_INITIALIZED; + mc->mc_flags &= ~C_EOF; + + if (IS_LEAF2(mp)) { + if (op == MDB_SET_RANGE || op == MDB_SET_KEY) { + key->mv_size = mc->mc_db->md_pad; + key->mv_data = LEAF2KEY(mp, mc->mc_ki[mc->mc_top], key->mv_size); + } + return MDB_SUCCESS; + } + + if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { + mdb_xcursor_init1(mc, leaf); + } + if (data) { + if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { + if (op == MDB_SET || op == MDB_SET_KEY || op == MDB_SET_RANGE) { + rc = mdb_cursor_first(&mc->mc_xcursor->mx_cursor, data, NULL); + } else { + int ex2, *ex2p; + if (op == MDB_GET_BOTH) { + ex2p = &ex2; + ex2 = 0; + } else { + ex2p = NULL; + } + rc = mdb_cursor_set(&mc->mc_xcursor->mx_cursor, data, NULL, MDB_SET_RANGE, ex2p); + if (rc != MDB_SUCCESS) + return rc; + } + } else if (op == MDB_GET_BOTH || op == MDB_GET_BOTH_RANGE) { + MDB_val olddata; + MDB_cmp_func *dcmp; + if ((rc = mdb_node_read(mc, leaf, &olddata)) != MDB_SUCCESS) + return rc; + dcmp = mc->mc_dbx->md_dcmp; +#if UINT_MAX < SIZE_MAX + if (dcmp == mdb_cmp_int && olddata.mv_size == sizeof(size_t)) + dcmp = mdb_cmp_clong; +#endif + rc = dcmp(data, &olddata); + if (rc) { + if (op == MDB_GET_BOTH || rc > 0) + return MDB_NOTFOUND; + rc = 0; + } + *data = olddata; + + } else { + if (mc->mc_xcursor) + mc->mc_xcursor->mx_cursor.mc_flags &= ~(C_INITIALIZED|C_EOF); + if ((rc = mdb_node_read(mc, leaf, data)) != MDB_SUCCESS) + return rc; + } + } + + /* The key already matches in all other cases */ + if (op == MDB_SET_RANGE || op == MDB_SET_KEY) + MDB_GET_KEY(leaf, key); + DPRINTF(("==> cursor placed on key [%s]", DKEY(key))); + + return rc; +} + +/** Move the cursor to the first item in the database. */ +static int +mdb_cursor_first(MDB_cursor *mc, MDB_val *key, MDB_val *data) +{ + int rc; + MDB_node *leaf; + + if (mc->mc_xcursor) + mc->mc_xcursor->mx_cursor.mc_flags &= ~(C_INITIALIZED|C_EOF); + + if (!(mc->mc_flags & C_INITIALIZED) || mc->mc_top) { + rc = mdb_page_search(mc, NULL, MDB_PS_FIRST); + if (rc != MDB_SUCCESS) + return rc; + } + mdb_cassert(mc, IS_LEAF(mc->mc_pg[mc->mc_top])); + + leaf = NODEPTR(mc->mc_pg[mc->mc_top], 0); + mc->mc_flags |= C_INITIALIZED; + mc->mc_flags &= ~C_EOF; + + mc->mc_ki[mc->mc_top] = 0; + + if (IS_LEAF2(mc->mc_pg[mc->mc_top])) { + key->mv_size = mc->mc_db->md_pad; + key->mv_data = LEAF2KEY(mc->mc_pg[mc->mc_top], 0, key->mv_size); + return MDB_SUCCESS; + } + + if (data) { + if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { + mdb_xcursor_init1(mc, leaf); + rc = mdb_cursor_first(&mc->mc_xcursor->mx_cursor, data, NULL); + if (rc) + return rc; + } else { + if ((rc = mdb_node_read(mc, leaf, data)) != MDB_SUCCESS) + return rc; + } + } + MDB_GET_KEY(leaf, key); + return MDB_SUCCESS; +} + +/** Move the cursor to the last item in the database. */ +static int +mdb_cursor_last(MDB_cursor *mc, MDB_val *key, MDB_val *data) +{ + int rc; + MDB_node *leaf; + + if (mc->mc_xcursor) + mc->mc_xcursor->mx_cursor.mc_flags &= ~(C_INITIALIZED|C_EOF); + + if (!(mc->mc_flags & C_INITIALIZED) || mc->mc_top) { + rc = mdb_page_search(mc, NULL, MDB_PS_LAST); + if (rc != MDB_SUCCESS) + return rc; + } + mdb_cassert(mc, IS_LEAF(mc->mc_pg[mc->mc_top])); + + mc->mc_ki[mc->mc_top] = NUMKEYS(mc->mc_pg[mc->mc_top]) - 1; + mc->mc_flags |= C_INITIALIZED|C_EOF; + leaf = NODEPTR(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top]); + + if (IS_LEAF2(mc->mc_pg[mc->mc_top])) { + key->mv_size = mc->mc_db->md_pad; + key->mv_data = LEAF2KEY(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top], key->mv_size); + return MDB_SUCCESS; + } + + if (data) { + if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { + mdb_xcursor_init1(mc, leaf); + rc = mdb_cursor_last(&mc->mc_xcursor->mx_cursor, data, NULL); + if (rc) + return rc; + } else { + if ((rc = mdb_node_read(mc, leaf, data)) != MDB_SUCCESS) + return rc; + } + } + + MDB_GET_KEY(leaf, key); + return MDB_SUCCESS; +} + +int +mdb_cursor_get(MDB_cursor *mc, MDB_val *key, MDB_val *data, + MDB_cursor_op op) +{ + int rc; + int exact = 0; + int (*mfunc)(MDB_cursor *mc, MDB_val *key, MDB_val *data); + + if (mc == NULL) + return EINVAL; + + if (mc->mc_txn->mt_flags & MDB_TXN_BLOCKED) + return MDB_BAD_TXN; + + switch (op) { + case MDB_GET_CURRENT: + if (!(mc->mc_flags & C_INITIALIZED)) { + rc = EINVAL; + } else { + MDB_page *mp = mc->mc_pg[mc->mc_top]; + int nkeys = NUMKEYS(mp); + if (!nkeys || mc->mc_ki[mc->mc_top] >= nkeys) { + mc->mc_ki[mc->mc_top] = nkeys; + rc = MDB_NOTFOUND; + break; + } + rc = MDB_SUCCESS; + if (IS_LEAF2(mp)) { + key->mv_size = mc->mc_db->md_pad; + key->mv_data = LEAF2KEY(mp, mc->mc_ki[mc->mc_top], key->mv_size); + } else { + MDB_node *leaf = NODEPTR(mp, mc->mc_ki[mc->mc_top]); + MDB_GET_KEY(leaf, key); + if (data) { + if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { + rc = mdb_cursor_get(&mc->mc_xcursor->mx_cursor, data, NULL, MDB_GET_CURRENT); + } else { + rc = mdb_node_read(mc, leaf, data); + } + } + } + } + break; + case MDB_GET_BOTH: + case MDB_GET_BOTH_RANGE: + if (data == NULL) { + rc = EINVAL; + break; + } + if (mc->mc_xcursor == NULL) { + rc = MDB_INCOMPATIBLE; + break; + } + /* FALLTHRU */ + case MDB_SET: + case MDB_SET_KEY: + case MDB_SET_RANGE: + if (key == NULL) { + rc = EINVAL; + } else { + rc = mdb_cursor_set(mc, key, data, op, + op == MDB_SET_RANGE ? NULL : &exact); + } + break; + case MDB_GET_MULTIPLE: + if (data == NULL || !(mc->mc_flags & C_INITIALIZED)) { + rc = EINVAL; + break; + } + if (!(mc->mc_db->md_flags & MDB_DUPFIXED)) { + rc = MDB_INCOMPATIBLE; + break; + } + rc = MDB_SUCCESS; + if (!(mc->mc_xcursor->mx_cursor.mc_flags & C_INITIALIZED) || + (mc->mc_xcursor->mx_cursor.mc_flags & C_EOF)) + break; + goto fetchm; + case MDB_NEXT_MULTIPLE: + if (data == NULL) { + rc = EINVAL; + break; + } + if (!(mc->mc_db->md_flags & MDB_DUPFIXED)) { + rc = MDB_INCOMPATIBLE; + break; + } + rc = mdb_cursor_next(mc, key, data, MDB_NEXT_DUP); + if (rc == MDB_SUCCESS) { + if (mc->mc_xcursor->mx_cursor.mc_flags & C_INITIALIZED) { + MDB_cursor *mx; +fetchm: + mx = &mc->mc_xcursor->mx_cursor; + data->mv_size = NUMKEYS(mx->mc_pg[mx->mc_top]) * + mx->mc_db->md_pad; + data->mv_data = METADATA(mx->mc_pg[mx->mc_top]); + mx->mc_ki[mx->mc_top] = NUMKEYS(mx->mc_pg[mx->mc_top])-1; + } else { + rc = MDB_NOTFOUND; + } + } + break; + case MDB_PREV_MULTIPLE: + if (data == NULL) { + rc = EINVAL; + break; + } + if (!(mc->mc_db->md_flags & MDB_DUPFIXED)) { + rc = MDB_INCOMPATIBLE; + break; + } + if (!(mc->mc_flags & C_INITIALIZED)) + rc = mdb_cursor_last(mc, key, data); + else + rc = MDB_SUCCESS; + if (rc == MDB_SUCCESS) { + MDB_cursor *mx = &mc->mc_xcursor->mx_cursor; + if (mx->mc_flags & C_INITIALIZED) { + rc = mdb_cursor_sibling(mx, 0); + if (rc == MDB_SUCCESS) + goto fetchm; + } else { + rc = MDB_NOTFOUND; + } + } + break; + case MDB_NEXT: + case MDB_NEXT_DUP: + case MDB_NEXT_NODUP: + rc = mdb_cursor_next(mc, key, data, op); + break; + case MDB_PREV: + case MDB_PREV_DUP: + case MDB_PREV_NODUP: + rc = mdb_cursor_prev(mc, key, data, op); + break; + case MDB_FIRST: + rc = mdb_cursor_first(mc, key, data); + break; + case MDB_FIRST_DUP: + mfunc = mdb_cursor_first; + mmove: + if (data == NULL || !(mc->mc_flags & C_INITIALIZED)) { + rc = EINVAL; + break; + } + if (mc->mc_xcursor == NULL) { + rc = MDB_INCOMPATIBLE; + break; + } + if (mc->mc_ki[mc->mc_top] >= NUMKEYS(mc->mc_pg[mc->mc_top])) { + mc->mc_ki[mc->mc_top] = NUMKEYS(mc->mc_pg[mc->mc_top]); + rc = MDB_NOTFOUND; + break; + } + { + MDB_node *leaf = NODEPTR(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top]); + if (!F_ISSET(leaf->mn_flags, F_DUPDATA)) { + MDB_GET_KEY(leaf, key); + rc = mdb_node_read(mc, leaf, data); + break; + } + } + if (!(mc->mc_xcursor->mx_cursor.mc_flags & C_INITIALIZED)) { + rc = EINVAL; + break; + } + rc = mfunc(&mc->mc_xcursor->mx_cursor, data, NULL); + break; + case MDB_LAST: + rc = mdb_cursor_last(mc, key, data); + break; + case MDB_LAST_DUP: + mfunc = mdb_cursor_last; + goto mmove; + default: + DPRINTF(("unhandled/unimplemented cursor operation %u", op)); + rc = EINVAL; + break; + } + + if (mc->mc_flags & C_DEL) + mc->mc_flags ^= C_DEL; + + return rc; +} + +/** Touch all the pages in the cursor stack. Set mc_top. + * Makes sure all the pages are writable, before attempting a write operation. + * @param[in] mc The cursor to operate on. + */ +static int +mdb_cursor_touch(MDB_cursor *mc) +{ + int rc = MDB_SUCCESS; + + if (mc->mc_dbi >= CORE_DBS && !(*mc->mc_dbflag & (DB_DIRTY|DB_DUPDATA))) { + /* Touch DB record of named DB */ + MDB_cursor mc2; + MDB_xcursor mcx; + if (TXN_DBI_CHANGED(mc->mc_txn, mc->mc_dbi)) + return MDB_BAD_DBI; + mdb_cursor_init(&mc2, mc->mc_txn, MAIN_DBI, &mcx); + rc = mdb_page_search(&mc2, &mc->mc_dbx->md_name, MDB_PS_MODIFY); + if (rc) + return rc; + *mc->mc_dbflag |= DB_DIRTY; + } + mc->mc_top = 0; + if (mc->mc_snum) { + do { + rc = mdb_page_touch(mc); + } while (!rc && ++(mc->mc_top) < mc->mc_snum); + mc->mc_top = mc->mc_snum-1; + } + return rc; +} + +/** Do not spill pages to disk if txn is getting full, may fail instead */ +#define MDB_NOSPILL 0x8000 + +int +mdb_cursor_put(MDB_cursor *mc, MDB_val *key, MDB_val *data, + unsigned int flags) +{ + MDB_env *env; + MDB_node *leaf = NULL; + MDB_page *fp, *mp, *sub_root = NULL; + uint16_t fp_flags; + MDB_val xdata, *rdata, dkey, olddata; + MDB_db dummy; + int do_sub = 0, insert_key, insert_data; + unsigned int mcount = 0, dcount = 0, nospill; + size_t nsize; + int rc, rc2; + unsigned int nflags; + DKBUF; + + if (mc == NULL || key == NULL) + return EINVAL; + + env = mc->mc_txn->mt_env; + + /* Check this first so counter will always be zero on any + * early failures. + */ + if (flags & MDB_MULTIPLE) { + dcount = data[1].mv_size; + data[1].mv_size = 0; + if (!F_ISSET(mc->mc_db->md_flags, MDB_DUPFIXED)) + return MDB_INCOMPATIBLE; + } + + nospill = flags & MDB_NOSPILL; + flags &= ~MDB_NOSPILL; + + if (mc->mc_txn->mt_flags & (MDB_TXN_RDONLY|MDB_TXN_BLOCKED)) + return (mc->mc_txn->mt_flags & MDB_TXN_RDONLY) ? EACCES : MDB_BAD_TXN; + + if (key->mv_size-1 >= ENV_MAXKEY(env)) + return MDB_BAD_VALSIZE; + +#if SIZE_MAX > MAXDATASIZE + if (data->mv_size > ((mc->mc_db->md_flags & MDB_DUPSORT) ? ENV_MAXKEY(env) : MAXDATASIZE)) + return MDB_BAD_VALSIZE; +#else + if ((mc->mc_db->md_flags & MDB_DUPSORT) && data->mv_size > ENV_MAXKEY(env)) + return MDB_BAD_VALSIZE; +#endif + + DPRINTF(("==> put db %d key [%s], size %"Z"u, data size %"Z"u", + DDBI(mc), DKEY(key), key ? key->mv_size : 0, data->mv_size)); + + dkey.mv_size = 0; + + if (flags == MDB_CURRENT) { + if (!(mc->mc_flags & C_INITIALIZED)) + return EINVAL; + rc = MDB_SUCCESS; + } else if (mc->mc_db->md_root == P_INVALID) { + /* new database, cursor has nothing to point to */ + mc->mc_snum = 0; + mc->mc_top = 0; + mc->mc_flags &= ~C_INITIALIZED; + rc = MDB_NO_ROOT; + } else { + int exact = 0; + MDB_val d2; + if (flags & MDB_APPEND) { + MDB_val k2; + rc = mdb_cursor_last(mc, &k2, &d2); + if (rc == 0) { + rc = mc->mc_dbx->md_cmp(key, &k2); + if (rc > 0) { + rc = MDB_NOTFOUND; + mc->mc_ki[mc->mc_top]++; + } else { + /* new key is <= last key */ + rc = MDB_KEYEXIST; + } + } + } else { + rc = mdb_cursor_set(mc, key, &d2, MDB_SET, &exact); + } + if ((flags & MDB_NOOVERWRITE) && rc == 0) { + DPRINTF(("duplicate key [%s]", DKEY(key))); + *data = d2; + return MDB_KEYEXIST; + } + if (rc && rc != MDB_NOTFOUND) + return rc; + } + + if (mc->mc_flags & C_DEL) + mc->mc_flags ^= C_DEL; + + /* Cursor is positioned, check for room in the dirty list */ + if (!nospill) { + if (flags & MDB_MULTIPLE) { + rdata = &xdata; + xdata.mv_size = data->mv_size * dcount; + } else { + rdata = data; + } + if ((rc2 = mdb_page_spill(mc, key, rdata))) + return rc2; + } + + if (rc == MDB_NO_ROOT) { + MDB_page *np; + /* new database, write a root leaf page */ + DPUTS("allocating new root leaf page"); + if ((rc2 = mdb_page_new(mc, P_LEAF, 1, &np))) { + return rc2; + } + mdb_cursor_push(mc, np); + mc->mc_db->md_root = np->mp_pgno; + mc->mc_db->md_depth++; + *mc->mc_dbflag |= DB_DIRTY; + if ((mc->mc_db->md_flags & (MDB_DUPSORT|MDB_DUPFIXED)) + == MDB_DUPFIXED) + np->mp_flags |= P_LEAF2; + mc->mc_flags |= C_INITIALIZED; + } else { + /* make sure all cursor pages are writable */ + rc2 = mdb_cursor_touch(mc); + if (rc2) + return rc2; + } + + insert_key = insert_data = rc; + if (insert_key) { + /* The key does not exist */ + DPRINTF(("inserting key at index %i", mc->mc_ki[mc->mc_top])); + if ((mc->mc_db->md_flags & MDB_DUPSORT) && + LEAFSIZE(key, data) > env->me_nodemax) + { + /* Too big for a node, insert in sub-DB. Set up an empty + * "old sub-page" for prep_subDB to expand to a full page. + */ + fp_flags = P_LEAF|P_DIRTY; + fp = env->me_pbuf; + fp->mp_pad = data->mv_size; /* used if MDB_DUPFIXED */ + fp->mp_lower = fp->mp_upper = (PAGEHDRSZ-PAGEBASE); + olddata.mv_size = PAGEHDRSZ; + goto prep_subDB; + } + } else { + /* there's only a key anyway, so this is a no-op */ + if (IS_LEAF2(mc->mc_pg[mc->mc_top])) { + char *ptr; + unsigned int ksize = mc->mc_db->md_pad; + if (key->mv_size != ksize) + return MDB_BAD_VALSIZE; + ptr = LEAF2KEY(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top], ksize); + memcpy(ptr, key->mv_data, ksize); +fix_parent: + /* if overwriting slot 0 of leaf, need to + * update branch key if there is a parent page + */ + if (mc->mc_top && !mc->mc_ki[mc->mc_top]) { + unsigned short dtop = 1; + mc->mc_top--; + /* slot 0 is always an empty key, find real slot */ + while (mc->mc_top && !mc->mc_ki[mc->mc_top]) { + mc->mc_top--; + dtop++; + } + if (mc->mc_ki[mc->mc_top]) + rc2 = mdb_update_key(mc, key); + else + rc2 = MDB_SUCCESS; + mc->mc_top += dtop; + if (rc2) + return rc2; + } + return MDB_SUCCESS; + } + +more: + leaf = NODEPTR(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top]); + olddata.mv_size = NODEDSZ(leaf); + olddata.mv_data = NODEDATA(leaf); + + /* DB has dups? */ + if (F_ISSET(mc->mc_db->md_flags, MDB_DUPSORT)) { + /* Prepare (sub-)page/sub-DB to accept the new item, + * if needed. fp: old sub-page or a header faking + * it. mp: new (sub-)page. offset: growth in page + * size. xdata: node data with new page or DB. + */ + unsigned i, offset = 0; + mp = fp = xdata.mv_data = env->me_pbuf; + mp->mp_pgno = mc->mc_pg[mc->mc_top]->mp_pgno; + + /* Was a single item before, must convert now */ + if (!F_ISSET(leaf->mn_flags, F_DUPDATA)) { + MDB_cmp_func *dcmp; + /* Just overwrite the current item */ + if (flags == MDB_CURRENT) + goto current; + dcmp = mc->mc_dbx->md_dcmp; +#if UINT_MAX < SIZE_MAX + if (dcmp == mdb_cmp_int && olddata.mv_size == sizeof(size_t)) + dcmp = mdb_cmp_clong; +#endif + /* does data match? */ + if (!dcmp(data, &olddata)) { + if (flags & (MDB_NODUPDATA|MDB_APPENDDUP)) + return MDB_KEYEXIST; + /* overwrite it */ + goto current; + } + + /* Back up original data item */ + dkey.mv_size = olddata.mv_size; + dkey.mv_data = memcpy(fp+1, olddata.mv_data, olddata.mv_size); + + /* Make sub-page header for the dup items, with dummy body */ + fp->mp_flags = P_LEAF|P_DIRTY|P_SUBP; + fp->mp_lower = (PAGEHDRSZ-PAGEBASE); + xdata.mv_size = PAGEHDRSZ + dkey.mv_size + data->mv_size; + if (mc->mc_db->md_flags & MDB_DUPFIXED) { + fp->mp_flags |= P_LEAF2; + fp->mp_pad = data->mv_size; + xdata.mv_size += 2 * data->mv_size; /* leave space for 2 more */ + } else { + xdata.mv_size += 2 * (sizeof(indx_t) + NODESIZE) + + (dkey.mv_size & 1) + (data->mv_size & 1); + } + fp->mp_upper = xdata.mv_size - PAGEBASE; + olddata.mv_size = xdata.mv_size; /* pretend olddata is fp */ + } else if (leaf->mn_flags & F_SUBDATA) { + /* Data is on sub-DB, just store it */ + flags |= F_DUPDATA|F_SUBDATA; + goto put_sub; + } else { + /* Data is on sub-page */ + fp = olddata.mv_data; + switch (flags) { + default: + if (!(mc->mc_db->md_flags & MDB_DUPFIXED)) { + offset = EVEN(NODESIZE + sizeof(indx_t) + + data->mv_size); + break; + } + offset = fp->mp_pad; + if (SIZELEFT(fp) < offset) { + offset *= 4; /* space for 4 more */ + break; + } + /* FALLTHRU: Big enough MDB_DUPFIXED sub-page */ + case MDB_CURRENT: + fp->mp_flags |= P_DIRTY; + COPY_PGNO(fp->mp_pgno, mp->mp_pgno); + mc->mc_xcursor->mx_cursor.mc_pg[0] = fp; + flags |= F_DUPDATA; + goto put_sub; + } + xdata.mv_size = olddata.mv_size + offset; + } + + fp_flags = fp->mp_flags; + if (NODESIZE + NODEKSZ(leaf) + xdata.mv_size > env->me_nodemax) { + /* Too big for a sub-page, convert to sub-DB */ + fp_flags &= ~P_SUBP; +prep_subDB: + if (mc->mc_db->md_flags & MDB_DUPFIXED) { + fp_flags |= P_LEAF2; + dummy.md_pad = fp->mp_pad; + dummy.md_flags = MDB_DUPFIXED; + if (mc->mc_db->md_flags & MDB_INTEGERDUP) + dummy.md_flags |= MDB_INTEGERKEY; + } else { + dummy.md_pad = 0; + dummy.md_flags = 0; + } + dummy.md_depth = 1; + dummy.md_branch_pages = 0; + dummy.md_leaf_pages = 1; + dummy.md_overflow_pages = 0; + dummy.md_entries = NUMKEYS(fp); + xdata.mv_size = sizeof(MDB_db); + xdata.mv_data = &dummy; + if ((rc = mdb_page_alloc(mc, 1, &mp))) + return rc; + offset = env->me_psize - olddata.mv_size; + flags |= F_DUPDATA|F_SUBDATA; + dummy.md_root = mp->mp_pgno; + sub_root = mp; + } + if (mp != fp) { + mp->mp_flags = fp_flags | P_DIRTY; + mp->mp_pad = fp->mp_pad; + mp->mp_lower = fp->mp_lower; + mp->mp_upper = fp->mp_upper + offset; + if (fp_flags & P_LEAF2) { + memcpy(METADATA(mp), METADATA(fp), NUMKEYS(fp) * fp->mp_pad); + } else { + memcpy((char *)mp + mp->mp_upper + PAGEBASE, (char *)fp + fp->mp_upper + PAGEBASE, + olddata.mv_size - fp->mp_upper - PAGEBASE); + memcpy((char *)(&mp->mp_ptrs), (char *)(&fp->mp_ptrs), NUMKEYS(fp) * sizeof(mp->mp_ptrs[0])); + for (i=0; i<NUMKEYS(fp); i++) + mp->mp_ptrs[i] += offset; + } + } + + rdata = &xdata; + flags |= F_DUPDATA; + do_sub = 1; + if (!insert_key) + mdb_node_del(mc, 0); + goto new_sub; + } +current: + /* LMDB passes F_SUBDATA in 'flags' to write a DB record */ + if ((leaf->mn_flags ^ flags) & F_SUBDATA) + return MDB_INCOMPATIBLE; + /* overflow page overwrites need special handling */ + if (F_ISSET(leaf->mn_flags, F_BIGDATA)) { + MDB_page *omp; + pgno_t pg; + int level, ovpages, dpages = OVPAGES(data->mv_size, env->me_psize); + + memcpy(&pg, olddata.mv_data, sizeof(pg)); + if ((rc2 = mdb_page_get(mc, pg, &omp, &level)) != 0) + return rc2; + ovpages = omp->mp_pages; + + /* Is the ov page large enough? */ + if (ovpages >= dpages) { + if (!(omp->mp_flags & P_DIRTY) && + (level || (env->me_flags & MDB_WRITEMAP))) + { + rc = mdb_page_unspill(mc->mc_txn, omp, &omp); + if (rc) + return rc; + level = 0; /* dirty in this txn or clean */ + } + /* Is it dirty? */ + if (omp->mp_flags & P_DIRTY) { + /* yes, overwrite it. Note in this case we don't + * bother to try shrinking the page if the new data + * is smaller than the overflow threshold. + */ + if (level > 1) { + /* It is writable only in a parent txn */ + size_t sz = (size_t) env->me_psize * ovpages, off; + MDB_page *np = mdb_page_malloc(mc->mc_txn, ovpages); + MDB_ID2 id2; + if (!np) + return ENOMEM; + id2.mid = pg; + id2.mptr = np; + /* Note - this page is already counted in parent's dirty_room */ + rc2 = mdb_mid2l_insert(mc->mc_txn->mt_u.dirty_list, &id2); + mdb_cassert(mc, rc2 == 0); + /* Currently we make the page look as with put() in the + * parent txn, in case the user peeks at MDB_RESERVEd + * or unused parts. Some users treat ovpages specially. + */ + if (!(flags & MDB_RESERVE)) { + /* Skip the part where LMDB will put *data. + * Copy end of page, adjusting alignment so + * compiler may copy words instead of bytes. + */ + off = (PAGEHDRSZ + data->mv_size) & -sizeof(size_t); + memcpy((size_t *)((char *)np + off), + (size_t *)((char *)omp + off), sz - off); + sz = PAGEHDRSZ; + } + memcpy(np, omp, sz); /* Copy beginning of page */ + omp = np; + } + SETDSZ(leaf, data->mv_size); + if (F_ISSET(flags, MDB_RESERVE)) + data->mv_data = METADATA(omp); + else + memcpy(METADATA(omp), data->mv_data, data->mv_size); + return MDB_SUCCESS; + } + } + if ((rc2 = mdb_ovpage_free(mc, omp)) != MDB_SUCCESS) + return rc2; + } else if (data->mv_size == olddata.mv_size) { + /* same size, just replace it. Note that we could + * also reuse this node if the new data is smaller, + * but instead we opt to shrink the node in that case. + */ + if (F_ISSET(flags, MDB_RESERVE)) + data->mv_data = olddata.mv_data; + else if (!(mc->mc_flags & C_SUB)) + memcpy(olddata.mv_data, data->mv_data, data->mv_size); + else { + memcpy(NODEKEY(leaf), key->mv_data, key->mv_size); + goto fix_parent; + } + return MDB_SUCCESS; + } + mdb_node_del(mc, 0); + } + + rdata = data; + +new_sub: + nflags = flags & NODE_ADD_FLAGS; + nsize = IS_LEAF2(mc->mc_pg[mc->mc_top]) ? key->mv_size : mdb_leaf_size(env, key, rdata); + if (SIZELEFT(mc->mc_pg[mc->mc_top]) < nsize) { + if (( flags & (F_DUPDATA|F_SUBDATA)) == F_DUPDATA ) + nflags &= ~MDB_APPEND; /* sub-page may need room to grow */ + if (!insert_key) + nflags |= MDB_SPLIT_REPLACE; + rc = mdb_page_split(mc, key, rdata, P_INVALID, nflags); + } else { + /* There is room already in this leaf page. */ + rc = mdb_node_add(mc, mc->mc_ki[mc->mc_top], key, rdata, 0, nflags); + if (rc == 0) { + /* Adjust other cursors pointing to mp */ + MDB_cursor *m2, *m3; + MDB_dbi dbi = mc->mc_dbi; + unsigned i = mc->mc_top; + MDB_page *mp = mc->mc_pg[i]; + + for (m2 = mc->mc_txn->mt_cursors[dbi]; m2; m2=m2->mc_next) { + if (mc->mc_flags & C_SUB) + m3 = &m2->mc_xcursor->mx_cursor; + else + m3 = m2; + if (m3 == mc || m3->mc_snum < mc->mc_snum || m3->mc_pg[i] != mp) continue; + if (m3->mc_ki[i] >= mc->mc_ki[i] && insert_key) { + m3->mc_ki[i]++; + } + XCURSOR_REFRESH(m3, i, mp); + } + } + } + + if (rc == MDB_SUCCESS) { + /* Now store the actual data in the child DB. Note that we're + * storing the user data in the keys field, so there are strict + * size limits on dupdata. The actual data fields of the child + * DB are all zero size. + */ + if (do_sub) { + int xflags, new_dupdata; + size_t ecount; +put_sub: + xdata.mv_size = 0; + xdata.mv_data = ""; + leaf = NODEPTR(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top]); + if (flags & MDB_CURRENT) { + xflags = MDB_CURRENT|MDB_NOSPILL; + } else { + mdb_xcursor_init1(mc, leaf); + xflags = (flags & MDB_NODUPDATA) ? + MDB_NOOVERWRITE|MDB_NOSPILL : MDB_NOSPILL; + } + if (sub_root) + mc->mc_xcursor->mx_cursor.mc_pg[0] = sub_root; + new_dupdata = (int)dkey.mv_size; + /* converted, write the original data first */ + if (dkey.mv_size) { + rc = mdb_cursor_put(&mc->mc_xcursor->mx_cursor, &dkey, &xdata, xflags); + if (rc) + goto bad_sub; + /* we've done our job */ + dkey.mv_size = 0; + } + if (!(leaf->mn_flags & F_SUBDATA) || sub_root) { + /* Adjust other cursors pointing to mp */ + MDB_cursor *m2; + MDB_xcursor *mx = mc->mc_xcursor; + unsigned i = mc->mc_top; + MDB_page *mp = mc->mc_pg[i]; + + for (m2 = mc->mc_txn->mt_cursors[mc->mc_dbi]; m2; m2=m2->mc_next) { + if (m2 == mc || m2->mc_snum < mc->mc_snum) continue; + if (!(m2->mc_flags & C_INITIALIZED)) continue; + if (m2->mc_pg[i] == mp) { + if (m2->mc_ki[i] == mc->mc_ki[i]) { + mdb_xcursor_init2(m2, mx, new_dupdata); + } else if (!insert_key) { + XCURSOR_REFRESH(m2, i, mp); + } + } + } + } + ecount = mc->mc_xcursor->mx_db.md_entries; + if (flags & MDB_APPENDDUP) + xflags |= MDB_APPEND; + rc = mdb_cursor_put(&mc->mc_xcursor->mx_cursor, data, &xdata, xflags); + if (flags & F_SUBDATA) { + void *db = NODEDATA(leaf); + memcpy(db, &mc->mc_xcursor->mx_db, sizeof(MDB_db)); + } + insert_data = mc->mc_xcursor->mx_db.md_entries - ecount; + } + /* Increment count unless we just replaced an existing item. */ + if (insert_data) + mc->mc_db->md_entries++; + if (insert_key) { + /* Invalidate txn if we created an empty sub-DB */ + if (rc) + goto bad_sub; + /* If we succeeded and the key didn't exist before, + * make sure the cursor is marked valid. + */ + mc->mc_flags |= C_INITIALIZED; + } + if (flags & MDB_MULTIPLE) { + if (!rc) { + mcount++; + /* let caller know how many succeeded, if any */ + data[1].mv_size = mcount; + if (mcount < dcount) { + data[0].mv_data = (char *)data[0].mv_data + data[0].mv_size; + insert_key = insert_data = 0; + goto more; + } + } + } + return rc; +bad_sub: + if (rc == MDB_KEYEXIST) /* should not happen, we deleted that item */ + rc = MDB_CORRUPTED; + } + mc->mc_txn->mt_flags |= MDB_TXN_ERROR; + return rc; +} + +int +mdb_cursor_del(MDB_cursor *mc, unsigned int flags) +{ + MDB_node *leaf; + MDB_page *mp; + int rc; + + if (mc->mc_txn->mt_flags & (MDB_TXN_RDONLY|MDB_TXN_BLOCKED)) + return (mc->mc_txn->mt_flags & MDB_TXN_RDONLY) ? EACCES : MDB_BAD_TXN; + + if (!(mc->mc_flags & C_INITIALIZED)) + return EINVAL; + + if (mc->mc_ki[mc->mc_top] >= NUMKEYS(mc->mc_pg[mc->mc_top])) + return MDB_NOTFOUND; + + if (!(flags & MDB_NOSPILL) && (rc = mdb_page_spill(mc, NULL, NULL))) + return rc; + + rc = mdb_cursor_touch(mc); + if (rc) + return rc; + + mp = mc->mc_pg[mc->mc_top]; + if (IS_LEAF2(mp)) + goto del_key; + leaf = NODEPTR(mp, mc->mc_ki[mc->mc_top]); + + if (F_ISSET(leaf->mn_flags, F_DUPDATA)) { + if (flags & MDB_NODUPDATA) { + /* mdb_cursor_del0() will subtract the final entry */ + mc->mc_db->md_entries -= mc->mc_xcursor->mx_db.md_entries - 1; + mc->mc_xcursor->mx_cursor.mc_flags &= ~C_INITIALIZED; + } else { + if (!F_ISSET(leaf->mn_flags, F_SUBDATA)) { + mc->mc_xcursor->mx_cursor.mc_pg[0] = NODEDATA(leaf); + } + rc = mdb_cursor_del(&mc->mc_xcursor->mx_cursor, MDB_NOSPILL); + if (rc) + return rc; + /* If sub-DB still has entries, we're done */ + if (mc->mc_xcursor->mx_db.md_entries) { + if (leaf->mn_flags & F_SUBDATA) { + /* update subDB info */ + void *db = NODEDATA(leaf); + memcpy(db, &mc->mc_xcursor->mx_db, sizeof(MDB_db)); + } else { + MDB_cursor *m2; + /* shrink fake page */ + mdb_node_shrink(mp, mc->mc_ki[mc->mc_top]); + leaf = NODEPTR(mp, mc->mc_ki[mc->mc_top]); + mc->mc_xcursor->mx_cursor.mc_pg[0] = NODEDATA(leaf); + /* fix other sub-DB cursors pointed at fake pages on this page */ + for (m2 = mc->mc_txn->mt_cursors[mc->mc_dbi]; m2; m2=m2->mc_next) { + if (m2 == mc || m2->mc_snum < mc->mc_snum) continue; + if (!(m2->mc_flags & C_INITIALIZED)) continue; + if (m2->mc_pg[mc->mc_top] == mp) { + XCURSOR_REFRESH(m2, mc->mc_top, mp); + } + } + } + mc->mc_db->md_entries--; + return rc; + } else { + mc->mc_xcursor->mx_cursor.mc_flags &= ~C_INITIALIZED; + } + /* otherwise fall thru and delete the sub-DB */ + } + + if (leaf->mn_flags & F_SUBDATA) { + /* add all the child DB's pages to the free list */ + rc = mdb_drop0(&mc->mc_xcursor->mx_cursor, 0); + if (rc) + goto fail; + } + } + /* LMDB passes F_SUBDATA in 'flags' to delete a DB record */ + else if ((leaf->mn_flags ^ flags) & F_SUBDATA) { + rc = MDB_INCOMPATIBLE; + goto fail; + } + + /* add overflow pages to free list */ + if (F_ISSET(leaf->mn_flags, F_BIGDATA)) { + MDB_page *omp; + pgno_t pg; + + memcpy(&pg, NODEDATA(leaf), sizeof(pg)); + if ((rc = mdb_page_get(mc, pg, &omp, NULL)) || + (rc = mdb_ovpage_free(mc, omp))) + goto fail; + } + +del_key: + return mdb_cursor_del0(mc); + +fail: + mc->mc_txn->mt_flags |= MDB_TXN_ERROR; + return rc; +} + +/** Allocate and initialize new pages for a database. + * Set #MDB_TXN_ERROR on failure. + * @param[in] mc a cursor on the database being added to. + * @param[in] flags flags defining what type of page is being allocated. + * @param[in] num the number of pages to allocate. This is usually 1, + * unless allocating overflow pages for a large record. + * @param[out] mp Address of a page, or NULL on failure. + * @return 0 on success, non-zero on failure. + */ +static int +mdb_page_new(MDB_cursor *mc, uint32_t flags, int num, MDB_page **mp) +{ + MDB_page *np; + int rc; + + if ((rc = mdb_page_alloc(mc, num, &np))) + return rc; + DPRINTF(("allocated new mpage %"Z"u, page size %u", + np->mp_pgno, mc->mc_txn->mt_env->me_psize)); + np->mp_flags = flags | P_DIRTY; + np->mp_lower = (PAGEHDRSZ-PAGEBASE); + np->mp_upper = mc->mc_txn->mt_env->me_psize - PAGEBASE; + + if (IS_BRANCH(np)) + mc->mc_db->md_branch_pages++; + else if (IS_LEAF(np)) + mc->mc_db->md_leaf_pages++; + else if (IS_OVERFLOW(np)) { + mc->mc_db->md_overflow_pages += num; + np->mp_pages = num; + } + *mp = np; + + return 0; +} + +/** Calculate the size of a leaf node. + * The size depends on the environment's page size; if a data item + * is too large it will be put onto an overflow page and the node + * size will only include the key and not the data. Sizes are always + * rounded up to an even number of bytes, to guarantee 2-byte alignment + * of the #MDB_node headers. + * @param[in] env The environment handle. + * @param[in] key The key for the node. + * @param[in] data The data for the node. + * @return The number of bytes needed to store the node. + */ +static size_t +mdb_leaf_size(MDB_env *env, MDB_val *key, MDB_val *data) +{ + size_t sz; + + sz = LEAFSIZE(key, data); + if (sz > env->me_nodemax) { + /* put on overflow page */ + sz -= data->mv_size - sizeof(pgno_t); + } + + return EVEN(sz + sizeof(indx_t)); +} + +/** Calculate the size of a branch node. + * The size should depend on the environment's page size but since + * we currently don't support spilling large keys onto overflow + * pages, it's simply the size of the #MDB_node header plus the + * size of the key. Sizes are always rounded up to an even number + * of bytes, to guarantee 2-byte alignment of the #MDB_node headers. + * @param[in] env The environment handle. + * @param[in] key The key for the node. + * @return The number of bytes needed to store the node. + */ +static size_t +mdb_branch_size(MDB_env *env, MDB_val *key) +{ + size_t sz; + + sz = INDXSIZE(key); + if (sz > env->me_nodemax) { + /* put on overflow page */ + /* not implemented */ + /* sz -= key->size - sizeof(pgno_t); */ + } + + return sz + sizeof(indx_t); +} + +/** Add a node to the page pointed to by the cursor. + * Set #MDB_TXN_ERROR on failure. + * @param[in] mc The cursor for this operation. + * @param[in] indx The index on the page where the new node should be added. + * @param[in] key The key for the new node. + * @param[in] data The data for the new node, if any. + * @param[in] pgno The page number, if adding a branch node. + * @param[in] flags Flags for the node. + * @return 0 on success, non-zero on failure. Possible errors are: + * <ul> + * <li>ENOMEM - failed to allocate overflow pages for the node. + * <li>MDB_PAGE_FULL - there is insufficient room in the page. This error + * should never happen since all callers already calculate the + * page's free space before calling this function. + * </ul> + */ +static int +mdb_node_add(MDB_cursor *mc, indx_t indx, + MDB_val *key, MDB_val *data, pgno_t pgno, unsigned int flags) +{ + unsigned int i; + size_t node_size = NODESIZE; + ssize_t room; + indx_t ofs; + MDB_node *node; + MDB_page *mp = mc->mc_pg[mc->mc_top]; + MDB_page *ofp = NULL; /* overflow page */ + void *ndata; + DKBUF; + + mdb_cassert(mc, mp->mp_upper >= mp->mp_lower); + + DPRINTF(("add to %s %spage %"Z"u index %i, data size %"Z"u key size %"Z"u [%s]", + IS_LEAF(mp) ? "leaf" : "branch", + IS_SUBP(mp) ? "sub-" : "", + mdb_dbg_pgno(mp), indx, data ? data->mv_size : 0, + key ? key->mv_size : 0, key ? DKEY(key) : "null")); + + if (IS_LEAF2(mp)) { + /* Move higher keys up one slot. */ + int ksize = mc->mc_db->md_pad, dif; + char *ptr = LEAF2KEY(mp, indx, ksize); + dif = NUMKEYS(mp) - indx; + if (dif > 0) + memmove(ptr+ksize, ptr, dif*ksize); + /* insert new key */ + memcpy(ptr, key->mv_data, ksize); + + /* Just using these for counting */ + mp->mp_lower += sizeof(indx_t); + mp->mp_upper -= ksize - sizeof(indx_t); + return MDB_SUCCESS; + } + + room = (ssize_t)SIZELEFT(mp) - (ssize_t)sizeof(indx_t); + if (key != NULL) + node_size += key->mv_size; + if (IS_LEAF(mp)) { + mdb_cassert(mc, key && data); + if (F_ISSET(flags, F_BIGDATA)) { + /* Data already on overflow page. */ + node_size += sizeof(pgno_t); + } else if (node_size + data->mv_size > mc->mc_txn->mt_env->me_nodemax) { + int ovpages = OVPAGES(data->mv_size, mc->mc_txn->mt_env->me_psize); + int rc; + /* Put data on overflow page. */ + DPRINTF(("data size is %"Z"u, node would be %"Z"u, put data on overflow page", + data->mv_size, node_size+data->mv_size)); + node_size = EVEN(node_size + sizeof(pgno_t)); + if ((ssize_t)node_size > room) + goto full; + if ((rc = mdb_page_new(mc, P_OVERFLOW, ovpages, &ofp))) + return rc; + DPRINTF(("allocated overflow page %"Z"u", ofp->mp_pgno)); + flags |= F_BIGDATA; + goto update; + } else { + node_size += data->mv_size; + } + } + node_size = EVEN(node_size); + if ((ssize_t)node_size > room) + goto full; + +update: + /* Move higher pointers up one slot. */ + for (i = NUMKEYS(mp); i > indx; i--) + mp->mp_ptrs[i] = mp->mp_ptrs[i - 1]; + + /* Adjust free space offsets. */ + ofs = mp->mp_upper - node_size; + mdb_cassert(mc, ofs >= mp->mp_lower + sizeof(indx_t)); + mp->mp_ptrs[indx] = ofs; + mp->mp_upper = ofs; + mp->mp_lower += sizeof(indx_t); + + /* Write the node data. */ + node = NODEPTR(mp, indx); + node->mn_ksize = (key == NULL) ? 