diff options
Diffstat (limited to 'src/libsystemd/sd-journal')
32 files changed, 14661 insertions, 0 deletions
diff --git a/src/libsystemd/sd-journal/audit-type.c b/src/libsystemd/sd-journal/audit-type.c new file mode 100644 index 0000000..122cdf5 --- /dev/null +++ b/src/libsystemd/sd-journal/audit-type.c @@ -0,0 +1,6 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ + +#include "audit-type.h" +#include "missing_audit.h" + +#include "audit_type-to-name.h" diff --git a/src/libsystemd/sd-journal/audit-type.h b/src/libsystemd/sd-journal/audit-type.h new file mode 100644 index 0000000..f2c4898 --- /dev/null +++ b/src/libsystemd/sd-journal/audit-type.h @@ -0,0 +1,22 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ +#pragma once + +#include <stdio.h> + +#include "alloc-util.h" +#include "macro.h" + +const char *audit_type_to_string(int type); +int audit_type_from_string(const char *s); + +/* This is inspired by DNS TYPEnnn formatting */ +#define audit_type_name_alloca(type) \ + ({ \ + const char *_s_; \ + _s_ = audit_type_to_string(type); \ + if (!_s_) { \ + _s_ = newa(char, STRLEN("AUDIT") + DECIMAL_STR_MAX(int)); \ + sprintf((char*) _s_, "AUDIT%04i", type); \ + } \ + _s_; \ + }) diff --git a/src/libsystemd/sd-journal/audit_type-to-name.awk b/src/libsystemd/sd-journal/audit_type-to-name.awk new file mode 100644 index 0000000..a859c44 --- /dev/null +++ b/src/libsystemd/sd-journal/audit_type-to-name.awk @@ -0,0 +1,14 @@ +# SPDX-License-Identifier: LGPL-2.1-or-later + +BEGIN{ + print "const char *audit_type_to_string(int type) {" + print " switch (type) {" +} +{ + printf " case AUDIT_%s: return \"%s\";\n", $1, $1 +} +END{ + print " default: return NULL;" + print " }" + print "}" +} diff --git a/src/libsystemd/sd-journal/catalog.c b/src/libsystemd/sd-journal/catalog.c new file mode 100644 index 0000000..5e7649c --- /dev/null +++ b/src/libsystemd/sd-journal/catalog.c @@ -0,0 +1,738 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ + +#include <errno.h> +#include <fcntl.h> +#include <locale.h> +#include <stdio.h> +#include <sys/mman.h> +#include <sys/stat.h> +#include <sys/types.h> +#include <unistd.h> + +#include "sd-id128.h" + +#include "alloc-util.h" +#include "catalog.h" +#include "conf-files.h" +#include "fd-util.h" +#include "fileio.h" +#include "hashmap.h" +#include "log.h" +#include "memory-util.h" +#include "mkdir.h" +#include "path-util.h" +#include "siphash24.h" +#include "sort-util.h" +#include "sparse-endian.h" +#include "strbuf.h" +#include "string-util.h" +#include "strv.h" +#include "tmpfile-util.h" + +const char * const catalog_file_dirs[] = { + "/usr/local/lib/systemd/catalog/", + "/usr/lib/systemd/catalog/", + NULL +}; + +#define CATALOG_SIGNATURE { 'R', 'H', 'H', 'H', 'K', 'S', 'L', 'P' } + +typedef struct CatalogHeader { + uint8_t signature[8]; /* "RHHHKSLP" */ + le32_t compatible_flags; + le32_t incompatible_flags; + le64_t header_size; + le64_t n_items; + le64_t catalog_item_size; +} CatalogHeader; + +typedef struct CatalogItem { + sd_id128_t id; + char language[32]; /* One byte is used for termination, so the maximum allowed + * length of the string is actually 31 bytes. */ + le64_t offset; +} CatalogItem; + +static void catalog_hash_func(const CatalogItem *i, struct siphash *state) { + siphash24_compress(&i->id, sizeof(i->id), state); + siphash24_compress_string(i->language, state); +} + +static int catalog_compare_func(const CatalogItem *a, const CatalogItem *b) { + unsigned k; + int r; + + for (k = 0; k < ELEMENTSOF(b->id.bytes); k++) { + r = CMP(a->id.bytes[k], b->id.bytes[k]); + if (r != 0) + return r; + } + + return strcmp(a->language, b->language); +} + +DEFINE_HASH_OPS(catalog_hash_ops, CatalogItem, catalog_hash_func, catalog_compare_func); + +static bool next_header(const char **s) { + const char *e; + + e = strchr(*s, '\n'); + + /* Unexpected end */ + if (!e) + return false; + + /* End of headers */ + if (e == *s) + return false; + + *s = e + 1; + return true; +} + +static const char *skip_header(const char *s) { + while (next_header(&s)) + ; + return s; +} + +static char *combine_entries(const char *one, const char *two) { + const char *b1, *b2; + size_t l1, l2, n; + char *dest, *p; + + /* Find split point of headers to body */ + b1 = skip_header(one); + b2 = skip_header(two); + + l1 = strlen(one); + l2 = strlen(two); + dest = new(char, l1 + l2 + 1); + if (!dest) { + log_oom(); + return NULL; + } + + p = dest; + + /* Headers from @one */ + n = b1 - one; + p = mempcpy(p, one, n); + + /* Headers from @two, these will only be found if not present above */ + n = b2 - two; + p = mempcpy(p, two, n); + + /* Body from @one */ + n = l1 - (b1 - one); + if (n > 0) + p = mempcpy(p, b1, n); + /* Body from @two */ + else { + n = l2 - (b2 - two); + p = mempcpy(p, b2, n); + } + + assert(p - dest <= (ptrdiff_t)(l1 + l2)); + p[0] = '\0'; + return dest; +} + +static int finish_item( + OrderedHashmap *h, + sd_id128_t id, + const