0 : key->mv_size; + node->mn_flags = flags; + if (IS_LEAF(mp)) + SETDSZ(node,data->mv_size); + else + SETPGNO(node,pgno); + + if (key) + memcpy(NODEKEY(node), key->mv_data, key->mv_size); + + if (IS_LEAF(mp)) { + ndata = NODEDATA(node); + if (ofp == NULL) { + if (F_ISSET(flags, F_BIGDATA)) + memcpy(ndata, data->mv_data, sizeof(pgno_t)); + else if (F_ISSET(flags, MDB_RESERVE)) + data->mv_data = ndata; + else + memcpy(ndata, data->mv_data, data->mv_size); + } else { + memcpy(ndata, &ofp->mp_pgno, sizeof(pgno_t)); + ndata = METADATA(ofp); + if (F_ISSET(flags, MDB_RESERVE)) + data->mv_data = ndata; + else + memcpy(ndata, data->mv_data, data->mv_size); + } + } + + return MDB_SUCCESS; + +full: + DPRINTF(("not enough room in page %"Z"u, got %u ptrs", + mdb_dbg_pgno(mp), NUMKEYS(mp))); + DPRINTF(("upper-lower = %u - %u = %"Z"d", mp->mp_upper,mp->mp_lower,room)); + DPRINTF(("node size = %"Z"u", node_size)); + mc->mc_txn->mt_flags |= MDB_TXN_ERROR; + return MDB_PAGE_FULL; +} + +/** Delete the specified node from a page. + * @param[in] mc Cursor pointing to the node to delete. + * @param[in] ksize The size of a node. Only used if the page is + * part of a #MDB_DUPFIXED database. + */ +static void +mdb_node_del(MDB_cursor *mc, int ksize) +{ + MDB_page *mp = mc->mc_pg[mc->mc_top]; + indx_t indx = mc->mc_ki[mc->mc_top]; + unsigned int sz; + indx_t i, j, numkeys, ptr; + MDB_node *node; + char *base; + + DPRINTF(("delete node %u on %s page %"Z"u", indx, + IS_LEAF(mp) ? "leaf" : "branch", mdb_dbg_pgno(mp))); + numkeys = NUMKEYS(mp); + mdb_cassert(mc, indx < numkeys); + + if (IS_LEAF2(mp)) { + int x = numkeys - 1 - indx; + base = LEAF2KEY(mp, indx, ksize); + if (x) + memmove(base, base + ksize, x * ksize); + mp->mp_lower -= sizeof(indx_t); + mp->mp_upper += ksize - sizeof(indx_t); + return; + } + + node = NODEPTR(mp, indx); + sz = NODESIZE + node->mn_ksize; + if (IS_LEAF(mp)) { + if (F_ISSET(node->mn_flags, F_BIGDATA)) + sz += sizeof(pgno_t); + else + sz += NODEDSZ(node); + } + sz = EVEN(sz); + + ptr = mp->mp_ptrs[indx]; + for (i = j = 0; i < numkeys; i++) { + if (i != indx) { + mp->mp_ptrs[j] = mp->mp_ptrs[i]; + if (mp->mp_ptrs[i] < ptr) + mp->mp_ptrs[j] += sz; + j++; + } + } + + base = (char *)mp + mp->mp_upper + PAGEBASE; + memmove(base + sz, base, ptr - mp->mp_upper); + + mp->mp_lower -= sizeof(indx_t); + mp->mp_upper += sz; +} + +/** Compact the main page after deleting a node on a subpage. + * @param[in] mp The main page to operate on. + * @param[in] indx The index of the subpage on the main page. + */ +static void +mdb_node_shrink(MDB_page *mp, indx_t indx) +{ + MDB_node *node; + MDB_page *sp, *xp; + char *base; + indx_t delta, nsize, len, ptr; + int i; + + node = NODEPTR(mp, indx); + sp = (MDB_page *)NODEDATA(node); + delta = SIZELEFT(sp); + nsize = NODEDSZ(node) - delta; + + /* Prepare to shift upward, set len = length(subpage part to shift) */ + if (IS_LEAF2(sp)) { + len = nsize; + if (nsize & 1) + return; /* do not make the node uneven-sized */ + } else { + xp = (MDB_page *)((char *)sp + delta); /* destination subpage */ + for (i = NUMKEYS(sp); --i >= 0; ) + xp->mp_ptrs[i] = sp->mp_ptrs[i] - delta; + len = PAGEHDRSZ; + } + sp->mp_upper = sp->mp_lower; + COPY_PGNO(sp->mp_pgno, mp->mp_pgno); + SETDSZ(node, nsize); + + /* Shift <lower nodes...initial part of subpage> upward */ + base = (char *)mp + mp->mp_upper + PAGEBASE; + memmove(base + delta, base, (char *)sp + len - base); + + ptr = mp->mp_ptrs[indx]; + for (i = NUMKEYS(mp); --i >= 0; ) { + if (mp->mp_ptrs[i] <= ptr) + mp->mp_ptrs[i] += delta; + } + mp->mp_upper += delta; +} + +/** Initial setup of a sorted-dups cursor. + * Sorted duplicates are implemented as a sub-database for the given key. + * The duplicate data items are actually keys of the sub-database. + * Operations on the duplicate data items are performed using a sub-cursor + * initialized when the sub-database is first accessed. This function does + * the preliminary setup of the sub-cursor, filling in the fields that + * depend only on the parent DB. + * @param[in] mc The main cursor whose sorted-dups cursor is to be initialized. + */ +static void +mdb_xcursor_init0(MDB_cursor *mc) +{ + MDB_xcursor *mx = mc->mc_xcursor; + + mx->mx_cursor.mc_xcursor = NULL; + mx->mx_cursor.mc_txn = mc->mc_txn; + mx->mx_cursor.mc_db = &mx->mx_db; + mx->mx_cursor.mc_dbx = &mx->mx_dbx; + mx->mx_cursor.mc_dbi = mc->mc_dbi; + mx->mx_cursor.mc_dbflag = &mx->mx_dbflag; + mx->mx_cursor.mc_snum = 0; + mx->mx_cursor.mc_top = 0; + mx->mx_cursor.mc_flags = C_SUB; + mx->mx_dbx.md_name.mv_size = 0; + mx->mx_dbx.md_name.mv_data = NULL; + mx->mx_dbx.md_cmp = mc->mc_dbx->md_dcmp; + mx->mx_dbx.md_dcmp = NULL; + mx->mx_dbx.md_rel = mc->mc_dbx->md_rel; +} + +/** Final setup of a sorted-dups cursor. + * Sets up the fields that depend on the data from the main cursor. + * @param[in] mc The main cursor whose sorted-dups cursor is to be initialized. + * @param[in] node The data containing the #MDB_db record for the + * sorted-dup database. + */ +static void +mdb_xcursor_init1(MDB_cursor *mc, MDB_node *node) +{ + MDB_xcursor *mx = mc->mc_xcursor; + + if (node->mn_flags & F_SUBDATA) { + memcpy(&mx->mx_db, NODEDATA(node), sizeof(MDB_db)); + mx->mx_cursor.mc_pg[0] = 0; + mx->mx_cursor.mc_snum = 0; + mx->mx_cursor.mc_top = 0; + mx->mx_cursor.mc_flags = C_SUB; + } else { + MDB_page *fp = NODEDATA(node); + mx->mx_db.md_pad = 0; + mx->mx_db.md_flags = 0; + mx->mx_db.md_depth = 1; + mx->mx_db.md_branch_pages = 0; + mx->mx_db.md_leaf_pages = 1; + mx->mx_db.md_overflow_pages = 0; + mx->mx_db.md_entries = NUMKEYS(fp); + COPY_PGNO(mx->mx_db.md_root, fp->mp_pgno); + mx->mx_cursor.mc_snum = 1; + mx->mx_cursor.mc_top = 0; + mx->mx_cursor.mc_flags = C_INITIALIZED|C_SUB; + mx->mx_cursor.mc_pg[0] = fp; + mx->mx_cursor.mc_ki[0] = 0; + if (mc->mc_db->md_flags & MDB_DUPFIXED) { + mx->mx_db.md_flags = MDB_DUPFIXED; + mx->mx_db.md_pad = fp->mp_pad; + if (mc->mc_db->md_flags & MDB_INTEGERDUP) + mx->mx_db.md_flags |= MDB_INTEGERKEY; + } + } + DPRINTF(("Sub-db -%u root page %"Z"u", mx->mx_cursor.mc_dbi, + mx->mx_db.md_root)); + mx->mx_dbflag = DB_VALID|DB_USRVALID|DB_DUPDATA; +#if UINT_MAX < SIZE_MAX + if (mx->mx_dbx.md_cmp == mdb_cmp_int && mx->mx_db.md_pad == sizeof(size_t)) + mx->mx_dbx.md_cmp = mdb_cmp_clong; +#endif +} + + +/** Fixup a sorted-dups cursor due to underlying update. + * Sets up some fields that depend on the data from the main cursor. + * Almost the same as init1, but skips initialization steps if the + * xcursor had already been used. + * @param[in] mc The main cursor whose sorted-dups cursor is to be fixed up. + * @param[in] src_mx The xcursor of an up-to-date cursor. + * @param[in] new_dupdata True if converting from a non-#F_DUPDATA item. + */ +static void +mdb_xcursor_init2(MDB_cursor *mc, MDB_xcursor *src_mx, int new_dupdata) +{ + MDB_xcursor *mx = mc->mc_xcursor; + + if (new_dupdata) { + mx->mx_cursor.mc_snum = 1; + mx->mx_cursor.mc_top = 0; + mx->mx_cursor.mc_flags |= C_INITIALIZED; + mx->mx_cursor.mc_ki[0] = 0; + mx->mx_dbflag = DB_VALID|DB_USRVALID|DB_DUPDATA; +#if UINT_MAX < SIZE_MAX + mx->mx_dbx.md_cmp = src_mx->mx_dbx.md_cmp; +#endif + } else if (!(mx->mx_cursor.mc_flags & C_INITIALIZED)) { + return; + } + mx->mx_db = src_mx->mx_db; + mx->mx_cursor.mc_pg[0] = src_mx->mx_cursor.mc_pg[0]; + DPRINTF(("Sub-db -%u root page %"Z"u", mx->mx_cursor.mc_dbi, + mx->mx_db.md_root)); +} + +/** Initialize a cursor for a given transaction and database. */ +static void +mdb_cursor_init(MDB_cursor *mc, MDB_txn *txn, MDB_dbi dbi, MDB_xcursor *mx) +{ + mc->mc_next = NULL; + mc->mc_backup = NULL; + mc->mc_dbi = dbi; + mc->mc_txn = txn; + mc->mc_db = &txn->mt_dbs[dbi]; + mc->mc_dbx = &txn->mt_dbxs[dbi]; + mc->mc_dbflag = &txn->mt_dbflags[dbi]; + mc->mc_snum = 0; + mc->mc_top = 0; + mc->mc_pg[0] = 0; + mc->mc_ki[0] = 0; + mc->mc_flags = 0; + if (txn->mt_dbs[dbi].md_flags & MDB_DUPSORT) { + mdb_tassert(txn, mx != NULL); + mc->mc_xcursor = mx; + mdb_xcursor_init0(mc); + } else { + mc->mc_xcursor = NULL; + } + if (*mc->mc_dbflag & DB_STALE) { + mdb_page_search(mc, NULL, MDB_PS_ROOTONLY); + } +} + +int +mdb_cursor_open(MDB_txn *txn, MDB_dbi dbi, MDB_cursor **ret) +{ + MDB_cursor *mc; + size_t size = sizeof(MDB_cursor); + + if (!ret || !TXN_DBI_EXIST(txn, dbi, DB_VALID)) + return EINVAL; + + if (txn->mt_flags & MDB_TXN_BLOCKED) + return MDB_BAD_TXN; + + if (dbi == FREE_DBI && !F_ISSET(txn->mt_flags, MDB_TXN_RDONLY)) + return EINVAL; + + if (txn->mt_dbs[dbi].md_flags & MDB_DUPSORT) + size += sizeof(MDB_xcursor); + + if ((mc = malloc(size)) != NULL) { + mdb_cursor_init(mc, txn, dbi, (MDB_xcursor *)(mc + 1)); + if (txn->mt_cursors) { + mc->mc_next = txn->mt_cursors[dbi]; + txn->mt_cursors[dbi] = mc; + mc->mc_flags |= C_UNTRACK; + } + } else { + return ENOMEM; + } + + *ret = mc; + + return MDB_SUCCESS; +} + +int +mdb_cursor_renew(MDB_txn *txn, MDB_cursor *mc) +{ + if (!mc || !TXN_DBI_EXIST(txn, mc->mc_dbi, DB_VALID)) + return EINVAL; + + if ((mc->mc_flags & C_UNTRACK) || txn->mt_cursors) + return EINVAL; + + if (txn->mt_flags & MDB_TXN_BLOCKED) + return MDB_BAD_TXN; + + mdb_cursor_init(mc, txn, mc->mc_dbi, mc->mc_xcursor); + return MDB_SUCCESS; +} + +/* Return the count of duplicate data items for the current key */ +int +mdb_cursor_count(MDB_cursor *mc, size_t *countp) +{ + MDB_node *leaf; + + if (mc == NULL || countp == NULL) + return EINVAL; + + if (mc->mc_xcursor == NULL) + return MDB_INCOMPATIBLE; + + if (mc->mc_txn->mt_flags & MDB_TXN_BLOCKED) + return MDB_BAD_TXN; + + if (!(mc->mc_flags & C_INITIALIZED)) + return EINVAL; + + if (!mc->mc_snum) + return MDB_NOTFOUND; + + if (mc->mc_flags & C_EOF) { + if (mc->mc_ki[mc->mc_top] >= NUMKEYS(mc->mc_pg[mc->mc_top])) + return MDB_NOTFOUND; + mc->mc_flags ^= C_EOF; + } + + leaf = NODEPTR(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top]); + if (!F_ISSET(leaf->mn_flags, F_DUPDATA)) { + *countp = 1; + } else { + if (!(mc->mc_xcursor->mx_cursor.mc_flags & C_INITIALIZED)) + return EINVAL; + + *countp = mc->mc_xcursor->mx_db.md_entries; + } + return MDB_SUCCESS; +} + +void +mdb_cursor_close(MDB_cursor *mc) +{ + if (mc && !mc->mc_backup) { + /* remove from txn, if tracked */ + if ((mc->mc_flags & C_UNTRACK) && mc->mc_txn->mt_cursors) { + MDB_cursor **prev = &mc->mc_txn->mt_cursors[mc->mc_dbi]; + while (*prev && *prev != mc) prev = &(*prev)->mc_next; + if (*prev == mc) + *prev = mc->mc_next; + } + free(mc); + } +} + +MDB_txn * +mdb_cursor_txn(MDB_cursor *mc) +{ + if (!mc) return NULL; + return mc->mc_txn; +} + +MDB_dbi +mdb_cursor_dbi(MDB_cursor *mc) +{ + return mc->mc_dbi; +} + +/** Replace the key for a branch node with a new key. + * Set #MDB_TXN_ERROR on failure. + * @param[in] mc Cursor pointing to the node to operate on. + * @param[in] key The new key to use. + * @return 0 on success, non-zero on failure. + */ +static int +mdb_update_key(MDB_cursor *mc, MDB_val *key) +{ + MDB_page *mp; + MDB_node *node; + char *base; + size_t len; + int delta, ksize, oksize; + indx_t ptr, i, numkeys, indx; + DKBUF; + + indx = mc->mc_ki[mc->mc_top]; + mp = mc->mc_pg[mc->mc_top]; + node = NODEPTR(mp, indx); + ptr = mp->mp_ptrs[indx]; +#if MDB_DEBUG + { + MDB_val k2; + char kbuf2[DKBUF_MAXKEYSIZE*2+1]; + k2.mv_data = NODEKEY(node); + k2.mv_size = node->mn_ksize; + DPRINTF(("update key %u (ofs %u) [%s] to [%s] on page %"Z"u", + indx, ptr, + mdb_dkey(&k2, kbuf2), + DKEY(key), + mp->mp_pgno)); + } +#endif + + /* Sizes must be 2-byte aligned. */ + ksize = EVEN(key->mv_size); + oksize = EVEN(node->mn_ksize); + delta = ksize - oksize; + + /* Shift node contents if EVEN(key length) changed. */ + if (delta) { + if (delta > 0 && SIZELEFT(mp) < delta) { + pgno_t pgno; + /* not enough space left, do a delete and split */ + DPRINTF(("Not enough room, delta = %d, splitting...", delta)); + pgno = NODEPGNO(node); + mdb_node_del(mc, 0); + return mdb_page_split(mc, key, NULL, pgno, MDB_SPLIT_REPLACE); + } + + numkeys = NUMKEYS(mp); + for (i = 0; i < numkeys; i++) { + if (mp->mp_ptrs[i] <= ptr) + mp->mp_ptrs[i] -= delta; + } + + base = (char *)mp + mp->mp_upper + PAGEBASE; + len = ptr - mp->mp_upper + NODESIZE; + memmove(base - delta, base, len); + mp->mp_upper -= delta; + + node = NODEPTR(mp, indx); + } + + /* But even if no shift was needed, update ksize */ + if (node->mn_ksize != key->mv_size) + node->mn_ksize = key->mv_size; + + if (key->mv_size) + memcpy(NODEKEY(node), key->mv_data, key->mv_size); + + return MDB_SUCCESS; +} + +static void +mdb_cursor_copy(const MDB_cursor *csrc, MDB_cursor *cdst); + +/** Perform \b act while tracking temporary cursor \b mn */ +#define WITH_CURSOR_TRACKING(mn, act) do { \ + MDB_cursor dummy, *tracked, **tp = &(mn).mc_txn->mt_cursors[mn.mc_dbi]; \ + if ((mn).mc_flags & C_SUB) { \ + dummy.mc_flags = C_INITIALIZED; \ + dummy.mc_xcursor = (MDB_xcursor *)&(mn); \ + tracked = &dummy; \ + } else { \ + tracked = &(mn); \ + } \ + tracked->mc_next = *tp; \ + *tp = tracked; \ + { act; } \ + *tp = tracked->mc_next; \ +} while (0) + +/** Move a node from csrc to cdst. + */ +static int +mdb_node_move(MDB_cursor *csrc, MDB_cursor *cdst, int fromleft) +{ + MDB_node *srcnode; + MDB_val key, data; + pgno_t srcpg; + MDB_cursor mn; + int rc; + unsigned short flags; + + DKBUF; + + /* Mark src and dst as dirty. */ + if ((rc = mdb_page_touch(csrc)) || + (rc = mdb_page_touch(cdst))) + return rc; + + if (IS_LEAF2(csrc->mc_pg[csrc->mc_top])) { + key.mv_size = csrc->mc_db->md_pad; + key.mv_data = LEAF2KEY(csrc->mc_pg[csrc->mc_top], csrc->mc_ki[csrc->mc_top], key.mv_size); + data.mv_size = 0; + data.mv_data = NULL; + srcpg = 0; + flags = 0; + } else { + srcnode = NODEPTR(csrc->mc_pg[csrc->mc_top], csrc->mc_ki[csrc->mc_top]); + mdb_cassert(csrc, !((size_t)srcnode & 1)); + srcpg = NODEPGNO(srcnode); + flags = srcnode->mn_flags; + if (csrc->mc_ki[csrc->mc_top] == 0 && IS_BRANCH(csrc->mc_pg[csrc->mc_top])) { + unsigned int snum = csrc->mc_snum; + MDB_node *s2; + /* must find the lowest key below src */ + rc = mdb_page_search_lowest(csrc); + if (rc) + return rc; + if (IS_LEAF2(csrc->mc_pg[csrc->mc_top])) { + key.mv_size = csrc->mc_db->md_pad; + key.mv_data = LEAF2KEY(csrc->mc_pg[csrc->mc_top], 0, key.mv_size); + } else { + s2 = NODEPTR(csrc->mc_pg[csrc->mc_top], 0); + key.mv_size = NODEKSZ(s2); + key.mv_data = NODEKEY(s2); + } + csrc->mc_snum = snum--; + csrc->mc_top = snum; + } else { + key.mv_size = NODEKSZ(srcnode); + key.mv_data = NODEKEY(srcnode); + } + data.mv_size = NODEDSZ(srcnode); + data.mv_data = NODEDATA(srcnode); + } + mn.mc_xcursor = NULL; + if (IS_BRANCH(cdst->mc_pg[cdst->mc_top]) && cdst->mc_ki[cdst->mc_top] == 0) { + unsigned int snum = cdst->mc_snum; + MDB_node *s2; + MDB_val bkey; + /* must find the lowest key below dst */ + mdb_cursor_copy(cdst, &mn); + rc = mdb_page_search_lowest(&mn); + if (rc) + return rc; + if (IS_LEAF2(mn.mc_pg[mn.mc_top])) { + bkey.mv_size = mn.mc_db->md_pad; + bkey.mv_data = LEAF2KEY(mn.mc_pg[mn.mc_top], 0, bkey.mv_size); + } else { + s2 = NODEPTR(mn.mc_pg[mn.mc_top], 0); + bkey.mv_size = NODEKSZ(s2); + bkey.mv_data = NODEKEY(s2); + } + mn.mc_snum = snum--; + mn.mc_top = snum; + mn.mc_ki[snum] = 0; + rc = mdb_update_key(&mn, &bkey); + if (rc) + return rc; + } + + DPRINTF(("moving %s node %u [%s] on page %"Z"u to node %u on page %"Z"u", + IS_LEAF(csrc->mc_pg[csrc->mc_top]) ? "leaf" : "branch", + csrc->mc_ki[csrc->mc_top], + DKEY(&key), + csrc->mc_pg[csrc->mc_top]->mp_pgno, + cdst->mc_ki[cdst->mc_top], cdst->mc_pg[cdst->mc_top]->mp_pgno)); + + /* Add the node to the destination page. + */ + rc = mdb_node_add(cdst, cdst->mc_ki[cdst->mc_top], &key, &data, srcpg, flags); + if (rc != MDB_SUCCESS) + return rc; + + /* Delete the node from the source page. + */ + mdb_node_del(csrc, key.mv_size); + + { + /* Adjust other cursors pointing to mp */ + MDB_cursor *m2, *m3; + MDB_dbi dbi = csrc->mc_dbi; + MDB_page *mpd, *mps; + + mps = csrc->mc_pg[csrc->mc_top]; + /* If we're adding on the left, bump others up */ + if (fromleft) { + mpd = cdst->mc_pg[csrc->mc_top]; + for (m2 = csrc->mc_txn->mt_cursors[dbi]; m2; m2=m2->mc_next) { + if (csrc->mc_flags & C_SUB) + m3 = &m2->mc_xcursor->mx_cursor; + else + m3 = m2; + if (!(m3->mc_flags & C_INITIALIZED) || m3->mc_top < csrc->mc_top) + continue; + if (m3 != cdst && + m3->mc_pg[csrc->mc_top] == mpd && + m3->mc_ki[csrc->mc_top] >= cdst->mc_ki[csrc->mc_top]) { + m3->mc_ki[csrc->mc_top]++; + } + if (m3 !=csrc && + m3->mc_pg[csrc->mc_top] == mps && + m3->mc_ki[csrc->mc_top] == csrc->mc_ki[csrc->mc_top]) { + m3->mc_pg[csrc->mc_top] = cdst->mc_pg[cdst->mc_top]; + m3->mc_ki[csrc->mc_top] = cdst->mc_ki[cdst->mc_top]; + m3->mc_ki[csrc->mc_top-1]++; + } + if (IS_LEAF(mps)) + XCURSOR_REFRESH(m3, csrc->mc_top, m3->mc_pg[csrc->mc_top]); + } + } else + /* Adding on the right, bump others down */ + { + for (m2 = csrc->mc_txn->mt_cursors[dbi]; m2; m2=m2->mc_next) { + if (csrc->mc_flags & C_SUB) + m3 = &m2->mc_xcursor->mx_cursor; + else + m3 = m2; + if (m3 == csrc) continue; + if (!(m3->mc_flags & C_INITIALIZED) || m3->mc_top < csrc->mc_top) + continue; + if (m3->mc_pg[csrc->mc_top] == mps) { + if (!m3->mc_ki[csrc->mc_top]) { + m3->mc_pg[csrc->mc_top] = cdst->mc_pg[cdst->mc_top]; + m3->mc_ki[csrc->mc_top] = cdst->mc_ki[cdst->mc_top]; + m3->mc_ki[csrc->mc_top-1]--; + } else { + m3->mc_ki[csrc->mc_top]--; + } + if (IS_LEAF(mps)) + XCURSOR_REFRESH(m3, csrc->mc_top, m3->mc_pg[csrc->mc_top]); + } + } + } + } + + /* Update the parent separators. + */ + if (csrc->mc_ki[csrc->mc_top] == 0) { + if (csrc->mc_ki[csrc->mc_top-1] != 0) { + if (IS_LEAF2(csrc->mc_pg[csrc->mc_top])) { + key.mv_data = LEAF2KEY(csrc->mc_pg[csrc->mc_top], 0, key.mv_size); + } else { + srcnode = NODEPTR(csrc->mc_pg[csrc->mc_top], 0); + key.mv_size = NODEKSZ(srcnode); + key.mv_data = NODEKEY(srcnode); + } + DPRINTF(("update separator for source page %"Z"u to [%s]", + csrc->mc_pg[csrc->mc_top]->mp_pgno, DKEY(&key))); + mdb_cursor_copy(csrc, &mn); + mn.mc_snum--; + mn.mc_top--; + /* We want mdb_rebalance to find mn when doing fixups */ + WITH_CURSOR_TRACKING(mn, + rc = mdb_update_key(&mn, &key)); + if (rc) + return rc; + } + if (IS_BRANCH(csrc->mc_pg[csrc->mc_top])) { + MDB_val nullkey; + indx_t ix = csrc->mc_ki[csrc->mc_top]; + nullkey.mv_size = 0; + csrc->mc_ki[csrc->mc_top] = 0; + rc = mdb_update_key(csrc, &nullkey); + csrc->mc_ki[csrc->mc_top] = ix; + mdb_cassert(csrc, rc == MDB_SUCCESS); + } + } + + if (cdst->mc_ki[cdst->mc_top] == 0) { + if (cdst->mc_ki[cdst->mc_top-1] != 0) { + if (IS_LEAF2(csrc->mc_pg[csrc->mc_top])) { + key.mv_data = LEAF2KEY(cdst->mc_pg[cdst->mc_top], 0, key.mv_size); + } else { + srcnode = NODEPTR(cdst->mc_pg[cdst->mc_top], 0); + key.mv_size = NODEKSZ(srcnode); + key.mv_data = NODEKEY(srcnode); + } + DPRINTF(("update separator for destination page %"Z"u to [%s]", + cdst->mc_pg[cdst->mc_top]->mp_pgno, DKEY(&key))); + mdb_cursor_copy(cdst, &mn); + mn.mc_snum--; + mn.mc_top--; + /* We want mdb_rebalance to find mn when doing fixups */ + WITH_CURSOR_TRACKING(mn, + rc = mdb_update_key(&mn, &key)); + if (rc) + return rc; + } + if (IS_BRANCH(cdst->mc_pg[cdst->mc_top])) { + MDB_val nullkey; + indx_t ix = cdst->mc_ki[cdst->mc_top]; + nullkey.mv_size = 0; + cdst->mc_ki[cdst->mc_top] = 0; + rc = mdb_update_key(cdst, &nullkey); + cdst->mc_ki[cdst->mc_top] = ix; + mdb_cassert(cdst, rc == MDB_SUCCESS); + } + } + + return MDB_SUCCESS; +} + +/** Merge one page into another. + * The nodes from the page pointed to by \b csrc will + * be copied to the page pointed to by \b cdst and then + * the \b csrc page will be freed. + * @param[in] csrc Cursor pointing to the source page. + * @param[in] cdst Cursor pointing to the destination page. + * @return 0 on success, non-zero on failure. + */ +static int +mdb_page_merge(MDB_cursor *csrc, MDB_cursor *cdst) +{ + MDB_page *psrc, *pdst; + MDB_node *srcnode; + MDB_val key, data; + unsigned nkeys; + int rc; + indx_t i, j; + + psrc = csrc->mc_pg[csrc->mc_top]; + pdst = cdst->mc_pg[cdst->mc_top]; + + DPRINTF(("merging page %"Z"u into %"Z"u", psrc->mp_pgno, pdst->mp_pgno)); + + mdb_cassert(csrc, csrc->mc_snum > 1); /* can't merge root page */ + mdb_cassert(csrc, cdst->mc_snum > 1); + + /* Mark dst as dirty. */ + if ((rc = mdb_page_touch(cdst))) + return rc; + + /* get dst page again now that we've touched it. */ + pdst = cdst->mc_pg[cdst->mc_top]; + + /* Move all nodes from src to dst. + */ + j = nkeys = NUMKEYS(pdst); + if (IS_LEAF2(psrc)) { + key.mv_size = csrc->mc_db->md_pad; + key.mv_data = METADATA(psrc); + for (i = 0; i < NUMKEYS(psrc); i++, j++) { + rc = mdb_node_add(cdst, j, &key, NULL, 0, 0); + if (rc != MDB_SUCCESS) + return rc; + key.mv_data = (char *)key.mv_data + key.mv_size; + } + } else { + for (i = 0; i < NUMKEYS(psrc); i++, j++) { + srcnode = NODEPTR(psrc, i); + if (i == 0 && IS_BRANCH(psrc)) { + MDB_cursor mn; + MDB_node *s2; + mdb_cursor_copy(csrc, &mn); + mn.mc_xcursor = NULL; + /* must find the lowest key below src */ + rc = mdb_page_search_lowest(&mn); + if (rc) + return rc; + if (IS_LEAF2(mn.mc_pg[mn.mc_top])) { + key.mv_size = mn.mc_db->md_pad; + key.mv_data = LEAF2KEY(mn.mc_pg[mn.mc_top], 0, key.mv_size); + } else { + s2 = NODEPTR(mn.mc_pg[mn.mc_top], 0); + key.mv_size = NODEKSZ(s2); + key.mv_data = NODEKEY(s2); + } + } else { + key.mv_size = srcnode->mn_ksize; + key.mv_data = NODEKEY(srcnode); + } + + data.mv_size = NODEDSZ(srcnode); + data.mv_data = NODEDATA(srcnode); + rc = mdb_node_add(cdst, j, &key, &data, NODEPGNO(srcnode), srcnode->mn_flags); + if (rc != MDB_SUCCESS) + return rc; + } + } + + DPRINTF(("dst page %"Z"u now has %u keys (%.1f%% filled)", + pdst->mp_pgno, NUMKEYS(pdst), + (float)PAGEFILL(cdst->mc_txn->mt_env, pdst) / 10)); + + /* Unlink the src page from parent and add to free list. + */ + csrc->mc_top--; + mdb_node_del(csrc, 0); + if (csrc->mc_ki[csrc->mc_top] == 0) { + key.mv_size = 0; + rc = mdb_update_key(csrc, &key); + if (rc) { + csrc->mc_top++; + return rc; + } + } + csrc->mc_top++; + + psrc = csrc->mc_pg[csrc->mc_top]; + /* If not operating on FreeDB, allow this page to be reused + * in this txn. Otherwise just add to free list. + */ + rc = mdb_page_loose(csrc, psrc); + if (rc) + return rc; + if (IS_LEAF(psrc)) + csrc->mc_db->md_leaf_pages--; + else + csrc->mc_db->md_branch_pages--; + { + /* Adjust other cursors pointing to mp */ + MDB_cursor *m2, *m3; + MDB_dbi dbi = csrc->mc_dbi; + unsigned int top = csrc->mc_top; + + for (m2 = csrc->mc_txn->mt_cursors[dbi]; m2; m2=m2->mc_next) { + if (csrc->mc_flags & C_SUB) + m3 = &m2->mc_xcursor->mx_cursor; + else + m3 = m2; + if (m3 == csrc) continue; + if (m3->mc_snum < csrc->mc_snum) continue; + if (m3->mc_pg[top] == psrc) { + m3->mc_pg[top] = pdst; + m3->mc_ki[top] += nkeys; + m3->mc_ki[top-1] = cdst->mc_ki[top-1]; + } else if (m3->mc_pg[top-1] == csrc->mc_pg[top-1] && + m3->mc_ki[top-1] > csrc->mc_ki[top-1]) { + m3->mc_ki[top-1]--; + } + if (IS_LEAF(psrc)) + XCURSOR_REFRESH(m3, top, m3->mc_pg[top]); + } + } + { + unsigned int snum = cdst->mc_snum; + uint16_t depth = cdst->mc_db->md_depth; + mdb_cursor_pop(cdst); + rc = mdb_rebalance(cdst); + /* Did the tree height change? */ + if (depth != cdst->mc_db->md_depth) + snum += cdst->mc_db->md_depth - depth; + cdst->mc_snum = snum; + cdst->mc_top = snum-1; + } + return rc; +} + +/** Copy the contents of a cursor. + * @param[in] csrc The cursor to copy from. + * @param[out] cdst The cursor to copy to. + */ +static void +mdb_cursor_copy(const MDB_cursor *csrc, MDB_cursor *cdst) +{ + unsigned int i; + + cdst->mc_txn = csrc->mc_txn; + cdst->mc_dbi = csrc->mc_dbi; + cdst->mc_db = csrc->mc_db; + cdst->mc_dbx = csrc->mc_dbx; + cdst->mc_snum = csrc->mc_snum; + cdst->mc_top = csrc->mc_top; + cdst->mc_flags = csrc->mc_flags; + + for (i=0; i<csrc->mc_snum; i++) { + cdst->mc_pg[i] = csrc->mc_pg[i]; + cdst->mc_ki[i] = csrc->mc_ki[i]; + } +} + +/** Rebalance the tree after a delete operation. + * @param[in] mc Cursor pointing to the page where rebalancing + * should begin. + * @return 0 on success, non-zero on failure. + */ +static int +mdb_rebalance(MDB_cursor *mc) +{ + MDB_node *node; + int rc, fromleft; + unsigned int ptop, minkeys, thresh; + MDB_cursor mn; + indx_t oldki; + + if (IS_BRANCH(mc->mc_pg[mc->mc_top])) { + minkeys = 2; + thresh = 1; + } else { + minkeys = 1; + thresh = FILL_THRESHOLD; + } + DPRINTF(("rebalancing %s page %"Z"u (has %u keys, %.1f%% full)", + IS_LEAF(mc->mc_pg[mc->mc_top]) ? "leaf" : "branch", + mdb_dbg_pgno(mc->mc_pg[mc->mc_top]), NUMKEYS(mc->mc_pg[mc->mc_top]), + (float)PAGEFILL(mc->mc_txn->mt_env, mc->mc_pg[mc->mc_top]) / 10)); + + if (PAGEFILL(mc->mc_txn->mt_env, mc->mc_pg[mc->mc_top]) >= thresh && + NUMKEYS(mc->mc_pg[mc->mc_top]) >= minkeys) { + DPRINTF(("no need to rebalance page %"Z"u, above fill threshold", + mdb_dbg_pgno(mc->mc_pg[mc->mc_top]))); + return MDB_SUCCESS; + } + + if (mc->mc_snum < 2) { + MDB_page *mp = mc->mc_pg[0]; + if (IS_SUBP(mp)) { + DPUTS("Can't rebalance a subpage, ignoring"); + return MDB_SUCCESS; + } + if (NUMKEYS(mp) == 0) { + DPUTS("tree is completely empty"); + mc->mc_db->md_root = P_INVALID; + mc->mc_db->md_depth = 0; + mc->mc_db->md_leaf_pages = 0; + rc = mdb_midl_append(&mc->mc_txn->mt_free_pgs, mp->mp_pgno); + if (rc) + return rc; + /* Adjust cursors pointing to mp */ + mc->mc_snum = 0; + mc->mc_top = 0; + mc->mc_flags &= ~C_INITIALIZED; + { + MDB_cursor *m2, *m3; + MDB_dbi dbi = mc->mc_dbi; + + for (m2 = mc->mc_txn->mt_cursors[dbi]; m2; m2=m2->mc_next) { + if (mc->mc_flags & C_SUB) + m3 = &m2->mc_xcursor->mx_cursor; + else + m3 = m2; + if (!(m3->mc_flags & C_INITIALIZED) || (m3->mc_snum < mc->mc_snum)) + continue; + if (m3->mc_pg[0] == mp) { + m3->mc_snum = 0; + m3->mc_top = 0; + m3->mc_flags &= ~C_INITIALIZED; + } + } + } + } else if (IS_BRANCH(mp) && NUMKEYS(mp) == 1) { + int i; + DPUTS("collapsing root page!"); + rc = mdb_midl_append(&mc->mc_txn->mt_free_pgs, mp->mp_pgno); + if (rc) + return rc; + mc->mc_db->md_root = NODEPGNO(NODEPTR(mp, 0)); + rc = mdb_page_get(mc, mc->mc_db->md_root, &mc->mc_pg[0], NULL); + if (rc) + return rc; + mc->mc_db->md_depth--; + mc->mc_db->md_branch_pages--; + mc->mc_ki[0] = mc->mc_ki[1]; + for (i = 1; i<mc->mc_db->md_depth; i++) { + mc->mc_pg[i] = mc->mc_pg[i+1]; + mc->mc_ki[i] = mc->mc_ki[i+1]; + } + { + /* Adjust other cursors pointing to mp */ + MDB_cursor *m2, *m3; + MDB_dbi dbi = mc->mc_dbi; + + for (m2 = mc->mc_txn->mt_cursors[dbi]; m2; m2=m2->mc_next) { + if (mc->mc_flags & C_SUB) + m3 = &m2->mc_xcursor->mx_cursor; + else + m3 = m2; + if (m3 == mc) continue; + if (!(m3->mc_flags & C_INITIALIZED)) + continue; + if (m3->mc_pg[0] == mp) { + for (i=0; i<mc->mc_db->md_depth; i++) { + m3->mc_pg[i] = m3->mc_pg[i+1]; + m3->mc_ki[i] = m3->mc_ki[i+1]; + } + m3->mc_snum--; + m3->mc_top--; + } + } + } + } else + DPUTS("root page doesn't need rebalancing"); + return MDB_SUCCESS; + } + + /* The parent (branch page) must have at least 2 pointers, + * otherwise the tree is invalid. + */ + ptop = mc->mc_top-1; + mdb_cassert(mc, NUMKEYS(mc->mc_pg[ptop]) > 1); + + /* Leaf page fill factor is below the threshold. + * Try to move keys from left or right neighbor, or + * merge with a neighbor page. + */ + + /* Find neighbors. + */ + mdb_cursor_copy(mc, &mn); + mn.mc_xcursor = NULL; + + oldki = mc->mc_ki[mc->mc_top]; + if (mc->mc_ki[ptop] == 0) { + /* We're the leftmost leaf in our parent. + */ + DPUTS("reading right neighbor"); + mn.mc_ki[ptop]++; + node = NODEPTR(mc->mc_pg[ptop], mn.mc_ki[ptop]); + rc = mdb_page_get(mc, NODEPGNO(node), &mn.mc_pg[mn.mc_top], NULL); + if (rc) + return rc; + mn.mc_ki[mn.mc_top] = 0; + mc->mc_ki[mc->mc_top] = NUMKEYS(mc->mc_pg[mc->mc_top]); + fromleft = 0; + } else { + /* There is at least one neighbor to the left. + */ + DPUTS("reading left neighbor"); + mn.mc_ki[ptop]--; + node = NODEPTR(mc->mc_pg[ptop], mn.mc_ki[ptop]); + rc = mdb_page_get(mc, NODEPGNO(node), &mn.mc_pg[mn.mc_top], NULL); + if (rc) + return rc; + mn.mc_ki[mn.mc_top] = NUMKEYS(mn.mc_pg[mn.mc_top]) - 1; + mc->mc_ki[mc->mc_top] = 0; + fromleft = 1; + } + + DPRINTF(("found neighbor page %"Z"u (%u keys, %.1f%% full)", + mn.mc_pg[mn.mc_top]->mp_pgno, NUMKEYS(mn.mc_pg[mn.mc_top]), + (float)PAGEFILL(mc->mc_txn->mt_env, mn.mc_pg[mn.mc_top]) / 10)); + + /* If the neighbor page is above threshold and has enough keys, + * move one key from it. Otherwise we should try to merge them. + * (A branch page must never have less than 2 keys.) + */ + if (PAGEFILL(mc->mc_txn->mt_env, mn.mc_pg[mn.mc_top]) >= thresh && NUMKEYS(mn.mc_pg[mn.mc_top]) > minkeys) { + rc = mdb_node_move(&mn, mc, fromleft); + if (fromleft) { + /* if we inserted on left, bump position up */ + oldki++; + } + } else { + if (!fromleft) { + rc = mdb_page_merge(&mn, mc); + } else { + oldki += NUMKEYS(mn.mc_pg[mn.mc_top]); + mn.mc_ki[mn.mc_top] += mc->mc_ki[mn.mc_top] + 1; + /* We want mdb_rebalance to find mn when doing fixups */ + WITH_CURSOR_TRACKING(mn, + rc = mdb_page_merge(mc, &mn)); + mdb_cursor_copy(&mn, mc); + } + mc->mc_flags &= ~C_EOF; + } + mc->mc_ki[mc->mc_top] = oldki; + return rc; +} + +/** Complete a delete operation started by #mdb_cursor_del(). */ +static int +mdb_cursor_del0(MDB_cursor *mc) +{ + int rc; + MDB_page *mp; + indx_t ki; + unsigned int nkeys; + MDB_cursor *m2, *m3; + MDB_dbi dbi = mc->mc_dbi; + + ki = mc->mc_ki[mc->mc_top]; + mp = mc->mc_pg[mc->mc_top]; + mdb_node_del(mc, mc->mc_db->md_pad); + mc->mc_db->md_entries--; + { + /* Adjust other cursors pointing to mp */ + for (m2 = mc->mc_txn->mt_cursors[dbi]; m2; m2=m2->mc_next) { + m3 = (mc->mc_flags & C_SUB) ? &m2->mc_xcursor->mx_cursor : m2; + if (! (m2->mc_flags & m3->mc_flags & C_INITIALIZED)) + continue; + if (m3 == mc || m3->mc_snum < mc->mc_snum) + continue; + if (m3->mc_pg[mc->mc_top] == mp) { + if (m3->mc_ki[mc->mc_top] == ki) { + m3->mc_flags |= C_DEL; + if (mc->mc_db->md_flags & MDB_DUPSORT) { + /* Sub-cursor referred into dataset which is gone */ + m3->mc_xcursor->mx_cursor.mc_flags &= ~(C_INITIALIZED|C_EOF); + } + continue; + } else if (m3->mc_ki[mc->mc_top] > ki) { + m3->mc_ki[mc->mc_top]--; + } + XCURSOR_REFRESH(m3, mc->mc_top, mp); + } + } + } + rc = mdb_rebalance(mc); + + if (rc == MDB_SUCCESS) { + /* DB is totally empty now, just bail out. + * Other cursors adjustments were already done + * by mdb_rebalance and aren't needed here. + */ + if (!mc->mc_snum) + return rc; + + mp = mc->mc_pg[mc->mc_top]; + nkeys = NUMKEYS(mp); + + /* Adjust other cursors pointing to mp */ + for (m2 = mc->mc_txn->mt_cursors[dbi]; !rc && m2; m2=m2->mc_next) { + m3 = (mc->mc_flags & C_SUB) ? &m2->mc_xcursor->mx_cursor : m2; + if (! (m2->mc_flags & m3->mc_flags & C_INITIALIZED)) + continue; + if (m3->mc_snum < mc->mc_snum) + continue; + if (m3->mc_pg[mc->mc_top] == mp) { + /* if m3 points past last node in page, find next sibling */ + if (m3->mc_ki[mc->mc_top] >= mc->mc_ki[mc->mc_top]) { + if (m3->mc_ki[mc->mc_top] >= nkeys) { + rc = mdb_cursor_sibling(m3, 1); + if (rc == MDB_NOTFOUND) { + m3->mc_flags |= C_EOF; + rc = MDB_SUCCESS; + continue; + } + } + if (mc->mc_db->md_flags & MDB_DUPSORT) { + MDB_node *node = NODEPTR(m3->mc_pg[m3->mc_top], m3->mc_ki[m3->mc_top]); + /* If this node has dupdata, it may need to be reinited + * because its data has moved. + * If the xcursor was not initd it must be reinited. + * Else if node points to a subDB, nothing is needed. + * Else (xcursor was initd, not a subDB) needs mc_pg[0] reset. + */ + if (node->mn_flags & F_DUPDATA) { + if (m3->mc_xcursor->mx_cursor.mc_flags & C_INITIALIZED) { + if (!