char *language, + char *payload, size_t payload_size) { + + _cleanup_free_ CatalogItem *i = NULL; + _cleanup_free_ char *combined = NULL; + char *prev; + + assert(h); + assert(payload); + assert(payload_size > 0); + + i = new0(CatalogItem, 1); + if (!i) + return log_oom(); + + i->id = id; + if (language) { + assert(strlen(language) > 1 && strlen(language) < 32); + strcpy(i->language, language); + } + + prev = ordered_hashmap_get(h, i); + if (prev) { + /* Already have such an item, combine them */ + combined = combine_entries(payload, prev); + if (!combined) + return log_oom(); + + if (ordered_hashmap_update(h, i, combined) < 0) + return log_oom(); + combined = NULL; + free(prev); + } else { + /* A new item */ + combined = memdup(payload, payload_size + 1); + if (!combined) + return log_oom(); + + if (ordered_hashmap_put(h, i, combined) < 0) + return log_oom(); + i = NULL; + combined = NULL; + } + + return 0; +} + +int catalog_file_lang(const char* filename, char **lang) { + char *beg, *end, *_lang; + + end = endswith(filename, ".catalog"); + if (!end) + return 0; + + beg = end - 1; + while (beg > filename && !IN_SET(*beg, '.', '/') && end - beg < 32) + beg--; + + if (*beg != '.' || end <= beg + 1) + return 0; + + _lang = strndup(beg + 1, end - beg - 1); + if (!_lang) + return -ENOMEM; + + *lang = _lang; + return 1; +} + +static int catalog_entry_lang( + const char* filename, + unsigned line, + const char* t, + const char* deflang, + char **ret) { + + size_t c; + char *z; + + c = strlen(t); + if (c < 2) + return log_error_errno(SYNTHETIC_ERRNO(EINVAL), + "[%s:%u] Language too short.", filename, line); + if (c > 31) + return log_error_errno(SYNTHETIC_ERRNO(EINVAL), + "[%s:%u] language too long.", filename, line); + + if (deflang) { + if (streq(t, deflang)) { + log_warning("[%s:%u] language specified unnecessarily", filename, line); + return 0; + } + + log_warning("[%s:%u] language differs from default for file", filename, line); + } + + z = strdup(t); + if (!z) + return -ENOMEM; + + *ret = z; + return 0; +} + +int catalog_import_file(OrderedHashmap *h, const char *path) { + _cleanup_fclose_ FILE *f = NULL; + _cleanup_free_ char *payload = NULL; + size_t payload_size = 0; + unsigned n = 0; + sd_id128_t id; + _cleanup_free_ char *deflang = NULL, *lang = NULL; + bool got_id = false, empty_line = true; + int r; + + assert(h); + assert(path); + + f = fopen(path, "re"); + if (!f) + return log_error_errno(errno, "Failed to open file %s: %m", path); + + r = catalog_file_lang(path, &deflang); + if (r < 0) + log_error_errno(r, "Failed to determine language for file %s: %m", path); + if (r == 1) + log_debug("File %s has language %s.", path, deflang); + + for (;;) { + _cleanup_free_ char *line = NULL; + size_t line_len; + + r = read_line(f, LONG_LINE_MAX, &line); + if (r < 0) + return log_error_errno(r, "Failed to read file %s: %m", path); + if (r == 0) + break; + + n++; + + if (isempty(line)) { + empty_line = true; + continue; + } + + if (strchr(COMMENTS, line[0])) + continue; + + if (empty_line && + strlen(line) >= 2+1+32 && + line[0] == '-' && + line[1] == '-' && + line[2] == ' ' && + IN_SET(line[2+1+32], ' ', '\0')) { + + bool with_language; + sd_id128_t jd; + + /* New entry */ + + with_language = line[2+1+32] != '\0'; + line[2+1+32] = '\0'; + + if (sd_id128_from_string(line + 2 + 1, &jd) >= 0) { + + if (got_id) { + if (payload_size == 0) + return log_error_errno(SYNTHETIC_ERRNO(EINVAL), + "[%s:%u] No payload text.", + path, + n); + + r = finish_item(h, id, lang ?: deflang, payload, payload_size); + if (r < 0) + return r; + + lang = mfree(lang); + payload_size = 0; + } + + if (with_language) { + char *t; + + t = strstrip(line + 2 + 1 + 32 + 1); + r = catalog_entry_lang(path, n, t, deflang, &lang); + if (r < 0) + return r; + } + + got_id = true; + empty_line = false; + id = jd; + + continue; + } + } + + /* Payload */ + if (!got_id) + return log_error_errno(SYNTHETIC_ERRNO(EINVAL), + "[%s:%u] Got payload before ID.", + path, n); + + line_len = strlen(line); + if (!GREEDY_REALLOC(payload, payload_size + (empty_line ? 1 : 0) + line_len + 1 + 1)) + return log_oom(); + + if (empty_line) + payload[payload_size++] = '\n'; + memcpy(payload + payload_size, line, line_len); + payload_size += line_len; + payload[payload_size++] = '\n'; + payload[payload_size] = '\0'; + + empty_line = false; + } + + if (got_id) { + if (payload_size == 0) + return log_error_errno(SYNTHETIC_ERRNO(EINVAL), + "[%s:%u] No payload text.", + path, n); + + r = finish_item(h, id, lang ?: deflang, payload, payload_size); + if (r < 0) + return r; + } + + return 0; +} + +static int64_t write_catalog( + const char *database, + struct strbuf *sb, + CatalogItem *items, + size_t n) { + + _cleanup_fclose_ FILE *w = NULL; + _cleanup_free_ char *p = NULL; + CatalogHeader header; + size_t k; + int r; + + r = mkdir_parents(database, 0755); + if (r < 0) + return log_error_errno(r, "Failed to create parent directories of %s: %m", database); + + r = fopen_temporary(database, &w, &p); + if (r < 0) + return log_error_errno(r, "Failed to open database for writing: %s: %m", + database); + + header = (CatalogHeader) { + .signature = CATALOG_SIGNATURE, + .header_size = htole64(CONST_ALIGN_TO(sizeof(CatalogHeader), 8)), + .catalog_item_size = htole64(sizeof(CatalogItem)), + .n_items = htole64(n), + }; + + r = -EIO; + + k = fwrite(&header, 1, sizeof(header), w); + if (k != sizeof(header)) { + log_error("%s: failed to write header.", p); + goto error; + } + + k = fwrite(items, 1, n * sizeof(CatalogItem), w); + if (k != n * sizeof(CatalogItem)) { + log_error("%s: failed to write database.", p); + goto error; + } + + k = fwrite(sb->buf, 1, sb->len, w); + if (k != sb->len) { + log_error("%s: failed to write strings.", p); + goto error; + } + + r = fflush_and_check(w); + if (r < 0) { + log_error_errno(r, "%s: failed to write database: %m", p); + goto error; + } + + (void) fchmod(fileno(w), 0644); + + if (rename(p, database) < 0) { + r = log_error_errno(errno, "rename (%s -> %s) failed: %m", p, database); + goto error; + } + + return ftello(w); + +error: + (void) unlink(p); + return r; +} + +int catalog_update(const char* database, const char* root, const char* const* dirs) { + _cleanup_strv_free_ char **files = NULL; + _cleanup_(strbuf_freep) struct strbuf *sb = NULL; + _cleanup_ordered_hashmap_free_free_free_ OrderedHashmap *h = NULL; + _cleanup_free_ CatalogItem *items = NULL; + ssize_t offset; + char *payload; + CatalogItem *i; + unsigned n; + int r; + int64_t sz; + + h = ordered_hashmap_new(&catalog_hash_ops); + sb = strbuf_new(); + if (!h || !sb) + return log_oom(); + + r = conf_files_list_strv(&files, ".catalog", root, 0, dirs); + if (r < 0) + return log_error_errno(r, "Failed to get catalog files: %m"); + + STRV_FOREACH(f, files) { + log_debug("Reading file '%s'", *f); + r = catalog_import_file(h, *f); + if (r < 0) + return log_error_errno(r, "Failed to import file '%s': %m", *f); + } + + if (ordered_hashmap_size(h) <= 0) { + log_info("No items in catalog."); + return 0; + } else + log_debug("Found %u items in catalog.", ordered_hashmap_size(h)); + + items = new(CatalogItem, ordered_hashmap_size(h)); + if (!items) + return log_oom(); + + n = 0; + ORDERED_HASHMAP_FOREACH_KEY(payload, i, h) { + log_trace("Found " SD_ID128_FORMAT_STR ", language %s", + SD_ID128_FORMAT_VAL(i->id), + isempty(i->language) ? "C" : i->language); + + offset = strbuf_add_string(sb, payload, strlen(payload)); + if (offset < 0) + return log_oom(); + + i->offset = htole64((uint64_t) offset); + items[n++] = *i; + } + + assert(n == ordered_hashmap_size(h)); + typesafe_qsort(items, n, catalog_compare_func); + + strbuf_complete(sb); + + sz = write_catalog(database, sb, items, n); + if (sz < 0) + return log_error_errno(sz, "Failed to write %s: %m", database); + + log_debug("%s: wrote %u items, with %zu bytes of strings, %"PRIi64" total size.", + database, n, sb->len, sz); + return 0; +} + +static int open_mmap(const char *database, int *_fd, struct stat *_st, void **_p) { + _cleanup_close_ int fd = -1; + const CatalogHeader *h; + struct stat st; + void *p; + + assert(_fd); + assert(_st); + assert(_p); + + fd = open(database, O_RDONLY|O_CLOEXEC); + if (fd < 0) + return -errno; + + if (fstat(fd, &st) < 0) + return -errno; + + if (st.st_size < (off_t) sizeof(CatalogHeader) || file_offset_beyond_memory_size(st.st_size)) + return -EINVAL; + + p = mmap(NULL, st.st_size, PROT_READ, MAP_SHARED, fd, 0); + if (p == MAP_FAILED) + return -errno; + + h = p; + if (memcmp(h->signature, (const uint8_t[]) CATALOG_SIGNATURE, sizeof(h->signature)) != 0 || + le64toh(h->header_size) < sizeof(CatalogHeader) || + le64toh(h->catalog_item_size) < sizeof(CatalogItem) || + h->incompatible_flags != 0 || + le64toh(h->n_items) <= 0 || + st.st_size < (off_t) (le64toh(h->header_size) + le64toh(h->catalog_item_size) * le64toh(h->n_items))) { + munmap(p, st.st_size); + return -EBADMSG; + } + + *_fd = TAKE_FD(fd); + *_st = st; + *_p = p; + + return 0; +} + +static const char *find_id(void *p, sd_id128_t id) { + CatalogItem *f = NULL, key = { .id = id }; + const CatalogHeader *h = p; + const char *loc; + + loc = setlocale(LC_MESSAGES, NULL); + if (!isempty(loc) && !STR_IN_SET(loc, "C", "POSIX")) { + size_t len; + + len = strcspn(loc, ".@"); + if (len > sizeof(key.language) - 1) + log_debug("LC_MESSAGES value too long, ignoring: \"%.