(node->mn_flags & F_SUBDATA)) + m3->mc_xcursor->mx_cursor.mc_pg[0] = NODEDATA(node); + } else { + mdb_xcursor_init1(m3, node); + m3->mc_xcursor->mx_cursor.mc_flags |= C_DEL; + } + } + } + } + } + } + mc->mc_flags |= C_DEL; + } + + if (rc) + mc->mc_txn->mt_flags |= MDB_TXN_ERROR; + return rc; +} + +int +mdb_del(MDB_txn *txn, MDB_dbi dbi, + MDB_val *key, MDB_val *data) +{ + if (!key || !TXN_DBI_EXIST(txn, dbi, DB_USRVALID)) + return EINVAL; + + if (txn->mt_flags & (MDB_TXN_RDONLY|MDB_TXN_BLOCKED)) + return (txn->mt_flags & MDB_TXN_RDONLY) ? EACCES : MDB_BAD_TXN; + + if (!F_ISSET(txn->mt_dbs[dbi].md_flags, MDB_DUPSORT)) { + /* must ignore any data */ + data = NULL; + } + + return mdb_del0(txn, dbi, key, data, 0); +} + +static int +mdb_del0(MDB_txn *txn, MDB_dbi dbi, + MDB_val *key, MDB_val *data, unsigned flags) +{ + MDB_cursor mc; + MDB_xcursor mx; + MDB_cursor_op op; + MDB_val rdata, *xdata; + int rc, exact = 0; + DKBUF; + + DPRINTF(("====> delete db %u key [%s]", dbi, DKEY(key))); + + mdb_cursor_init(&mc, txn, dbi, &mx); + + if (data) { + op = MDB_GET_BOTH; + rdata = *data; + xdata = &rdata; + } else { + op = MDB_SET; + xdata = NULL; + flags |= MDB_NODUPDATA; + } + rc = mdb_cursor_set(&mc, key, xdata, op, &exact); + if (rc == 0) { + /* let mdb_page_split know about this cursor if needed: + * delete will trigger a rebalance; if it needs to move + * a node from one page to another, it will have to + * update the parent's separator key(s). If the new sepkey + * is larger than the current one, the parent page may + * run out of space, triggering a split. We need this + * cursor to be consistent until the end of the rebalance. + */ + mc.mc_flags |= C_UNTRACK; + mc.mc_next = txn->mt_cursors[dbi]; + txn->mt_cursors[dbi] = &mc; + rc = mdb_cursor_del(&mc, flags); + txn->mt_cursors[dbi] = mc.mc_next; + } + return rc; +} + +/** Split a page and insert a new node. + * Set #MDB_TXN_ERROR on failure. + * @param[in,out] mc Cursor pointing to the page and desired insertion index. + * The cursor will be updated to point to the actual page and index where + * the node got inserted after the split. + * @param[in] newkey The key for the newly inserted node. + * @param[in] newdata The data for the newly inserted node. + * @param[in] newpgno The page number, if the new node is a branch node. + * @param[in] nflags The #NODE_ADD_FLAGS for the new node. + * @return 0 on success, non-zero on failure. + */ +static int +mdb_page_split(MDB_cursor *mc, MDB_val *newkey, MDB_val *newdata, pgno_t newpgno, + unsigned int nflags) +{ + unsigned int flags; + int rc = MDB_SUCCESS, new_root = 0, did_split = 0; + indx_t newindx; + pgno_t pgno = 0; + int i, j, split_indx, nkeys, pmax; + MDB_env *env = mc->mc_txn->mt_env; + MDB_node *node; + MDB_val sepkey, rkey, xdata, *rdata = &xdata; + MDB_page *copy = NULL; + MDB_page *mp, *rp, *pp; + int ptop; + MDB_cursor mn; + DKBUF; + + mp = mc->mc_pg[mc->mc_top]; + newindx = mc->mc_ki[mc->mc_top]; + nkeys = NUMKEYS(mp); + + DPRINTF(("-----> splitting %s page %"Z"u and adding [%s] at index %i/%i", + IS_LEAF(mp) ? "leaf" : "branch", mp->mp_pgno, + DKEY(newkey), mc->mc_ki[mc->mc_top], nkeys)); + + /* Create a right sibling. */ + if ((rc = mdb_page_new(mc, mp->mp_flags, 1, &rp))) + return rc; + rp->mp_pad = mp->mp_pad; + DPRINTF(("new right sibling: page %"Z"u", rp->mp_pgno)); + + /* Usually when splitting the root page, the cursor + * height is 1. But when called from mdb_update_key, + * the cursor height may be greater because it walks + * up the stack while finding the branch slot to update. + */ + if (mc->mc_top < 1) { + if ((rc = mdb_page_new(mc, P_BRANCH, 1, &pp))) + goto done; + /* shift current top to make room for new parent */ + for (i=mc->mc_snum; i>0; i--) { + mc->mc_pg[i] = mc->mc_pg[i-1]; + mc->mc_ki[i] = mc->mc_ki[i-1]; + } + mc->mc_pg[0] = pp; + mc->mc_ki[0] = 0; + mc->mc_db->md_root = pp->mp_pgno; + DPRINTF(("root split! new root = %"Z"u", pp->mp_pgno)); + new_root = mc->mc_db->md_depth++; + + /* Add left (implicit) pointer. */ + if ((rc = mdb_node_add(mc, 0, NULL, NULL, mp->mp_pgno, 0)) != MDB_SUCCESS) { + /* undo the pre-push */ + mc->mc_pg[0] = mc->mc_pg[1]; + mc->mc_ki[0] = mc->mc_ki[1]; + mc->mc_db->md_root = mp->mp_pgno; + mc->mc_db->md_depth--; + goto done; + } + mc->mc_snum++; + mc->mc_top++; + ptop = 0; + } else { + ptop = mc->mc_top-1; + DPRINTF(("parent branch page is %"Z"u", mc->mc_pg[ptop]->mp_pgno)); + } + + mdb_cursor_copy(mc, &mn); + mn.mc_xcursor = NULL; + mn.mc_pg[mn.mc_top] = rp; + mn.mc_ki[ptop] = mc->mc_ki[ptop]+1; + + if (nflags & MDB_APPEND) { + mn.mc_ki[mn.mc_top] = 0; + sepkey = *newkey; + split_indx = newindx; + nkeys = 0; + } else { + + split_indx = (nkeys+1) / 2; + + if (IS_LEAF2(rp)) { + char *split, *ins; + int x; + unsigned int lsize, rsize, ksize; + /* Move half of the keys to the right sibling */ + x = mc->mc_ki[mc->mc_top] - split_indx; + ksize = mc->mc_db->md_pad; + split = LEAF2KEY(mp, split_indx, ksize); + rsize = (nkeys - split_indx) * ksize; + lsize = (nkeys - split_indx) * sizeof(indx_t); + mp->mp_lower -= lsize; + rp->mp_lower += lsize; + mp->mp_upper += rsize - lsize; + rp->mp_upper -= rsize - lsize; + sepkey.mv_size = ksize; + if (newindx == split_indx) { + sepkey.mv_data = newkey->mv_data; + } else { + sepkey.mv_data = split; + } + if (x<0) { + ins = LEAF2KEY(mp, mc->mc_ki[mc->mc_top], ksize); + memcpy(rp->mp_ptrs, split, rsize); + sepkey.mv_data = rp->mp_ptrs; + memmove(ins+ksize, ins, (split_indx - mc->mc_ki[mc->mc_top]) * ksize); + memcpy(ins, newkey->mv_data, ksize); + mp->mp_lower += sizeof(indx_t); + mp->mp_upper -= ksize - sizeof(indx_t); + } else { + if (x) + memcpy(rp->mp_ptrs, split, x * ksize); + ins = LEAF2KEY(rp, x, ksize); + memcpy(ins, newkey->mv_data, ksize); + memcpy(ins+ksize, split + x * ksize, rsize - x * ksize); + rp->mp_lower += sizeof(indx_t); + rp->mp_upper -= ksize - sizeof(indx_t); + mc->mc_ki[mc->mc_top] = x; + } + } else { + int psize, nsize, k; + /* Maximum free space in an empty page */ + pmax = env->me_psize - PAGEHDRSZ; + if (IS_LEAF(mp)) + nsize = mdb_leaf_size(env, newkey, newdata); + else + nsize = mdb_branch_size(env, newkey); + nsize = EVEN(nsize); + + /* grab a page to hold a temporary copy */ + copy = mdb_page_malloc(mc->mc_txn, 1); + if (copy == NULL) { + rc = ENOMEM; + goto done; + } + copy->mp_pgno = mp->mp_pgno; + copy->mp_flags = mp->mp_flags; + copy->mp_lower = (PAGEHDRSZ-PAGEBASE); + copy->mp_upper = env->me_psize - PAGEBASE; + + /* prepare to insert */ + for (i=0, j=0; i<nkeys; i++) { + if (i == newindx) { + copy->mp_ptrs[j++] = 0; + } + copy->mp_ptrs[j++] = mp->mp_ptrs[i]; + } + + /* When items are relatively large the split point needs + * to be checked, because being off-by-one will make the + * difference between success or failure in mdb_node_add. + * + * It's also relevant if a page happens to be laid out + * such that one half of its nodes are all "small" and + * the other half of its nodes are "large." If the new + * item is also "large" and falls on the half with + * "large" nodes, it also may not fit. + * + * As a final tweak, if the new item goes on the last + * spot on the page (and thus, onto the new page), bias + * the split so the new page is emptier than the old page. + * This yields better packing during sequential inserts. + */ + if (nkeys < 20 || nsize > pmax/16 || newindx >= nkeys) { + /* Find split point */ + psize = 0; + if (newindx <= split_indx || newindx >= nkeys) { + i = 0; j = 1; + k = newindx >= nkeys ? nkeys : split_indx+1+IS_LEAF(mp); + } else { + i = nkeys; j = -1; + k = split_indx-1; + } + for (; i!=k; i+=j) { + if (i == newindx) { + psize += nsize; + node = NULL; + } else { + node = (MDB_node *)((char *)mp + copy->mp_ptrs[i] + PAGEBASE); + psize += NODESIZE + NODEKSZ(node) + sizeof(indx_t); + if (IS_LEAF(mp)) { + if (F_ISSET(node->mn_flags, F_BIGDATA)) + psize += sizeof(pgno_t); + else + psize += NODEDSZ(node); + } + psize = EVEN(psize); + } + if (psize > pmax || i == k-j) { + split_indx = i + (j<0); + break; + } + } + } + if (split_indx == newindx) { + sepkey.mv_size = newkey->mv_size; + sepkey.mv_data = newkey->mv_data; + } else { + node = (MDB_node *)((char *)mp + copy->mp_ptrs[split_indx] + PAGEBASE); + sepkey.mv_size = node->mn_ksize; + sepkey.mv_data = NODEKEY(node); + } + } + } + + DPRINTF(("separator is %d [%s]", split_indx, DKEY(&sepkey))); + + /* Copy separator key to the parent. + */ + if (SIZELEFT(mn.mc_pg[ptop]) < mdb_branch_size(env, &sepkey)) { + int snum = mc->mc_snum; + mn.mc_snum--; + mn.mc_top--; + did_split = 1; + /* We want other splits to find mn when doing fixups */ + WITH_CURSOR_TRACKING(mn, + rc = mdb_page_split(&mn, &sepkey, NULL, rp->mp_pgno, 0)); + if (rc) + goto done; + + /* root split? */ + if (mc->mc_snum > snum) { + ptop++; + } + /* Right page might now have changed parent. + * Check if left page also changed parent. + */ + if (mn.mc_pg[ptop] != mc->mc_pg[ptop] && + mc->mc_ki[ptop] >= NUMKEYS(mc->mc_pg[ptop])) { + for (i=0; i<ptop; i++) { + mc->mc_pg[i] = mn.mc_pg[i]; + mc->mc_ki[i] = mn.mc_ki[i]; + } + mc->mc_pg[ptop] = mn.mc_pg[ptop]; + if (mn.mc_ki[ptop]) { + mc->mc_ki[ptop] = mn.mc_ki[ptop] - 1; + } else { + /* find right page's left sibling */ + mc->mc_ki[ptop] = mn.mc_ki[ptop]; + mdb_cursor_sibling(mc, 0); + } + } + } else { + mn.mc_top--; + rc = mdb_node_add(&mn, mn.mc_ki[ptop], &sepkey, NULL, rp->mp_pgno, 0); + mn.mc_top++; + } + if (rc != MDB_SUCCESS) { + goto done; + } + if (nflags & MDB_APPEND) { + mc->mc_pg[mc->mc_top] = rp; + mc->mc_ki[mc->mc_top] = 0; + rc = mdb_node_add(mc, 0, newkey, newdata, newpgno, nflags); + if (rc) + goto done; + for (i=0; i<mc->mc_top; i++) + mc->mc_ki[i] = mn.mc_ki[i]; + } else if (!IS_LEAF2(mp)) { + /* Move nodes */ + mc->mc_pg[mc->mc_top] = rp; + i = split_indx; + j = 0; + do { + if (i == newindx) { + rkey.mv_data = newkey->mv_data; + rkey.mv_size = newkey->mv_size; + if (IS_LEAF(mp)) { + rdata = newdata; + } else + pgno = newpgno; + flags = nflags; + /* Update index for the new key. */ + mc->mc_ki[mc->mc_top] = j; + } else { + node = (MDB_node *)((char *)mp + copy->mp_ptrs[i] + PAGEBASE); + rkey.mv_data = NODEKEY(node); + rkey.mv_size = node->mn_ksize; + if (IS_LEAF(mp)) { + xdata.mv_data = NODEDATA(node); + xdata.mv_size = NODEDSZ(node); + rdata = &xdata; + } else + pgno = NODEPGNO(node); + flags = node->mn_flags; + } + + if (!IS_LEAF(mp) && j == 0) { + /* First branch index doesn't need key data. */ + rkey.mv_size = 0; + } + + rc = mdb_node_add(mc, j, &rkey, rdata, pgno, flags); + if (rc) + goto done; + if (i == nkeys) { + i = 0; + j = 0; + mc->mc_pg[mc->mc_top] = copy; + } else { + i++; + j++; + } + } while (i != split_indx); + + nkeys = NUMKEYS(copy); + for (i=0; i<nkeys; i++) + mp->mp_ptrs[i] = copy->mp_ptrs[i]; + mp->mp_lower = copy->mp_lower; + mp->mp_upper = copy->mp_upper; + memcpy(NODEPTR(mp, nkeys-1), NODEPTR(copy, nkeys-1), + env->me_psize - copy->mp_upper - PAGEBASE); + + /* reset back to original page */ + if (newindx < split_indx) { + mc->mc_pg[mc->mc_top] = mp; + } else { + mc->mc_pg[mc->mc_top] = rp; + mc->mc_ki[ptop]++; + /* Make sure mc_ki is still valid. + */ + if (mn.mc_pg[ptop] != mc->mc_pg[ptop] && + mc->mc_ki[ptop] >= NUMKEYS(mc->mc_pg[ptop])) { + for (i=0; i<=ptop; i++) { + mc->mc_pg[i] = mn.mc_pg[i]; + mc->mc_ki[i] = mn.mc_ki[i]; + } + } + } + if (nflags & MDB_RESERVE) { + node = NODEPTR(mc->mc_pg[mc->mc_top], mc->mc_ki[mc->mc_top]); + if (!(node->mn_flags & F_BIGDATA)) + newdata->mv_data = NODEDATA(node); + } + } else { + if (newindx >= split_indx) { + mc->mc_pg[mc->mc_top] = rp; + mc->mc_ki[ptop]++; + /* Make sure mc_ki is still valid. + */ + if (mn.mc_pg[ptop] != mc->mc_pg[ptop] && + mc->mc_ki[ptop] >= NUMKEYS(mc->mc_pg[ptop])) { + for (i=0; i<=ptop; i++) { + mc->mc_pg[i] = mn.mc_pg[i]; + mc->mc_ki[i] = mn.mc_ki[i]; + } + } + } + } + + { + /* Adjust other cursors pointing to mp */ + MDB_cursor *m2, *m3; + MDB_dbi dbi = mc->mc_dbi; + nkeys = NUMKEYS(mp); + + for (m2 = mc->mc_txn->mt_cursors[dbi]; m2; m2=m2->mc_next) { + if (mc->mc_flags & C_SUB) + m3 = &m2->mc_xcursor->mx_cursor; + else + m3 = m2; + if (m3 == mc) + continue; + if (!(m2->mc_flags & m3->mc_flags & C_INITIALIZED)) + continue; + if (new_root) { + int k; + /* sub cursors may be on different DB */ + if (m3->mc_pg[0] != mp) + continue; + /* root split */ + for (k=new_root; k>=0; k--) { + m3->mc_ki[k+1] = m3->mc_ki[k]; + m3->mc_pg[k+1] = m3->mc_pg[k]; + } + if (m3->mc_ki[0] >= nkeys) { + m3->mc_ki[0] = 1; + } else { + m3->mc_ki[0] = 0; + } + m3->mc_pg[0] = mc->mc_pg[0]; + m3->mc_snum++; + m3->mc_top++; + } + if (m3->mc_top >= mc->mc_top && m3->mc_pg[mc->mc_top] == mp) { + if (m3->mc_ki[mc->mc_top] >= newindx && !(nflags & MDB_SPLIT_REPLACE)) + m3->mc_ki[mc->mc_top]++; + if (m3->mc_ki[mc->mc_top] >= nkeys) { + m3->mc_pg[mc->mc_top] = rp; + m3->mc_ki[mc->mc_top] -= nkeys; + for (i=0; i<mc->mc_top; i++) { + m3->mc_ki[i] = mn.mc_ki[i]; + m3->mc_pg[i] = mn.mc_pg[i]; + } + } + } else if (!did_split && m3->mc_top >= ptop && m3->mc_pg[ptop] == mc->mc_pg[ptop] && + m3->mc_ki[ptop] >= mc->mc_ki[ptop]) { + m3->mc_ki[ptop]++; + } + if (IS_LEAF(mp)) + XCURSOR_REFRESH(m3, mc->mc_top, m3->mc_pg[mc->mc_top]); + } + } + DPRINTF(("mp left: %d, rp left: %d", SIZELEFT(mp), SIZELEFT(rp))); + +done: + if (copy) /* tmp page */ + mdb_page_free(env, copy); + if (rc) + mc->mc_txn->mt_flags |= MDB_TXN_ERROR; + return rc; +} + +int +mdb_put(MDB_txn *txn, MDB_dbi dbi, + MDB_val *key, MDB_val *data, unsigned int flags) +{ + MDB_cursor mc; + MDB_xcursor mx; + int rc; + + if (!key || !data || !TXN_DBI_EXIST(txn, dbi, DB_USRVALID)) + return EINVAL; + + if (flags & ~(MDB_NOOVERWRITE|MDB_NODUPDATA|MDB_RESERVE|MDB_APPEND|MDB_APPENDDUP)) + return EINVAL; + + if (txn->mt_flags & (MDB_TXN_RDONLY|MDB_TXN_BLOCKED)) + return (txn->mt_flags & MDB_TXN_RDONLY) ? EACCES : MDB_BAD_TXN; + + mdb_cursor_init(&mc, txn, dbi, &mx); + mc.mc_next = txn->mt_cursors[dbi]; + txn->mt_cursors[dbi] = &mc; + rc = mdb_cursor_put(&mc, key, data, flags); + txn->mt_cursors[dbi] = mc.mc_next; + return rc; +} + +#ifndef MDB_WBUF +#define MDB_WBUF (1024*1024) +#endif +#define MDB_EOF 0x10 /**< #mdb_env_copyfd1() is done reading */ + + /** State needed for a double-buffering compacting copy. */ +typedef struct mdb_copy { + MDB_env *mc_env; + MDB_txn *mc_txn; + pthread_mutex_t mc_mutex; + pthread_cond_t mc_cond; /**< Condition variable for #mc_new */ + char *mc_wbuf[2]; + char *mc_over[2]; + int mc_wlen[2]; + int mc_olen[2]; + pgno_t mc_next_pgno; + HANDLE mc_fd; + int mc_toggle; /**< Buffer number in provider */ + int mc_new; /**< (0-2 buffers to write) | (#MDB_EOF at end) */ + /** Error code. Never cleared if set. Both threads can set nonzero + * to fail the copy. Not mutex-protected, LMDB expects atomic int. + */ + volatile int mc_error; +} mdb_copy; + + /** Dedicated writer thread for compacting copy. */ +static THREAD_RET ESECT CALL_CONV +mdb_env_copythr(void *arg) +{ + mdb_copy *my = arg; + char *ptr; + int toggle = 0, wsize, rc; +#ifdef _WIN32 + DWORD len; +#define DO_WRITE(rc, fd, ptr, w2, len) rc = WriteFile(fd, ptr, w2, &len, NULL) +#else + int len; +#define DO_WRITE(rc, fd, ptr, w2, len) len = write(fd, ptr, w2); rc = (len >= 0) +#ifdef SIGPIPE + sigset_t set; + sigemptyset(&set); + sigaddset(&set, SIGPIPE); + if ((rc = pthread_sigmask(SIG_BLOCK, &set, NULL)) != 0) + my->mc_error = rc; +#endif +#endif + + pthread_mutex_lock(&my->mc_mutex); + for(;;) { + while (!my->mc_new) + pthread_cond_wait(&my->mc_cond, &my->mc_mutex); + if (my->mc_new == 0 + MDB_EOF) /* 0 buffers, just EOF */ + break; + wsize = my->mc_wlen[toggle]; + ptr = my->mc_wbuf[toggle]; +again: + rc = MDB_SUCCESS; + while (wsize > 0 && !my->mc_error) { + DO_WRITE(rc, my->mc_fd, ptr, wsize, len); + if (!rc) { + rc = ErrCode(); +#if defined(SIGPIPE) && !defined(_WIN32) + if (rc == EPIPE) { + /* Collect the pending SIGPIPE, otherwise at least OS X + * gives it to the process on thread-exit (ITS#8504). + */ + int tmp; + sigwait(&set, &tmp); + } +#endif + break; + } else if (len > 0) { + rc = MDB_SUCCESS; + ptr += len; + wsize -= len; + continue; + } else { + rc = EIO; + break; + } + } + if (rc) { + my->mc_error = rc; + } + /* If there's an overflow page tail, write it too */ + if (my->mc_olen[toggle]) { + wsize = my->mc_olen[toggle]; + ptr = my->mc_over[toggle]; + my->mc_olen[toggle] = 0; + goto again; + } + my->mc_wlen[toggle] = 0; + toggle ^= 1; + /* Return the empty buffer to provider */ + my->mc_new--; + pthread_cond_signal(&my->mc_cond); + } + pthread_mutex_unlock(&my->mc_mutex); + return (THREAD_RET)0; +#undef DO_WRITE +} + + /** Give buffer and/or #MDB_EOF to writer thread, await unused buffer. + * + * @param[in] my control structure. + * @param[in] adjust (1 to hand off 1 buffer) | (MDB_EOF when ending). + */ +static int ESECT +mdb_env_cthr_toggle(mdb_copy *my, int adjust) +{ + pthread_mutex_lock(&my->mc_mutex); + my->mc_new += adjust; + pthread_cond_signal(&my->mc_cond); + while (my->mc_new & 2) /* both buffers in use */ + pthread_cond_wait(&my->mc_cond, &my->mc_mutex); + pthread_mutex_unlock(&my->mc_mutex); + + my->mc_toggle ^= (adjust & 1); + /* Both threads reset mc_wlen, to be safe from threading errors */ + my->mc_wlen[my->mc_toggle] = 0; + return my->mc_error; +} + + /** Depth-first tree traversal for compacting copy. + * @param[in] my control structure. + * @param[in,out] pg database root. + * @param[in] flags includes #F_DUPDATA if it is a sorted-duplicate sub-DB. + */ +static int ESECT +mdb_env_cwalk(mdb_copy *my, pgno_t *pg, int flags) +{ + MDB_cursor mc = {0}; + MDB_node *ni; + MDB_page *mo, *mp, *leaf; + char *buf, *ptr; + int rc, toggle; + unsigned int i; + + /* Empty DB, nothing to do */ + if (*pg == P_INVALID) + return MDB_SUCCESS; + + mc.mc_snum = 1; + mc.mc_txn = my->mc_txn; + + rc = mdb_page_get(&mc, *pg, &mc.mc_pg[0], NULL); + if (rc) + return rc; + rc = mdb_page_search_root(&mc, NULL, MDB_PS_FIRST); + if (rc) + return rc; + + /* Make cursor pages writable */ + buf = ptr = malloc(my->mc_env->me_psize * mc.mc_snum); + if (buf == NULL) + return ENOMEM; + + for (i=0; i<mc.mc_top; i++) { + mdb_page_copy((MDB_page *)ptr, mc.mc_pg[i], my->mc_env->me_psize); + mc.mc_pg[i] = (MDB_page *)ptr; + ptr += my->mc_env->me_psize; + } + + /* This is writable space for a leaf page. Usually not needed. */ + leaf = (MDB_page *)ptr; + + toggle = my->mc_toggle; + while (mc.mc_snum > 0) { + unsigned n; + mp = mc.mc_pg[mc.mc_top]; + n = NUMKEYS(mp); + + if (IS_LEAF(mp)) { + if (!IS_LEAF2(mp) && !(flags & F_DUPDATA)) { + for (i=0; i<n; i++) { + ni = NODEPTR(mp, i); + if (ni->mn_flags & F_BIGDATA) { + MDB_page *omp; + pgno_t pg; + + /* Need writable leaf */ + if (mp != leaf) { + mc.mc_pg[mc.mc_top] = leaf; + mdb_page_copy(leaf, mp, my->mc_env->me_psize); + mp = leaf; + ni = NODEPTR(mp, i); + } + + memcpy(&pg, NODEDATA(ni), sizeof(pg)); + memcpy(NODEDATA(ni), &my->mc_next_pgno, sizeof(pgno_t)); + rc = mdb_page_get(&mc, pg, &omp, NULL); + if (rc) + goto done; + if (my->mc_wlen[toggle] >= MDB_WBUF) { + rc = mdb_env_cthr_toggle(my, 1); + if (rc) + goto done; + toggle = my->mc_toggle; + } + mo = (MDB_page *)(my->mc_wbuf[toggle] + my->mc_wlen[toggle]); + memcpy(mo, omp, my->mc_env->me_psize); + mo->mp_pgno = my->mc_next_pgno; + my->mc_next_pgno += omp->mp_pages; + my->mc_wlen[toggle] += my->mc_env->me_psize; + if (omp->mp_pages > 1) { + my->mc_olen[toggle] = my->mc_env->me_psize * (omp->mp_pages - 1); + my->mc_over[toggle] = (char *)omp + my->mc_env->me_psize; + rc = mdb_env_cthr_toggle(my, 1); + if (rc) + goto done; + toggle = my->mc_toggle; + } + } else if (ni->mn_flags & F_SUBDATA) { + MDB_db db; + + /* Need writable leaf */ + if (mp != leaf) { + mc.mc_pg[mc.mc_top] = leaf; + mdb_page_copy(leaf, mp, my->mc_env->me_psize); + mp = leaf; + ni = NODEPTR(mp, i); + } + + memcpy(&db, NODEDATA(ni), sizeof(db)); + my->mc_toggle = toggle; + rc = mdb_env_cwalk(my, &db.md_root, ni->mn_flags & F_DUPDATA); + if (rc) + goto done; + toggle = my->mc_toggle; + memcpy(NODEDATA(ni), &db, sizeof(db)); + } + } + } + } else { + mc.mc_ki[mc.mc_top]++; + if (mc.mc_ki[mc.mc_top] < n) { + pgno_t pg; +again: + ni = NODEPTR(mp, mc.mc_ki[mc.mc_top]); + pg = NODEPGNO(ni); + rc = mdb_page_get(&mc, pg, &mp, NULL); + if (rc) + goto done; + mc.mc_top++; + mc.mc_snum++; + mc.mc_ki[mc.mc_top] = 0; + if (IS_BRANCH(mp)) { + /* Whenever we advance to a sibling branch page, + * we must proceed all the way down to its first leaf. + */ + mdb_page_copy(mc.mc_pg[mc.mc_top], mp, my->mc_env->me_psize); + goto again; + } else + mc.mc_pg[mc.mc_top] = mp; + continue; + } + } + if (my->mc_wlen[toggle] >= MDB_WBUF) { + rc = mdb_env_cthr_toggle(my, 1); + if (rc) + goto done; + toggle = my->mc_toggle; + } + mo = (MDB_page *)(my->mc_wbuf[toggle] + my->mc_wlen[toggle]); + mdb_page_copy(mo, mp, my->mc_env->me_psize); + mo->mp_pgno = my->mc_next_pgno++; + my->mc_wlen[toggle] += my->mc_env->me_psize; + if (mc.mc_top) { + /* Update parent if there is one */ + ni = NODEPTR(mc.mc_pg[mc.mc_top-1], mc.mc_ki[mc.mc_top-1]); + SETPGNO(ni, mo->mp_pgno); + mdb_cursor_pop(&mc); + } else { + /* Otherwise we're done */ + *pg = mo->mp_pgno; + break; + } + } +done: + free(buf); + return rc; +} + + /** Copy environment with compaction. */ +static int ESECT +mdb_env_copyfd1(MDB_env *env, HANDLE fd) +{ + MDB_meta *mm; + MDB_page *mp; + mdb_copy my = {0}; + MDB_txn *txn = NULL; + pthread_t thr; + pgno_t root, new_root; + int rc = MDB_SUCCESS; + +#ifdef _WIN32 + if (!(my.mc_mutex = CreateMutex(NULL, FALSE, NULL)) || + !(my.mc_cond = CreateEvent(NULL, FALSE, FALSE, NULL))) { + rc = ErrCode(); + goto done; + } + my.mc_wbuf[0] = _aligned_malloc(MDB_WBUF*2, env->me_os_psize); + if (my.mc_wbuf[0] == NULL) { + /* _aligned_malloc() sets errno, but we use Windows error codes */ + rc = ERROR_NOT_ENOUGH_MEMORY; + goto done; + } +#else + if ((rc = pthread_mutex_init(&my.mc_mutex, NULL)) != 0) + return rc; + if ((rc = pthread_cond_init(&my.mc_cond, NULL)) != 0) + goto done2; +#ifdef HAVE_MEMALIGN + my.mc_wbuf[0] = memalign(env->me_os_psize, MDB_WBUF*2); + if (my.mc_wbuf[0] == NULL) { + rc = errno; + goto done; + } +#else + { + void *p; + if ((rc = posix_memalign(&p, env->me_os_psize, MDB_WBUF*2)) != 0) + goto done; + my.mc_wbuf[0] = p; + } +#endif +#endif + memset(my.mc_wbuf[0], 0, MDB_WBUF*2); + my.mc_wbuf[1] = my.mc_wbuf[0] + MDB_WBUF; + my.mc_next_pgno = NUM_METAS; + my.mc_env = env; + my.mc_fd = fd; + rc = THREAD_CREATE(thr, mdb_env_copythr, &my); + if (rc) + goto done; + + rc = mdb_txn_begin(env, NULL, MDB_RDONLY, &txn); + if (rc) + goto finish; + + mp = (MDB_page *)my.mc_wbuf[0]; + memset(mp, 0, NUM_METAS * env->me_psize); + mp->mp_pgno = 0; + mp->mp_flags = P_META; + mm = (MDB_meta *)METADATA(mp); + mdb_env_init_meta0(env, mm); + mm->mm_address = env->me_metas[0]->mm_address; + + mp = (MDB_page *)(my.mc_wbuf[0] + env->me_psize); + mp->mp_pgno = 1; + mp->mp_flags = P_META; + *(MDB_meta *)METADATA(mp) = *mm; + mm = (MDB_meta *)METADATA(mp); + + /* Set metapage 1 with current main DB */ + root = new_root = txn->mt_dbs[MAIN_DBI].md_root; + if (root != P_INVALID) { + /* Count free pages + freeDB pages. Subtract from last_pg + * to find the new last_pg, which also becomes the new root. + */ + MDB_ID freecount = 0; + MDB_cursor mc; + MDB_val key, data; + mdb_cursor_init(&mc, txn, FREE_DBI, NULL); + while ((rc = mdb_cursor_get(&mc, &key, &data, MDB_NEXT)) == 0) + freecount += *(MDB_ID *)data.mv_data; + if (rc != MDB_NOTFOUND) + goto finish; + freecount += txn->mt_dbs[FREE_DBI].md_branch_pages + + txn->mt_dbs[FREE_DBI].md_leaf_pages + + txn->mt_dbs[FREE_DBI].md_overflow_pages; + + new_root = txn->mt_next_pgno - 1 - freecount; + mm->mm_last_pg = new_root; + mm->mm_dbs[MAIN_DBI] = txn->mt_dbs[MAIN_DBI]; + mm->mm_dbs[MAIN_DBI].md_root = new_root; + } else { + /* When the DB is empty, handle it specially to + * fix any breakage like page leaks from ITS#8174. + */ + mm->mm_dbs[MAIN_DBI].md_flags = txn->mt_dbs[MAIN_DBI].md_flags; + } + if (root != P_INVALID || mm->mm_dbs[MAIN_DBI].md_flags) { + mm->mm_txnid = 1; /* use metapage 1 */ + } + + my.mc_wlen[0] = env->me_psize * NUM_METAS; + my.mc_txn = txn; + rc = mdb_env_cwalk(&my, &root, 0); + if (rc == MDB_SUCCESS && root != new_root) { + rc = MDB_INCOMPATIBLE; /* page leak or corrupt DB */ + } + +finish: + if (rc) + my.mc_error = rc; + mdb_env_cthr_toggle(&my, 1 | MDB_EOF); + rc = THREAD_FINISH(thr); + mdb_txn_abort(txn); + +done: +#ifdef _WIN32 + if (my.mc_wbuf[0]) _aligned_free(my.mc_wbuf[0]); + if (my.mc_cond) CloseHandle(my.mc_cond); + if (my.mc_mutex) CloseHandle(my.mc_mutex); +#else + free(my.mc_wbuf[0]); + pthread_cond_destroy(&my.mc_cond); +done2: + pthread_mutex_destroy(&my.mc_mutex); +#endif + return rc ? rc : my.mc_error; +} + + /** Copy environment as-is. */ +static int ESECT +mdb_env_copyfd0(MDB_env *env, HANDLE fd) +{ + MDB_txn *txn = NULL; + mdb_mutexref_t wmutex = NULL; + int rc; + size_t wsize, w3; + char *ptr; +#ifdef _WIN32 + DWORD len, w2; +#define DO_WRITE(rc, fd, ptr, w2, len) rc = WriteFile(fd, ptr, w2, &len, NULL) +#else + ssize_t len; + size_t w2; +#define DO_WRITE(rc, fd, ptr, w2, len) len = write(fd, ptr, w2); rc = (len >= 0) +#endif + + /* Do the lock/unlock of the reader mutex before starting the + * write txn. Otherwise other read txns could block writers. + */ + rc = mdb_txn_begin(env, NULL, MDB_RDONLY, &txn); + if (rc) + return rc; + + if (env->me_txns) { + /* We must start the actual read txn after blocking writers */ + mdb_txn_end(txn, MDB_END_RESET_TMP); + + /* Temporarily block writers until we snapshot the meta pages */ + wmutex = env->me_wmutex; + if (LOCK_MUTEX(rc, env, wmutex)) + goto leave; + + rc = mdb_txn_renew0(txn); + if (rc) { + UNLOCK_MUTEX(wmutex); + goto leave; + } + } + + wsize = env->me_psize * NUM_METAS; + ptr = env->me_map; + w2 = wsize; + while (w2 > 0) { + DO_WRITE(rc, fd, ptr, w2, len); + if (!rc) { + rc = ErrCode(); + break; + } else if (len > 0) { + rc = MDB_SUCCESS; + ptr += len; + w2 -= len; + continue; + } else { + /* Non-blocking or async handles are not supported */ + rc = EIO; + break; + } + } + if (wmutex) + UNLOCK_MUTEX(wmutex); + + if (rc) + goto leave; + + w3 = txn->mt_next_pgno * env->me_psize; + { + size_t fsize = 0; + if ((rc = mdb_fsize(env->me_fd, &fsize))) + goto leave; + if (w3 > fsize) + w3 = fsize; + } + wsize = w3 - wsize; + while (wsize > 0) { + if (wsize > MAX_WRITE) + w2 = MAX_WRITE; + else + w2 = wsize; + DO_WRITE(rc, fd, ptr, w2, len); + if (!rc) { + rc = ErrCode(); + break; + } else if (len > 0) { + rc = MDB_SUCCESS; + ptr += len; + wsize -= len; + continue; + } else { + rc = EIO; + break; + } + } + +leave: + mdb_txn_abort(txn); + return rc; +} + +int ESECT +mdb_env_copyfd2(MDB_env *env, HANDLE fd, unsigned int flags) +{ + if (flags & MDB_CP_COMPACT) + return mdb_env_copyfd1(env, fd); + else + return mdb_env_copyfd0(env, fd); +} + +int ESECT +mdb_env_copyfd(MDB_env *env, HANDLE fd) +{ + return mdb_env_copyfd2(env, fd, 0); +} + +int ESECT +mdb_env_copy2(MDB_env *env, const char *path, unsigned int flags) +{ + int rc; + MDB_name fname; + HANDLE newfd = INVALID_HANDLE_VALUE; + + rc = mdb_fname_init(path, env->me_flags | MDB_NOLOCK, &fname); + if (rc == MDB_SUCCESS) { + rc = mdb_fopen(env, &fname, MDB_O_COPY, 0666, &newfd); + mdb_fname_destroy(fname); + } + if (rc == MDB_SUCCESS) { + rc = mdb_env_copyfd2(env, newfd, flags); + if (close(newfd) < 0 && rc == MDB_SUCCESS) + rc = ErrCode(); + } + return rc; +} + +int ESECT +mdb_env_copy(MDB_env *env, const char *path) +{ + return mdb_env_copy2(env, path, 0); +} + +int ESECT +mdb_env_set_flags(MDB_env *env, unsigned int flag, int onoff) +{ + if (flag & ~CHANGEABLE) + return EINVAL; + if (onoff) + env->me_flags |= flag; + else + env->me_flags &= ~flag; + return MDB_SUCCESS; +} + +int ESECT +mdb_env_get_flags(MDB_env *env, unsigned int *arg) +{ + if (!env || !arg) + return EINVAL; + + *arg = env->me_flags & (CHANGEABLE|CHANGELESS); + return MDB_SUCCESS; +} + +int ESECT +mdb_env_set_userctx(MDB_env *env, void *ctx) +{ + if (!env) + return EINVAL; + env->me_userctx = ctx; + return MDB_SUCCESS; +} + +void * ESECT +mdb_env_get_userctx(MDB_env *env) +{ + return env ? env->me_userctx : NULL; +} + +int ESECT +mdb_env_set_assert(MDB_env *env, MDB_assert_func *func) +{ + if (!env) + return EINVAL; +#ifndef NDEBUG + env->me_assert_func = func; +#endif + return MDB_SUCCESS; +} + +int ESECT +mdb_env_get_path(MDB_env *env, const char **arg) +{ + if (!env || !arg) + return EINVAL; + + *arg = env->me_path; + return MDB_SUCCESS; +} + +int ESECT +mdb_env_get_fd(MDB_env *env, mdb_filehandle_t *arg) +{ + if (!env || !arg) + return EINVAL; + + *arg = env->me_fd; + return MDB_SUCCESS; +} + +/** Common code for #mdb_stat() and #mdb_env_stat(). + * @param[in] env the environment to operate in. + * @param[in] db the #MDB_db record containing the stats to return. + * @param[out] arg the address of an #MDB_stat structure to receive the stats. + * @return 0, this function always succeeds. + */ +static int ESECT +mdb_stat0(MDB_env *env, MDB_db *db, MDB_stat *arg) +{ + arg->ms_psize = env->me_psize; + arg->ms_depth = db->md_depth; + arg->ms_branch_pages = db->md_branch_pages; + arg->ms_leaf_pages = db->md_leaf_pages; + arg->ms_overflow_pages = db->md_overflow_pages; + arg->ms_entries = db->md_entries; + + return MDB_SUCCESS; +} + +int ESECT +mdb_env_stat(MDB_env *env, MDB_stat *arg) +{ + MDB_meta *meta; + + if (env == NULL || arg == NULL) + return EINVAL; + + meta = mdb_env_pick_meta(env); + + return mdb_stat0(env, &meta->mm_dbs[MAIN_DBI], arg); +} + +int ESECT +mdb_env_info(MDB_env *env, MDB_envinfo *arg) +{ + MDB_meta *meta; + + if (env == NULL || arg == NULL) + return EINVAL; + + meta = mdb_env_pick_meta(env); + arg->me_mapaddr = meta->mm_address; + arg->me_last_pgno = meta->mm_last_pg; + arg->me_last_txnid = meta->mm_txnid; + + arg->me_mapsize = env->me_mapsize; + arg->me_maxreaders = env->me_maxreaders; + arg->me_numreaders = env->me_txns ? env->me_txns->mti_numreaders : 0; + return MDB_SUCCESS; +} + +/** Set the default comparison functions for a database. + * Called immediately after a database is opened to set the defaults. + * The user can then override them with #mdb_set_compare() or + * #mdb_set_dupsort(). + * @param[in] txn A transaction handle returned by #mdb_txn_begin() + * @param[in] dbi A database handle returned by #mdb_dbi_open() + */ +static void +mdb_default_cmp(MDB_txn *txn, MDB_dbi dbi) +{ + uint16_t f = txn->mt_dbs[dbi].md_flags; + + txn->mt_dbxs[dbi].md_cmp = + (f & MDB_REVERSEKEY) ? mdb_cmp_memnr : + (f & MDB_INTEGERKEY) ? mdb_cmp_cint : mdb_cmp_memn; + + txn->mt_dbxs[dbi].md_dcmp = + !(f & MDB_DUPSORT) ? 0 : + ((f & MDB_INTEGERDUP) + ? ((f & MDB_DUPFIXED) ? mdb_cmp_int : mdb_cmp_cint) + : ((f & MDB_REVERSEDUP) ? mdb_cmp_memnr : mdb_cmp_memn)); +} + +int mdb_dbi_open(MDB_txn *txn, const char *name, unsigned int flags, MDB_dbi *dbi) +{ + MDB_val key, data; + MDB_dbi i; + MDB_cursor mc; + MDB_db dummy; + int rc, dbflag, exact; + unsigned int unused = 0, seq; + char *namedup; + size_t len; + + if (flags & ~VALID_FLAGS) + return EINVAL; + if (txn->mt_flags & MDB_TXN_BLOCKED) + return MDB_BAD_TXN; + + /* main DB? */ + if (!name) { + *dbi = MAIN_DBI; + if (flags & PERSISTENT_FLAGS) { + uint16_t f2 = flags & PERSISTENT_FLAGS; + /* make sure flag changes get committed */ + if ((txn->mt_dbs[MAIN_DBI].md_flags | f2) != txn->mt_dbs[MAIN_DBI].md_flags) { + txn->mt_dbs[MAIN_DBI].md_flags |= f2; + txn->mt_flags |= MDB_TXN_DIRTY; + } + } + mdb_default_cmp(txn, MAIN_DBI); + return MDB_SUCCESS; + } + + if (txn->mt_dbxs[MAIN_DBI].md_cmp == NULL) { + mdb_default_cmp(txn, MAIN_DBI); + } + + /* Is the DB already open? */ + len = strlen(name); + for (i=CORE_DBS; i<txn->mt_numdbs; i++) { + if (!txn->mt_dbxs[i].md_name.mv_size) { + /* Remember this free slot */ + if (!unused) unused = i; + continue; + } + if (len == txn->mt_dbxs[i].md_name.mv_size && + !strncmp(name, txn->mt_dbxs[i].md_name.mv_data, len)) { + *dbi = i; + return MDB_SUCCESS; + } + } + + /* If no free slot and max hit, fail */ + if (!unused && txn->mt_numdbs >= txn->mt_env->me_maxdbs) + return MDB_DBS_FULL; + + /* Cannot mix named databases with some mainDB flags */ + if (txn->mt_dbs[MAIN_DBI].md_flags & (MDB_DUPSORT|MDB_INTEGERKEY)) + return (flags & MDB_CREATE) ? MDB_INCOMPATIBLE : MDB_NOTFOUND; + + /* Find the DB info */ + dbflag = DB_NEW|DB_VALID|DB_USRVALID; + exact = 0; + key.mv_size = len; + key.mv_data = (void *)name; + mdb_cursor_init(&mc, txn, MAIN_DBI, NULL); + rc = mdb_cursor_set(&mc, &key, &data, MDB_SET, &exact); + if (rc == MDB_SUCCESS) { + /* make sure this is actually a DB */ + MDB_node *node = NODEPTR(mc.mc_pg[mc.mc_top], mc.mc_ki[mc.mc_top]); + if ((node->mn_flags & (F_DUPDATA|F_SUBDATA)) != F_SUBDATA) + return MDB_INCOMPATIBLE; + } else { + if (rc != MDB_NOTFOUND || !