*s\"", (int) len, loc); + else { + strncpy(key.language, loc, len); + key.language[len] = '\0'; + + f = bsearch(&key, + (const uint8_t*) p + le64toh(h->header_size), + le64toh(h->n_items), + le64toh(h->catalog_item_size), + (comparison_fn_t) catalog_compare_func); + if (!f) { + char *e; + + e = strchr(key.language, '_'); + if (e) { + *e = 0; + f = bsearch(&key, + (const uint8_t*) p + le64toh(h->header_size), + le64toh(h->n_items), + le64toh(h->catalog_item_size), + (comparison_fn_t) catalog_compare_func); + } + } + } + } + + if (!f) { + zero(key.language); + f = bsearch(&key, + (const uint8_t*) p + le64toh(h->header_size), + le64toh(h->n_items), + le64toh(h->catalog_item_size), + (comparison_fn_t) catalog_compare_func); + } + + if (!f) + return NULL; + + return (const char*) p + + le64toh(h->header_size) + + le64toh(h->n_items) * le64toh(h->catalog_item_size) + + le64toh(f->offset); +} + +int catalog_get(const char* database, sd_id128_t id, char **_text) { + _cleanup_close_ int fd = -1; + void *p = NULL; + struct stat st = {}; + char *text = NULL; + int r; + const char *s; + + assert(_text); + + r = open_mmap(database, &fd, &st, &p); + if (r < 0) + return r; + + s = find_id(p, id); + if (!s) { + r = -ENOENT; + goto finish; + } + + text = strdup(s); + if (!text) { + r = -ENOMEM; + goto finish; + } + + *_text = text; + r = 0; + +finish: + if (p) + munmap(p, st.st_size); + + return r; +} + +static char *find_header(const char *s, const char *header) { + + for (;;) { + const char *v; + + v = startswith(s, header); + if (v) { + v += strspn(v, WHITESPACE); + return strndup(v, strcspn(v, NEWLINE)); + } + + if (!next_header(&s)) + return NULL; + } +} + +static void dump_catalog_entry(FILE *f, sd_id128_t id, const char *s, bool oneline) { + if (oneline) { + _cleanup_free_ char *subject = NULL, *defined_by = NULL; + + subject = find_header(s, "Subject:"); + defined_by = find_header(s, "Defined-By:"); + + fprintf(f, SD_ID128_FORMAT_STR " %s: %s\n", + SD_ID128_FORMAT_VAL(id), + strna(defined_by), strna(subject)); + } else + fprintf(f, "-- " SD_ID128_FORMAT_STR "\n%s\n", + SD_ID128_FORMAT_VAL(id), s); +} + +int catalog_list(FILE *f, const char *database, bool oneline) { + _cleanup_close_ int fd = -1; + void *p = NULL; + struct stat st; + const CatalogHeader *h; + const CatalogItem *items; + int r; + unsigned n; + sd_id128_t last_id; + bool last_id_set = false; + + r = open_mmap(database, &fd, &st, &p); + if (r < 0) + return r; + + h = p; + items = (const CatalogItem*) ((const uint8_t*) p + le64toh(h->header_size)); + + for (n = 0; n < le64toh(h->n_items); n++) { + const char *s; + + if (last_id_set && sd_id128_equal(last_id, items[n].id)) + continue; + + assert_se(s = find_id(p, items[n].id)); + + dump_catalog_entry(f, items[n].id, s, oneline); + + last_id_set = true; + last_id = items[n].id; + } + + munmap(p, st.st_size); + + return 0; +} + +int catalog_list_items(FILE *f, const char *database, bool oneline, char **items) { + int r = 0; + + STRV_FOREACH(item, items) { + sd_id128_t id; + int k; + _cleanup_free_ char *msg = NULL; + + k = sd_id128_from_string(*item, &id); + if (k < 0) { + log_error_errno(k, "Failed to parse id128 '%s': %m", *item); + if (r == 0) + r = k; + continue; + } + + k = catalog_get(database, id, &msg); + if (k < 0) { + log_full_errno(k == -ENOENT ? LOG_NOTICE : LOG_ERR, k, + "Failed to retrieve catalog entry for '%s': %m", *item); + if (r == 0) + r = k; + continue; + } + + dump_catalog_entry(f, id, msg, oneline); + } + + return r; +} diff --git a/src/libsystemd/sd-journal/catalog.h b/src/libsystemd/sd-journal/catalog.h new file mode 100644 index 0000000..df27869 --- /dev/null +++ b/src/libsystemd/sd-journal/catalog.h @@ -0,0 +1,19 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ +#pragma once + +#include <stdbool.h> +#include <stdio.h> + +#include "sd-id128.h" + +#include "hashmap.h" +#include "strbuf.h" + +int catalog_import_file(OrderedHashmap *h, const char *path); +int catalog_update(const char* database, const char* root, const char* const* dirs); +int catalog_get(const char* database, sd_id128_t id, char **data); +int catalog_list(FILE *f, const char* database, bool oneline); +int catalog_list_items(FILE *f, const char* database, bool oneline, char **items); +int catalog_file_lang(const char *filename, char **lang); +extern const char * const catalog_file_dirs[]; +extern const struct hash_ops catalog_hash_ops; diff --git a/src/libsystemd/sd-journal/fsprg.c b/src/libsystemd/sd-journal/fsprg.c new file mode 100644 index 0000000..ab4cef9 --- /dev/null +++ b/src/libsystemd/sd-journal/fsprg.c @@ -0,0 +1,381 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later + * + * fsprg v0.1 - (seekable) forward-secure pseudorandom generator + * Copyright © 2012 B. Poettering + * Contact: fsprg@point-at-infinity.org + * + * This library is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * This library 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 + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with this library; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA + * 02110-1301 USA + */ + +/* + * See "Practical Secure Logging: Seekable Sequential Key Generators" + * by G. A. Marson, B. Poettering for details: + * + * http://eprint.iacr.org/2013/397 + */ + +#include <string.h> + +#include "fsprg.h" +#include "gcrypt-util.h" +#include "memory-util.h" + +#define ISVALID_SECPAR(secpar) (((secpar) % 16 == 0) && ((secpar) >= 