(flags & MDB_CREATE)) + return rc; + if (F_ISSET(txn->mt_flags, MDB_TXN_RDONLY)) + return EACCES; + } + + /* Done here so we cannot fail after creating a new DB */ + if ((namedup = strdup(name)) == NULL) + return ENOMEM; + + if (rc) { + /* MDB_NOTFOUND and MDB_CREATE: Create new DB */ + data.mv_size = sizeof(MDB_db); + data.mv_data = &dummy; + memset(&dummy, 0, sizeof(dummy)); + dummy.md_root = P_INVALID; + dummy.md_flags = flags & PERSISTENT_FLAGS; + WITH_CURSOR_TRACKING(mc, + rc = mdb_cursor_put(&mc, &key, &data, F_SUBDATA)); + dbflag |= DB_DIRTY; + } + + if (rc) { + free(namedup); + } else { + /* Got info, register DBI in this txn */ + unsigned int slot = unused ? unused : txn->mt_numdbs; + txn->mt_dbxs[slot].md_name.mv_data = namedup; + txn->mt_dbxs[slot].md_name.mv_size = len; + txn->mt_dbxs[slot].md_rel = NULL; + txn->mt_dbflags[slot] = dbflag; + /* txn-> and env-> are the same in read txns, use + * tmp variable to avoid undefined assignment + */ + seq = ++txn->mt_env->me_dbiseqs[slot]; + txn->mt_dbiseqs[slot] = seq; + + memcpy(&txn->mt_dbs[slot], data.mv_data, sizeof(MDB_db)); + *dbi = slot; + mdb_default_cmp(txn, slot); + if (!unused) { + txn->mt_numdbs++; + } + } + + return rc; +} + +int ESECT +mdb_stat(MDB_txn *txn, MDB_dbi dbi, MDB_stat *arg) +{ + if (!arg || !TXN_DBI_EXIST(txn, dbi, DB_VALID)) + return EINVAL; + + if (txn->mt_flags & MDB_TXN_BLOCKED) + return MDB_BAD_TXN; + + if (txn->mt_dbflags[dbi] & DB_STALE) { + MDB_cursor mc; + MDB_xcursor mx; + /* Stale, must read the DB's root. cursor_init does it for us. */ + mdb_cursor_init(&mc, txn, dbi, &mx); + } + return mdb_stat0(txn->mt_env, &txn->mt_dbs[dbi], arg); +} + +void mdb_dbi_close(MDB_env *env, MDB_dbi dbi) +{ + char *ptr; + if (dbi < CORE_DBS || dbi >= env->me_maxdbs) + return; + ptr = env->me_dbxs[dbi].md_name.mv_data; + /* If there was no name, this was already closed */ + if (ptr) { + env->me_dbxs[dbi].md_name.mv_data = NULL; + env->me_dbxs[dbi].md_name.mv_size = 0; + env->me_dbflags[dbi] = 0; + env->me_dbiseqs[dbi]++; + free(ptr); + } +} + +int mdb_dbi_flags(MDB_txn *txn, MDB_dbi dbi, unsigned int *flags) +{ + /* We could return the flags for the FREE_DBI too but what's the point? */ + if (!TXN_DBI_EXIST(txn, dbi, DB_USRVALID)) + return EINVAL; + *flags = txn->mt_dbs[dbi].md_flags & PERSISTENT_FLAGS; + return MDB_SUCCESS; +} + +/** Add all the DB's pages to the free list. + * @param[in] mc Cursor on the DB to free. + * @param[in] subs non-Zero to check for sub-DBs in this DB. + * @return 0 on success, non-zero on failure. + */ +static int +mdb_drop0(MDB_cursor *mc, int subs) +{ + int rc; + + rc = mdb_page_search(mc, NULL, MDB_PS_FIRST); + if (rc == MDB_SUCCESS) { + MDB_txn *txn = mc->mc_txn; + MDB_node *ni; + MDB_cursor mx; + unsigned int i; + + /* DUPSORT sub-DBs have no ovpages/DBs. Omit scanning leaves. + * This also avoids any P_LEAF2 pages, which have no nodes. + * Also if the DB doesn't have sub-DBs and has no overflow + * pages, omit scanning leaves. + */ + if ((mc->mc_flags & C_SUB) || + (!subs && !mc->mc_db->md_overflow_pages)) + mdb_cursor_pop(mc); + + mdb_cursor_copy(mc, &mx); + while (mc->mc_snum > 0) { + MDB_page *mp = mc->mc_pg[mc->mc_top]; + unsigned n = NUMKEYS(mp); + if (IS_LEAF(mp)) { + for (i=0; i<n; i++) { + ni = NODEPTR(mp, i); + if (ni->mn_flags & F_BIGDATA) { + MDB_page *omp; + pgno_t pg; + memcpy(&pg, NODEDATA(ni), sizeof(pg)); + rc = mdb_page_get(mc, pg, &omp, NULL); + if (rc != 0) + goto done; + mdb_cassert(mc, IS_OVERFLOW(omp)); + rc = mdb_midl_append_range(&txn->mt_free_pgs, + pg, omp->mp_pages); + if (rc) + goto done; + mc->mc_db->md_overflow_pages -= omp->mp_pages; + if (!mc->mc_db->md_overflow_pages && !subs) + break; + } else if (subs && (ni->mn_flags & F_SUBDATA)) { + mdb_xcursor_init1(mc, ni); + rc = mdb_drop0(&mc->mc_xcursor->mx_cursor, 0); + if (rc) + goto done; + } + } + if (!subs && !mc->mc_db->md_overflow_pages) + goto pop; + } else { + if ((rc = mdb_midl_need(&txn->mt_free_pgs, n)) != 0) + goto done; + for (i=0; i<n; i++) { + pgno_t pg; + ni = NODEPTR(mp, i); + pg = NODEPGNO(ni); + /* free it */ + mdb_midl_xappend(txn->mt_free_pgs, pg); + } + } + if (!mc->mc_top) + break; + mc->mc_ki[mc->mc_top] = i; + rc = mdb_cursor_sibling(mc, 1); + if (rc) { + if (rc != MDB_NOTFOUND) + goto done; + /* no more siblings, go back to beginning + * of previous level. + */ +pop: + mdb_cursor_pop(mc); + mc->mc_ki[0] = 0; + for (i=1; i<mc->mc_snum; i++) { + mc->mc_ki[i] = 0; + mc->mc_pg[i] = mx.mc_pg[i]; + } + } + } + /* free it */ + rc = mdb_midl_append(&txn->mt_free_pgs, mc->mc_db->md_root); +done: + if (rc) + txn->mt_flags |= MDB_TXN_ERROR; + } else if (rc == MDB_NOTFOUND) { + rc = MDB_SUCCESS; + } + mc->mc_flags &= ~C_INITIALIZED; + return rc; +} + +int mdb_drop(MDB_txn *txn, MDB_dbi dbi, int del) +{ + MDB_cursor *mc, *m2; + int rc; + + if ((unsigned)del > 1 || !TXN_DBI_EXIST(txn, dbi, DB_USRVALID)) + return EINVAL; + + if (F_ISSET(txn->mt_flags, MDB_TXN_RDONLY)) + return EACCES; + + if (TXN_DBI_CHANGED(txn, dbi)) + return MDB_BAD_DBI; + + rc = mdb_cursor_open(txn, dbi, &mc); + if (rc) + return rc; + + rc = mdb_drop0(mc, mc->mc_db->md_flags & MDB_DUPSORT); + /* Invalidate the dropped DB's cursors */ + for (m2 = txn->mt_cursors[dbi]; m2; m2 = m2->mc_next) + m2->mc_flags &= ~(C_INITIALIZED|C_EOF); + if (rc) + goto leave; + + /* Can't delete the main DB */ + if (del && dbi >= CORE_DBS) { + rc = mdb_del0(txn, MAIN_DBI, &mc->mc_dbx->md_name, NULL, F_SUBDATA); + if (!rc) { + txn->mt_dbflags[dbi] = DB_STALE; + mdb_dbi_close(txn->mt_env, dbi); + } else { + txn->mt_flags |= MDB_TXN_ERROR; + } + } else { + /* reset the DB record, mark it dirty */ + txn->mt_dbflags[dbi] |= DB_DIRTY; + txn->mt_dbs[dbi].md_depth = 0; + txn->mt_dbs[dbi].md_branch_pages = 0; + txn->mt_dbs[dbi].md_leaf_pages = 0; + txn->mt_dbs[dbi].md_overflow_pages = 0; + txn->mt_dbs[dbi].md_entries = 0; + txn->mt_dbs[dbi].md_root = P_INVALID; + + txn->mt_flags |= MDB_TXN_DIRTY; + } +leave: + mdb_cursor_close(mc); + return rc; +} + +int mdb_set_compare(MDB_txn *txn, MDB_dbi dbi, MDB_cmp_func *cmp) +{ + if (!TXN_DBI_EXIST(txn, dbi, DB_USRVALID)) + return EINVAL; + + txn->mt_dbxs[dbi].md_cmp = cmp; + return MDB_SUCCESS; +} + +int mdb_set_dupsort(MDB_txn *txn, MDB_dbi dbi, MDB_cmp_func *cmp) +{ + if (!TXN_DBI_EXIST(txn, dbi, DB_USRVALID)) + return EINVAL; + + txn->mt_dbxs[dbi].md_dcmp = cmp; + return MDB_SUCCESS; +} + +int mdb_set_relfunc(MDB_txn *txn, MDB_dbi dbi, MDB_rel_func *rel) +{ + if (!TXN_DBI_EXIST(txn, dbi, DB_USRVALID)) + return EINVAL; + + txn->mt_dbxs[dbi].md_rel = rel; + return MDB_SUCCESS; +} + +int mdb_set_relctx(MDB_txn *txn, MDB_dbi dbi, void *ctx) +{ + if (!TXN_DBI_EXIST(txn, dbi, DB_USRVALID)) + return EINVAL; + + txn->mt_dbxs[dbi].md_relctx = ctx; + return MDB_SUCCESS; +} + +int ESECT +mdb_env_get_maxkeysize(MDB_env *env) +{ + return ENV_MAXKEY(env); +} + +int ESECT +mdb_reader_list(MDB_env *env, MDB_msg_func *func, void *ctx) +{ + unsigned int i, rdrs; + MDB_reader *mr; + char buf[64]; + int rc = 0, first = 1; + + if (!env || !func) + return -1; + if (!env->me_txns) { + return func("(no reader locks)\n", ctx); + } + rdrs = env->me_txns->mti_numreaders; + mr = env->me_txns->mti_readers; + for (i=0; i<rdrs; i++) { + if (mr[i].mr_pid) { + txnid_t txnid = mr[i].mr_txnid; + sprintf(buf, txnid == (txnid_t)-1 ? + "%10d %"Z"x -\n" : "%10d %"Z"x %"Z"u\n", + (int)mr[i].mr_pid, (size_t)mr[i].mr_tid, txnid); + if (first) { + first = 0; + rc = func(" pid thread txnid\n", ctx); + if (rc < 0) + break; + } + rc = func(buf, ctx); + if (rc < 0) + break; + } + } + if (first) { + rc = func("(no active readers)\n", ctx); + } + return rc; +} + +/** Insert pid into list if not already present. + * return -1 if already present. + */ +static int ESECT +mdb_pid_insert(MDB_PID_T *ids, MDB_PID_T pid) +{ + /* binary search of pid in list */ + unsigned base = 0; + unsigned cursor = 1; + int val = 0; + unsigned n = ids[0]; + + while( 0 < n ) { + unsigned pivot = n >> 1; + cursor = base + pivot + 1; + val = pid - ids[cursor]; + + if( val < 0 ) { + n = pivot; + + } else if ( val > 0 ) { + base = cursor; + n -= pivot + 1; + + } else { + /* found, so it's a duplicate */ + return -1; + } + } + + if( val > 0 ) { + ++cursor; + } + ids[0]++; + for (n = ids[0]; n > cursor; n--) + ids[n] = ids[n-1]; + ids[n] = pid; + return 0; +} + +int ESECT +mdb_reader_check(MDB_env *env, int *dead) +{ + if (!env) + return EINVAL; + if (dead) + *dead = 0; + return env->me_txns ? mdb_reader_check0(env, 0, dead) : MDB_SUCCESS; +} + +/** As #mdb_reader_check(). \b rlocked is set if caller locked #me_rmutex. */ +static int ESECT +mdb_reader_check0(MDB_env *env, int rlocked, int *dead) +{ + mdb_mutexref_t rmutex = rlocked ? NULL : env->me_rmutex; + unsigned int i, j, rdrs; + MDB_reader *mr; + MDB_PID_T *pids, pid; + int rc = MDB_SUCCESS, count = 0; + + rdrs = env->me_txns->mti_numreaders; + pids = malloc((rdrs+1) * sizeof(MDB_PID_T)); + if (!pids) + return ENOMEM; + pids[0] = 0; + mr = env->me_txns->mti_readers; + for (i=0; i<rdrs; i++) { + pid = mr[i].mr_pid; + if (pid && pid != env->me_pid) { + if (mdb_pid_insert(pids, pid) == 0) { + if (!mdb_reader_pid(env, Pidcheck, pid)) { + /* Stale reader found */ + j = i; + if (rmutex) { + if ((rc = LOCK_MUTEX0(rmutex)) != 0) { + if ((rc = mdb_mutex_failed(env, rmutex, rc))) + break; + rdrs = 0; /* the above checked all readers */ + } else { + /* Recheck, a new process may have reused pid */ + if (mdb_reader_pid(env, Pidcheck, pid)) + j = rdrs; + } + } + for (; j<rdrs; j++) + if (mr[j].mr_pid == pid) { + DPRINTF(("clear stale reader pid %u txn %"Z"d", + (unsigned) pid, mr[j].mr_txnid)); + mr[j].mr_pid = 0; + count++; + } + if (rmutex) + UNLOCK_MUTEX(rmutex); + } + } + } + } + free(pids); + if (dead) + *dead = count; + return rc; +} + +#ifdef MDB_ROBUST_SUPPORTED +/** Handle #LOCK_MUTEX0() failure. + * Try to repair the lock file if the mutex owner died. + * @param[in] env the environment handle + * @param[in] mutex LOCK_MUTEX0() mutex + * @param[in] rc LOCK_MUTEX0() error (nonzero) + * @return 0 on success with the mutex locked, or an error code on failure. + */ +static int ESECT +mdb_mutex_failed(MDB_env *env, mdb_mutexref_t mutex, int rc) +{ + int rlocked, rc2; + MDB_meta *meta; + + if (rc == MDB_OWNERDEAD) { + /* We own the mutex. Clean up after dead previous owner. */ + rc = MDB_SUCCESS; + rlocked = (mutex == env->me_rmutex); + if (!rlocked) { + /* Keep mti_txnid updated, otherwise next writer can + * overwrite data which latest meta page refers to. + */ + meta = mdb_env_pick_meta(env); + env->me_txns->mti_txnid = meta->mm_txnid; + /* env is hosed if the dead thread was ours */ + if (env->me_txn) { + env->me_flags |= MDB_FATAL_ERROR; + env->me_txn = NULL; + rc = MDB_PANIC; + } + } + DPRINTF(("%cmutex owner died, %s", (rlocked ? 'r' : 'w'), + (rc ? "this process' env is hosed" : "recovering"))); + rc2 = mdb_reader_check0(env, rlocked, NULL); + if (rc2 == 0) + rc2 = mdb_mutex_consistent(mutex); + if (rc || (rc = rc2)) { + DPRINTF(("LOCK_MUTEX recovery failed, %s", mdb_strerror(rc))); + UNLOCK_MUTEX(mutex); + } + } else { +#ifdef _WIN32 + rc = ErrCode(); +#endif + DPRINTF(("LOCK_MUTEX failed, %s", mdb_strerror(rc))); + } + + return rc; +} +#endif /* MDB_ROBUST_SUPPORTED */ + +#if defined(_WIN32) +/** Convert \b src to new wchar_t[] string with room for \b xtra extra chars */ +static int ESECT +utf8_to_utf16(const char *src, MDB_name *dst, int xtra) +{ + int rc, need = 0; + wchar_t *result = NULL; + for (;;) { /* malloc result, then fill it in */ + need = MultiByteToWideChar(CP_UTF8, 0, src, -1, result, need); + if (!need) { + rc = ErrCode(); + free(result); + return rc; + } + if (!result) { + result = malloc(sizeof(wchar_t) * (need + xtra)); + if (!result) + return ENOMEM; + continue; + } + dst->mn_alloced = 1; + dst->mn_len = need - 1; + dst->mn_val = result; + return MDB_SUCCESS; + } +} +#endif /* defined(_WIN32) */ +/** @} */ diff --git a/src/contrib/lmdb/midl.c b/src/contrib/lmdb/midl.c new file mode 100644 index 0000000..7b2b77e --- /dev/null +++ b/src/contrib/lmdb/midl.c @@ -0,0 +1,359 @@ +/** @file midl.c + * @brief ldap bdb back-end ID List functions */ +/* $OpenLDAP$ */ +/* This work is part of OpenLDAP Software <http://www.openldap.org/>. + * + * Copyright 2000-2018 The OpenLDAP Foundation. + * Portions Copyright 2001-2018 Howard Chu, Symas Corp. + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted only as authorized by the OpenLDAP + * Public License. + * + * A copy of this license is available in the file LICENSE in the + * top-level directory of the distribution or, alternatively, at + * <http://www.OpenLDAP.org/license.html>. + */ + +#include <limits.h> +#include <string.h> +#include <stdlib.h> +#include <errno.h> +#include <sys/types.h> +#include "midl.h" + +/** @defgroup internal LMDB Internals + * @{ + */ +/** @defgroup idls ID List Management + * @{ + */ +#define CMP(x,y) ( (x) < (y) ? -1 : (x) > (y) ) + +unsigned mdb_midl_search( MDB_IDL ids, MDB_ID id ) +{ + /* + * binary search of id in ids + * if found, returns position of id + * if not found, returns first position greater than id + */ + unsigned base = 0; + unsigned cursor = 1; + int val = 0; + unsigned n = ids[0]; + + while( 0 < n ) { + unsigned pivot = n >> 1; + cursor = base + pivot + 1; + val = CMP( ids[cursor], id ); + + if( val < 0 ) { + n = pivot; + + } else if ( val > 0 ) { + base = cursor; + n -= pivot + 1; + + } else { + return cursor; + } + } + + if( val > 0 ) { + ++cursor; + } + return cursor; +} + +#if 0 /* superseded by append/sort */ +int mdb_midl_insert( MDB_IDL ids, MDB_ID id ) +{ + unsigned x, i; + + x = mdb_midl_search( ids, id ); + assert( x > 0 ); + + if( x < 1 ) { + /* internal error */ + return -2; + } + + if ( x <= ids[0] && ids[x] == id ) { + /* duplicate */ + assert(0); + return -1; + } + + if ( ++ids[0] >= MDB_IDL_DB_MAX ) { + /* no room */ + --ids[0]; + return -2; + + } else { + /* insert id */ + for (i=ids[0]; i>x; i--) + ids[i] = ids[i-1]; + ids[x] = id; + } + + return 0; +} +#endif + +MDB_IDL mdb_midl_alloc(int num) +{ + MDB_IDL ids = malloc((num+2) * sizeof(MDB_ID)); + if (ids) { + *ids++ = num; + *ids = 0; + } + return ids; +} + +void mdb_midl_free(MDB_IDL ids) +{ + if (ids) + free(ids-1); +} + +void mdb_midl_shrink( MDB_IDL *idp ) +{ + MDB_IDL ids = *idp; + if (*(--ids) > MDB_IDL_UM_MAX && + (ids = realloc(ids, (MDB_IDL_UM_MAX+2) * sizeof(MDB_ID)))) + { + *ids++ = MDB_IDL_UM_MAX; + *idp = ids; + } +} + +static int mdb_midl_grow( MDB_IDL *idp, int num ) +{ + MDB_IDL idn = *idp-1; + /* grow it */ + idn = realloc(idn, (*idn + num + 2) * sizeof(MDB_ID)); + if (!idn) + return ENOMEM; + *idn++ += num; + *idp = idn; + return 0; +} + +int mdb_midl_need( MDB_IDL *idp, unsigned num ) +{ + MDB_IDL ids = *idp; + num += ids[0]; + if (num > ids[-1]) { + num = (num + num/4 + (256 + 2)) & -256; + if (!(ids = realloc(ids-1, num * sizeof(MDB_ID)))) + return ENOMEM; + *ids++ = num - 2; + *idp = ids; + } + return 0; +} + +int mdb_midl_append( MDB_IDL *idp, MDB_ID id ) +{ + MDB_IDL ids = *idp; + /* Too big? */ + if (ids[0] >= ids[-1]) { + if (mdb_midl_grow(idp, MDB_IDL_UM_MAX)) + return ENOMEM; + ids = *idp; + } + ids[0]++; + ids[ids[0]] = id; + return 0; +} + +int mdb_midl_append_list( MDB_IDL *idp, MDB_IDL app ) +{ + MDB_IDL ids = *idp; + /* Too big? */ + if (ids[0] + app[0] >= ids[-1]) { + if (mdb_midl_grow(idp, app[0])) + return ENOMEM; + ids = *idp; + } + memcpy(&ids[ids[0]+1], &app[1], app[0] * sizeof(MDB_ID)); + ids[0] += app[0]; + return 0; +} + +int mdb_midl_append_range( MDB_IDL *idp, MDB_ID id, unsigned n ) +{ + MDB_ID *ids = *idp, len = ids[0]; + /* Too big? */ + if (len + n > ids[-1]) { + if (mdb_midl_grow(idp, n | MDB_IDL_UM_MAX)) + return ENOMEM; + ids = *idp; + } + ids[0] = len + n; + ids += len; + while (n) + ids[n--] = id++; + return 0; +} + +void mdb_midl_xmerge( MDB_IDL idl, MDB_IDL merge ) +{ + MDB_ID old_id, merge_id, i = merge[0], j = idl[0], k = i+j, total = k; + idl[0] = (MDB_ID)-1; /* delimiter for idl scan below */ + old_id = idl[j]; + while (i) { + merge_id = merge[i--]; + for (; old_id < merge_id; old_id = idl[--j]) + idl[k--] = old_id; + idl[k--] = merge_id; + } + idl[0] = total; +} + +/* Quicksort + Insertion sort for small arrays */ + +#define SMALL 8 +#define MIDL_SWAP(a,b) { itmp=(a); (a)=(b); (b)=itmp; } + +void +mdb_midl_sort( MDB_IDL ids ) +{ + /* Max possible depth of int-indexed tree * 2 items/level */ + int istack[sizeof(int)*CHAR_BIT * 2]; + int i,j,k,l,ir,jstack; + MDB_ID a, itmp; + + ir = (int)ids[0]; + l = 1; + jstack = 0; + for(;;) { + if (ir - l < SMALL) { /* Insertion sort */ + for (j=l+1;j<=ir;j++) { + a = ids[j]; + for (i=j-1;i>=1;i--) { + if (ids[i] >= a) break; + ids[i+1] = ids[i]; + } + ids[i+1] = a; + } + if (jstack == 0) break; + ir = istack[jstack--]; + l = istack[jstack--]; + } else { + k = (l + ir) >> 1; /* Choose median of left, center, right */ + MIDL_SWAP(ids[k], ids[l+1]); + if (ids[l] < ids[ir]) { + MIDL_SWAP(ids[l], ids[ir]); + } + if (ids[l+1] < ids[ir]) { + MIDL_SWAP(ids[l+1], ids[ir]); + } + if (ids[l] < ids[l+1]) { + MIDL_SWAP(ids[l], ids[l+1]); + } + i = l+1; + j = ir; + a = ids[l+1]; + for(;;) { + do i++; while(ids[i] > a); + do j--; while(ids[j] < a); + if (j < i) break; + MIDL_SWAP(ids[i],ids[j]); + } + ids[l+1] = ids[j]; + ids[j] = a; + jstack += 2; + if (ir-i+1 >= j-l) { + istack[jstack] = ir; + istack[jstack-1] = i; + ir = j-1; + } else { + istack[jstack] = j-1; + istack[jstack-1] = l; + l = i; + } + } + } +} + +unsigned mdb_mid2l_search( MDB_ID2L ids, MDB_ID id ) +{ + /* + * binary search of id in ids + * if found, returns position of id + * if not found, returns first position greater than id + */ + unsigned base = 0; + unsigned cursor = 1; + int val = 0; + unsigned n = (unsigned)ids[0].mid; + + while( 0 < n ) { + unsigned pivot = n >> 1; + cursor = base + pivot + 1; + val = CMP( id, ids[cursor].mid ); + + if( val < 0 ) { + n = pivot; + + } else if ( val > 0 ) { + base = cursor; + n -= pivot + 1; + + } else { + return cursor; + } + } + + if( val > 0 ) { + ++cursor; + } + return cursor; +} + +int mdb_mid2l_insert( MDB_ID2L ids, MDB_ID2 *id ) +{ + unsigned x, i; + + x = mdb_mid2l_search( ids, id->mid ); + + if( x < 1 ) { + /* internal error */ + return -2; + } + + if ( x <= ids[0].mid && ids[x].mid == id->mid ) { + /* duplicate */ + return -1; + } + + if ( ids[0].mid >= MDB_IDL_UM_MAX ) { + /* too big */ + return -2; + + } else { + /* insert id */ + ids[0].mid++; + for (i=(unsigned)ids[0].mid; i>x; i--) + ids[i] = ids[i-1]; + ids[x] = *id; + } + + return 0; +} + +int mdb_mid2l_append( MDB_ID2L ids, MDB_ID2 *id ) +{ + /* Too big? */ + if (ids[0].mid >= MDB_IDL_UM_MAX) { + return -2; + } + ids[0].mid++; + ids[ids[0].mid] = *id; + return 0; +} + +/** @} */ +/** @} */ diff --git a/src/contrib/lmdb/midl.h b/src/contrib/lmdb/midl.h new file mode 100644 index 0000000..1aa374c --- /dev/null +++ b/src/contrib/lmdb/midl.h @@ -0,0 +1,186 @@ +/** @file midl.h + * @brief LMDB ID List header file. + * + * This file was originally part of back-bdb but has been + * modified for use in libmdb. Most of the macros defined + * in this file are unused, just left over from the original. + * + * This file is only used internally in libmdb and its definitions + * are not exposed publicly. + */ +/* $OpenLDAP$ */ +/* This work is part of OpenLDAP Software <http://www.openldap.org/>. + * + * Copyright 2000-2018 The OpenLDAP Foundation. + * Portions Copyright 2001-2018 Howard Chu, Symas Corp. + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted only as authorized by the OpenLDAP + * Public License. + * + * A copy of this license is available in the file LICENSE in the + * top-level directory of the distribution or, alternatively, at + * <http://www.OpenLDAP.org/license.html>. + */ + +#ifndef _MDB_MIDL_H_ +#define _MDB_MIDL_H_ + +#include <stddef.h> + +#ifdef __cplusplus +extern "C" { +#endif + +/** @defgroup internal LMDB Internals + * @{ + */ + +/** @defgroup idls ID List Management + * @{ + */ + /** A generic unsigned ID number. These were entryIDs in back-bdb. + * Preferably it should have the same size as a pointer. + */ +typedef size_t MDB_ID; + + /** An IDL is an ID List, a sorted array of IDs. The first + * element of the array is a counter for how many actual + * IDs are in the list. In the original back-bdb code, IDLs are + * sorted in ascending order. For libmdb IDLs are sorted in + * descending order. + */ +typedef MDB_ID *MDB_IDL; + +/* IDL sizes - likely should be even bigger + * limiting factors: sizeof(ID), thread stack size + */ +#define MDB_IDL_LOGN 16 /* DB_SIZE is 2^16, UM_SIZE is 2^17 */ +#define MDB_IDL_DB_SIZE (1<<MDB_IDL_LOGN) +#define MDB_IDL_UM_SIZE (1<<(MDB_IDL_LOGN+1)) + +#define MDB_IDL_DB_MAX (MDB_IDL_DB_SIZE-1) +#define MDB_IDL_UM_MAX (MDB_IDL_UM_SIZE-1) + +#define MDB_IDL_SIZEOF(ids) (((ids)[0]+1) * sizeof(MDB_ID)) +#define MDB_IDL_IS_ZERO(ids) ( (ids)[0] == 0 ) +#define MDB_IDL_CPY( dst, src ) (memcpy( dst, src, MDB_IDL_SIZEOF( src ) )) +#define MDB_IDL_FIRST( ids ) ( (ids)[1] ) +#define MDB_IDL_LAST( ids ) ( (ids)[(ids)[0]] ) + + /** Current max length of an #mdb_midl_alloc()ed IDL */ +#define MDB_IDL_ALLOCLEN( ids ) ( (ids)[-1] ) + + /** Append ID to IDL. The IDL must be big enough. */ +#define mdb_midl_xappend(idl, id) do { \ + MDB_ID *xidl = (idl), xlen = ++(xidl[0]); \ + xidl[xlen] = (id); \ + } while (0) + + /** Search for an ID in an IDL. + * @param[in] ids The IDL to search. + * @param[in] id The ID to search for. + * @return The index of the first ID greater than or equal to \b id. + */ +unsigned mdb_midl_search( MDB_IDL ids, MDB_ID id ); + + /** Allocate an IDL. + * Allocates memory for an IDL of the given size. + * @return IDL on success, NULL on failure. + */ +MDB_IDL mdb_midl_alloc(int num); + + /** Free an IDL. + * @param[in] ids The IDL to free. + */ +void mdb_midl_free(MDB_IDL ids); + + /** Shrink an IDL. + * Return the IDL to the default size if it has grown larger. + * @param[in,out] idp Address of the IDL to shrink. + */ +void mdb_midl_shrink(MDB_IDL *idp); + + /** Make room for num additional elements in an IDL. + * @param[in,out] idp Address of the IDL. + * @param[in] num Number of elements to make room for. + * @return 0 on success, ENOMEM on failure. + */ +int mdb_midl_need(MDB_IDL *idp, unsigned num); + + /** Append an ID onto an IDL. + * @param[in,out] idp Address of the IDL to append to. + * @param[in] id The ID to append. + * @return 0 on success, ENOMEM if the IDL is too large. + */ +int mdb_midl_append( MDB_IDL *idp, MDB_ID id ); + + /** Append an IDL onto an IDL. + * @param[in,out] idp Address of the IDL to append to. + * @param[in] app The IDL to append. + * @return 0 on success, ENOMEM if the IDL is too large. + */ +int mdb_midl_append_list( MDB_IDL *idp, MDB_IDL app ); + + /** Append an ID range onto an IDL. + * @param[in,out] idp Address of the IDL to append to. + * @param[in] id The lowest ID to append. + * @param[in] n Number of IDs to append. + * @return 0 on success, ENOMEM if the IDL is too large. + */ +int mdb_midl_append_range( MDB_IDL *idp, MDB_ID id, unsigned n ); + + /** Merge an IDL onto an IDL. The destination IDL must be big enough. + * @param[in] idl The IDL to merge into. + * @param[in] merge The IDL to merge. + */ +void mdb_midl_xmerge( MDB_IDL idl, MDB_IDL merge ); + + /** Sort an IDL. + * @param[in,out] ids The IDL to sort. + */ +void mdb_midl_sort( MDB_IDL ids ); + + /** An ID2 is an ID/pointer pair. + */ +typedef struct MDB_ID2 { + MDB_ID mid; /**< The ID */ + void *mptr; /**< The pointer */ +} MDB_ID2; + + /** An ID2L is an ID2 List, a sorted array of ID2s. + * The first element's \b mid member is a count of how many actual + * elements are in the array. The \b mptr member of the first element is unused. + * The array is sorted in ascending order by \b mid. + */ +typedef MDB_ID2 *MDB_ID2L; + + /** Search for an ID in an ID2L. + * @param[in] ids The ID2L to search. + * @param[in] id The ID to search for. + * @return The index of the first ID2 whose \b mid member is greater than or equal to \b id. + */ +unsigned mdb_mid2l_search( MDB_ID2L ids, MDB_ID id ); + + + /** Insert an ID2 into a ID2L. + * @param[in,out] ids The ID2L to insert into. + * @param[in] id The ID2 to insert. + * @return 0 on success, -1 if the ID was already present in the ID2L. + */ +int mdb_mid2l_insert( MDB_ID2L ids, MDB_ID2 *id ); + + /** Append an ID2 into a ID2L. + * @param[in,out] ids The ID2L to append into. + * @param[in] id The ID2 to append. + * @return 0 on success, -2 if the ID2L is too big. + */ +int mdb_mid2l_append( MDB_ID2L ids, MDB_ID2 *id ); + +/** @} */ +/** @} */ +#ifdef __cplusplus +} +#endif +#endif /* _MDB_MIDL_H_ */ diff --git a/src/contrib/macros.h b/src/contrib/macros.h new file mode 100644 index 0000000..1db2241 --- /dev/null +++ b/src/contrib/macros.h @@ -0,0 +1,43 @@ +/* Copyright (C) 2018 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ +/*! + * \brief Common macros. + */ + +#pragma once + +/*! \brief Eliminate compiler warning with unused parameters. */ +#define UNUSED(param) (void)(param) + +#ifndef MIN +/*! \brief Type-safe minimum macro. */ +#define MIN(a, b) \ + ({ __typeof__ (a) _a = (a); __typeof__ (b) _b = (b); _a < _b ? _a : _b; }) + +/*! \brief Type-safe maximum macro. */ +#define MAX(a, b) \ + ({ __typeof__ (a) _a = (a); __typeof__ (b) _b = (b); _a > _b ? _a : _b; }) +#endif + +#ifndef likely +/*! \brief Optimize for x to be true value. */ +#define likely(x) __builtin_expect((x), 1) +#endif + +#ifndef unlikely +/*! \brief Optimize for x to be false value. */ +#define unlikely(x) __builtin_expect((x), 0) +#endif diff --git a/src/contrib/mempattern.c b/src/contrib/mempattern.c new file mode 100644 index 0000000..74add91 --- /dev/null +++ b/src/contrib/mempattern.c @@ -0,0 +1,122 @@ +/* Copyright (C) 2017 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ + +#include <stdlib.h> + +#include "contrib/mempattern.h" +#include "contrib/string.h" +#include "contrib/ucw/mempool.h" + +static void mm_nofree(void *p) +{ + /* nop */ +} + +static void *mm_malloc(void *ctx, size_t n) +{ + (void)ctx; + return malloc(n); +} + +void *mm_alloc(knot_mm_t *mm, size_t size) +{ + if (mm) { + return mm->alloc(mm->ctx, size); + } else { + return malloc(size); + } +} + +void *mm_calloc(knot_mm_t *mm, size_t nmemb, size_t size) +{ + if (nmemb == 0 || size == 0) { + return NULL; + } + if (mm) { + size_t total_size = nmemb * size; + if (total_size / nmemb != size) { // Overflow check + return NULL; + } + void *mem = mm_alloc(mm, total_size); + if (mem == NULL) { + return NULL; + } + return memzero(mem, total_size); + } else { + return calloc(nmemb, size); + } +} + +void *mm_realloc(knot_mm_t *mm, void *what, size_t size, size_t prev_size) +{ + if (mm) { + void *p = mm->alloc(mm->ctx, size); + if (p == NULL) { + return NULL; + } else { + if (what) { + memcpy(p, what, + prev_size < size ? prev_size : size); + } + mm_free(mm, what); + return p; + } + } else { + return realloc(what, size); + } +} + +char *mm_strdup(knot_mm_t *mm, const char *s) +{ + if (s == NULL) { + return NULL; + } + if (mm) { + size_t len = strlen(s) + 1; + void *mem = mm_alloc(mm, len); + if (mem == NULL) { + return NULL; + } + return memcpy(mem, s, len); + } else { + return strdup(s); + } +} + +void mm_free(knot_mm_t *mm, void *what) +{ + if (mm) { + if (mm->free) { + mm->free(what); + } + } else { + free(what); + } +} + +void mm_ctx_init(knot_mm_t *mm) +{ + mm->ctx = NULL; + mm->alloc = mm_malloc; + mm->free = free; +} + +void mm_ctx_mempool(knot_mm_t *mm, size_t chunk_size) +{ + mm->ctx = mp_new(chunk_size); + mm->alloc = (knot_mm_alloc_t)mp_alloc; + mm->free = mm_nofree; +} diff --git a/src/contrib/mempattern.h b/src/contrib/mempattern.h new file mode 100644 index 0000000..1c5d9a2 --- /dev/null +++ b/src/contrib/mempattern.h @@ -0,0 +1,46 @@ +/* Copyright (C) 2018 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ +/*! + * \brief Memory allocation related functions. + */ + +#pragma once + +#include "libknot/mm_ctx.h" + +/*! \brief Default memory block size. */ +#define MM_DEFAULT_BLKSIZE 4096 + +/*! \brief Allocs using 'mm' if any, uses system malloc() otherwise. */ +void *mm_alloc(knot_mm_t *mm, size_t size); + +/*! \brief Callocs using 'mm' if any, uses system calloc() otherwise. */ +void *mm_calloc(knot_mm_t *mm, size_t nmemb, size_t size); + +/*! \brief Reallocs using 'mm' if any, uses system realloc() otherwise. */ +void *mm_realloc(knot_mm_t *mm, void *what, size_t size, size_t prev_size); + +/*! \brief Strdups using 'mm' if any, uses system strdup() otherwise. */ +char *mm_strdup(knot_mm_t *mm, const char *s); + +/*! \brief Free using 'mm' if any, uses system free() otherwise. */ +void mm_free(knot_mm_t *mm, void *what); + +/*! \brief Initialize default memory allocation context. */ +void mm_ctx_init(knot_mm_t *mm); + +/*! \brief Memory pool context. */ +void mm_ctx_mempool(knot_mm_t *mm, size_t chunk_size); diff --git a/src/contrib/net.c b/src/contrib/net.c new file mode 100644 index 0000000..81fc8ef --- /dev/null +++ b/src/contrib/net.c @@ -0,0 +1,598 @@ +/* Copyright (C) 2011 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ + +#include <assert.h> +#include <errno.h> +#include <fcntl.h> +#include <netinet/in.h> +#include <poll.h> +#include <stdbool.h> +#include <sys/socket.h> +#include <sys/uio.h> +#include <unistd.h> + +#include "libknot/errcode.h" +#include "contrib/net.h" +#include "contrib/sockaddr.h" + +/* + * OS X doesn't support MSG_NOSIGNAL. Use SO_NOSIGPIPE socket option instead. + */ +#if defined(__APPLE__) && !defined(MSG_NOSIGNAL) +# define MSG_NOSIGNAL 0 +# define osx_block_sigpipe(sock) sockopt_enable(sock, SOL_SOCKET, SO_NOSIGPIPE) +#else +# define osx_block_sigpipe(sock) KNOT_EOK +#endif + +/*! + * \brief Enable socket option. + */ +static int sockopt_enable(int sock, int level, int optname) +{ + const int enable = 1; + if (setsockopt(sock, level, optname, &enable, sizeof(enable)) != 0) { + return knot_map_errno(); + } + + return KNOT_EOK; +} + +/*! + * \brief Create a non-blocking socket. + * + * Prefer SOCK_NONBLOCK if available to save one fcntl() syscall. + * + */ +static int socket_create(int family, int type, int proto) +{ +#ifdef SOCK_NONBLOCK + type |= SOCK_NONBLOCK; +#endif + int sock = socket(family, type, proto); + if (sock < 0) { + return knot_map_errno(); + } + +#ifndef SOCK_NONBLOCK + if (fcntl(sock, F_SETFL, O_NONBLOCK) != 0) { + int ret = knot_map_errno(); + close(sock); + return ret; + } +#endif + + int ret = osx_block_sigpipe(sock); + if (ret != KNOT_EOK) { + return ret; + } + + return sock; +} + +int net_unbound_socket(int type, const struct sockaddr *sa) +{ + if (sa == NULL) { + return KNOT_EINVAL; + } + + /* Create socket. */ + return socket_create(sa->sa_family, type, 0); +} + +struct option { + int level; + int name; +}; + +/*! + * \brief Get setsock option for binding non-local address. + */ +static const struct option *nonlocal_option(int family) +{ + static const struct option ipv4 = { + #if defined(IP_FREEBIND) + IPPROTO_IP, IP_FREEBIND + #elif defined(IP_BINDANY) + IPPROTO_IP, IP_BINDANY + #else + 0, 0 + #endif + }; + + static const struct option ipv6 = { + #if defined(IP_FREEBIND) + IPPROTO_IP, IP_FREEBIND + #elif defined(IPV6_BINDANY) + IPPROTO_IPV6, IPV6_BINDANY + #else + 0, 0 + #endif + + }; + + switch (family) { + case AF_INET: return &ipv4; + case AF_INET6: return &ipv6; + default: + return NULL; + } +} + +static int enable_nonlocal(int sock, int family) +{ + const struct option *opt = nonlocal_option(family); + if (opt == NULL || opt->name == 0) { + return KNOT_ENOTSUP; + } + + return sockopt_enable(sock, opt->level, opt->name); +} + +static int enable_reuseport(int sock) +{ +#ifdef ENABLE_REUSEPORT + return sockopt_enable(sock, SOL_SOCKET, SO_REUSEPORT); +#else + return KNOT_ENOTSUP; +#endif +} + +static void unlink_unix_socket(const struct sockaddr *addr) +{ + char path[SOCKADDR_STRLEN] = { 0 }; + sockaddr_tostr(path, sizeof(path), addr); + unlink(path); +} + +int net_bound_socket(int type, const struct sockaddr *sa, enum net_flags flags) +{ + /* Create socket. */ + int sock = net_unbound_socket(type, sa); + if (sock < 0) { + return sock; + } + + /* Unlink UNIX sock if exists. */ + if (sa->sa_family == AF_UNIX) { + unlink_unix_socket(sa); + } + + /* Reuse old address if taken. */ + int ret = sockopt_enable(sock, SOL_SOCKET, SO_REUSEADDR); + if (ret != KNOT_EOK) { + close(sock); + return ret; + } + + /* Don't bind IPv4 for IPv6 any address. */ + if (sa->sa_family == AF_INET6) { + ret = sockopt_enable(sock, IPPROTO_IPV6, IPV6_V6ONLY); + if (ret != KNOT_EOK) { + close(sock); + return ret; + } + } + + /* Allow bind to non-local address. */ + if (flags & NET_BIND_NONLOCAL) { + ret = enable_nonlocal(sock, sa->sa_family); + if (ret != KNOT_EOK) { + close(sock); + return ret; + } + } + + /* Allow to bind the same address by multiple threads. */ + if (flags & NET_BIND_MULTIPLE) { + ret = enable_reuseport(sock); + if (ret != KNOT_EOK) { + close(sock); + return ret; + } + } + + /* Bind to specified address. */ + ret = bind(sock, sa, sockaddr_len(sa)); + if (ret < 0) { + ret = knot_map_errno(); + close(sock); + return ret; + } + + return sock; +} + +int net_connected_socket(int type, const struct sockaddr *dst_addr, + const struct sockaddr *src_addr) +{ + if (dst_addr == NULL) { + return KNOT_EINVAL; + } + + /* Check port. */ + if (sockaddr_port(dst_addr) == 0) { + return KNOT_NET_EADDR; + } + + /* Bind to specific source address - if set. */ + int sock = -1; + if (src_addr && src_addr->sa_family != AF_UNSPEC) { + sock = net_bound_socket(type, src_addr, 0); + } else { + sock = net_unbound_socket(type, dst_addr); + } + if (sock < 0) { + return sock; + } + + /* Connect to destination. */ + const struct sockaddr *sa = (const struct sockaddr *)dst_addr; + int ret = connect(sock, sa, sockaddr_len(sa)); + if (ret != 0 && errno != EINPROGRESS) { + ret = knot_map_errno(); + close(sock); + return ret; + } + + return sock; +} + +bool net_is_connected(int sock) +{ + struct sockaddr_storage ss; + socklen_t len = sizeof(ss); + return (getpeername(sock, (struct sockaddr *)&ss, &len) == 0); +} + +int net_socktype(int sock) +{ + int type; + socklen_t size = sizeof(type); + + if (getsockopt(sock, SOL_SOCKET, SO_TYPE, &type, &size) == 0) { + return type; + } else { + return AF_UNSPEC; + } +} + +bool net_is_stream(int sock) +{ + return net_socktype(sock) == SOCK_STREAM; +} + +int net_accept(int sock, struct sockaddr_storage *addr) +{ + socklen_t sa_len = sizeof(*addr); + + int remote = -1; + +#if defined(HAVE_ACCEPT4) && defined(SOCK_NONBLOCK) + remote = accept4(sock, (struct sockaddr *)addr, &sa_len, SOCK_NONBLOCK); + if (remote < 0) { + return knot_map_errno(); + } +#else + remote = accept(sock, (struct sockaddr *)addr, &sa_len); + if (fcntl(remote, F_SETFL, O_NONBLOCK) != 0) { + int error = knot_map_errno(); + close(remote); + return error; + } +#endif + + return remote; +} + +/* -- I/O interface handling partial -------------------------------------- */ + +/*! + * \brief Perform \a poll() on one socket. + */ +static int poll_one(int fd, int events, int timeout_ms) +{ + struct pollfd pfd = { + .fd = fd, + .events = events + }; + + return poll(&pfd, 1, timeout_ms); +} + +/*! + * \brief Check if we should wait for I/O readiness. + * + * \param error \a errno set by the failed I/O operation. + */ +static bool io_should_wait(int error) +{ + /* socket data not ready */ + if (error == EAGAIN || error == EWOULDBLOCK) { + return true; + } + +#ifndef __linux__ + /* FreeBSD: connection in progress */ + if (error == ENOTCONN) { + return true; + } +#endif + + return false; +} + +/*! + * \brief I/O operation callbacks. + */ +struct io { + ssize_t (*process)(int sockfd, struct msghdr *msg); + int (*wait)(int sockfd, int timeout_ms); +}; + +/*! + * \brief Get total size of I/O vector in a message. + */ +static size_t msg_iov_len(const struct msghdr *msg) +{ + size_t total = 0; + + for (int i = 0; i < msg->msg_iovlen; i++) { + total += msg->msg_iov[i].iov_len; + } + + return total; +} + +/*! + * \brief Shift processed data out of message IO vectors. + */ +static void msg_iov_shift(struct msghdr *msg, size_t done) +{ + struct iovec *iov = msg->msg_iov; + int iovlen = msg->msg_iovlen; + + for (int i = 0; i < iovlen && done > 0; i++) { + if (iov[i].iov_len > done) { + iov[i].iov_base += done; + iov[i].iov_len -= done; + done = 0; + } else { + done -= iov[i].iov_len; + msg->msg_iov += 1; + msg->msg_iovlen -= 1; + } + } + + assert(done == 0); +} + +/*! + * \brief Perform an I/O operation with a socket with waiting. + * + * \param oneshot If set, doesn't wait until the buffer is fully processed. + * + */ +static ssize_t io_exec(const struct io *io, int fd, struct msghdr *msg, + bool oneshot, int timeout_ms) +{ + size_t done = 0; + size_t total = msg_iov_len(msg); + + for (;;) { + /* Perform I/O. */ + ssize_t ret = io->process(fd, msg); + if (ret == -1 && errno == EINTR) { + continue; + } + if (ret > 0) { + done += ret; + if (oneshot || done == total) { + break; + } + msg_iov_shift(msg, ret); + } + + /* Wait for data readiness. */ + if (ret > 0 || (ret == -1 && io_should_wait(errno))) { + do { + ret = io->wait(fd, timeout_ms); + } while (ret == -1 && errno == EINTR); + if (ret == 1) { + continue; + } else if (ret == 0) { + return KNOT_ETIMEOUT; + } + } + + /* Disconnected or error. */ + return KNOT_ECONN; + } + + return done; +} + +static ssize_t recv_process(int fd, struct msghdr *msg) +{ + return recvmsg(fd, msg, MSG_DONTWAIT | MSG_NOSIGNAL); +} + +static int recv_wait(int fd, int timeout_ms) +{ + return poll_one(fd, POLLIN, timeout_ms); +} + +static ssize_t recv_data(int sock, struct msghdr *msg, bool oneshot, int timeout_ms) +{ + static const struct io RECV_IO = { + .process = recv_process, + .wait = recv_wait + }; + + return io_exec(&RECV_IO, sock, msg, oneshot, timeout_ms); +} + +static ssize_t send_process(int fd, struct msghdr *msg) +{ + return sendmsg(fd, msg, MSG_NOSIGNAL); +} + +static int send_wait(int fd, int timeout_ms) +{ + return poll_one(fd, POLLOUT, timeout_ms); +} + +static ssize_t send_data(int sock, struct msghdr *msg, int timeout_ms) +{ + static const struct io SEND_IO = { + .process = send_process, + .wait = send_wait + }; + + return io_exec(&SEND_IO, sock, msg, false, timeout_ms); +} + +/* -- generic stream and datagram I/O -------------------------------------- */ + +ssize_t net_send(int sock, const uint8_t *buffer, size_t size, + const struct sockaddr *addr, int timeout_ms) +{ + if (sock < 0 || buffer == NULL) { + return KNOT_EINVAL; + } + + struct iovec iov = { 0 }; + iov.iov_base = (void *)buffer; + iov.iov_len = size; + + struct msghdr msg = { 0 }; + msg.msg_name = (void *)addr; + msg.msg_namelen = sockaddr_len(addr); + msg.msg_iov = &iov; + msg.msg_iovlen = 1; + + int ret = send_data(sock, &msg, timeout_ms); + if (ret < 0) { + return ret; + } else if (ret != size) { + return KNOT_ECONN; + } + + return ret; +} + +ssize_t net_recv(int sock, uint8_t *buffer, size_t size, + struct sockaddr_storage *addr, int timeout_ms) +{ + if (sock < 0 || buffer == NULL) { + return KNOT_EINVAL; + } + + struct iovec iov = { 0 }; + iov.iov_base = buffer; + iov.iov_len = size; + + struct msghdr msg = { 0 }; + msg.msg_name = (void *)addr; + msg.msg_namelen = addr ? sizeof(*addr) : 0; + msg.msg_iov = &iov; + msg.msg_iovlen = 1; + + return recv_data(sock, &msg, true, timeout_ms); +} + +ssize_t net_dgram_send(int sock, const uint8_t *buffer, size_t size, + const struct sockaddr *addr) +{ + return net_send(sock, buffer, size, addr, 0); +} + +ssize_t net_dgram_recv(int sock, uint8_t *buffer, size_t size, int timeout_ms) +{ + return net_recv(sock, buffer, size, NULL, timeout_ms); +} + +ssize_t net_stream_send(int sock, const uint8_t *buffer, size_t size, int timeout_ms) +{ + return net_send(sock, buffer, size, NULL, timeout_ms); +} + +ssize_t net_stream_recv(int sock, uint8_t *buffer, size_t size, int timeout_ms) +{ + return net_recv(sock, buffer, size, NULL, timeout_ms); +} + +/* -- DNS specific I/O ----------------------------------------------------- */ + +ssize_t net_dns_tcp_send(int sock, const uint8_t *buffer, size_t size, int timeout_ms) +{ + if (sock < 0 || buffer == NULL || size > UINT16_MAX) { + return KNOT_EINVAL; + } + + struct iovec iov[2]; + uint16_t pktsize = htons(size); + iov[0].iov_base = &pktsize; + iov[0].iov_len = sizeof(uint16_t); + iov[1].iov_base = (void *)buffer; + iov[1].iov_len = size; + + struct msghdr msg = { 0 }; + msg.msg_iov = iov; + msg.msg_iovlen = 2; + + ssize_t ret = send_data(sock, &msg, timeout_ms); + if (ret < 0) { + return ret; + } + + return size; /* Do not count the size prefix. */ +} + +ssize_t net_dns_tcp_recv(int sock, uint8_t *buffer, size_t size, int timeout_ms) +{ + if (sock < 0 || buffer == NULL) { + return KNOT_EINVAL; + } + + struct iovec iov = { 0 }; + struct msghdr msg = { 0 }; + msg.msg_iov = &iov; + msg.msg_iovlen = 1; + + /* Receive size. */ + uint16_t pktsize = 0; + iov.iov_base = &pktsize; + iov.iov_len = sizeof(pktsize); + int ret = recv_data(sock, &msg, false, timeout_ms); + if (ret != sizeof(pktsize)) { + return ret; + } + + pktsize = ntohs(pktsize); + + /* Check packet size */ + if (size < pktsize) { + return KNOT_ESPACE; + } + + /* Receive payload. */ + msg.msg_iov = &iov; + msg.msg_iovlen = 1; + iov.iov_base = buffer; + iov.iov_len = pktsize; + return recv_data(sock, &msg, false, timeout_ms); +} diff --git a/src/contrib/net.h b/src/contrib/net.h new file mode 100644 index 0000000..1b6b527 --- /dev/null +++ b/src/contrib/net.h @@ -0,0 +1,185 @@ +/* Copyright (C) 2015 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. +*/ + +#pragma once + +#include <stdbool.h> +#include <stdint.h> +#include <sys/socket.h> + +/*! + * \brief Network interface flags. + */ +enum net_flags { + NET_BIND_NONLOCAL = (1 << 0), //!< Allow to bind unavailable address. + NET_BIND_MULTIPLE = (1 << 1), //!< Allow to bind address multiple times. +}; + +/*! + * \brief Create unbound socket of given family and type. + * + * \note The socket is set to non-blocking mode. + * + * \param type Socket transport type (SOCK_STREAM, SOCK_DGRAM). + * \param sa Socket address. + * + * \return socket or error code + */ +int net_unbound_socket(int type, const struct sockaddr *sa); + +/*! + * \brief Create socket bound to given address. + * + * The socket is set to non-blocking mode. + * + * \param type Socket transport type (SOCK_STREAM, SOCK_DGRAM). + * \param sa Socket address. + * \param flags Socket binding options. + * + * \return socket or error code + */ +int net_bound_socket(int type, const struct sockaddr *sa, enum net_flags flags); + +/*! + * \brief Create socket connected (asynchronously) to destination address. + * + * \note The socket is set to non-blocking mode. + * + * \param type Socket transport type (SOCK_STREAM, SOCK_DGRAM). + * \param dst_addr Destination address. + * \param src_addr Source address (can be NULL). + * + * \return socket or error code + */ +int net_connected_socket(int type, const struct sockaddr *dst_addr, + const struct sockaddr *src_addr); + +/*! + * \brief Return true if the socket is fully connected. + * + * \param sock Socket. + * + * \return true if connected + */ +bool net_is_connected(int sock); + +/*! + * \brief Get socket type (e.g. \a SOCK_STREAM). + * + * \param sock Socket. + */ +int net_socktype(int sock); + +/*! + * \brief Check if socket is a SOCK_STREAM socket. + */ +bool net_is_stream(int sock); + +/*! + * \brief Accept a connection on a listening socket. + * + * \brief The socket is set to non-blocking mode. + * + * \param sock Socket + * \param addr Remote address (can be NULL). + * + * \return socket or error code + */ +int net_accept(int sock, struct sockaddr_storage *addr); + +/*! + * \brief Send a message on a socket. + * + * The socket can be SOCK_STREAM or SOCK_DGRAM. + * + * The implementation handles partial-writes automatically. + * + * \param[in] sock Socket. + * \param[in] buffer Message buffer. + * \param[in] size Size of the message. + * \param[in] addr Remote address (ignored for SOCK_STREAM). + * \param[in] timeout_ms Write timeout in miliseconds (-1 for infinity, + * not valid for SOCK_DGRAM). + * + * \return Number of bytes sent or negative error code. + */ +ssize_t net_send(int sock, const uint8_t *buffer, size_t size, + const struct sockaddr *addr, int timeout_ms); + +/*! + * \brief Receive a message from a socket. + * + * \param[in] sock Socket. + * \param[out] buffer Receiving buffer. + * \param[in] size Capacity of the receiving buffer. + * \param[out] addr Remote address (can be NULL). + * \param[in] timeout_ms Read timeout in miliseconds (-1 for infinity). + * + * \return Number of bytes read or negative error code. + */ +ssize_t net_recv(int sock, uint8_t *buffer, size_t size, + struct sockaddr_storage *addr, int timeout_ms); + +/*! + * \brief Send a message on a SOCK_DGRAM socket. + * + * \see net_send + */ +ssize_t net_dgram_send(int sock, const uint8_t *buffer, size_t size, + const struct sockaddr *addr); + +/*! + * \brief Receive a message from a SOCK_DGRAM socket. + * + * \see net_recv + */ +ssize_t net_dgram_recv(int sock, uint8_t *buffer, size_t size, int timeout_ms); + +/*! + * \brief Send a message on a SOCK_STREAM socket. + * + * \see net_send + */ +ssize_t net_stream_send(int sock, const uint8_t *buffer, size_t size, int timeout_ms); + +/*! + * \brief Receive a message from a SOCK_STREAM socket. + * + * \see net_recv + */ +ssize_t net_stream_recv(int sock, uint8_t *buffer, size_t size, int timeout_ms); + +/*! + * \brief Send a DNS message on a TCP socket. + * + * The outgoing message is prefixed with a two-byte value carrying the DNS + * message size according to the specification. These two bytes are not + * reflected in the return value. + * + * \see net_send + */ +ssize_t net_dns_tcp_send(int sock, const uint8_t *buffer, size_t size, int timeout_ms); + +/*! + * \brief Receive a DNS message from a TCP socket. + * + * The first two bytes of the incoming message are interpreted as a DNS message + * size according to the specification. These two bytes are not included in + * the returned size. Only a complete DNS message is retrieved. + * + * \see net_recv + */ +ssize_t net_dns_tcp_recv(int sock, uint8_t *buffer, size_t size, int timeout_ms); diff --git a/src/contrib/openbsd/LICENSE b/src/contrib/openbsd/LICENSE new file mode 100644 index 0000000..e9a1aaa --- /dev/null +++ b/src/contrib/openbsd/LICENSE @@ -0,0 +1,2 @@ +../licenses/0BSD +../licenses/BSD-3-Clause
\ No newline at end of file diff --git a/src/contrib/openbsd/siphash.c b/src/contrib/openbsd/siphash.c new file mode 100644 index 0000000..26b8cfc --- /dev/null +++ b/src/contrib/openbsd/siphash.c @@ -0,0 +1,176 @@ +/* $OpenBSD: siphash.c,v 1.6 2017/04/12 17:41:49 deraadt Exp $ */ + +/*- + * Copyright (c) 2013 Andre Oppermann <andre@FreeBSD.org> + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. The name of the author may not be used to endorse or promote + * products derived from this software without specific prior written + * permission. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + */ + +/* + * SipHash is a family of PRFs SipHash-c-d where the integer parameters c and d + * are the number of compression rounds and the number of finalization rounds. + * A compression round is identical to a finalization round and this round + * function is called SipRound. Given a 128-bit key k and a (possibly empty) + * byte string m, SipHash-c-d returns a 64-bit value SipHash-c-d(k; m). + * + * Implemented from the paper "SipHash: a fast short-input PRF", 2012.09.18, + * by Jean-Philippe Aumasson and Daniel J. Bernstein, + * Permanent Document ID b9a943a805fbfc6fde808af9fc0ecdfa + * https://131002.net/siphash/siphash.pdf + * https://131002.net/siphash/ + */ + +#include <string.h> + +#include "libknot/endian.h" +#include "contrib/string.h" +#include "contrib/openbsd/siphash.h" + +static void SipHash_CRounds(SIPHASH_CTX *, int); +static void SipHash_Rounds(SIPHASH_CTX *, int); + +void +SipHash_Init(SIPHASH_CTX *ctx, const SIPHASH_KEY *key) +{ + uint64_t k0, k1; + + k0 = le64toh(key->k0); + k1 = le64toh(key->k1); + + ctx->v[0] = 0x736f6d6570736575ULL ^ k0; + ctx->v[1] = 0x646f72616e646f6dULL ^ k1; + ctx->v[2] = 0x6c7967656e657261ULL ^ k0; + ctx->v[3] = 0x7465646279746573ULL ^ k1; + + memset(ctx->buf, 0, sizeof(ctx->buf)); + ctx->bytes = 0; +} + +void +SipHash_Update(SIPHASH_CTX *ctx, int rc, int rf, const void *src, size_t len) +{ + const uint8_t *ptr = src; + size_t left, used; + + if (len == 0) + return; + + used = ctx->bytes % sizeof(ctx->buf); + ctx->bytes += len; + + if (used > 0) { + left = sizeof(ctx->buf) - used; + + if (len >= left) { + memcpy(&ctx->buf[used], ptr, left); + SipHash_CRounds(ctx, rc); + len -= left; + ptr += left; + } else { + memcpy(&ctx->buf[used], ptr, len); + return; + } + } + + while (len >= sizeof(ctx->buf)) { + memcpy(ctx->buf, ptr, sizeof(ctx->buf)); + SipHash_CRounds(ctx, rc); + len -= sizeof(ctx->buf); + ptr += sizeof(ctx->buf); + } + + if (len > 0) + memcpy(&ctx->buf, ptr, len); +} + +uint64_t +SipHash_End(SIPHASH_CTX *ctx, int rc, int rf) +{ + uint64_t r; + size_t left, used; + + used = ctx->bytes % sizeof(ctx->buf); + left = sizeof(ctx->buf) - used; + memset(&ctx->buf[used], 0, left - 1); + ctx->buf[7] = ctx->bytes; + + SipHash_CRounds(ctx, rc); + ctx->v[2] ^= 0xff; + SipHash_Rounds(ctx, rf); + + r = (ctx->v[0] ^ ctx->v[1]) ^ (ctx->v[2] ^ ctx->v[3]); + memzero(ctx, sizeof(*ctx)); + return htole64(r); +} + +uint64_t +SipHash(const SIPHASH_KEY *key, int rc, int rf, const void *src, size_t len) +{ + SIPHASH_CTX ctx; + + SipHash_Init(&ctx, key); + SipHash_Update(&ctx, rc, rf, src, len); + return (SipHash_End(&ctx, rc, rf)); +} + +#define SIP_ROTL(x, b) ((x) << (b)) | ( (x) >> (64 - (b))) + +static void +SipHash_Rounds(SIPHASH_CTX *ctx, int rounds) +{ + while (rounds--) { + ctx->v[0] += ctx->v[1]; + ctx->v[2] += ctx->v[3]; + ctx->v[1] = SIP_ROTL(ctx->v[1], 13); + ctx->v[3] = SIP_ROTL(ctx->v[3], 16); + + ctx->v[1] ^= ctx->v[0]; + ctx->v[3] ^= ctx->v[2]; + ctx->v[0] = SIP_ROTL(ctx->v[0], 32); + + ctx->v[2] += ctx->v[1]; + ctx->v[0] += ctx->v[3]; + ctx->v[1] = SIP_ROTL(ctx->v[1], 17); + ctx->v[3] = SIP_ROTL(ctx->v[3], 21); + + ctx->v[1] ^= ctx->v[2]; + ctx->v[3] ^= ctx->v[0]; + ctx->v[2] = SIP_ROTL(ctx->v[2], 32); + } +} + +static void +SipHash_CRounds(SIPHASH_CTX *ctx, int rounds) +{ + uint64_t tmp; + + memcpy(&tmp, ctx->buf, sizeof(tmp)); + uint64_t m = le64toh(tmp); + + ctx->v[3] ^= m; + SipHash_Rounds(ctx, rounds); + ctx->v[0] ^= m; +} diff --git a/src/contrib/openbsd/siphash.h b/src/contrib/openbsd/siphash.h new file mode 100644 index 0000000..d551fe8 --- /dev/null +++ b/src/contrib/openbsd/siphash.h @@ -0,0 +1,83 @@ +/*- + * Copyright (c) 2013 Andre Oppermann <andre@FreeBSD.org> + * All rights reserved. + * + * Redistribution and use in source and binary forms, with or without + * modification, are permitted provided that the following conditions + * are met: + * 1. Redistributions of source code must retain the above copyright + * notice, this list of conditions and the following disclaimer. + * 2. Redistributions in binary form must reproduce the above copyright + * notice, this list of conditions and the following disclaimer in the + * documentation and/or other materials provided with the distribution. + * 3. The name of the author may not be used to endorse or promote + * products derived from this software without specific prior written + * permission. + * + * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND + * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE + * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE + * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE + * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL + * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS + * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) + * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT + * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY + * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF + * SUCH DAMAGE. + * + * $OpenBSD: siphash.h,v 1.3 2015/02/20 11:51:03 tedu Exp $ + */ + +/* + * SipHash is a family of pseudorandom functions (a.k.a. keyed hash functions) + * optimized for speed on short messages returning a 64bit hash/digest value. + * + * The number of rounds is defined during the initialization: + * SipHash24_Init() for the fast and resonable strong version + * SipHash48_Init() for the strong version (half as fast) + * + * struct SIPHASH_CTX ctx; + * SipHash24_Init(&ctx); + * SipHash_SetKey(&ctx, "16bytes long key"); + * SipHash_Update(&ctx, pointer_to_string, length_of_string); + * SipHash_End(&ctx); + */ + +#ifndef _SIPHASH_H_ +#define _SIPHASH_H_ + +#include <stddef.h> +#include <stdint.h> + +#define SIPHASH_BLOCK_LENGTH 8 +#define SIPHASH_KEY_LENGTH 16 +#define SIPHASH_DIGEST_LENGTH 8 + +typedef struct _SIPHASH_CTX { + uint64_t v[4]; + uint8_t buf[SIPHASH_BLOCK_LENGTH]; + uint32_t bytes; +} SIPHASH_CTX; + +typedef struct { + uint64_t k0; + uint64_t k1; +} SIPHASH_KEY; + +void SipHash_Init(SIPHASH_CTX *, const SIPHASH_KEY *); +void SipHash_Update(SIPHASH_CTX *, int, int, const void *, size_t); +uint64_t SipHash_End(SIPHASH_CTX *, int, int); +uint64_t SipHash(const SIPHASH_KEY *, int, int, const void *, size_t); + +#define SipHash24_Init(_c, _k) SipHash_Init((_c), (_k)) +#define SipHash24_Update(_c, _p, _l) SipHash_Update((_c), 2, 4, (_p), (_l)) +#define SipHash24_End(_d) SipHash_End((_d), 2, 4) +#define SipHash24(_k, _p, _l) SipHash((_k), 2, 4, (_p), (_l)) + +#define SipHash48_Init(_c, _k) SipHash_Init((_c), (_k)) +#define SipHash48_Update(_c, _p, _l) SipHash_Update((_c), 4, 8, (_p), (_l)) +#define SipHash48_End(_d) SipHash_End((_d), 4, 8) +#define SipHash48(_k, _p, _l) SipHash((_k), 4, 8, (_p), (_l)) + +#endif /* _SIPHASH_H_ */ diff --git a/src/contrib/openbsd/strlcat.c b/src/contrib/openbsd/strlcat.c new file mode 100644 index 0000000..1409062 --- /dev/null +++ b/src/contrib/openbsd/strlcat.c @@ -0,0 +1,48 @@ +/* + * Copyright (c) 1998 Todd C. Miller <Todd.Miller@courtesan.com> + * + * Permission to use, copy, modify, and distribute this software for any + * purpose with or without fee is hereby granted, provided that the above + * copyright notice and this permission notice appear in all copies. + * + * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES + * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF + * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR + * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES + * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN + * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF + * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. + */ + +#include <sys/types.h> +#include <string.h> + +#include "contrib/openbsd/strlcat.h" + +size_t +knot_strlcat(char *dst, const char *src, size_t siz) +{ + char *d = dst; + const char *s = src; + size_t n = siz; + size_t dlen; + + /* Find the end of dst and adjust bytes left but don't go past end */ + while (n-- != 0 && *d != '\0') + d++; + dlen = d - dst; + n = siz - dlen; + + if (n == 0) + return(dlen + strlen(s)); + while (*s != '\0') { + if (n != 1) { + *d++ = *s; + n--; + } + s++; + } + *d = '\0'; + + return(dlen + (s - src)); /* count does not include NUL */ +} diff --git a/src/contrib/openbsd/strlcat.h b/src/contrib/openbsd/strlcat.h new file mode 100644 index 0000000..7016069 --- /dev/null +++ b/src/contrib/openbsd/strlcat.h @@ -0,0 +1,31 @@ +/* + * Copyright (c) 1998 Todd C. Miller <Todd.Miller@courtesan.com> + * + * Permission to use, copy, modify, and distribute this software for any + * purpose with or without fee is hereby granted, provided that the above + * copyright notice and this permission notice appear in all copies. + * + * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES + * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF + * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR + * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES + * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN + * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF + * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. + */ + +#pragma once + +#ifndef HAVE_STRLCAT +#define strlcat(dst, src, size) knot_strlcat(dst, src, size) +#endif + +/* + * Appends src to string dst of size siz (unlike strncat, siz is the + * full size of dst, not space left). At most siz-1 characters + * will be copied. Always NUL terminates (unless siz <= strlen(dst)). + * Returns strlen(src) + MIN(siz, strlen(initial dst)). + * If retval >= siz, truncation occurred. + */ +size_t +knot_strlcat(char *dst, const char *src, size_t siz); diff --git a/src/contrib/openbsd/strlcpy.c b/src/contrib/openbsd/strlcpy.c new file mode 100644 index 0000000..eafc0e4 --- /dev/null +++ b/src/contrib/openbsd/strlcpy.c @@ -0,0 +1,46 @@ +/* + * Copyright (c) 1998 Todd C. Miller <Todd.Miller@courtesan.com> + * + * Permission to use, copy, modify, and distribute this software for any + * purpose with or without fee is hereby granted, provided that the above + * copyright notice and this permission notice appear in all copies. + * + * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES + * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF + * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR + * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES + * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN + * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF + * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. + */ + +#include <sys/types.h> +#include <string.h> + +#include "contrib/openbsd/strlcpy.h" + +size_t +knot_strlcpy(char *dst, const char *src, size_t siz) +{ + char *d = dst; + const char *s = src; + size_t n = siz; + + /* Copy as many bytes as will fit */ + if (n != 0) { + while (--n != 0) { + if ((*d++ = *s++) == '\0') + break; + } + } + + /* Not enough room in dst, add NUL and traverse rest of src */ + if (n == 0) { + if (siz != 0) + *d = '\0'; /* NUL-terminate dst */ + while (*s++) + ; + } + + return(s - src - 1); /* count does not include NUL */ +} diff --git a/src/contrib/openbsd/strlcpy.h b/src/contrib/openbsd/strlcpy.h new file mode 100644 index 0000000..6421068 --- /dev/null +++ b/src/contrib/openbsd/strlcpy.h @@ -0,0 +1,29 @@ +/* + * Copyright (c) 1998 Todd C. Miller <Todd.Miller@courtesan.com> + * + * Permission to use, copy, modify, and distribute this software for any + * purpose with or without fee is hereby granted, provided that the above + * copyright notice and this permission notice appear in all copies. + * + * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES + * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF + * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR + * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES + * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN + * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF + * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. + */ + +#pragma once + +#ifndef HAVE_STRLCPY +#define strlcpy(dst, src, size) knot_strlcpy(dst, src, size) +#endif + +/* + * Copy src to string dst of size siz. At most siz-1 characters + * will be copied. Always NUL terminates (unless siz == 0). + * Returns strlen(src); if retval >= siz, truncation occurred. + */ +size_t +knot_strlcpy(char *dst, const char *src, size_t siz); diff --git a/src/contrib/qp-trie/trie.c b/src/contrib/qp-trie/trie.c new file mode 100644 index 0000000..10f9919 --- /dev/null +++ b/src/contrib/qp-trie/trie.c @@ -0,0 +1,835 @@ +/* Copyright (C) 2016 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + + The code originated from https://github.com/fanf2/qp/blob/master/qp.c + at revision 5f6d93753. + */ + +#include <assert.h> +#include <stdlib.h> +#include <string.h> + +#include "contrib/qp-trie/trie.h" +#include "contrib/macros.h" +#include "contrib/mempattern.h" +#include "libknot/errcode.h" + +#if defined(__i386) || defined(__x86_64) || defined(_M_IX86) \ + || (defined(__BYTE_ORDER__) && defined(__ORDER_LITTLE_ENDIAN) \ + && __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__) + + /*! + * \brief Use a pointer alignment hack to save memory. + * + * When on, isbranch() relies on the fact that in leaf_t the first pointer + * is aligned on multiple of 4 bytes and that the flags bitfield is + * overlaid over the lowest two bits of that pointer. + * Neither is really guaranteed by the C standards; the second part should + * be OK with x86_64 ABI and most likely any other little-endian platform. + * It would be possible to manipulate the right bits portably, but it would + * complicate the code nontrivially. C++ doesn't even guarantee type-punning. + * In debug mode we check this works OK when creating a new trie instance. + */ + #define FLAGS_HACK 1 +#else + #define FLAGS_HACK 0 +#endif + +typedef unsigned char byte; +typedef unsigned int uint; +typedef uint bitmap_t; /*! Bit-maps, using the range of 1<<0 to 1<<16 (inclusive). */ + +typedef struct { + uint32_t len; // 32 bits are enough for key lengths; probably even 16 bits would be. + char chars[]; +} tkey_t; + +/*! \brief Leaf of trie. */ +typedef struct { + #if !FLAGS_HACK + byte flags; + #endif + tkey_t *key; /*!< The pointer must be aligned to 4-byte multiples! */ + trie_val_t val; +} leaf_t; + +/*! \brief A trie node is either leaf_t or branch_t. */ +typedef union node node_t; + +/*! + * \brief Branch node of trie. + * + * - The flags distinguish whether the node is a leaf_t (0), or a branch + * testing the more-important nibble (1) or the less-important one (2). + * - It stores the index of the byte that the node tests. The combined + * value (index*4 + flags) increases in branch nodes as you go deeper + * into the trie. All the keys below a branch are identical up to the + * nibble identified by the branch. Indices have to be stored because + * we skip any branch nodes that would have a single child. + * (Consequently, the skipped parts of key have to be validated in a leaf.) + * - The bitmap indicates which subtries are present. The present child nodes + * are stored in the twigs array (with no holes between them). + * - To simplify storing keys that are prefixes of each other, the end-of-string + * position is treated as another nibble value, ordered before all others. + * That affects the bitmap and twigs fields. + * + * \note The branch nodes are never allocated individually, but they are + * always part of either the root node or the twigs array of the parent. + */ +typedef struct { + #if FLAGS_HACK + uint32_t flags : 2, + bitmap : 17; /*!