16) && ((secpar) <= 16384)) +#define VALIDATE_SECPAR(secpar) assert(ISVALID_SECPAR(secpar)); + +#define RND_HASH GCRY_MD_SHA256 +#define RND_GEN_P 0x01 +#define RND_GEN_Q 0x02 +#define RND_GEN_X 0x03 + +#pragma GCC diagnostic ignored "-Wpointer-arith" +/* TODO: remove void* arithmetic and this work-around */ + +/******************************************************************************/ + +static void mpi_export(void *buf, size_t buflen, const gcry_mpi_t x) { + unsigned len; + size_t nwritten; + + assert(gcry_mpi_cmp_ui(x, 0) >= 0); + len = (gcry_mpi_get_nbits(x) + 7) / 8; + assert(len <= buflen); + memzero(buf, buflen); + gcry_mpi_print(GCRYMPI_FMT_USG, buf + (buflen - len), len, &nwritten, x); + assert(nwritten == len); +} + +static gcry_mpi_t mpi_import(const void *buf, size_t buflen) { + gcry_mpi_t h; + _unused_ unsigned len; + + assert_se(gcry_mpi_scan(&h, GCRYMPI_FMT_USG, buf, buflen, NULL) == 0); + len = (gcry_mpi_get_nbits(h) + 7) / 8; + assert(len <= buflen); + assert(gcry_mpi_cmp_ui(h, 0) >= 0); + + return h; +} + +static void uint64_export(void *buf, size_t buflen, uint64_t x) { + assert(buflen == 8); + ((uint8_t*) buf)[0] = (x >> 56) & 0xff; + ((uint8_t*) buf)[1] = (x >> 48) & 0xff; + ((uint8_t*) buf)[2] = (x >> 40) & 0xff; + ((uint8_t*) buf)[3] = (x >> 32) & 0xff; + ((uint8_t*) buf)[4] = (x >> 24) & 0xff; + ((uint8_t*) buf)[5] = (x >> 16) & 0xff; + ((uint8_t*) buf)[6] = (x >> 8) & 0xff; + ((uint8_t*) buf)[7] = (x >> 0) & 0xff; +} + +_pure_ static uint64_t uint64_import(const void *buf, size_t buflen) { + assert(buflen == 8); + return + (uint64_t)(((uint8_t*) buf)[0]) << 56 | + (uint64_t)(((uint8_t*) buf)[1]) << 48 | + (uint64_t)(((uint8_t*) buf)[2]) << 40 | + (uint64_t)(((uint8_t*) buf)[3]) << 32 | + (uint64_t)(((uint8_t*) buf)[4]) << 24 | + (uint64_t)(((uint8_t*) buf)[5]) << 16 | + (uint64_t)(((uint8_t*) buf)[6]) << 8 | + (uint64_t)(((uint8_t*) buf)[7]) << 0; +} + +/* deterministically generate from seed/idx a string of buflen pseudorandom bytes */ +static void det_randomize(void *buf, size_t buflen, const void *seed, size_t seedlen, uint32_t idx) { + gcry_md_hd_t hd, hd2; + size_t olen, cpylen; + gcry_error_t err; + uint32_t ctr; + + olen = gcry_md_get_algo_dlen(RND_HASH); + err = gcry_md_open(&hd, RND_HASH, 0); + assert_se(gcry_err_code(err) == GPG_ERR_NO_ERROR); /* This shouldn't happen */ + gcry_md_write(hd, seed, seedlen); + gcry_md_putc(hd, (idx >> 24) & 0xff); + gcry_md_putc(hd, (idx >> 16) & 0xff); + gcry_md_putc(hd, (idx >> 8) & 0xff); + gcry_md_putc(hd, (idx >> 0) & 0xff); + + for (ctr = 0; buflen; ctr++) { + err = gcry_md_copy(&hd2, hd); + assert_se(gcry_err_code(err) == GPG_ERR_NO_ERROR); /* This shouldn't happen */ + gcry_md_putc(hd2, (ctr >> 24) & 0xff); + gcry_md_putc(hd2, (ctr >> 16) & 0xff); + gcry_md_putc(hd2, (ctr >> 8) & 0xff); + gcry_md_putc(hd2, (ctr >> 0) & 0xff); + gcry_md_final(hd2); + cpylen = (buflen < olen) ? buflen : olen; + memcpy(buf, gcry_md_read(hd2, RND_HASH), cpylen); + gcry_md_close(hd2); + buf += cpylen; + buflen -= cpylen; + } + gcry_md_close(hd); +} + +/* deterministically generate from seed/idx a prime of length `bits' that is 3 (mod 4) */ +static gcry_mpi_t genprime3mod4(int bits, const void *seed, size_t seedlen, uint32_t idx) { + size_t buflen = bits / 8; + uint8_t buf[buflen]; + gcry_mpi_t p; + + assert(bits % 8 == 0); + assert(buflen > 0); + + det_randomize(buf, buflen, seed, seedlen, idx); + buf[0] |= 0xc0; /* set upper two bits, so that n=pq has maximum size */ + buf[buflen - 1] |= 0x03; /* set lower two bits, to have result 3 (mod 4) */ + + p = mpi_import(buf, buflen); + while (gcry_prime_check(p, 0)) + gcry_mpi_add_ui(p, p, 4); + + return p; +} + +/* deterministically generate from seed/idx a quadratic residue (mod n) */ +static gcry_mpi_t gensquare(const gcry_mpi_t n, const void *seed, size_t seedlen, uint32_t idx, unsigned secpar) { + size_t buflen = secpar / 8; + uint8_t buf[buflen]; + gcry_mpi_t x; + + det_randomize(buf, buflen, seed, seedlen, idx); + buf[0] &= 0x7f; /* clear upper bit, so that we have x < n */ + x = mpi_import(buf, buflen); + assert(gcry_mpi_cmp(x, n) < 0); + gcry_mpi_mulm(x, x, x, n); + return x; +} + +/* compute 2^m (mod phi(p)), for a prime p */ +static gcry_mpi_t twopowmodphi(uint64_t m, const gcry_mpi_t p) { + gcry_mpi_t phi, r; + int n; + + phi = gcry_mpi_new(0); + gcry_mpi_sub_ui(phi, p, 1); + + /* count number of used bits in m */ + for (n = 0; (1ULL << n) <= m; n++) + ; + + r = gcry_mpi_new(0); + gcry_mpi_set_ui(r, 1); + while (n) { /* square and multiply algorithm for fast exponentiation */ + n--; + gcry_mpi_mulm(r, r, r, phi); + if (m & ((uint64_t)1 << n)) { + gcry_mpi_add(r, r, r); + if (gcry_mpi_cmp(r, phi) >= 0) + gcry_mpi_sub(r, r, phi); + } + } + + gcry_mpi_release(phi); + return r; +} + +/* Decompose $x \in Z_n$ into $(xp,xq) \in Z_p \times Z_q$ using Chinese Remainder Theorem */ +static void CRT_decompose(gcry_mpi_t *xp, gcry_mpi_t *xq, const gcry_mpi_t x, const gcry_mpi_t p, const gcry_mpi_t q) { + *xp = gcry_mpi_new(0); + *xq = gcry_mpi_new(0); + gcry_mpi_mod(*xp, x, p); + gcry_mpi_mod(*xq, x, q); +} + +/* Compose $(xp,xq) \in Z_p \times Z_q$ into $x \in Z_n$ using Chinese Remainder Theorem */ +static void CRT_compose(gcry_mpi_t *x, const gcry_mpi_t xp, const gcry_mpi_t xq, const gcry_mpi_t p, const gcry_mpi_t q) { + gcry_mpi_t a, u; + + a = gcry_mpi_new(0); + u = gcry_mpi_new(0); + *x = gcry_mpi_new(0); + gcry_mpi_subm(a, xq, xp, q); + gcry_mpi_invm(u, p, q); + gcry_mpi_mulm(a, a, u, q); /* a = (xq - xp) / p (mod q) */ + gcry_mpi_mul(*x, p, a); + gcry_mpi_add(*x, *x, xp); /* x = p * ((xq - xp) / p mod q) + xp */ + gcry_mpi_release(a); + gcry_mpi_release(u); +} + +/******************************************************************************/ + +size_t FSPRG_mskinbytes(unsigned _secpar) { + VALIDATE_SECPAR(_secpar); + return 2 + 2 * (_secpar / 2) / 8; /* to store header,p,q */ +} + +size_t FSPRG_mpkinbytes(unsigned _secpar) { + VALIDATE_SECPAR(_secpar); + return 2 + _secpar / 8; /* to store header,n */ +} + +size_t FSPRG_stateinbytes(unsigned _secpar) { + VALIDATE_SECPAR(_secpar); + return 2 + 2 * _secpar / 8 + 8; /* to store header,n,x,epoch */ +} + +static void store_secpar(void *buf, uint16_t secpar) { + secpar = secpar / 16 - 1; + ((uint8_t*) buf)[0] = (secpar >> 8) & 0xff; + ((uint8_t*) buf)[1] = (secpar >> 0) & 0xff; +} + +static uint16_t read_secpar(const void *buf) { + uint16_t secpar; + secpar = + (uint16_t)(((uint8_t*) buf)[0]) << 8 | + (uint16_t)(((uint8_t*) buf)[1]) << 0; + return 16 * (secpar + 1); +} + +void FSPRG_GenMK(void *msk, void *mpk, const void *seed, size_t seedlen, unsigned _secpar) { + uint8_t iseed[FSPRG_RECOMMENDED_SEEDLEN]; + gcry_mpi_t n, p, q; + uint16_t secpar; + + VALIDATE_SECPAR(_secpar); + secpar = _secpar; + + initialize_libgcrypt(false); + + if (!seed) { + gcry_randomize(iseed, FSPRG_RECOMMENDED_SEEDLEN, GCRY_STRONG_RANDOM); + seed = iseed; + seedlen = FSPRG_RECOMMENDED_SEEDLEN; + } + + p = genprime3mod4(secpar / 2, seed, seedlen, RND_GEN_P); + q = genprime3mod4(secpar / 2, seed, seedlen, RND_GEN_Q); + + if (msk) { + store_secpar(msk + 0, secpar); + mpi_export(msk + 2 + 0 * (secpar / 2) / 8, (secpar / 2) / 8, p); + mpi_export(msk + 2 + 1 * (secpar / 2) / 8, (secpar / 2) / 8, q); + } + + if (mpk) { + n = gcry_mpi_new(0); + gcry_mpi_mul(n, p, q); + assert(gcry_mpi_get_nbits(n) == secpar); + + store_secpar(mpk + 0, secpar); + mpi_export(mpk + 2, secpar / 8, n); + + gcry_mpi_release(n); + } + + gcry_mpi_release(p); + gcry_mpi_release(q); +} + +void FSPRG_GenState0(void *state, const void *mpk, const void *seed, size_t seedlen) { + gcry_mpi_t n, x; + uint16_t secpar; + + initialize_libgcrypt(false); + + secpar = read_secpar(mpk + 0); + n = mpi_import(mpk + 2, secpar / 8); + x = gensquare(n, seed, seedlen, RND_GEN_X, secpar); + + memcpy(state, mpk, 2 + secpar / 8); + mpi_export(state + 2 + 1 * secpar / 8, secpar / 8, x); + memzero(state + 2 + 2 * secpar / 8, 8); + + gcry_mpi_release(n); + gcry_mpi_release(x); +} + +void FSPRG_Evolve(void *state) { + gcry_mpi_t n, x; + uint16_t secpar; + uint64_t epoch; + + initialize_libgcrypt(false); + + secpar = read_secpar(state + 0); + n = mpi_import(state + 2 + 0 * secpar / 8, secpar / 8); + x = mpi_import(state + 2 + 1 * secpar / 8, secpar / 8); + epoch = uint64_import(state + 2 + 2 * secpar / 8, 8); + + gcry_mpi_mulm(x, x, x, n); + epoch++; + + mpi_export(state + 2 + 1 * secpar / 8, secpar / 8, x); + uint64_export(state + 2 + 2 * secpar / 8, 8, epoch); + + gcry_mpi_release(n); + gcry_mpi_release(x); +} + +uint64_t FSPRG_GetEpoch(const void *state) { + uint16_t secpar; + secpar = read_secpar(state + 0); + return uint64_import(state + 2 + 2 * secpar / 8, 8); +} + +void FSPRG_Seek(void *state, uint64_t epoch, const void *msk, const void *seed, size_t seedlen) { + gcry_mpi_t p, q, n, x, xp, xq, kp, kq, xm; + uint16_t secpar; + + initialize_libgcrypt(false); + + secpar = read_secpar(msk + 0); + p = mpi_import(msk + 2 + 0 * (secpar / 2) / 8, (secpar / 2) / 8); + q = mpi_import(msk + 2 + 1 * (secpar / 2) / 8, (secpar / 2) / 8); + + n = gcry_mpi_new(0); + gcry_mpi_mul(n, p, q); + + x = gensquare(n, seed, seedlen, RND_GEN_X, secpar); + CRT_decompose(&xp, &xq, x, p, q); /* split (mod n) into (mod p) and (mod q) using CRT */ + + kp = twopowmodphi(epoch, p); /* compute 2^epoch (mod phi(p)) */ + kq = twopowmodphi(epoch, q); /* compute 2^epoch (mod phi(q)) */ + + gcry_mpi_powm(xp, xp, kp, p); /* compute x^(2^epoch) (mod p) */ + gcry_mpi_powm(xq, xq, kq, q); /* compute x^(2^epoch) (mod q) */ + + CRT_compose(&xm, xp, xq, p, q); /* combine (mod p) and (mod q) to (mod n) using CRT */ + + store_secpar(state + 0, secpar); + mpi_export(state + 2 + 0 * secpar / 8, secpar / 8, n); + mpi_export(state + 2 + 1 * secpar / 8, secpar / 8, xm); + uint64_export(state + 2 + 2 * secpar / 8, 