< The first bitmap bit is for end-of-string child. */ + #else + byte flags; + uint32_t bitmap; + #endif + uint32_t index; + node_t *twigs; +} branch_t; + +union node { + leaf_t leaf; + branch_t branch; +}; + +struct trie { + node_t root; // undefined when weight == 0, see empty_root() + size_t weight; + knot_mm_t mm; +}; + +/*! \brief Make the root node empty (debug-only). */ +static inline void empty_root(node_t *root) { +#ifndef NDEBUG + *root = (node_t){ .branch = { + .flags = 3, // invalid value that fits + .bitmap = 0, + .index = -1, + .twigs = NULL + } }; +#endif +} + +/*! \brief Check that unportable code works OK (debug-only). */ +static void assert_portability(void) { +#if FLAGS_HACK + assert(((union node){ .leaf = { + .key = ((void *)NULL) + 1, + .val = NULL + } }).branch.flags == 1); +#endif +} + +/*! \brief Propagate error codes. */ +#define ERR_RETURN(x) \ + do { \ + int err_code_ = x; \ + if (unlikely(err_code_ != KNOT_EOK)) \ + return err_code_; \ + } while (false) + +/*! + * \brief Count the number of set bits. + * + * \TODO This implementation may be relatively slow on some HW. + */ +static uint bitmap_weight(bitmap_t w) +{ + assert((w & ~((1 << 17) - 1)) == 0); // using the least-important 17 bits + return __builtin_popcount(w); +} + +/*! \brief Test flags to determine type of this node. */ +static bool isbranch(const node_t *t) +{ + uint f = t->branch.flags; + assert(f <= 2); + return f != 0; +} + +/*! \brief Make a bitmask for testing a branch bitmap. */ +static bitmap_t nibbit(byte k, uint flags) +{ + uint shift = (2 - flags) << 2; + uint nibble = (k >> shift) & 0xf; + return 1 << (nibble + 1/*because of prefix keys*/); +} + +/*! \brief Extract a nibble from a key and turn it into a bitmask. */ +static bitmap_t twigbit(node_t *t, const char *key, uint32_t len) +{ + assert(isbranch(t)); + uint i = t->branch.index; + + if (i >= len) + return 1 << 0; // leaf position + + return nibbit((byte)key[i], t->branch.flags); +} + +/*! \brief Test if a branch node has a child indicated by a bitmask. */ +static bool hastwig(node_t *t, bitmap_t bit) +{ + assert(isbranch(t)); + return t->branch.bitmap & bit; +} + +/*! \brief Compute offset of an existing child in a branch node. */ +static uint twigoff(node_t *t, bitmap_t b) +{ + assert(isbranch(t)); + return bitmap_weight(t->branch.bitmap & (b - 1)); +} + +/*! \brief Get pointer to a particular child of a branch node. */ +static node_t* twig(node_t *t, uint i) +{ + assert(isbranch(t)); + return &t->branch.twigs[i]; +} + +/*! + * \brief For a branch nod, compute offset of a child and child count. + * + * Having this separate might be meaningful for performance optimization. + */ +#define TWIGOFFMAX(off, max, t, b) do { \ + off = twigoff(t, b); \ + max = bitmap_weight(t->branch.bitmap); \ + } while(0) + +/*! \brief Simple string comparator. */ +static int key_cmp(const char *k1, uint32_t k1_len, const char *k2, uint32_t k2_len) +{ + int ret = memcmp(k1, k2, MIN(k1_len, k2_len)); + if (ret != 0) { + return ret; + } + + /* Key string is equal, compare lengths. */ + if (k1_len == k2_len) { + return 0; + } else if (k1_len < k2_len) { + return -1; + } else { + return 1; + } +} + +trie_t* trie_create(knot_mm_t *mm) +{ + assert_portability(); + trie_t *trie = mm_alloc(mm, sizeof(trie_t)); + if (trie != NULL) { + empty_root(&trie->root); + trie->weight = 0; + if (mm != NULL) + trie->mm = *mm; + else + mm_ctx_init(&trie->mm); + } + return trie; +} + +/*! \brief Free anything under the trie node, except for the passed pointer itself. */ +static void clear_trie(node_t *trie, knot_mm_t *mm) +{ + if (!isbranch(trie)) { + mm_free(mm, trie->leaf.key); + } else { + branch_t *b = &trie->branch; + int len = bitmap_weight(b->bitmap); + for (int i = 0; i < len; ++i) + clear_trie(b->twigs + i, mm); + mm_free(mm, b->twigs); + } +} + +void trie_free(trie_t *tbl) +{ + if (tbl == NULL) + return; + if (tbl->weight) + clear_trie(&tbl->root, &tbl->mm); + mm_free(&tbl->mm, tbl); +} + +void trie_clear(trie_t *tbl) +{ + assert(tbl); + if (!tbl->weight) + return; + clear_trie(&tbl->root, &tbl->mm); + empty_root(&tbl->root); + tbl->weight = 0; +} + +size_t trie_weight(const trie_t *tbl) +{ + assert(tbl); + return tbl->weight; +} + +trie_val_t* trie_get_try(trie_t *tbl, const char *key, uint32_t len) +{ + assert(tbl); + if (!tbl->weight) + return NULL; + node_t *t = &tbl->root; + while (isbranch(t)) { + __builtin_prefetch(t->branch.twigs); + bitmap_t b = twigbit(t, key, len); + if (!hastwig(t, b)) + return NULL; + t = twig(t, twigoff(t, b)); + } + if (key_cmp(key, len, t->leaf.key->chars, t->leaf.key->len) != 0) + return NULL; + return &t->leaf.val; +} + +int trie_del(trie_t *tbl, const char *key, uint32_t len, trie_val_t *val) +{ + assert(tbl); + if (!tbl->weight) + return KNOT_ENOENT; + node_t *t = &tbl->root; // current and parent node + branch_t *p = NULL; + bitmap_t b = 0; + while (isbranch(t)) { + __builtin_prefetch(t->branch.twigs); + b = twigbit(t, key, len); + if (!hastwig(t, b)) + return KNOT_ENOENT; + p = &t->branch; + t = twig(t, twigoff(t, b)); + } + if (key_cmp(key, len, t->leaf.key->chars, t->leaf.key->len) != 0) + return KNOT_ENOENT; + mm_free(&tbl->mm, t->leaf.key); + if (val != NULL) + *val = t->leaf.val; // we return trie_val_t directly when deleting + --tbl->weight; + if (unlikely(!p)) { // whole trie was a single leaf + assert(tbl->weight == 0); + empty_root(&tbl->root); + return KNOT_EOK; + } + // remove leaf t as child of p + int ci = t - p->twigs, // child index via pointer arithmetic + cc = bitmap_weight(p->bitmap); // child count + assert(ci >= 0 && ci < cc); + + if (cc == 2) { // collapse binary node p: move the other child to this node + node_t *twigs = p->twigs; + (*(node_t *)p) = twigs[1 - ci]; // it might be a leaf or branch + mm_free(&tbl->mm, twigs); + return KNOT_EOK; + } + memmove(p->twigs + ci, p->twigs + ci + 1, sizeof(node_t) * (cc - ci - 1)); + p->bitmap &= ~b; + node_t *twigs = mm_realloc(&tbl->mm, p->twigs, sizeof(node_t) * (cc - 1), + sizeof(node_t) * cc); + if (likely(twigs != NULL)) + p->twigs = twigs; + /* We can ignore mm_realloc failure, only beware that next time + * the prev_size passed to it wouldn't be correct; TODO? */ + return KNOT_EOK; +} + +/*! + * \brief Stack of nodes, storing a path down a trie. + * + * The structure also serves directly as the public trie_it_t type, + * in which case it always points to the current leaf, unless we've finished + * (i.e. it->len == 0). + */ +typedef struct trie_it { + node_t* *stack; /*!< The stack; malloc is used directly instead of mm. */ + uint32_t len; /*!< Current length of the stack. */ + uint32_t alen; /*!< Allocated/available length of the stack. */ + /*! \brief Initial storage for \a stack; it should fit in most use cases. */ + node_t* stack_init[2000 / sizeof(node_t *)]; +} nstack_t; + +/*! \brief Create a node stack containing just the root (or empty). */ +static void ns_init(nstack_t *ns, trie_t *tbl) +{ + assert(tbl); + ns->stack = ns->stack_init; + ns->alen = sizeof(ns->stack_init) / sizeof(ns->stack_init[0]); + if (tbl->weight) { + ns->len = 1; + ns->stack[0] = &tbl->root; + } else { + ns->len = 0; + } +} + +/*! \brief Free inside of the stack, i.e. not the passed pointer itself. */ +static void ns_cleanup(nstack_t *ns) +{ + assert(ns && ns->stack); + if (likely(ns->stack == ns->stack_init)) + return; + free(ns->stack); + #ifndef NDEBUG + ns->stack = NULL; + ns->alen = 0; + #endif +} + +/*! \brief Allocate more space for the stack. */ +static int ns_longer_alloc(nstack_t *ns) +{ + ns->alen *= 2; + size_t new_size = sizeof(nstack_t) + ns->alen * sizeof(node_t *); + node_t **st; + if (ns->stack == ns->stack_init) { + st = malloc(new_size); + if (st != NULL) + memcpy(st, ns->stack, ns->len * sizeof(node_t *)); + } else { + st = realloc(ns->stack, new_size); + } + if (st == NULL) + return KNOT_ENOMEM; + ns->stack = st; + return KNOT_EOK; +} + +/*! \brief Ensure the node stack can be extended by one. */ +static inline int ns_longer(nstack_t *ns) +{ + // get a longer stack if needed + if (likely(ns->len < ns->alen)) + return KNOT_EOK; + return ns_longer_alloc(ns); // hand-split the part suitable for inlining +} + +/*! + * \brief Find the "branching point" as if searching for a key. + * + * The whole path to the point is kept on the passed stack; + * always at least the root will remain on the top of it. + * Beware: the precise semantics of this function is rather tricky. + * The top of the stack will contain: the corresponding leaf if exact match is found; + * or the immediate node below a branching-point-on-edge or the branching-point itself. + * + * \param info Set position of the point of first mismatch (in index and flags). + * \param first Set the value of the first non-matching character (from trie), + * optionally; end-of-string character has value -256 (that's why it's int). + * Note: the character is converted to *unsigned* char (i.e. 0..255), + * as that's the ordering used in the trie. + * + * \return KNOT_EOK or KNOT_ENOMEM. + */ +static int ns_find_branch(nstack_t *ns, const char *key, uint32_t len, + branch_t *info, int *first) +{ + assert(ns && ns->len && info); + // First find some leaf with longest matching prefix. + while (isbranch(ns->stack[ns->len - 1])) { + ERR_RETURN(ns_longer(ns)); + node_t *t = ns->stack[ns->len - 1]; + __builtin_prefetch(t->branch.twigs); + bitmap_t b = twigbit(t, key, len); + // Even if our key is missing from this branch we need to + // keep iterating down to a leaf. It doesn't matter which + // twig we choose since the keys are all the same up to this + // index. Note that blindly using twigoff(t, b) can cause + // an out-of-bounds index if it equals twigmax(t). + uint i = hastwig(t, b) ? twigoff(t, b) : 0; + ns->stack[ns->len++] = twig(t, i); + } + tkey_t *lkey = ns->stack[ns->len-1]->leaf.key; + // Find index of the first char that differs. + uint32_t index = 0; + while (index < MIN(len,lkey->len)) { + if (key[index] != lkey->chars[index]) + break; + else + ++index; + } + info->index = index; + if (first) + *first = lkey->len > index ? (unsigned char)lkey->chars[index] : -256; + // Find flags: which half-byte has matched. + uint flags; + if (index == len && len == lkey->len) { // found equivalent key + info->flags = flags = 0; + goto success; + } + if (likely(index < MIN(len,lkey->len))) { + byte k2 = (byte)lkey->chars[index]; + byte k1 = (byte)key[index]; + flags = ((k1 ^ k2) & 0xf0) ? 1 : 2; + } else { // one is prefix of another + flags = 1; + } + info->flags = flags; + // now go up the trie from the current leaf + branch_t *t; + do { + if (unlikely(ns->len == 1)) + goto success; // only the root stays on the stack + t = (branch_t*)ns->stack[ns->len - 2]; + if (t->index < index || (t->index == index && t->flags < flags)) + goto success; + --ns->len; + } while (true); +success: + #ifndef NDEBUG // invariants on successful return + assert(ns->len); + if (isbranch(ns->stack[ns->len - 1])) { + t = &ns->stack[ns->len - 1]->branch; + assert(t->index > index || (t->index == index && t->flags >= flags)); + } + if (ns->len > 1) { + t = &ns->stack[ns->len - 2]->branch; + assert(t->index < index || (t->index == index + && (t->flags < flags || (t->flags == 1 && flags == 0)))); + } + #endif + return KNOT_EOK; +} + +/*! + * \brief Advance the node stack to the last leaf in the subtree. + * + * \return KNOT_EOK or KNOT_ENOMEM. + */ +static int ns_last_leaf(nstack_t *ns) +{ + assert(ns); + do { + ERR_RETURN(ns_longer(ns)); + node_t *t = ns->stack[ns->len - 1]; + if (!isbranch(t)) + return KNOT_EOK; + int lasti = bitmap_weight(t->branch.bitmap) - 1; + assert(lasti >= 0); + ns->stack[ns->len++] = twig(t, lasti); + } while (true); +} + +/*! + * \brief Advance the node stack to the first leaf in the subtree. + * + * \return KNOT_EOK or KNOT_ENOMEM. + */ +static int ns_first_leaf(nstack_t *ns) +{ + assert(ns && ns->len); + do { + ERR_RETURN(ns_longer(ns)); + node_t *t = ns->stack[ns->len - 1]; + if (!isbranch(t)) + return KNOT_EOK; + ns->stack[ns->len++] = twig(t, 0); + } while (true); +} + +/*! + * \brief Advance the node stack to the leaf that is previous to the current node. + * + * \note Prefix leaf under the current node DOES count (if present; perhaps questionable). + * \return KNOT_EOK on success, KNOT_ENOENT on not-found, or possibly KNOT_ENOMEM. + */ +static int ns_prev_leaf(nstack_t *ns) +{ + assert(ns && ns->len > 0); + + node_t *t = ns->stack[ns->len - 1]; + if (hastwig(t, 1 << 0)) { // the prefix leaf + t = twig(t, 0); + ERR_RETURN(ns_longer(ns)); + ns->stack[ns->len++] = t; + return KNOT_EOK; + } + + do { + if (ns->len < 2) + return KNOT_ENOENT; // root without empty key has no previous leaf + t = ns->stack[ns->len - 1]; + node_t *p = ns->stack[ns->len - 2]; + int pindex = t - p->branch.twigs; // index in parent via pointer arithmetic + assert(pindex >= 0 && pindex <= 16); + if (pindex > 0) { // t isn't the first child -> go down the previous one + ns->stack[ns->len - 1] = twig(p, pindex - 1); + return ns_last_leaf(ns); + } + // we've got to go up again + --ns->len; + } while (true); +} + +/*! + * \brief Advance the node stack to the leaf that is successor to the current node. + * + * \note Prefix leaf or anything else under the current node DOES count. + * \return KNOT_EOK on success, KNOT_ENOENT on not-found, or possibly KNOT_ENOMEM. + */ +static int ns_next_leaf(nstack_t *ns) +{ + assert(ns && ns->len > 0); + + node_t *t = ns->stack[ns->len - 1]; + if (isbranch(t)) + return ns_first_leaf(ns); + do { + if (ns->len < 2) + return KNOT_ENOENT; // not found, as no more parent is available + t = ns->stack[ns->len - 1]; + node_t *p = ns->stack[ns->len - 2]; + int pindex = t - p->branch.twigs; // index in parent via pointer arithmetic + assert(pindex >= 0 && pindex <= 16); + int pcount = bitmap_weight(p->branch.bitmap); + if (pindex + 1 < pcount) { // t isn't the last child -> go down the next one + ns->stack[ns->len - 1] = twig(p, pindex + 1); + return ns_first_leaf(ns); + } + // we've got to go up again + --ns->len; + } while (true); +} + +int trie_get_leq(trie_t *tbl, const char *key, uint32_t len, trie_val_t **val) +{ + assert(tbl && val); + *val = NULL; // so on failure we can just return; + if (tbl->weight == 0) + return KNOT_ENOENT; + { // Intentionally un-indented; until end of function, to bound cleanup attr. + // First find a key with longest-matching prefix + __attribute__((cleanup(ns_cleanup))) + nstack_t ns_local; + ns_init(&ns_local, tbl); + nstack_t *ns = &ns_local; + branch_t bp; + int un_leaf; // first unmatched character in the leaf + ERR_RETURN(ns_find_branch(ns, key, len, &bp, &un_leaf)); + int un_key = bp.index < len ? (unsigned char)key[bp.index] : -256; + node_t *t = ns->stack[ns->len - 1]; + if (bp.flags == 0) { // found exact match + *val = &t->leaf.val; + return KNOT_EOK; + } + // Get t: the last node on matching path + if (isbranch(t) && t->branch.index == bp.index && t->branch.flags == bp.flags) { + // t is OK + } else { + // the top of the stack was the first unmatched node -> step up + if (ns->len == 1) { + // root was unmatched already + if (un_key < un_leaf) + return KNOT_ENOENT; + ERR_RETURN(ns_last_leaf(ns)); + goto success; + } + --ns->len; + t = ns->stack[ns->len - 1]; + } + // Now we re-do the first "non-matching" step in the trie + // but try the previous child if key was less (it may not exist) + bitmap_t b = twigbit(t, key, len); + int i = hastwig(t, b) + ? twigoff(t, b) - (un_key < un_leaf) + : twigoff(t, b) - 1 /*twigoff returns successor when !hastwig*/; + if (i >= 0) { + ERR_RETURN(ns_longer(ns)); + ns->stack[ns->len++] = twig(t, i); + ERR_RETURN(ns_last_leaf(ns)); + } else { + ERR_RETURN(ns_prev_leaf(ns)); + } +success: + assert(!isbranch(ns->stack[ns->len - 1])); + *val = &ns->stack[ns->len - 1]->leaf.val; + return 1; + } +} + +/*! \brief Initialize a new leaf, copying the key, and returning failure code. */ +static int mk_leaf(node_t *leaf, const char *key, uint32_t len, knot_mm_t *mm) +{ + tkey_t *k = mm_alloc(mm, sizeof(tkey_t) + len); + #if FLAGS_HACK + assert(((uintptr_t)k) % 4 == 0); // we need an aligned pointer + #endif + if (unlikely(!k)) + return KNOT_ENOMEM; + k->len = len; + memcpy(k->chars, key, len); + leaf->leaf = (leaf_t){ + #if !FLAGS_HACK + .flags = 0, + #endif + .val = NULL, + .key = k + }; + return KNOT_EOK; +} + +trie_val_t* trie_get_ins(trie_t *tbl, const char *key, uint32_t len) +{ + assert(tbl); + // First leaf in an empty tbl? + if (unlikely(!tbl->weight)) { + if (unlikely(mk_leaf(&tbl->root, key, len, &tbl->mm))) + return NULL; + ++tbl->weight; + return &tbl->root.leaf.val; + } + { // Intentionally un-indented; until end of function, to bound cleanup attr. + // Find the branching-point + __attribute__((cleanup(ns_cleanup))) + nstack_t ns_local; + ns_init(&ns_local, tbl); + nstack_t *ns = &ns_local; + branch_t bp; // branch-point: index and flags signifying the longest common prefix + int k2; // the first unmatched character in the leaf + if (unlikely(ns_find_branch(ns, key, len, &bp, &k2))) + return NULL; + node_t *t = ns->stack[ns->len - 1]; + if (bp.flags == 0) // the same key was already present + return &t->leaf.val; + node_t leaf; + if (unlikely(mk_leaf(&leaf, key, len, &tbl->mm))) + return NULL; + + if (isbranch(t) && bp.index == t->branch.index && bp.flags == t->branch.flags) { + // The node t needs a new leaf child. + bitmap_t b1 = twigbit(t, key, len); + assert(!hastwig(t, b1)); + uint s, m; TWIGOFFMAX(s, m, t, b1); // new child position and original child count + node_t *twigs = mm_realloc(&tbl->mm, t->branch.twigs, + sizeof(node_t) * (m + 1), sizeof(node_t) * m); + if (unlikely(!twigs)) + goto err_leaf; + memmove(twigs + s + 1, twigs + s, sizeof(node_t) * (m - s)); + twigs[s] = leaf; + t->branch.twigs = twigs; + t->branch.bitmap |= b1; + ++tbl->weight; + return &twigs[s].leaf.val; + } else { + // We need to insert a new binary branch with leaf at *t. + // Note: it works the same for the case where we insert above root t. + #ifndef NDEBUG + if (ns->len > 1) { + node_t *pt = ns->stack[ns->len - 2]; + assert(hastwig(pt, twigbit(pt, key, len))); + } + #endif + node_t *twigs = mm_alloc(&tbl->mm, sizeof(node_t) * 2); + if (unlikely(!twigs)) + goto err_leaf; + node_t t2 = *t; // Save before overwriting t. + t->branch.flags = bp.flags; + t->branch.index = bp.index; + t->branch.twigs = twigs; + bitmap_t b1 = twigbit(t, key, len); + bitmap_t b2 = unlikely(k2 == -256) ? (1 << 0) : nibbit(k2, bp.flags); + t->branch.bitmap = b1 | b2; + *twig(t, twigoff(t, b1)) = leaf; + *twig(t, twigoff(t, b2)) = t2; + ++tbl->weight; + return &twig(t, twigoff(t, b1))->leaf.val; + }; +err_leaf: + mm_free(&tbl->mm, leaf.leaf.key); + return NULL; + } +} + +/*! \brief Apply a function to every trie_val_t*, in order; a recursive solution. */ +static int apply_trie(node_t *t, int (*f)(trie_val_t *, void *), void *d) +{ + assert(t); + if (!isbranch(t)) + return f(&t->leaf.val, d); + int child_count = bitmap_weight(t->branch.bitmap); + for (int i = 0; i < child_count; ++i) + ERR_RETURN(apply_trie(twig(t, i), f, d)); + return KNOT_EOK; +} + +int trie_apply(trie_t *tbl, int (*f)(trie_val_t *, void *), void *d) +{ + assert(tbl && f); + if (!tbl->weight) + return KNOT_EOK; + return apply_trie(&tbl->root, f, d); +} + +/* These are all thin wrappers around static Tns* functions. */ +trie_it_t* trie_it_begin(trie_t *tbl) +{ + assert(tbl); + trie_it_t *it = malloc(sizeof(nstack_t)); + if (!it) + return NULL; + ns_init(it, tbl); + if (it->len == 0) // empty tbl + return it; + if (ns_first_leaf(it)) { + ns_cleanup(it); + free(it); + return NULL; + } + return it; +} + +void trie_it_next(trie_it_t *it) +{ + assert(it && it->len); + if (ns_next_leaf(it) != KNOT_EOK) + it->len = 0; +} + +bool trie_it_finished(trie_it_t *it) +{ + assert(it); + return it->len == 0; +} + +void trie_it_free(trie_it_t *it) +{ + if (!it) + return; + ns_cleanup(it); + free(it); +} + +const char* trie_it_key(trie_it_t *it, size_t *len) +{ + assert(it && it->len); + node_t *t = it->stack[it->len - 1]; + assert(!isbranch(t)); + tkey_t *key = t->leaf.key; + if (len) + *len = key->len; + return key->chars; +} + +trie_val_t* trie_it_val(trie_it_t *it) +{ + assert(it && it->len); + node_t *t = it->stack[it->len - 1]; + assert(!isbranch(t)); + return &t->leaf.val; +} diff --git a/src/contrib/qp-trie/trie.h b/src/contrib/qp-trie/trie.h new file mode 100644 index 0000000..28ea571 --- /dev/null +++ b/src/contrib/qp-trie/trie.h @@ -0,0 +1,112 @@ +/* Copyright (C) 2017 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ + +#pragma once + +#include <stdbool.h> +#include <stdint.h> + +#include "libknot/mm_ctx.h" + +/*! + * \brief Native API of QP-tries: + * + * - keys are char strings, not necessarily zero-terminated, + * the structure copies the contents of the passed keys + * - values are void* pointers, typically you get an ephemeral pointer to it + * - key lengths are limited by 2^32-1 ATM + */ + +/*! \brief Element value. */ +typedef void* trie_val_t; + +/*! \brief Opaque structure holding a QP-trie. */ +typedef struct trie trie_t; + +/*! \brief Opaque type for holding a QP-trie iterator. */ +typedef struct trie_it trie_it_t; + +/*! \brief Create a trie instance. */ +trie_t* trie_create(knot_mm_t *mm); + +/*! \brief Free a trie instance. */ +void trie_free(trie_t *tbl); + +/*! \brief Clear a trie instance (make it empty). */ +void trie_clear(trie_t *tbl); + +/*! \brief Return the number of keys in the trie. */ +size_t trie_weight(const trie_t *tbl); + +/*! \brief Search the trie, returning NULL on failure. */ +trie_val_t* trie_get_try(trie_t *tbl, const char *key, uint32_t len); + +/*! \brief Search the trie, inserting NULL trie_val_t on failure. */ +trie_val_t* trie_get_ins(trie_t *tbl, const char *key, uint32_t len); + +/*! + * \brief Search for less-or-equal element. + * + * \param tbl Trie. + * \param key Searched key. + * \param len Key length. + * \param val Must be valid; it will be set to NULL if not found or errored. + * \return KNOT_EOK for exact match, 1 for previous, KNOT_ENOENT for not-found, + * or KNOT_E*. + */ +int trie_get_leq(trie_t *tbl, const char *key, uint32_t len, trie_val_t **val); + +/*! + * \brief Apply a function to every trie_val_t, in order. + * + * \return KNOT_EOK if success or KNOT_E* if error. + */ +int trie_apply(trie_t *tbl, int (*f)(trie_val_t *, void *), void *d); + +/*! + * \brief Remove an item, returning KNOT_EOK if succeeded or KNOT_ENOENT if not found. + * + * If val!=NULL and deletion succeeded, the deleted value is set. + */ +int trie_del(trie_t *tbl, const char *key, uint32_t len, trie_val_t *val); + +/*! \brief Create a new iterator pointing to the first element (if any). */ +trie_it_t* trie_it_begin(trie_t *tbl); + +/*! + * \brief Advance the iterator to the next element. + * + * Iteration is in ascending lexicographical order. + * In particular, the empty string would be considered as the very first. + */ +void trie_it_next(trie_it_t *it); + +/*! \brief Test if the iterator has gone past the last element. */ +bool trie_it_finished(trie_it_t *it); + +/*! \brief Free any resources of the iterator. It's OK to call it on NULL. */ +void trie_it_free(trie_it_t *it); + +/*! + * \brief Return pointer to the key of the current element. + * + * \note The len is uint32_t internally but size_t is better for our usage + * as it is without an additional type conversion. + */ +const char* trie_it_key(trie_it_t *it, size_t *len); + +/*! \brief Return pointer to the value of the current element (writable). */ +trie_val_t* trie_it_val(trie_it_t *it); diff --git a/src/contrib/sockaddr.c b/src/contrib/sockaddr.c new file mode 100644 index 0000000..d569292 --- /dev/null +++ b/src/contrib/sockaddr.c @@ -0,0 +1,351 @@ +/* Copyright (C) 2016 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ + +#include <stdio.h> +#include <stdlib.h> +#include <string.h> +#include <netdb.h> + +#include "libknot/consts.h" +#include "libknot/errcode.h" +#include "contrib/sockaddr.h" +#include "contrib/openbsd/strlcpy.h" +#include "contrib/macros.h" + +int sockaddr_len(const struct sockaddr *sa) +{ + if (sa == NULL) { + return 0; + } + + switch(sa->sa_family) { + case AF_INET: + return sizeof(struct sockaddr_in); + case AF_INET6: + return sizeof(struct sockaddr_in6); + case AF_UNIX: + return sizeof(struct sockaddr_un); + default: + return 0; + } +} + +static int cmp_ipv4(const struct sockaddr_in *a, const struct sockaddr_in *b) +{ + if (a->sin_addr.s_addr < b->sin_addr.s_addr) { + return -1; + } else if (a->sin_addr.s_addr > b->sin_addr.s_addr) { + return 1; + } else { + return a->sin_port - b->sin_port; + } +} + +static int cmp_ipv6(const struct sockaddr_in6 *a, const struct sockaddr_in6 *b) +{ + int ret = memcmp(&a->sin6_addr, &b->sin6_addr, sizeof(struct in6_addr)); + if (ret == 0) { + ret = a->sin6_port - b->sin6_port; + } + + return ret; +} + +static int cmp_unix(const struct sockaddr_un *a, const struct sockaddr_un *b) +{ + int len_a = strnlen(a->sun_path, sizeof(a->sun_path)); + int len_b = strnlen(b->sun_path, sizeof(b->sun_path)); + int len_min = len_a <= len_b ? len_a : len_b; + + int ret = strncmp(a->sun_path, b->sun_path, len_min); + if (ret == 0) { + ret = len_a - len_b; + } + + return ret; +} + +int sockaddr_cmp(const struct sockaddr *a, const struct sockaddr *b) +{ + if (a->sa_family != b->sa_family) { + return (int)a->sa_family - (int)b->sa_family; + } + + switch (a->sa_family) { + case AF_UNSPEC: + return 0; + case AF_INET: + return cmp_ipv4((struct sockaddr_in *)a, (struct sockaddr_in *)b); + case AF_INET6: + return cmp_ipv6((struct sockaddr_in6 *)a, (struct sockaddr_in6 *)b); + case AF_UNIX: + return cmp_unix((struct sockaddr_un *)a, (struct sockaddr_un *)b); + default: + return 1; + } +} + +int sockaddr_set(struct sockaddr_storage *ss, int family, const char *straddr, int port) +{ + if (ss == NULL || straddr == NULL) { + return KNOT_EINVAL; + } + + /* Set family and port. */ + memset(ss, 0, sizeof(*ss)); + ss->ss_family = family; + sockaddr_port_set((struct sockaddr *)ss, port); + + /* Initialize address depending on address family. */ + if (family == AF_INET6) { + struct sockaddr_in6 *ipv6 = (struct sockaddr_in6 *)ss; + if (inet_pton(family, straddr, &ipv6->sin6_addr) < 1) { + return KNOT_ERROR; + } + return KNOT_EOK; + } else if (family == AF_INET) { + struct sockaddr_in *ipv4 = (struct sockaddr_in *)ss; + if (inet_pton(family, straddr, &ipv4->sin_addr) < 1) { + return KNOT_ERROR; + } + return KNOT_EOK; + } else if (family == AF_UNIX) { + struct sockaddr_un *un = (struct sockaddr_un *)ss; + size_t ret = strlcpy(un->sun_path, straddr, sizeof(un->sun_path)); + if (ret >= sizeof(un->sun_path)) { + return KNOT_ESPACE; + } + return KNOT_EOK; + } + + return KNOT_EINVAL; +} + +void *sockaddr_raw(const struct sockaddr *sa, size_t *addr_size) +{ + if (sa == NULL || addr_size == NULL) { + return NULL; + } + + if (sa->sa_family == AF_INET) { + struct sockaddr_in *ipv4 = (struct sockaddr_in *)sa; + *addr_size = sizeof(ipv4->sin_addr); + return &ipv4->sin_addr; + } else if (sa->sa_family == AF_INET6) { + struct sockaddr_in6 *ipv6 = (struct sockaddr_in6 *)sa; + *addr_size = sizeof(ipv6->sin6_addr); + return &ipv6->sin6_addr; + } else { + return NULL; + } +} + +int sockaddr_set_raw(struct sockaddr_storage *ss, int family, + const uint8_t *raw_addr, size_t raw_addr_size) +{ + if (ss == NULL || raw_addr == NULL) { + return KNOT_EINVAL; + } + + memset(ss, 0, sizeof(*ss)); + ss->ss_family = family; + + size_t sa_size = 0; + void *sa_data = sockaddr_raw((struct sockaddr *)ss, &sa_size); + if (sa_data == NULL || sa_size != raw_addr_size) { + return KNOT_EINVAL; + } + + memcpy(sa_data, raw_addr, sa_size); + + return KNOT_EOK; +} + +int sockaddr_tostr(char *buf, size_t maxlen, const struct sockaddr *sa) +{ + if (sa == NULL || buf == NULL) { + return KNOT_EINVAL; + } + + const char *out = NULL; + + /* Convert network address string. */ + if (sa->sa_family == AF_INET6) { + const struct sockaddr_in6 *s = (const struct sockaddr_in6 *)sa; + out = inet_ntop(sa->sa_family, &s->sin6_addr, buf, maxlen); + } else if (sa->sa_family == AF_INET) { + const struct sockaddr_in *s = (const struct sockaddr_in *)sa; + out = inet_ntop(sa->sa_family, &s->sin_addr, buf, maxlen); + } else if (sa->sa_family == AF_UNIX) { + const struct sockaddr_un *s = (const struct sockaddr_un *)sa; + size_t ret = strlcpy(buf, s->sun_path, maxlen); + out = (ret < maxlen) ? buf : NULL; + } else { + return KNOT_EINVAL; + } + + if (out == NULL) { + *buf = '\0'; + return KNOT_ESPACE; + } + + /* Write separator and port. */ + int written = strlen(buf); + int port = sockaddr_port(sa); + if (port > 0) { + int ret = snprintf(&buf[written], maxlen - written, "@%d", port); + if (ret <= 0 || (size_t)ret >= maxlen - written) { + *buf = '\0'; + return KNOT_ESPACE; + } + + written += ret; + } + + return written; +} + +int sockaddr_port(const struct sockaddr *sa) +{ + if (sa == NULL) { + return KNOT_EINVAL; + } + + if (sa->sa_family == AF_INET6) { + return ntohs(((struct sockaddr_in6 *)sa)->sin6_port); + } else if (sa->sa_family == AF_INET) { + return ntohs(((struct sockaddr_in *)sa)->sin_port); + } else { + return KNOT_EINVAL; + } +} + +void sockaddr_port_set(struct sockaddr *sa, uint16_t port) +{ + if (sa == NULL) { + return; + } + + if (sa->sa_family == AF_INET6) { + ((struct sockaddr_in6 *)sa)->sin6_port = htons(port); + } else if (sa->sa_family == AF_INET) { + ((struct sockaddr_in *)sa)->sin_port = htons(port); + } +} + +char *sockaddr_hostname(void) +{ + /* Fetch hostname. */ + char host[KNOT_DNAME_MAXLEN + 1] = { '\0' }; + if (gethostname(host, sizeof(host)) != 0) { + return NULL; + } + /* Just to be sure. */ + host[sizeof(host) - 1] = '\0'; + + /* Fetch canonical name for this address/DNS. */ + struct addrinfo hints, *info = NULL; + memset(&hints, 0, sizeof hints); + hints.ai_family = AF_UNSPEC; + hints.ai_socktype = SOCK_DGRAM; + hints.ai_flags = AI_CANONNAME; + if (getaddrinfo(host, "domain", &hints, &info) != 0) { + return NULL; + } + + /* Fetch first valid hostname. */ + char *hname = NULL; + struct addrinfo *p = NULL; + for (p = info; p != NULL; p = p->ai_next) { + if (p->ai_canonname) { + hname = strdup(p->ai_canonname); + break; + } + } + + /* No valid hostname found, resort to gethostname() result */ + if (hname == NULL) { + hname = strdup(host); + } + + freeaddrinfo(info); + return hname; +} + +bool sockaddr_is_any(const struct sockaddr *sa) +{ + if (sa == NULL) { + return false; + } + + if (sa->sa_family == AF_INET) { + const struct sockaddr_in *ipv4 = (struct sockaddr_in *)sa; + return ipv4->sin_addr.s_addr == INADDR_ANY; + } + + if (sa->sa_family == AF_INET6) { + const struct sockaddr_in6 *ipv6 = (struct sockaddr_in6 *)sa; + return memcmp(&ipv6->sin6_addr, &in6addr_any, sizeof(ipv6->sin6_addr)) == 0; + } + + return false; +} + +bool sockaddr_net_match(const struct sockaddr *ss1, + const struct sockaddr *ss2, + unsigned prefix) +{ + if (ss1 == NULL || ss2 == NULL) { + return false; + } + + if (ss1->sa_family != ss2->sa_family) { + return false; + } + + size_t raw_len = 0; + const uint8_t *raw_1 = sockaddr_raw(ss1, &raw_len); + const uint8_t *raw_2 = sockaddr_raw(ss2, &raw_len); + + prefix = MIN(prefix, raw_len * 8); + unsigned bytes = prefix / 8; + unsigned bits = prefix % 8; + + /* Compare full bytes. */ + if (memcmp(raw_1, raw_2, bytes) != 0) { + return false; + } + + /* Compare last partial byte. */ + return bits == 0 || + (raw_1[bytes] >> (8 - bits) == raw_2[bytes] >> (8 - bits)); +} + +bool sockaddr_range_match(const struct sockaddr *sa, + const struct sockaddr *ss_min, + const struct sockaddr *ss_max) +{ + if (sa == NULL || ss_min == NULL || ss_max == NULL) { + return false; + } + + if (ss_min->sa_family != ss_max->sa_family || + ss_min->sa_family != sa->sa_family) { + return false; + } + + return sockaddr_cmp(sa, ss_min) >= 0 && sockaddr_cmp(sa, ss_max) <= 0; +} diff --git a/src/contrib/sockaddr.h b/src/contrib/sockaddr.h new file mode 100644 index 0000000..0fbf1ab --- /dev/null +++ b/src/contrib/sockaddr.h @@ -0,0 +1,160 @@ +/* Copyright (C) 2017 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ + +#pragma once + +/* BSD IPv6 */ +#ifndef __POSIX_VISIBLE +#define __POSIX_VISIBLE 200112 +#endif + +#include <stdbool.h> +#include <sys/socket.h> +#include <sys/types.h> +#include <sys/un.h> +#include <netinet/in.h> +#include <arpa/inet.h> +#include <stdint.h> +#include <unistd.h> + +/* Subnet maximum prefix length. */ +#define IPV4_PREFIXLEN 32 +#define IPV6_PREFIXLEN 128 + +/* Address string "address[@port]" maximum length. */ +#define SOCKADDR_STRLEN_EXT (1 + 6) /* '@', 5 digits number, \0 */ +#define SOCKADDR_STRLEN (sizeof(struct sockaddr_un) + SOCKADDR_STRLEN_EXT) + +/*! + * \brief Calculate current structure length based on address family. + * + * \param sa Socket address. + * + * \return Number of bytes or error code. + */ +int sockaddr_len(const struct sockaddr *sa); + +/*! + * \brief Compare addresses. + * + * \return like memcmp(3) + */ +int sockaddr_cmp(const struct sockaddr *k1, const struct sockaddr *k2); + +/*! + * \brief Set address and port. + * + * \param ss Socket address. + * \param family Address family. + * \param straddr IP address in string format. + * \param port Port. + * + * \return KNOT_EOK on success or an error code. + */ +int sockaddr_set(struct sockaddr_storage *ss, int family, const char *straddr, int port); + +/*! + * \brief Return raw network address in network byte order. + * + * \param[in] sa Socket address. + * \param[out] addr_size Address length. + * + * \return Pointer to binary buffer of size addr_size. + */ +void *sockaddr_raw(const struct sockaddr *sa, size_t *addr_size); + +/*! + * \brief Set raw address. + * + * \param ss Socket address. + * \param family Address family. + * \param raw_addr IP address in binary format. + * \param raw_addr_size Size of the binary address. + * + * \return KNOT_EOK on success or an error code. + */ +int sockaddr_set_raw(struct sockaddr_storage *ss, int family, + const uint8_t *raw_addr, size_t raw_addr_size); + +/*! + * \brief Return string representation of socket address. + * + * \note String format: \<address>[@<port>], f.e. '127.0.0.1@53' + * + * \param buf Destination for string representation. + * \param maxlen Maximum number of written bytes. + * \param sa Socket address. + * + * \return Number of bytes written on success, error code on failure. + */ +int sockaddr_tostr(char *buf, size_t maxlen, const struct sockaddr *sa); + +/*! + * \brief Return port number from address. + * + * \param sa Socket address. + * + * \return Port number or error code. + */ +int sockaddr_port(const struct sockaddr *sa); + +/*! + * \brief Set port number. + * + * \param sa Socket address. + * \param port Port to set. + */ +void sockaddr_port_set(struct sockaddr *sa, uint16_t port); + +/*! + * \brief Get host FQDN address. + * + * \return Hostname string or NULL. + */ +char *sockaddr_hostname(void); + +/*! + * \brief Check if address is ANY address. + * + * \param sa Socket address. + */ +bool sockaddr_is_any(const struct sockaddr *sa); + +/*! + * \brief Check if two addresses match the given network prefix. + * + * \param sa1 First address. + * \param sa2 Second address. + * \param prefix Prefix length. + * + * \return True on match. + */ +bool sockaddr_net_match(const struct sockaddr *sa1, + const struct sockaddr *sa2, + unsigned prefix); + +/*! + * \brief Check if the address is within the given address range (inclusive). + * + * \param sa Address to check. + * \param sa_min Minimum address. + * \param sa_max Maximum address. + * + * \return True on match. + */ +bool sockaddr_range_match(const struct sockaddr *sa, + const struct sockaddr *sa_min, + const struct sockaddr *sa_max); diff --git a/src/contrib/string.c b/src/contrib/string.c new file mode 100644 index 0000000..6e0eea9 --- /dev/null +++ b/src/contrib/string.c @@ -0,0 +1,112 @@ +/* Copyright (C) 2017 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. +*/ + +#include <stdarg.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> + +#include "contrib/string.h" +#include "contrib/ctype.h" + +uint8_t *memdup(const uint8_t *data, size_t data_size) +{ + uint8_t *result = (uint8_t *)malloc(data_size); + if (!result) { + return NULL; + } + + return memcpy(result, data, data_size); +} + +char *sprintf_alloc(const char *fmt, ...) +{ + char *strp = NULL; + va_list ap; + + va_start(ap, fmt); + int ret = vasprintf(&strp, fmt, ap); + va_end(ap); + + if (ret < 0) { + return NULL; + } + return strp; +} + +char *strcdup(const char *s1, const char *s2) +{ + if (!s1 || !s2) { + return NULL; + } + + size_t s1len = strlen(s1); + size_t s2len = strlen(s2); + size_t nlen = s1len + s2len + 1; + + char* dst = malloc(nlen); + if (dst == NULL) { + return NULL; + } + + memcpy(dst, s1, s1len); + memcpy(dst + s1len, s2, s2len + 1); + return dst; +} + +char *strstrip(const char *str) +{ + // leading white-spaces + const char *scan = str; + while (is_space(scan[0])) { + scan += 1; + } + + // trailing white-spaces + size_t len = strlen(scan); + while (len > 0 && is_space(scan[len - 1])) { + len -= 1; + } + + char *trimmed = malloc(len + 1); + if (!trimmed) { + return NULL; + } + + memcpy(trimmed, scan, len); + trimmed[len] = '\0'; + + return trimmed; +} + +int const_time_memcmp(const void *s1, const void *s2, size_t n) +{ + volatile uint8_t equal = 0; + + for (size_t i = 0; i < n; i++) { + equal |= ((uint8_t *)s1)[i] ^ ((uint8_t *)s2)[i]; + } + + return equal; +} + +typedef void *(*memset_t)(void *, int, size_t); +static volatile memset_t volatile_memset = memset; + +void *memzero(void *s, size_t n) +{ + return volatile_memset(s, 0, n); +} diff --git a/src/contrib/string.h b/src/contrib/string.h new file mode 100644 index 0000000..19c74c7 --- /dev/null +++ b/src/contrib/string.h @@ -0,0 +1,82 @@ +/* Copyright (C) 2018 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ +/*! + * \brief String manipulations. + */ + +#pragma once + +#include <stddef.h> +#include <stdint.h> + +/*! + * \brief Create a copy of a binary buffer. + * + * Like \c strdup, but for binary data. + */ +uint8_t *memdup(const uint8_t *data, size_t data_size); + +/*! + * \brief Format string and take care of allocating memory. + * + * \note sprintf(3) manual page reference implementation. + * + * \param fmt Message format. + * \return formatted message or NULL. + */ +char *sprintf_alloc(const char *fmt, ...); + +/*! + * \brief Create new string from a concatenation of s1 and s2. + * + * \param s1 First string. + * \param s2 Second string. + * + * \retval Newly allocated string on success. + * \retval NULL on error. + */ +char *strcdup(const char *s1, const char *s2); + +/*! + * \brief Create a copy of a string skipping leading and trailing white spaces. + * + * \return Newly allocated string, NULL in case of error. + */ +char *strstrip(const char *str); + +/*! + * \brief Compare data in time based on string length. + * This function just checks for (in)equality not for relation + * + * \param s1 The first address to compare. + * \param s2 The second address to compare. + * \param n The size of memory to compare. + * + * \return Non zero on difference and zero if the buffers are identical. + */ +int const_time_memcmp(const void *s1, const void *s2, size_t n); + +/*! + * \brief Fill memory with zeroes. + * + * Inspired by OPENSSL_cleanse. Such a memset shouldn't be optimized out. + * + * \param s The address to fill. + * \param n The size of memory to fill. + * + * \return Pointer to the memory. + */ +void *memzero(void *s, size_t n); diff --git a/src/contrib/strtonum.h b/src/contrib/strtonum.h new file mode 100644 index 0000000..6cfe8bb --- /dev/null +++ b/src/contrib/strtonum.h @@ -0,0 +1,120 @@ +/* Copyright (C) 2018 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. +*/ + +#pragma once + +#include <assert.h> +#include <errno.h> +#include <inttypes.h> +#include <limits.h> +#include <stddef.h> +#include <stdint.h> + +#include "libknot/errcode.h" +#include "contrib/ctype.h" + +inline static int intmax_from_str(const char *src, intmax_t *dst, + intmax_t min, intmax_t max) +{ + if (!is_digit(*src) && *src != '-' && *src != '+') { + return KNOT_EINVAL; + } + + errno = 0; + char *end = NULL; + intmax_t result = strtoimax(src, &end, 10); + + if (errno == ERANGE) { + return KNOT_ERANGE; + } + + if (src == end || *end != '\0') { + return KNOT_EINVAL; + } + + if (result < min || result > max) { + return KNOT_ERANGE; + } + + *dst = result; + return KNOT_EOK; +} + +inline static int uintmax_from_str(const char *src, uintmax_t *dst, + uintmax_t min, uintmax_t max) +{ + if (!is_digit(*src) && *src != '+') { + return KNOT_EINVAL; + } + + errno = 0; + char *end = NULL; + uintmax_t result = strtoumax(src, &end, 10); + + if (errno == ERANGE) { + return KNOT_ERANGE; + } + + if (src == end || *end != '\0') { + return KNOT_EINVAL; + } + + if (result < min || result > max) { + return KNOT_ERANGE; + } + + *dst = result; + return KNOT_EOK; +} + +#define CONVERT(prefix, type, min, max, src, dst) \ +{ \ + assert(src && dst); \ + prefix##max_t value; \ + int result = prefix##max_from_str(src, &value, min, max); \ + if (result != KNOT_EOK) { \ + return result; \ + } \ + *dst = (type)value; \ + return KNOT_EOK; \ +} + +inline static int str_to_int(const char *src, int *dst, int min, int max) +{ + CONVERT(int, int, min, max, src, dst); +} + +inline static int str_to_u8(const char *src, uint8_t *dst) +{ + CONVERT(uint, uint8_t, 0, UINT8_MAX, src, dst); +} + +inline static int str_to_u16(const char *src, uint16_t *dst) +{ + CONVERT(uint, uint16_t, 0, UINT16_MAX, src, dst); +} + +inline static int str_to_u32(const char *src, uint32_t *dst) +{ + CONVERT(uint, uint32_t, 0, UINT32_MAX, src, dst); +} + +inline static int str_to_size(const char *src, size_t *dst, size_t min, size_t max) +{ + CONVERT(uint, size_t, min, max, src, dst); +} + +#undef CONVERT diff --git a/src/contrib/time.c b/src/contrib/time.c new file mode 100644 index 0000000..8c32c2d --- /dev/null +++ b/src/contrib/time.c @@ -0,0 +1,405 @@ +/* Copyright (C) 2017 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ + +#include <assert.h> +#include <limits.h> +#include <stdbool.h> +#include <stdio.h> +#include <stdlib.h> +#include <string.h> + +#include "contrib/time.h" +#include "contrib/ctype.h" +#ifndef HAVE_CLOCK_GETTIME + #include <sys/time.h> +#endif + +struct timespec time_now(void) +{ + struct timespec result = { 0 }; + +#ifdef HAVE_CLOCK_GETTIME + clock_gettime(CLOCK_MONOTONIC, &result); +#else // OS X < Sierra fallback. + struct timeval tmp = { 0 }; + gettimeofday(&tmp, NULL); + result.tv_sec = tmp.tv_sec; + result.tv_nsec = 1000 * tmp.tv_usec; +#endif + + return result; +} + +struct timespec time_diff(const struct timespec *begin, const struct timespec *end) +{ + struct timespec result = { 0 }; + + if (end->tv_nsec >= begin->tv_nsec) { + result.tv_sec = end->tv_sec - begin->tv_sec; + result.tv_nsec = end->tv_nsec - begin->tv_nsec; + } else { + result.tv_sec = end->tv_sec - begin->tv_sec - 1; + result.tv_nsec = 1000000000 - begin->tv_nsec + end->tv_nsec; + } + + return result; +} + +double time_diff_ms(const struct timespec *begin, const struct timespec *end) +{ + struct timespec result = time_diff(begin, end); + + return (result.tv_sec * 1e3) + (result.tv_nsec / 1e6); +} + +typedef struct { + const char *format; + const char *timespec; + const char *parsed; + knot_timediff_t offset; + char offset_sign; + char offset_unit; + struct tm calendar; + int error; +} time_ctx_t; + +// After casting (struct tm) to (int []), we can use indexes... +static int calendar_index(char ind) +{ + switch (ind) { + case 'Y': return 5; + case 'M': return 4; + case 'D': return 3; + case 'h': return 2; + case 'm': return 1; + case 's': return 0; + default: assert(0); return 6; + } +} + +static size_t calendar_digits(int index) +{ + return index == 5 ? 4 : 2; +} + +static size_t unit_value(char unit) +{ + size_t val = 1; + switch (unit) { + case 'M': + return 3600 * 24 * 30; + case 'Y': + val *= 365; + // FALLTHROUGH + case 'D': + val *= 24; + // FALLTHROUGH + case 'h': + val *= 60; + // FALLTHROUGH + case 'm': + val *= 60; + // FALLTHROUGH + case 's': + default: + return val; + } +} + +static knot_time_t time_ctx_finalize(time_ctx_t *ctx) +{ + if (ctx->offset_sign) { + ctx->offset *= unit_value(ctx->offset_unit); + return knot_time_add(knot_time(), (ctx->offset_sign == '-' ? -1 : 1) * ctx->offset); + } else if (ctx->offset) { + return (knot_time_t)ctx->offset; + } else if (ctx->calendar.tm_year != 0) { + ctx->calendar.tm_isdst = -1; + ctx->calendar.tm_year -= 1900; + ctx->calendar.tm_mon -= 1; + // Set UTC timezone before using mktime + putenv("TZ=UTC"); + tzset(); + return (knot_time_t)mktime(&ctx->calendar); + } else { + return (knot_time_t)0; + } +} + +static void time_ctx_reset(time_ctx_t *ctx) +{ + ctx->parsed = ctx->timespec; + ctx->offset = 0; + ctx->offset_sign = 0; + memset(&ctx->calendar, 0, sizeof(ctx->calendar)); + ctx->error = 0; +} + +static void parse_quote(time_ctx_t *ctx) +{ + while (*ctx->format != '|' && *ctx->format != '\0') { + if (*ctx->format == '\'') { + ctx->format++; + return; + } + if (*ctx->format++ != *ctx->parsed++) { + ctx->error = -1; + return; + } + } + ctx->error = -2; + return; +} + +static void parse_offset(time_ctx_t *ctx) +{ + ctx->offset = 0; + ctx->error = -1; + while (is_digit(*ctx->parsed)) { + ctx->offset *= 10; + ctx->offset += *ctx->parsed++ - '0'; + ctx->error = 0; + } +} + +static void parse_calendar(time_ctx_t *ctx, int index) +{ + int *cal_arr = (int *)&ctx->calendar; + cal_arr[index] = 0; + for (size_t i = 0; i < calendar_digits(index); i++) { + if (!is_digit(*ctx->parsed)) { + ctx->error = -1; + return; + } + cal_arr[index] *= 10; + cal_arr[index] += *ctx->parsed++ - '0'; + } +} + +static void parse_sign(time_ctx_t *ctx) +{ + char sign1 = *(ctx->format - 1), sign2 = *ctx->format; + + bool use_sign2 = (sign2 == '+' || sign2 == '-'); + + bool allow_plus = (sign1 == '+' || (sign1 == '-' && sign2 == '+')); + bool allow_minus = (sign1 == '-' || (sign1 == '+' && sign2 == '-')); + assert(sign1 == '+' || sign1 == '-'); + + if ((*ctx->parsed == '+' && allow_plus) || (*ctx->parsed == '-' && allow_minus)) { + ctx->offset_sign = *ctx->parsed++; + ctx->format += (use_sign2 ? 1 : 0); + } else { + ctx->error = -11; + } +} + +static void parse_unit1(time_ctx_t *ctx) +{ + char u = *ctx->parsed++; + switch (u) { + case 'Y': + case 'M': + case 'D': + case 'h': + case 'm': + case 's': + ctx->offset_unit = u; + break; + default: + ctx->error = -1; + } +} + +static void parse_unit2(time_ctx_t *ctx) +{ + char u = *ctx->parsed++; + switch (u) { + case 'y': + case 'd': + ctx->offset_unit = toupper((unsigned char)u); + break; + case 'h': + case 's': + ctx->offset_unit = u; + break; + case 'm': + switch (*ctx->parsed++) { + case 'o': + ctx->offset_unit = 'M'; + break; + case 'i': + ctx->offset_unit = 'm'; + break; + default: + ctx->error = -1; + } + break; + default: + ctx->error = -1; + } +} + +int knot_time_parse(const char *format, const char *timespec, knot_time_t *time) +{ + if (format == NULL || timespec == NULL || time == NULL) { + return -1; + } + + time_ctx_t ctx = { + .format = format, + .timespec = timespec, + .parsed = timespec, + .offset = 0, + .offset_sign = 0, + // we hope that .calendar is zeroed by default + .error = 0, + }; + + while (ctx.error == 0 && *ctx.format != '\0') { + switch (*ctx.format++) { + case '|': + if (*ctx.parsed == '\0') { + *time = time_ctx_finalize(&ctx); + return 0; + } else { + time_ctx_reset(&ctx); + } + break; + case '\'': + parse_quote(&ctx); + break; + case '#': + parse_offset(&ctx); + break; + case 'Y': + case 'M': + case 'D': + case 'h': + case 'm': + case 's': + parse_calendar(&ctx, calendar_index(*(ctx.format - 1))); + break; + case '+': + case '-': + parse_sign(&ctx); + break; + case 'U': + parse_unit1(&ctx); + break; + case 'u': + parse_unit2(&ctx); + break; + default: + return -1; + } + + if (ctx.error < 0) { + while (*ctx.format != '|' && *ctx.format != '\0') { + ctx.format++; + } + time_ctx_reset(&ctx); + ctx.error = (*ctx.format == '\0' ? -1 : 0); + } + } + + if (ctx.error == 0 && *ctx.parsed == '\0') { + *time = time_ctx_finalize(&ctx); + return 0; + } + return -1; +} + +static char *unit_names_mixed[] = { "Y", "M", "D", "h", "m", "s" }; +static char *unit_names_lower[] = { "y", "mo", "d", "h", "mi", "s" }; +static size_t unit_sizes[] = { 3600*24*365, 3600*24*30, 3600*24, 3600, 60, 1 }; +static const size_t unit_count = 6; + +static int print_unit(char *dst, size_t dst_len, char *unit_names[unit_count], + size_t max_units, knot_time_t time) +{ + int ret; + if (time == 0) { + ret = snprintf(dst, dst_len, "0"); + return (ret < 0 || ret >= dst_len ? -1 : 0); + } + knot_timediff_t diff = knot_time_diff(time, knot_time()); + if (dst_len-- < 1) { + return -1; + } + *dst++ = (diff < 0 ? '-' : '+'); + if (diff < 0) { + diff = -diff; + } else if (diff == 0) { + ret = snprintf(dst, dst_len, "0%s", unit_names[unit_count - 1]); + return (ret < 0 || ret >= dst_len ? -1 : 0); + } + size_t curr_unit = 0, used_units = 0; + while (curr_unit < unit_count && used_units < max_units) { + if (diff >= unit_sizes[curr_unit]) { + ret = snprintf(dst, dst_len, "%"KNOT_TIMEDIFF_PRINTF"%s", + diff / unit_sizes[curr_unit], + unit_names[curr_unit]); + if (ret < 0 || ret >= dst_len) { + return -1; + } + dst += ret; + dst_len -= ret; + used_units++; + diff %= unit_sizes[curr_unit]; + } + curr_unit++; + } + return 0; +} + +int knot_time_print(knot_time_print_t format, knot_time_t time, char *dst, size_t dst_len) +{ + if (dst == NULL) { + return -1; + } + + int ret; + switch (format) { + case TIME_PRINT_UNIX: + ret = snprintf(dst, dst_len, "%"KNOT_TIME_PRINTF, time); + return ((ret >= 0 && ret < dst_len) ? 0 : -1); + case TIME_PRINT_ISO8601: + if (time > LONG_MAX) { + return -1; + } + + // Set timezone to UTC before using timezone dependent functions + putenv("TZ=UTC"); + tzset(); + + struct tm lt; + time_t tt = (time_t)time; + ret = (localtime_r(&tt, <) == NULL ? -1 : + strftime(dst, dst_len, "%Y-%m-%dT%H:%M:%SZ", <)); + return (ret > 0 ? 0 : -1); + case TIME_PRINT_RELSEC: + ret = snprintf(dst, dst_len, "%+"KNOT_TIMEDIFF_PRINTF, + knot_time_diff(time, knot_time())); + return ((ret >= 0 && ret < dst_len) ? 0 : -1); + case TIME_PRINT_HUMAN_MIXED: + return print_unit(dst, dst_len, unit_names_mixed, unit_count, time); + case TIME_PRINT_HUMAN_LOWER: + return print_unit(dst, dst_len, unit_names_lower, unit_count, time); + default: + return -1; + } +} diff --git a/src/contrib/time.h b/src/contrib/time.h new file mode 100644 index 0000000..55b8727 --- /dev/null +++ b/src/contrib/time.h @@ -0,0 +1,171 @@ +/* Copyright (C) 2017 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ + +#pragma once + +#include <stdint.h> +#include <time.h> +#include <inttypes.h> + +/*! + * \brief Specify output format for knot_time_print(). + */ +typedef enum { + TIME_PRINT_UNIX, // numeric UNIX time + TIME_PRINT_ISO8601, // 2016-12-31T23:59:00 + TIME_PRINT_RELSEC, // relative +6523 + TIME_PRINT_HUMAN_MIXED, // relative with mixed-case units + TIME_PRINT_HUMAN_LOWER, // relative with lower-case units +} knot_time_print_t; + +/*! + * \brief Get current time. + */ +struct timespec time_now(void); + +/*! + * \brief Get time elapsed between two events. + */ +struct timespec time_diff(const struct timespec *begin, const struct timespec *end); + +/*! + * \brief Get time elapsed between two events in miliseconds. + */ +double time_diff_ms(const struct timespec *begin, const struct timespec *end); + +/*! + * \brief Data type for keeping UNIX timestamps. + * + * This is because time_t can be 32-bit on some systems, which is bad. + * Zero value represents infinity. + */ +typedef uint64_t knot_time_t; + +/*! + * \brief Data type for keeping time differences. + */ +typedef int64_t knot_timediff_t; + +#define KNOT_TIMEDIFF_MIN INT64_MIN +#define KNOT_TIMEDIFF_MAX INT64_MAX + +#define KNOT_TIME_PRINTF PRIu64 +#define KNOT_TIMEDIFF_PRINTF PRId64 + +/*! + * \brief Returns current time sice epoch. + */ +inline static knot_time_t knot_time(void) +{ + return (knot_time_t)time(NULL); +} + +/*! + * \brief Compare two timestamps. + * + * \return 0 if equal, -1 if the former is smaller (=earlier), 1 else. + */ +inline static int knot_time_cmp(knot_time_t a, knot_time_t b) +{ + return (a == b ? 0 : 1) * ((a && b) == 0 ? -1 : 1) * (a < b ? -1 : 1); +} + +/*! + * \brief Return the smaller (=earlier) from given two timestamps. + */ +inline static knot_time_t knot_time_min(knot_time_t a, knot_time_t b) +{ + if ((a && b) == 0) { + return a + b; + } else { + return (a < b ? a : b); + } +} + +/*! + * \brief Return the difference between two timestamps (to "minus" from). + * + * \note If both are zero (=infinity), KNOT_TIMEDIFF_MAX is returned. + */ +inline static knot_timediff_t knot_time_diff(knot_time_t to, knot_time_t from) +{ + if ((to && from) == 0) { + return (to > from ? KNOT_TIMEDIFF_MIN : KNOT_TIMEDIFF_MAX); + } else { + return (knot_timediff_t)to - (knot_timediff_t)from; + } +} + +/*! + * \brief Add a time difference to timestamp. + */ +inline static knot_time_t knot_time_add(knot_time_t since, knot_timediff_t howlong) +{ + return (since != 0 ? since + howlong : since); +} + +/*! + * \brief Convert uint32_t-encoded timestamp to knot_time_t. + * + * In RRSIG rdata, there are inception and expiration timestamps in uint32_t format. + * One shall use 'serial arithmetics' to decode them. + * + * \todo However it needs time(now) context which is slow to obtain, so we don't do it + * for now. Please fix this in next 100 years. + */ +inline static knot_time_t knot_time_from_u32(uint32_t u32time) +{ + return (knot_time_t)u32time; +} + +/*! + * \brief Parse a text-formatted timestamp to knot_time_t using format specification. + * + * \param format The timestamp text format specification. + * \param timespec Text-formatted timestamp. + * \param time The parsed timestamp. + * + * The format specification basics: + * <format 1>|<format 2> - The pipe sign separates two time format specifications. Leftmost + * specification matching the timespec is used. + * '<a string>' - Matches exactly <a string> (not containing apostrophes) in timespec. + * # - Hashtag matches for a number in timespec, stands for either a UNIX timestamp, + * or, within a context of an unit, as a number of such units. + * Y, M, D, h, m, s - Matches a number, stands for a number of years, months, days, hours, + * minutes and seconds, respectively. + * +, - - The + and - signs declaring that following timespec is relative to "now". + * A single sign can be used to limit the timestamp being in future or in past, + * or both +- allow the timestamp to select any (just one) of them. + * U - Matches one of Y, M, D, h, m, s in the timespec standing for a time unit. + * u - Like U, but the unit in the timestamp is from: y, mo, d, h, mi, s. + * + * \retval -1 An error occurred, out_time has no sense. + * \return 0 OK, timestamp parsed successfully. + */ +int knot_time_parse(const char *format, const char *timespec, knot_time_t *time); + +/*! + * \brief Print the timestamp in specified format into a string buffer. + * + * \param format The timestamp text format specification. + * \param time The timestamp to be printed. + * \param dst The destination buffer pointer with text-formatted timestamp. + * \param dst_len The destination buffer length. + * + * \retval -1 An error occurred, the buffer may be filled with nonsense. + * \return 0 OK, timestamp printed successfully. + */ +int knot_time_print(knot_time_print_t format, knot_time_t time, char *dst, size_t dst_len); diff --git a/src/contrib/tolower.h b/src/contrib/tolower.h new file mode 100644 index 0000000..8f55182 --- /dev/null +++ b/src/contrib/tolower.h @@ -0,0 +1,51 @@ +/* Copyright (C) 2018 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ +/*! + * \brief Table for converting ASCII characters to lowercase. + */ + +#pragma once + +#include <stdint.h> + +/*! + * \brief Converts binary character to lowercase. + * + * \param c Character code. + * + * \return \a c converted to lowercase (or \a c if not applicable). + */ +static inline uint8_t knot_tolower(uint8_t c) { + const uint8_t *tolower_table = (uint8_t *) + "\x00\x01\x02\x03\x04\x05\x06\x07\x08\x09\x0A\x0B\x0C\x0D\x0E\x0F" + "\x10\x11\x12\x13\x14\x15\x16\x17\x18\x19\x1A\x1B\x1C\x1D\x1E\x1F" + "\x20\x21\x22\x23\x24\x25\x26\x27\x28\x29\x2A\x2B\x2C\x2D\x2E\x2F" + "\x30\x31\x32\x33\x34\x35\x36\x37\x38\x39\x3A\x3B\x3C\x3D\x3E\x3F" + "\x40\x61\x62\x63\x64\x65\x66\x67\x68\x69\x6A\x6B\x6C\x6D\x6E\x6F" + "\x70\x71\x72\x73\x74\x75\x76\x77\x78\x79\x7A\x5B\x5C\x5D\x5E\x5F" + "\x60\x61\x62\x63\x64\x65\x66\x67\x68\x69\x6A\x6B\x6C\x6D\x6E\x6F" + "\x70\x71\x72\x73\x74\x75\x76\x77\x78\x79\x7A\x7B\x7C\x7D\x7E\x7F" + "\x80\x81\x82\x83\x84\x85\x86\x87\x88\x89\x8A\x8B\x8C\x8D\x8E\x8F" + "\x90\x91\x92\x93\x94\x95\x96\x97\x98\x99\x9A\x9B\x9C\x9D\x9E\x9F" + "\xA0\xA1\xA2\xA3\xA4\xA5\xA6\xA7\xA8\xA9\xAA\xAB\xAC\xAD\xAE\xAF" + "\xB0\xB1\xB2\xB3\xB4\xB5\xB6\xB7\xB8\xB9\xBA\xBB\xBC\xBD\xBE\xBF" + "\xC0\xC1\xC2\xC3\xC4\xC5\xC6\xC7\xC8\xC9\xCA\xCB\xCC\xCD\xCE\xCF" + "\xD0\xD1\xD2\xD3\xD4\xD5\xD6\xD7\xD8\xD9\xDA\xDB\xDC\xDD\xDE\xDF" + "\xE0\xE1\xE2\xE3\xE4\xE5\xE6\xE7\xE8\xE9\xEA\xEB\xEC\xED\xEE\xEF" + "\xF0\xF1\xF2\xF3\xF4\xF5\xF6\xF7\xF8\xF9\xFA\xFB\xFC\xFD\xFE\xFF"; + + return tolower_table[c]; +} diff --git a/src/contrib/trim.h b/src/contrib/trim.h new file mode 100644 index 0000000..c1b83fa --- /dev/null +++ b/src/contrib/trim.h @@ -0,0 +1,35 @@ +/* Copyright (C) 2018 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. +*/ +/*! + * \brief Heap memory trimmer. + */ + +#pragma once + +#ifdef HAVE_MALLOC_TRIM +#include <malloc.h> +#endif + +/*! + * \brief Trim excess heap memory. + */ +static inline void mem_trim(void) +{ +#ifdef HAVE_MALLOC_TRIM + malloc_trim(0); +#endif + return; +} diff --git a/src/contrib/ucw/LICENSE b/src/contrib/ucw/LICENSE new file mode 100644 index 0000000..b463d57 --- /dev/null +++ b/src/contrib/ucw/LICENSE @@ -0,0 +1 @@ +../licenses/LGPL-2.0
\ No newline at end of file diff --git a/src/contrib/ucw/array-sort.h b/src/contrib/ucw/array-sort.h new file mode 100644 index 0000000..1ff1377 --- /dev/null +++ b/src/contrib/ucw/array-sort.h @@ -0,0 +1,195 @@ +/* + * UCW Library -- Universal Simple Array Sorter + * + * (c) 2003--2008 Martin Mares <mj@ucw.cz> + * + * This software may be freely distributed and used according to the terms + * of the GNU Lesser General Public License. + */ + +#pragma once + +#include "contrib/macros.h" + +/* + * This is not a normal header file, it's a generator of sorting + * routines. Each time you include it with parameters set in the + * corresponding preprocessor macros, it generates an array sorter + * with the parameters given. + * + * You might wonder why the heck do we implement our own array sorter + * instead of using qsort(). The primary reason is that qsort handles + * only continuous arrays, but we need to sort array-like data structures + * where the only way to access elements is by using an indexing macro. + * Besides that, we are more than 2 times faster. + * + * So much for advocacy, there are the parameters (those marked with [*] + * are mandatory): + * + * ASORT_PREFIX(x) [*] add a name prefix (used on all global names + * defined by the sorter) + * ASORT_KEY_TYPE [*] data type of a single array entry key + * ASORT_ELT(i) returns the key of i-th element; if this macro is not + * defined, the function gets a pointer to an array to be sorted + * ASORT_LT(x,y) x < y for ASORT_KEY_TYPE (default: "x<y") + * ASORT_SWAP(i,j) swap i-th and j-th element (default: assume _ELT + * is an l-value and swap just the keys) + * ASORT_THRESHOLD threshold for switching between quicksort and insertsort + * ASORT_EXTRA_ARGS extra arguments for the sort function (they are always + * visible in all the macros supplied above), starts with comma + * + * After including this file, a function ASORT_PREFIX(sort)(unsigned array_size) + * or ASORT_PREFIX(sort)(ASORT_KEY_TYPE *array, unsigned array_size) [if ASORT_ELT + * is not defined] is declared and all parameter macros are automatically + * undef'd. + */ + +#ifndef ASORT_LT +#define ASORT_LT(x,y) ((x) < (y)) +#endif + +#ifndef ASORT_SWAP +#define ASORT_SWAP(i,j) do { ASORT_KEY_TYPE tmp = ASORT_ELT(i); ASORT_ELT(i)=ASORT_ELT(j); ASORT_ELT(j)=tmp; } while (0) +#endif + +#ifndef ASORT_THRESHOLD +#define ASORT_THRESHOLD 8 /* Guesswork and experimentation */ +#endif + +#ifndef ASORT_EXTRA_ARGS +#define ASORT_EXTRA_ARGS +#endif + +#ifndef ASORT_ELT +#define ASORT_ARRAY_ARG ASORT_KEY_TYPE *array, +#define ASORT_ELT(i) array[i] +#else +#define ASORT_ARRAY_ARG +#endif + +/** + * The generated sorting function. If `ASORT_ELT` macro is not provided, the + * @ASORT_ARRAY_ARG is equal to `ASORT_KEY_TYPE *array` and is the array to be + * sorted. If the macro is provided, this parameter is omitted. In that case, + * you can sort global variables or pass your structure by @ASORT_EXTRA_ARGS. + **/ +static void ASORT_PREFIX(sort)(ASORT_ARRAY_ARG unsigned array_size ASORT_EXTRA_ARGS) +{ + struct stk { int l, r; } stack[8*sizeof(unsigned)]; + int l, r, left, right, m; + unsigned sp = 0; + ASORT_KEY_TYPE pivot; + + if (array_size <= 1) + return; + + /* QuickSort with optimizations a'la Sedgewick, but stop at ASORT_THRESHOLD */ + + left = 0; + right = array_size - 1; + for(;;) + { + l = left; + r = right; + m = (l+r)/2; + if (ASORT_LT(ASORT_ELT(m), ASORT_ELT(l))) + ASORT_SWAP(l,m); + if (ASORT_LT(ASORT_ELT(r), ASORT_ELT(m))) + { + ASORT_SWAP(m,r); + if (ASORT_LT(ASORT_ELT(m), ASORT_ELT(l))) + ASORT_SWAP(l,m); + } + pivot = ASORT_ELT(m); + do + { + while (ASORT_LT(ASORT_ELT(l), pivot)) + l++; + while (ASORT_LT(pivot, ASORT_ELT(r))) + r--; + if (l < r) + { + ASORT_SWAP(l,r); + l++; + r--; + } + else if (l == r) + { + l++; + r--; + } + } + while (l <= r); + if ((r - left) >= ASORT_THRESHOLD && (right - l) >= ASORT_THRESHOLD) + { + /* Both partitions ok => push the larger one */ + if ((r - left) > (right - l)) + { + stack[sp].l = left; + stack[sp].r = r; + left = l; + } + else + { + stack[sp].l = l; + stack[sp].r = right; + right = r; + } + sp++; + } + else if ((r - left) >= ASORT_THRESHOLD) + { + /* Left partition OK, right undersize */ + right = r; + } + else if ((right - l) >= ASORT_THRESHOLD) + { + /* Right partition OK, left undersize */ + left = l; + } + else + { + /* Both partitions undersize => pop */ + if (!sp) + break; + sp--; + left = stack[sp].l; + right = stack[sp].r; + } + } + + /* + * We have a partially sorted array, finish by insertsort. Inspired + * by qsort() in GNU libc. + */ + + /* Find minimal element which will serve as a barrier */ + r = MIN(array_size, ASORT_THRESHOLD); + m = 0; + for (l=1; l<r; l++) + if (ASORT_LT(ASORT_ELT(l),ASORT_ELT(m))) + m = l; + ASORT_SWAP(0,m); + + /* Insertion sort */ + for (m=1; m<(int)array_size; m++) + { + l=m; + while (ASORT_LT(ASORT_ELT(m),ASORT_ELT(l-1))) + l--; + while (l < m) + { + ASORT_SWAP(l,m); + l++; + } + } +} + +#undef ASORT_PREFIX +#undef ASORT_KEY_TYPE +#undef ASORT_ELT +#undef ASORT_LT +#undef ASORT_SWAP +#undef ASORT_THRESHOLD +#undef ASORT_EXTRA_ARGS +#undef ASORT_ARRAY_ARG diff --git a/src/contrib/ucw/binsearch.h b/src/contrib/ucw/binsearch.h new file mode 100644 index 0000000..b791d39 --- /dev/null +++ b/src/contrib/ucw/binsearch.h @@ -0,0 +1,50 @@ +/* + * UCW Library -- Generic Binary Search + * + * (c) 2005 Martin Mares <mj@ucw.cz> + * + * This software may be freely distributed and used according to the terms + * of the GNU Lesser General Public License. + */ + +#pragma once + +/*** + * [[defs]] + * Definitions + * ----------- + ***/ + +/** + * Find the first element not lower than \p x in the sorted array \p ary of \p N elements (non-decreasing order). + * Returns the index of the found element or \p N if no exists. Uses `ary_lt_x(ary,i,x)` to compare the i'th element with \p x. + * The time complexity is `O(log(N))`. + **/ +#define BIN_SEARCH_FIRST_GE_CMP(ary, N, ary_lt_x, x, ...) ({ \ + unsigned l = 0, r = (N); \ + while (l < r) \ + { \ + unsigned m = (l+r)/2; \ + if (ary_lt_x(ary, m, x, __VA_ARGS__)) \ + l = m+1; \ + else \ + r = m; \ + } \ + l; \ +}) + +/** + * The default comparison macro for \ref BIN_SEARCH_FIRST_GE_CMP(). + **/ +#define ARY_LT_NUM(ary,i,x) (ary)[i] < (x) + +/** + * Same as \ref BIN_SEARCH_FIRST_GE_CMP(), but uses the default `<` operator for comparisons. + **/ +#define BIN_SEARCH_FIRST_GE(ary,N,x) BIN_SEARCH_FIRST_GE_CMP(ary,N,ARY_LT_NUM,x) + +/** + * Search the sorted array \p ary of \p N elements (non-decreasing) for the first occurrence of \p x. + * Returns the index or -1 if no such element exists. Uses the `<` operator for comparisons. + **/ +#define BIN_SEARCH_EQ(ary,N,x) ({ int i = BIN_SEARCH_FIRST_GE(ary,N,x); if (i >= (N) || (ary)[i] != (x)) i=-1; i; }) diff --git a/src/contrib/ucw/heap.c b/src/contrib/ucw/heap.c new file mode 100644 index 0000000..d7ed18e --- /dev/null +++ b/src/contrib/ucw/heap.c @@ -0,0 +1,166 @@ +/* + * Binary heap + * + * (c) 2012 Ondrej Filip <feela@network.cz> + * + * This software may be freely distributed and used according to the terms + * of the GNU Lesser General Public License. + */ + +/*** + * Introduction + * ------------ + * + * Binary heap is a simple data structure, which for example supports efficient insertions, deletions + * and access to the minimal inserted item. We define several macros for such operations. + * Note that because of simplicity of heaps, we have decided to define direct macros instead + * of a <<generic:,macro generator>> as for several other data structures in the Libucw. + * + * A heap is represented by a number of elements and by an array of values. Beware that we + * index this array from one, not from zero as do the standard C arrays. + * + * Most macros use these parameters: + * + * - @num - a variable (signed or unsigned integer) with the number of elements + * - @heap - a C array of type @type; the heap is stored in `heap[1] .. heap[num]`; `heap[0]` is unused + * + * A valid heap must follow these rules: + * + * - `num >= 0` + * - `heap[i] >= heap[i / 2]` for each `i` in `[2, num]` + * + * The first element `heap[1]` is always lower or equal to all other elements. + ***/ + +#include <string.h> +#include <stdlib.h> +#include "contrib/ucw/heap.h" + +static inline void heap_swap(heap_val_t **e1, heap_val_t **e2) +{ + if (e1 == e2) return; /* Stack tmp should be faster than tmpelem. */ + heap_val_t *tmp = *e1; /* Even faster than 2-XOR nowadays. */ + *e1 = *e2; + *e2 = tmp; + int pos = (*e1)->pos; + (*e1)->pos = (*e2)->pos; + (*e2)->pos = pos; +} + +int heap_init(struct heap *h, int (*cmp)(void *, void *), int init_size) +{ + int isize = init_size ? init_size : INITIAL_HEAP_SIZE; + + h->num = 0; + h->max_size = isize; + h->cmp = cmp; + h->data = malloc((isize + 1) * sizeof(heap_val_t*)); /* Temp element unused. */ + + return h->data ? 