8, epoch); + + gcry_mpi_release(p); + gcry_mpi_release(q); + gcry_mpi_release(n); + gcry_mpi_release(x); + gcry_mpi_release(xp); + gcry_mpi_release(xq); + gcry_mpi_release(kp); + gcry_mpi_release(kq); + gcry_mpi_release(xm); +} + +void FSPRG_GetKey(const void *state, void *key, size_t keylen, uint32_t idx) { + uint16_t secpar; + + initialize_libgcrypt(false); + + secpar = read_secpar(state + 0); + det_randomize(key, keylen, state + 2, 2 * secpar / 8 + 8, idx); +} diff --git a/src/libsystemd/sd-journal/fsprg.h b/src/libsystemd/sd-journal/fsprg.h new file mode 100644 index 0000000..dfe2d79 --- /dev/null +++ b/src/libsystemd/sd-journal/fsprg.h @@ -0,0 +1,62 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ +#pragma once + +/* + * fsprg v0.1 - (seekable) forward-secure pseudorandom generator + * Copyright © 2012 B. Poettering + * Contact: fsprg@point-at-infinity.org + * + * This library is free software; you can redistribute it and/or + * modify it under the terms of the GNU Lesser General Public + * License as published by the Free Software Foundation; either + * version 2.1 of the License, or (at your option) any later version. + * + * This library 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 + * Lesser General Public License for more details. + * + * You should have received a copy of the GNU Lesser General Public + * License along with this library; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA + * 02110-1301 USA + */ + +#include <inttypes.h> +#include <sys/types.h> + +#include "macro.h" +#include "util.h" + +#ifdef __cplusplus +extern "C" { +#endif + +#define FSPRG_RECOMMENDED_SECPAR 1536 +#define FSPRG_RECOMMENDED_SEEDLEN (96/8) + +size_t FSPRG_mskinbytes(unsigned secpar) _const_; +size_t FSPRG_mpkinbytes(unsigned secpar) _const_; +size_t FSPRG_stateinbytes(unsigned secpar) _const_; + +/* Setup msk and mpk. Providing seed != NULL makes this algorithm deterministic. */ +void FSPRG_GenMK(void *msk, void *mpk, const void *seed, size_t seedlen, unsigned secpar); + +/* Initialize state deterministically in dependence on seed. */ +/* Note: in case one wants to run only one GenState0 per GenMK it is safe to use + the same seed for both GenMK and GenState0. +*/ +void FSPRG_GenState0(void *state, const void *mpk, const void *seed, size_t seedlen); + +void FSPRG_Evolve(void *state); + +uint64_t FSPRG_GetEpoch(const void *state) _pure_; + +/* Seek to any arbitrary state (by providing msk together with seed from GenState0). */ +void FSPRG_Seek(void *state, uint64_t epoch, const void *msk, const void *seed, size_t seedlen); + +void FSPRG_GetKey(const void *state, void *key, size_t keylen, uint32_t idx); + +#ifdef __cplusplus +} +#endif diff --git a/src/libsystemd/sd-journal/generate-audit_type-list.sh b/src/libsystemd/sd-journal/generate-audit_type-list.sh new file mode 100755 index 0000000..3851ea1 --- /dev/null +++ b/src/libsystemd/sd-journal/generate-audit_type-list.sh @@ -0,0 +1,17 @@ +#!/usr/bin/env bash +# SPDX-License-Identifier: LGPL-2.1-or-later +set -eu +set -o pipefail + +cpp="${1:?}" +shift + +includes=() +for i in "$@"; do + includes+=(-include "$i") +done + +$cpp -dM "${includes[@]}" - </dev/null | \ + grep -vE 'AUDIT_.*(FIRST|LAST)_' | \ + sed -r -n 's/^#define\s+AUDIT_(\w+)\s+([0-9]{4})\s*$$/\1\t\2/p' | \ + sort -k2 diff --git a/src/libsystemd/sd-journal/journal-authenticate.c b/src/libsystemd/sd-journal/journal-authenticate.c new file mode 100644 index 0000000..1cb8943 --- /dev/null +++ b/src/libsystemd/sd-journal/journal-authenticate.c @@ -0,0 +1,541 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ + +#include <fcntl.h> +#include <sys/mman.h> + +#include "fd-util.h" +#include "fsprg.h" +#include "gcrypt-util.h" +#include "hexdecoct.h" +#include "journal-authenticate.h" +#include "journal-def.h" +#include "journal-file.h" +#include "memory-util.h" +#include "time-util.h" + +static uint64_t journal_file_tag_seqnum(JournalFile *f) { + uint64_t r; + + assert(f); + + r = le64toh(f->header->n_tags) + 1; + f->header->n_tags = htole64(r); + + return r; +} + +int journal_file_append_tag(JournalFile *f) { + Object *o; + uint64_t p; + int r; + + assert(f); + + if (!JOURNAL_HEADER_SEALED(f->header)) + return 0; + + if (!f->hmac_running) + return 0; + + assert(f->hmac); + + r = journal_file_append_object(f, OBJECT_TAG, sizeof(struct TagObject), &o, &p); + if (r < 0) + return r; + + o->tag.seqnum = htole64(journal_file_tag_seqnum(f)); + o->tag.epoch = htole64(FSPRG_GetEpoch(f->fsprg_state)); + + log_debug("Writing tag %"PRIu64" for epoch %"PRIu64"", + le64toh(o->tag.seqnum), + FSPRG_GetEpoch(f->fsprg_state)); + + /* Add the tag object itself, so that we can protect its + * header. This will exclude the actual hash value in it */ + r = journal_file_hmac_put_object(f, OBJECT_TAG, o, p); + if (r < 0) + return r; + + /* Get the HMAC tag and store it in the object */ + memcpy(o->tag.tag, gcry_md_read(f->hmac, 0), TAG_LENGTH); + f->hmac_running = false; + + return 0; +} + +int