1 : 0; +} + +void heap_deinit(struct heap *h) +{ + free(h->data); + memset(h, 0, sizeof(*h)); +} + +static inline void _heap_bubble_down(struct heap *h, int e) +{ + int e1; + for (;;) + { + e1 = 2*e; + if(e1 > h->num) break; + if((h->cmp(*HELEMENT(h, e),*HELEMENT(h,e1)) < 0) && (e1 == h->num || (h->cmp(*HELEMENT(h, e),*HELEMENT(h,e1+1)) < 0))) break; + if((e1 != h->num) && (h->cmp(*HELEMENT(h, e1+1), *HELEMENT(h,e1)) < 0)) e1++; + heap_swap(HELEMENT(h,e),HELEMENT(h,e1)); + e = e1; + } +} + +static inline void _heap_bubble_up(struct heap *h, int e) +{ + int e1; + while (e > 1) + { + e1 = e/2; + if(h->cmp(*HELEMENT(h, e1),*HELEMENT(h,e)) < 0) break; + heap_swap(HELEMENT(h,e),HELEMENT(h,e1)); + e = e1; + } + +} + +static void heap_increase(struct heap *h, int pos, heap_val_t *e) +{ + *HELEMENT(h, pos) = e; + e->pos = pos; + _heap_bubble_down(h, pos); +} + +static void heap_decrease(struct heap *h, int pos, heap_val_t *e) +{ + *HELEMENT(h, pos) = e; + e->pos = pos; + _heap_bubble_up(h, pos); +} + +void heap_replace(struct heap *h, int pos, heap_val_t *e) +{ + if (h->cmp(*HELEMENT(h, pos),e) < 0) { + heap_increase(h, pos, e); + } else { + heap_decrease(h, pos, e); + } +} + +void heap_delmin(struct heap *h) +{ + if(h->num == 0) return; + if(h->num > 1) + { + heap_swap(HHEAD(h),HELEMENT(h,h->num)); + } + (*HELEMENT(h, h->num))->pos = 0; + --h->num; + _heap_bubble_down(h, 1); +} + +int heap_insert(struct heap *h, heap_val_t *e) +{ + if(h->num == h->max_size) + { + h->max_size = h->max_size * HEAP_INCREASE_STEP; + h->data = realloc(h->data, (h->max_size + 1) * sizeof(heap_val_t*)); + if (!h->data) { + return 0; + } + } + + h->num++; + *HELEMENT(h,h->num) = e; + e->pos = h->num; + _heap_bubble_up(h,h->num); + return 1; +} + +int heap_find(struct heap *h, heap_val_t *elm) +{ + return ((struct heap_val *) elm)->pos; +} + +void heap_delete(struct heap *h, int e) +{ + heap_swap(HELEMENT(h, e), HELEMENT(h, h->num)); + (*HELEMENT(h, h->num))->pos = 0; + h->num--; + if(h->cmp(*HELEMENT(h, e), *HELEMENT(h, h->num + 1)) < 0) _heap_bubble_up(h, e); + else _heap_bubble_down(h, e); + + if ((h->num > INITIAL_HEAP_SIZE) && (h->num < h->max_size / HEAP_DECREASE_THRESHOLD)) + { + h->max_size = h->max_size / HEAP_INCREASE_STEP; + h->data = realloc(h->data, (h->max_size + 1) * sizeof(heap_val_t*)); + } +} diff --git a/src/contrib/ucw/heap.h b/src/contrib/ucw/heap.h new file mode 100644 index 0000000..7419b34 --- /dev/null +++ b/src/contrib/ucw/heap.h @@ -0,0 +1,46 @@ +/* Copyright (C) 2011 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + * + * This program is free software: you can redistribute it and/or modify + * it under the terms of the GNU General Public License as published by + * the Free Software Foundation, either version 3 of the License, or + * (at your option) any later version. + * + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with this program. If not, see <http://www.gnu.org/licenses/>. + */ + +#pragma once + +struct heap_val { + int pos; +}; + +typedef struct heap_val heap_val_t; + +struct heap { + int num; /* Number of elements */ + int max_size; /* Size of allocated memory */ + int (*cmp)(void *, void *); + heap_val_t **data; +}; /* Array follows */ + +#define INITIAL_HEAP_SIZE 512 /* initial heap size */ +#define HEAP_INCREASE_STEP 2 /* multiplier for each inflation, keep conservative */ +#define HEAP_DECREASE_THRESHOLD 2 /* threshold for deflation, keep conservative */ +#define HELEMENT(h,num) ((h)->data + (num)) +#define HHEAD(h) HELEMENT((h), 1) +#define EMPTY_HEAP(h) ((h)->num == 0) /* h->num == 0 */ + +int heap_init(struct heap *, int (*cmp)(void *, void *), int); +void heap_deinit(struct heap *); + +void heap_delmin(struct heap *); +int heap_insert(struct heap *, heap_val_t *); +int heap_find(struct heap *, heap_val_t *); +void heap_delete(struct heap *, int); +void heap_replace(struct heap *h, int pos, heap_val_t *); diff --git a/src/contrib/ucw/lists.c b/src/contrib/ucw/lists.c new file mode 100644 index 0000000..8a9fa96 --- /dev/null +++ b/src/contrib/ucw/lists.c @@ -0,0 +1,235 @@ +/* + * BIRD Library -- Linked Lists + * + * (c) 1998 Martin Mares <mj@ucw.cz> + * (c) 2015, 2017 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + * + * Can be freely distributed and used under the terms of the GNU GPL. + */ + +/** + * DOC: Linked lists + * + * The BIRD library provides a set of functions for operating on linked + * lists. The lists are internally represented as standard doubly linked + * lists with synthetic head and tail which makes all the basic operations + * run in constant time and contain no extra end-of-list checks. Each list + * is described by a &list structure, nodes can have any format as long + * as they start with a &node structure. If you want your nodes to belong + * to multiple lists at once, you can embed multiple &node structures in them + * and use the SKIP_BACK() macro to calculate a pointer to the start of the + * structure from a &node pointer, but beware of obscurity. + * + * There also exist safe linked lists (&slist, &snode and all functions + * being prefixed with |s_|) which support asynchronous walking very + * similar to that used in the &fib structure. + */ + +#include <stdlib.h> +#include <string.h> +#include "contrib/ucw/lists.h" +#include "contrib/mempattern.h" + +/** + * add_tail - append a node to a list + * \p l: linked list + * \p n: list node + * + * add_tail() takes a node \p n and appends it at the end of the list \p l. + */ +void +add_tail(list_t *l, node_t *n) +{ + node_t *z = l->tail; + + n->next = (node_t *) &l->null; + n->prev = z; + z->next = n; + l->tail = n; +} + +/** + * add_head - prepend a node to a list + * \p l: linked list + * \p n: list node + * + * add_head() takes a node \p n and prepends it at the start of the list \p l. + */ +void +add_head(list_t *l, node_t *n) +{ + node_t *z = l->head; + + n->next = z; + n->prev = (node_t *) &l->head; + z->prev = n; + l->head = n; +} + +/** + * insert_node - insert a node to a list + * \p n: a new list node + * \p after: a node of a list + * + * Inserts a node \p n to a linked list after an already inserted + * node \p after. + */ +void +insert_node(node_t *n, node_t *after) +{ + node_t *z = after->next; + + n->next = z; + n->prev = after; + after->next = n; + z->prev = n; +} + +/** + * rem_node - remove a node from a list + * \p n: node to be removed + * + * Removes a node \p n from the list it's linked in. + */ +void +rem_node(node_t *n) +{ + node_t *z = n->prev; + node_t *x = n->next; + + z->next = x; + x->prev = z; + n->prev = 0; + n->next = 0; +} + +/** + * init_list - create an empty list + * \p l: list + * + * init_list() takes a &list structure and initializes its + * fields, so that it represents an empty list. + */ +void +init_list(list_t *l) +{ + l->head = (node_t *) &l->null; + l->null = NULL; + l->tail = (node_t *) &l->head; +} + +/** + * add_tail_list - concatenate two lists + * \p to: destination list + * \p l: source list + * + * This function appends all elements of the list \p l to + * the list \p to in constant time. + */ +void +add_tail_list(list_t *to, list_t *l) +{ + node_t *p = to->tail; + node_t *q = l->head; + + p->next = q; + q->prev = p; + q = l->tail; + q->next = (node_t *) &to->null; + to->tail = q; +} + +/** + * list_dup - duplicate list + * \p to: destination list + * \p l: source list + * + * This function duplicates all elements of the list \p l to + * the list \p to in linear time. + * + * This function only works with a homogenous item size. + */ +void list_dup(list_t *dst, list_t *src, size_t itemsz) +{ + node_t *n = 0; + WALK_LIST(n, *src) { + node_t *i = malloc(itemsz); + memcpy(i, n, itemsz); + add_tail(dst, i); + } +} + +/** + * list_size - gets number of nodes + * \p l: list + * + * This function counts nodes in list \p l and returns this number. + */ +size_t list_size(const list_t *l) +{ + size_t count = 0; + + node_t *n = 0; + WALK_LIST(n, *l) { + count++; + } + + return count; +} + +/** + * ptrlist_add - add pointer to pointer list + * \p to: destination list + * \p val: added pointer + * \p mm: memory context + */ +ptrnode_t *ptrlist_add(list_t *to, void *val, knot_mm_t *mm) +{ + ptrnode_t *node = mm_alloc(mm , sizeof(ptrnode_t)); + if (node == NULL) { + return NULL; + } else { + node->d = val; + } + add_tail(to, &node->n); + return node; +} + +/** + * ptrlist_free - free all nodes in pointer list + * \p list: list nodes + * \p mm: memory context + */ +void ptrlist_free(list_t *list, knot_mm_t *mm) +{ + node_t *n = NULL, *nxt = NULL; + WALK_LIST_DELSAFE(n, nxt, *list) { + mm_free(mm, n); + } + init_list(list); +} + +/** + * ptrlist_rem - remove pointer node + * \p val: pointer to remove + * \p mm: memory context + */ +void ptrlist_rem(ptrnode_t *node, knot_mm_t *mm) +{ + rem_node(&node->n); + mm_free(mm, node); +} + +/** + * ptrlist_deep_free - free all nodes incl referenced data + * \p list: list nodes + * \p mm: memory context + */ +void ptrlist_deep_free(list_t *l, knot_mm_t *mm) +{ + ptrnode_t *n; + WALK_LIST(n, *l) { + mm_free(mm, n->d); + } + ptrlist_free(l, mm); +} diff --git a/src/contrib/ucw/lists.h b/src/contrib/ucw/lists.h new file mode 100644 index 0000000..922e152 --- /dev/null +++ b/src/contrib/ucw/lists.h @@ -0,0 +1,84 @@ +/* + * BIRD Library -- Linked Lists + * + * (c) 1998 Martin Mares <mj@ucw.cz> + * (c) 2015, 2017 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + * + * Can be freely distributed and used under the terms of the GNU GPL. + */ + +#pragma once + +/* + * I admit the list structure is very tricky and also somewhat awkward, + * but it's both efficient and easy to manipulate once one understands the + * basic trick: The list head always contains two synthetic nodes which are + * always present in the list: the head and the tail. But as the `next' + * entry of the tail and the `prev' entry of the head are both NULL, the + * nodes can overlap each other: + * + * head head_node.next + * null head_node.prev tail_node.next + * tail tail_node.prev + */ + +#include <string.h> +#include "libknot/mm_ctx.h" + +typedef struct node { + struct node *next, *prev; +} node_t; + +typedef struct list { /* In fact two overlayed nodes */ + struct node *head, *null, *tail; +} list_t; + +#define NODE (node_t *) +#define HEAD(list) ((void *)((list).head)) +#define TAIL(list) ((void *)((list).tail)) +#define WALK_LIST(n,list) for(n=HEAD(list);(NODE (n))->next; \ + n=(void *)((NODE (n))->next)) +#define WALK_LIST_DELSAFE(n,nxt,list) \ + for(n=HEAD(list); (nxt=(void *)((NODE (n))->next)); n=(void *) nxt) +/* WALK_LIST_FIRST supposes that called code removes each processed node */ +#define WALK_LIST_FIRST(n,list) \ + while(n=HEAD(list), (NODE (n))->next) +#define WALK_LIST_BACKWARDS(n,list) for(n=TAIL(list);(NODE (n))->prev; \ + n=(void *)((NODE (n))->prev)) +#define WALK_LIST_BACKWARDS_DELSAFE(n,prv,list) \ + for(n=TAIL(list); prv=(void *)((NODE (n))->prev); n=(void *) prv) + +#define EMPTY_LIST(list) (!(list).head->next) + +/*! \brief Free every node in the list. */ +#define WALK_LIST_FREE(list) \ + do { \ + node_t *n=0,*nxt=0; \ + WALK_LIST_DELSAFE(n,nxt,list) { \ + free(n); \ + } \ + init_list(&list); \ + } while(0) + +void add_tail(list_t *, node_t *); +void add_head(list_t *, node_t *); +void rem_node(node_t *); +void add_tail_list(list_t *, list_t *); +void init_list(list_t *); +void insert_node(node_t *, node_t *); +void list_dup(list_t *dst, list_t *src, size_t itemsz); +size_t list_size(const list_t *); + +/*! + * \brief Generic pointer list implementation. + */ +typedef struct ptrnode { + node_t n; + void *d; +} ptrnode_t; + +ptrnode_t *ptrlist_add(list_t *, void *, knot_mm_t *); +void ptrlist_free(list_t *, knot_mm_t *); +void ptrlist_rem(ptrnode_t *node, knot_mm_t *mm); +void ptrlist_deep_free(list_t *, knot_mm_t *); + diff --git a/src/contrib/ucw/mempool.c b/src/contrib/ucw/mempool.c new file mode 100644 index 0000000..bc41345 --- /dev/null +++ b/src/contrib/ucw/mempool.c @@ -0,0 +1,322 @@ +/* + * UCW Library -- Memory Pools (One-Time Allocation) + * + * (c) 1997--2001 Martin Mares <mj@ucw.cz> + * (c) 2007 Pavel Charvat <pchar@ucw.cz> + * (c) 2015, 2017 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + * + * This software may be freely distributed and used according to the terms + * of the GNU Lesser General Public License. + */ + +#undef LOCAL_DEBUG + +#include <string.h> +#include <stdlib.h> +#include <stdio.h> +#include <assert.h> +#include "contrib/asan.h" +#include "contrib/macros.h" +#include "contrib/ucw/mempool.h" + +/** \todo This shouldn't be precalculated, but computed on load. */ +#define CPU_PAGE_SIZE 4096 + +/** Align an integer \p s to the nearest higher multiple of \p a (which should be a power of two) **/ +#define ALIGN_TO(s, a) (((s)+a-1)&~(a-1)) +#define MP_CHUNK_TAIL ALIGN_TO(sizeof(struct mempool_chunk), CPU_STRUCT_ALIGN) +#define MP_SIZE_MAX (~0U - MP_CHUNK_TAIL - CPU_PAGE_SIZE) +#define DBG(s, ...) + +/** \note Imported MMAP backend from bigalloc.c */ +#define CONFIG_UCW_POOL_IS_MMAP +#ifdef CONFIG_UCW_POOL_IS_MMAP +#include <sys/mman.h> +static void * +page_alloc(uint64_t len) +{ + if (!len) { + return NULL; + } + if (len > SIZE_MAX) { + return NULL; + } + assert(!(len & (CPU_PAGE_SIZE-1))); + uint8_t *p = mmap(NULL, len, PROT_READ | PROT_WRITE, MAP_PRIVATE | MAP_ANON, -1, 0); + if (p == (uint8_t*) MAP_FAILED) { + return NULL; + } + return p; +} + +static void +page_free(void *start, uint64_t len) +{ + assert(!(len & (CPU_PAGE_SIZE-1))); + assert(!((uintptr_t) start & (CPU_PAGE_SIZE-1))); + munmap(start, len); +} +#endif + +struct mempool_chunk { + struct mempool_chunk *next; + unsigned size; +}; + +static unsigned +mp_align_size(unsigned size) +{ +#ifdef CONFIG_UCW_POOL_IS_MMAP + return ALIGN_TO(size + MP_CHUNK_TAIL, CPU_PAGE_SIZE) - MP_CHUNK_TAIL; +#else + return ALIGN_TO(size, CPU_STRUCT_ALIGN); +#endif +} + +void +mp_init(struct mempool *pool, unsigned chunk_size) +{ + chunk_size = mp_align_size(MAX(sizeof(struct mempool), chunk_size)); + *pool = (struct mempool) { + .chunk_size = chunk_size, + .threshold = chunk_size >> 1, + .last_big = &pool->last_big + }; +} + +static void * +mp_new_big_chunk(unsigned size) +{ + uint8_t *data = malloc(size + MP_CHUNK_TAIL); + if (!data) { + return NULL; + } + ASAN_POISON_MEMORY_REGION(data, size); + struct mempool_chunk *chunk = (struct mempool_chunk *)(data + size); + chunk->size = size; + return chunk; +} + +static void +mp_free_big_chunk(struct mempool_chunk *chunk) +{ + void *ptr = (uint8_t *)chunk - chunk->size; + ASAN_UNPOISON_MEMORY_REGION(ptr, chunk->size); + free(ptr); +} + +static void * +mp_new_chunk(unsigned size) +{ +#ifdef CONFIG_UCW_POOL_IS_MMAP + uint8_t *data = page_alloc(size + MP_CHUNK_TAIL); + if (!data) { + return NULL; + } + ASAN_POISON_MEMORY_REGION(data, size); + struct mempool_chunk *chunk = (struct mempool_chunk *)(data + size); + chunk->size = size; + return chunk; +#else + return mp_new_big_chunk(size); +#endif +} + +static void +mp_free_chunk(struct mempool_chunk *chunk) +{ +#ifdef CONFIG_UCW_POOL_IS_MMAP + uint8_t *data = (uint8_t *)chunk - chunk->size; + ASAN_UNPOISON_MEMORY_REGION(data, chunk->size); + page_free(data, chunk->size + MP_CHUNK_TAIL); +#else + mp_free_big_chunk(chunk); +#endif +} + +struct mempool * +mp_new(unsigned chunk_size) +{ + chunk_size = mp_align_size(MAX(sizeof(struct mempool), chunk_size)); + struct mempool_chunk *chunk = mp_new_chunk(chunk_size); + struct mempool *pool = (void *)chunk - chunk_size; + ASAN_UNPOISON_MEMORY_REGION(pool, sizeof(*pool)); + DBG("Creating mempool %p with %u bytes long chunks", pool, chunk_size); + chunk->next = NULL; + ASAN_POISON_MEMORY_REGION(chunk, sizeof(struct mempool_chunk)); + *pool = (struct mempool) { + .state = { .free = { chunk_size - sizeof(*pool) }, .last = { chunk } }, + .chunk_size = chunk_size, + .threshold = chunk_size >> 1, + .last_big = &pool->last_big + }; + return pool; +} + +static void +mp_free_chain(struct mempool_chunk *chunk) +{ + while (chunk) { + ASAN_UNPOISON_MEMORY_REGION(chunk, sizeof(struct mempool_chunk)); + struct mempool_chunk *next = chunk->next; + mp_free_chunk(chunk); + chunk = next; + } +} + +static void +mp_free_big_chain(struct mempool_chunk *chunk) +{ + while (chunk) { + ASAN_UNPOISON_MEMORY_REGION(chunk, sizeof(struct mempool_chunk)); + struct mempool_chunk *next = chunk->next; + mp_free_big_chunk(chunk); + chunk = next; + } +} + +void +mp_delete(struct mempool *pool) +{ + if (pool == NULL) { + return; + } + DBG("Deleting mempool %p", pool); + mp_free_big_chain(pool->state.last[1]); + mp_free_chain(pool->unused); + mp_free_chain(pool->state.last[0]); // can contain the mempool structure +} + +void +mp_flush(struct mempool *pool) +{ + mp_free_big_chain(pool->state.last[1]); + struct mempool_chunk *chunk = pool->state.last[0], *next; + while (chunk) { + ASAN_UNPOISON_MEMORY_REGION(chunk, sizeof(struct mempool_chunk)); + if ((uint8_t *)chunk - chunk->size == (uint8_t *)pool) { + break; + } + next = chunk->next; + chunk->next = pool->unused; + ASAN_POISON_MEMORY_REGION(chunk, sizeof(struct mempool_chunk)); + pool->unused = chunk; + chunk = next; + } + pool->state.last[0] = chunk; + if (chunk) { + pool->state.free[0] = chunk->size - sizeof(*pool); + ASAN_POISON_MEMORY_REGION(chunk, sizeof(struct mempool_chunk)); + } else { + pool->state.free[0] = 0; + } + pool->state.last[1] = NULL; + pool->state.free[1] = 0; + pool->last_big = &pool->last_big; +} + +static void +mp_stats_chain(struct mempool_chunk *chunk, struct mempool_stats *stats, unsigned idx) +{ + struct mempool_chunk *next; + while (chunk) { + ASAN_UNPOISON_MEMORY_REGION(chunk, sizeof(struct mempool_chunk)); + stats->chain_size[idx] += chunk->size + sizeof(*chunk); + stats->chain_count[idx]++; + next = chunk->next; + ASAN_POISON_MEMORY_REGION(chunk, sizeof(struct mempool_chunk)); + chunk = next; + } + stats->total_size += stats->chain_size[idx]; +} + +void +mp_stats(struct mempool *pool, struct mempool_stats *stats) +{ + bzero(stats, sizeof(*stats)); + mp_stats_chain(pool->state.last[0], stats, 0); + mp_stats_chain(pool->state.last[1], stats, 1); + mp_stats_chain(pool->unused, stats, 2); +} + +uint64_t +mp_total_size(struct mempool *pool) +{ + struct mempool_stats stats; + mp_stats(pool, &stats); + return stats.total_size; +} + +static void * +mp_alloc_internal(struct mempool *pool, unsigned size) +{ + struct mempool_chunk *chunk; + if (size <= pool->threshold) { + pool->idx = 0; + if (pool->unused) { + chunk = pool->unused; + ASAN_UNPOISON_MEMORY_REGION(chunk, sizeof(struct mempool_chunk)); + pool->unused = chunk->next; + } else { + chunk = mp_new_chunk(pool->chunk_size); + } + chunk->next = pool->state.last[0]; + ASAN_POISON_MEMORY_REGION(chunk, sizeof(struct mempool_chunk)); + pool->state.last[0] = chunk; + pool->state.free[0] = pool->chunk_size - size; + return (uint8_t *)chunk - pool->chunk_size; + } else if (size <= MP_SIZE_MAX) { + pool->idx = 1; + unsigned aligned = ALIGN_TO(size, CPU_STRUCT_ALIGN); + chunk = mp_new_big_chunk(aligned); + if (!chunk) { + return NULL; + } + chunk->next = pool->state.last[1]; + ASAN_POISON_MEMORY_REGION(chunk, sizeof(struct mempool_chunk)); + pool->state.last[1] = chunk; + pool->state.free[1] = aligned - size; + return pool->last_big = (uint8_t *)chunk - aligned; + } else { + fprintf(stderr, "Cannot allocate %u bytes from a mempool", size); + assert(0); + return NULL; + } +} + +void * +mp_alloc(struct mempool *pool, unsigned size) +{ + unsigned avail = pool->state.free[0] & ~(CPU_STRUCT_ALIGN - 1); + void *ptr = NULL; + if (size <= avail) { + pool->state.free[0] = avail - size; + ptr = (uint8_t*)pool->state.last[0] - avail; + } else { + ptr = mp_alloc_internal(pool, size); + } + ASAN_UNPOISON_MEMORY_REGION(ptr, size); + return ptr; +} + +void * +mp_alloc_noalign(struct mempool *pool, unsigned size) +{ + void *ptr = NULL; + if (size <= pool->state.free[0]) { + ptr = (uint8_t*)pool->state.last[0] - pool->state.free[0]; + pool->state.free[0] -= size; + } else { + ptr = mp_alloc_internal(pool, size); + } + ASAN_UNPOISON_MEMORY_REGION(ptr, size); + return ptr; +} + +void * +mp_alloc_zero(struct mempool *pool, unsigned size) +{ + void *ptr = mp_alloc(pool, size); + bzero(ptr, size); + return ptr; +} diff --git a/src/contrib/ucw/mempool.h b/src/contrib/ucw/mempool.h new file mode 100644 index 0000000..c5a4fa8 --- /dev/null +++ b/src/contrib/ucw/mempool.h @@ -0,0 +1,124 @@ +/* + * UCW Library -- Memory Pools + * + * (c) 1997--2005 Martin Mares <mj@ucw.cz> + * (c) 2007 Pavel Charvat <pchar@ucw.cz> + * (c) 2015, 2017 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + * + * This software may be freely distributed and used according to the terms + * of the GNU Lesser General Public License. + */ + +#pragma once + +#include <string.h> +#include <stdint.h> + +#define CPU_STRUCT_ALIGN (sizeof(void*)) + +/*** + * [[defs]] + * Definitions + * ----------- + ***/ + +/** + * Memory pool state (see mp_push(), ...). + * You should use this one as an opaque handle only, the insides are internal. + **/ +struct mempool_state { + unsigned free[2]; + void *last[2]; +}; + +/** + * Memory pool. + * You should use this one as an opaque handle only, the insides are internal. + **/ +struct mempool { + struct mempool_state state; + void *unused, *last_big; + unsigned chunk_size, threshold, idx; +}; + +struct mempool_stats { /** Mempool statistics. See mp_stats(). **/ + uint64_t total_size; /** Real allocated size in bytes. */ + unsigned chain_count[3]; /** Number of allocated chunks in small/big/unused chains. */ + unsigned chain_size[3]; /** Size of allocated chunks in small/big/unused chains. */ +}; + +/*** + * [[basic]] + * Basic manipulation + * ------------------ + ***/ + +/** + * Initialize a given mempool structure. + * \p chunk_size must be in the interval `[1, UINT_MAX / 2]`. + * It will allocate memory by this large chunks and take + * memory to satisfy requests from them. + * + * Memory pools can be treated as <<trans:respools,resources>>, see <<trans:res_mempool()>>. + **/ +void mp_init(struct mempool *pool, unsigned chunk_size); + +/** + * Allocate and initialize a new memory pool. + * See \ref mp_init() for \p chunk_size limitations. + * + * The new mempool structure is allocated on the new mempool. + * + * Memory pools can be treated as <<trans:respools,resources>>, see <<trans:res_mempool()>>. + **/ +struct mempool *mp_new(unsigned chunk_size); + +/** + * Cleanup mempool initialized by mp_init or mp_new. + * Frees all the memory allocated by this mempool and, + * if created by \ref mp_new(), the \p pool itself. + **/ +void mp_delete(struct mempool *pool); + +/** + * Frees all data on a memory pool, but leaves it working. + * It can keep some of the chunks allocated to serve + * further allocation requests. Leaves the \p pool alive, + * even if it was created with \ref mp_new(). + **/ +void mp_flush(struct mempool *pool); + +/** + * Compute some statistics for debug purposes. + * See the definition of the <<struct_mempool_stats,mempool_stats structure>>. + **/ +void mp_stats(struct mempool *pool, struct mempool_stats *stats); +uint64_t mp_total_size(struct mempool *pool); /** How many bytes were allocated by the pool. **/ + +/*** + * [[alloc]] + * Allocation routines + * ------------------- + ***/ + +/** + * The function allocates new \p size bytes on a given memory pool. + * If the \p size is zero, the resulting pointer is undefined, + * but it may be safely reallocated or used as the parameter + * to other functions below. + * + * The resulting pointer is always aligned to a multiple of + * `CPU_STRUCT_ALIGN` bytes and this condition remains true also + * after future reallocations. + **/ +void *mp_alloc(struct mempool *pool, unsigned size); + +/** + * The same as \ref mp_alloc(), but the result may be unaligned. + **/ +void *mp_alloc_noalign(struct mempool *pool, unsigned size); + +/** + * The same as \ref mp_alloc(), but fills the newly allocated memory with zeroes. + **/ +void *mp_alloc_zero(struct mempool *pool, unsigned size); diff --git a/src/contrib/wire_ctx.h b/src/contrib/wire_ctx.h new file mode 100644 index 0000000..25ff143 --- /dev/null +++ b/src/contrib/wire_ctx.h @@ -0,0 +1,361 @@ +/* Copyright (C) 2018 CZ.NIC, z.s.p.o. <knot-dns@labs.nic.cz> + + This program is free software: you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation, either version 3 of the License, or + (at your option) any later version. + + This program is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see <http://www.gnu.org/licenses/>. + */ + +#pragma once + +#include <assert.h> +#include <stdbool.h> +#include <stddef.h> +#include <stdint.h> +#include <string.h> +#include <sys/types.h> + +#include "libknot/endian.h" +#include "libknot/errcode.h" + +/*! + * \brief Struct to keep the wire context. + */ +typedef struct wire_ctx { + size_t size; + uint8_t *wire; + uint8_t *position; + int error; + bool readonly; +} wire_ctx_t; + +/*! + * \brief Initialize wire context. + */ +static inline wire_ctx_t wire_ctx_init(uint8_t *data, size_t size) +{ + assert(data); + + wire_ctx_t result = { + .size = size, + .wire = data, + .position = data, + .error = KNOT_EOK, + .readonly = false + }; + + return result; +} + +/*! + * \brief Initialize read only wire context. + * + * \note No write is performed, and error is set to KNOT_EACCES. + * + */ +static inline wire_ctx_t wire_ctx_init_const(const uint8_t *data, size_t size) +{ + assert(data); + + wire_ctx_t result = wire_ctx_init((uint8_t *)data, size); + result.readonly = true; + + return result; +} + +/*! + * \brief Gets actual position. + * + * \return position from the begin. + */ +static inline size_t wire_ctx_offset(wire_ctx_t *ctx) +{ + assert(ctx); + + return ctx->position - ctx->wire; +} + +/*! + * \brief Set position offset from the begin. + * + * \param offset Wire offset (starts from 0). + * + * \note Noop if previous error. + */ +static inline void wire_ctx_set_offset(wire_ctx_t *ctx, size_t offset) +{ + assert(ctx); + + if (ctx->error != KNOT_EOK) { + return; + } + + if (offset > ctx->size) { + ctx->error = KNOT_ERANGE; + return; + } + + ctx->position = ctx->wire + offset; +} + +/*! + * \brief Gets available bytes. + * + * \return Number of bytes to end. + */ +static inline size_t wire_ctx_available(wire_ctx_t *ctx) +{ + assert(ctx); + + return ctx->size - wire_ctx_offset(ctx); +} + +/*! + * \brief Add offset to the current position. + * + * \note Noop if previous error. + */ +static inline void wire_ctx_skip(wire_ctx_t *ctx, ssize_t offset) +{ + assert(ctx); + + if (ctx->error != KNOT_EOK) { + return; + } + + // Check for out of scope skip. + if (offset >= 0) { + if (offset > wire_ctx_available(ctx)) { + ctx->error = KNOT_ERANGE; + return; + } + } else { + if (-offset > wire_ctx_offset(ctx)) { + ctx->error = KNOT_ERANGE; + return; + } + } + + ctx->position += offset; +} + +/*! + * \brief Check the context if reading is possible. + */ +static inline int wire_ctx_can_read(wire_ctx_t *ctx, size_t size) +{ + assert(ctx); + + if (ctx->error != KNOT_EOK) { + return ctx->error; + } + + if (wire_ctx_available(ctx) < size) { + return KNOT_EFEWDATA; + } + + return KNOT_EOK; +} + +/*! + * \brief Check the context if writing is possible. + */ +static inline int wire_ctx_can_write(wire_ctx_t *ctx, size_t size) +{ + assert(ctx); + + if (ctx->error != KNOT_EOK) { + return ctx->error; + } + + if (ctx->readonly) { + return KNOT_EACCES; + } + + if (wire_ctx_available(ctx) < size) { + return KNOT_ESPACE; + } + + return KNOT_EOK; +} + + +static inline void wire_ctx_read(wire_ctx_t *ctx, void *data, size_t size) +{ + assert(ctx); + assert(data); + + if (ctx->error != KNOT_EOK) { + /* Avoid leaving data uninitialized. */ + memset(data, 0, size); + return; + } + + int ret = wire_ctx_can_read(ctx, size); + if (ret != KNOT_EOK) { + ctx->error = ret; + memset(data, 0, size); + return; + } + + memcpy(data, ctx->position, size); + ctx->position += size; +} + +static inline uint8_t wire_ctx_read_u8(wire_ctx_t *ctx) +{ + uint8_t result; + wire_ctx_read(ctx, &result, sizeof(result)); + + return result; +} + +static inline uint16_t wire_ctx_read_u16(wire_ctx_t *ctx) +{ + uint16_t result; + wire_ctx_read(ctx, &result, sizeof(result)); + + return be16toh(result); +} + +static inline uint32_t wire_ctx_read_u32(wire_ctx_t *ctx) +{ + uint32_t result; + wire_ctx_read(ctx, &result, sizeof(result)); + + return be32toh(result); +} + +static inline uint64_t wire_ctx_read_u48(wire_ctx_t *ctx) +{ + /* This case is slightly tricky. */ + uint64_t result = 0; + wire_ctx_read(ctx, (uint8_t *)&result + 1, 6); + + return be64toh(result) >> 8; +} + +static inline uint64_t wire_ctx_read_u64(wire_ctx_t *ctx) +{ + uint64_t result; + wire_ctx_read(ctx, &result, sizeof(result)); + + return be64toh(result); +} + + +static inline void wire_ctx_write(wire_ctx_t *ctx, const void *data, size_t size) +{ + assert(ctx); + + if (ctx->error != KNOT_EOK) { + return; + } + + if (size == 0) { + return; + } + + assert(data); + + int ret = wire_ctx_can_write(ctx, size); + if (ret != KNOT_EOK) { + ctx->error = ret; + return; + } + + memcpy(ctx->position, data, size); + ctx->position += size; +} + +static inline void wire_ctx_write_u8(wire_ctx_t *ctx, uint8_t value) +{ + wire_ctx_write(ctx, &value, sizeof(value)); +} + +static inline void wire_ctx_write_u16(wire_ctx_t *ctx, uint16_t value) +{ + uint16_t beval = htobe16(value); + wire_ctx_write(ctx, &beval, sizeof(beval)); +} + +static inline void wire_ctx_write_u32(wire_ctx_t *ctx, uint32_t value) +{ + uint32_t beval = htobe32(value); + wire_ctx_write(ctx, &beval, sizeof(beval)); +} + +static inline void wire_ctx_write_u48(wire_ctx_t *ctx, uint64_t value) +{ + /* This case is slightly tricky. */ + uint64_t swapped = htobe64(value << 8); + wire_ctx_write(ctx, (uint8_t *)&swapped + 1, 6); +} + +static inline void wire_ctx_write_u64(wire_ctx_t *ctx, uint64_t value) +{ + uint64_t beval = htobe64(value); + wire_ctx_write(ctx, &beval, sizeof(beval)); +} + + +static inline void wire_ctx_memset(wire_ctx_t *dst, int value, size_t size) +{ + assert(dst); + + if (dst->error != KNOT_EOK) { + return; + } + + if (size == 0) { + return; + } + + int ret = wire_ctx_can_write(dst, size); + if (ret != KNOT_EOK) { + dst->error = ret; + return; + } + + memset(dst->position, value, size); + dst->position += size; +} + +static inline void wire_ctx_clear(wire_ctx_t *ctx, size_t size) +{ + wire_ctx_memset(ctx, 0, size); +} + +static inline void wire_ctx_copy(wire_ctx_t *dst, wire_ctx_t *src, size_t size) +{ + assert(dst); + assert(src); + + if (size == 0 || dst->error != KNOT_EOK) { + return; + } + + if (wire_ctx_can_read(src, size) != KNOT_EOK) { + dst->error = KNOT_EFEWDATA; + return; + } + + int ret = wire_ctx_can_write(dst, size); + if (ret != KNOT_EOK) { + dst->error = ret; + return; + } + + memcpy(dst->position, src->position, size); + dst->position += size; + src->position += size; +} + |