journal_file_hmac_start(JournalFile *f) { + uint8_t key[256 / 8]; /* Let's pass 256 bit from FSPRG to HMAC */ + gcry_error_t err; + + assert(f); + + if (!JOURNAL_HEADER_SEALED(f->header)) + return 0; + + if (f->hmac_running) + return 0; + + /* Prepare HMAC for next cycle */ + gcry_md_reset(f->hmac); + FSPRG_GetKey(f->fsprg_state, key, sizeof(key), 0); + err = gcry_md_setkey(f->hmac, key, sizeof(key)); + if (gcry_err_code(err) != GPG_ERR_NO_ERROR) + return log_debug_errno(SYNTHETIC_ERRNO(EIO), + "gcry_md_setkey() failed with error code: %s", + gcry_strerror(err)); + + f->hmac_running = true; + + return 0; +} + +static int journal_file_get_epoch(JournalFile *f, uint64_t realtime, uint64_t *epoch) { + uint64_t t; + + assert(f); + assert(epoch); + assert(JOURNAL_HEADER_SEALED(f->header)); + + if (f->fss_start_usec == 0 || + f->fss_interval_usec == 0) + return -EOPNOTSUPP; + + if (realtime < f->fss_start_usec) + return -ESTALE; + + t = realtime - f->fss_start_usec; + t = t / f->fss_interval_usec; + + *epoch = t; + return 0; +} + +static int journal_file_fsprg_need_evolve(JournalFile *f, uint64_t realtime) { + uint64_t goal, epoch; + int r; + assert(f); + + if (!JOURNAL_HEADER_SEALED(f->header)) + return 0; + + r = journal_file_get_epoch(f, realtime, &goal); + if (r < 0) + return r; + + epoch = FSPRG_GetEpoch(f->fsprg_state); + if (epoch > goal) + return -ESTALE; + + return epoch != goal; +} + +int journal_file_fsprg_evolve(JournalFile *f, uint64_t realtime) { + uint64_t goal, epoch; + int r; + + assert(f); + + if (!JOURNAL_HEADER_SEALED(f->header)) + return 0; + + r = journal_file_get_epoch(f, realtime, &goal); + if (r < 0) + return r; + + epoch = FSPRG_GetEpoch(f->fsprg_state); + if (epoch < goal) + log_debug("Evolving FSPRG key from epoch %"PRIu64" to %"PRIu64".", epoch, goal); + + for (;;) { + if (epoch > goal) + return -ESTALE; + if (epoch == goal) + return 0; + + FSPRG_Evolve(f->fsprg_state); + epoch = FSPRG_GetEpoch(f->fsprg_state); + } +} + +int journal_file_fsprg_seek(JournalFile *f, uint64_t goal) { + void *msk; + uint64_t epoch; + + assert(f); + + if (!JOURNAL_HEADER_SEALED(f->header)) + return 0; + + assert(f->fsprg_seed); + + if (f->fsprg_state) { + /* Cheaper... */ + + epoch = FSPRG_GetEpoch(f->fsprg_state); + if (goal == epoch) + return 0; + + if (goal == epoch+1) { + FSPRG_Evolve(f->fsprg_state); + return 0; + } + } else { + f->fsprg_state_size = FSPRG_stateinbytes(FSPRG_RECOMMENDED_SECPAR); + f->fsprg_state = malloc(f->fsprg_state_size); + if (!f->fsprg_state) + return -ENOMEM; + } + + log_debug("Seeking FSPRG key to %"PRIu64".", goal); + + msk = alloca_safe(FSPRG_mskinbytes(FSPRG_RECOMMENDED_SECPAR)); + FSPRG_GenMK(msk, NULL, f->fsprg_seed, f->fsprg_seed_size, FSPRG_RECOMMENDED_SECPAR); + FSPRG_Seek(f->fsprg_state, goal, msk, f->fsprg_seed, f->fsprg_seed_size); + return 0; +} + +int journal_file_maybe_append_tag(JournalFile *f, uint64_t realtime) { + int r; + + assert(f); + + if (!JOURNAL_HEADER_SEALED(f->header)) + return 0; + + if (realtime <= 0) + realtime = now(CLOCK_REALTIME); + + r = journal_file_fsprg_need_evolve(f, realtime); + if (r <= 0) + return 0; + + r = journal_file_append_tag(f); + if (r < 0) + return r; + + r = journal_file_fsprg_evolve(f, realtime); + if (r < 0) + return r; + + return 0; +} + +int journal_file_hmac_put_object(JournalFile *f, ObjectType type, Object *o, uint64_t p) { + int r; + + assert(f); + + if (!JOURNAL_HEADER_SEALED(f->header)) + return 0; + + r = journal_file_hmac_start(f); + if (r < 0) + return r; + + if (!o) { + r = journal_file_move_to_object(f, type, p, &o); + if (r < 0) + return r; + } else { + if (type > OBJECT_UNUSED && o->object.type != type) + return -EBADMSG; + } + + gcry_md_write(f->hmac, o, offsetof(ObjectHeader, payload)); + + switch (o->object.type) { + + case OBJECT_DATA: + /* All but hash and payload are mutable */ + gcry_md_write(f->hmac, &o->data.hash, sizeof(o->data.hash)); + gcry_md_write(f->hmac, journal_file_data_payload_field(f, o), le64toh(o->object.size) - journal_file_data_payload_offset(f)); + break; + + case OBJECT_FIELD: + /* Same here */ + gcry_md_write(f->hmac, &o->field.hash, sizeof(o->field.hash)); + gcry_md_write(f->hmac, o->field.payload, le64toh(o->object.size) - offsetof(Object, field.payload)); + break; + + case OBJECT_ENTRY: + /* All */ + gcry_md_write(f->hmac, &o->entry.seqnum, le64toh(o->object.size) - offsetof(Object, entry.seqnum)); + break; + + case OBJECT_FIELD_HASH_TABLE: + case OBJECT_DATA_HASH_TABLE: + case OBJECT_ENTRY_ARRAY: + /* Nothing: everything is mutable */ + break; + + case OBJECT_TAG: + /* All but the tag itself */ + gcry_md_write(f->hmac, &o->tag.seqnum, sizeof(o->tag.seqnum)); + gcry_md_write(f->hmac, &o->tag.epoch, sizeof(o->tag.epoch)); + break; + default: + return -EINVAL; + } + + return 0; +} + +int journal_file_hmac_put_header(JournalFile *f) { + int r; + + assert(f); + + if (!JOURNAL_HEADER_SEALED(f->header)) + return 0; + + r = journal_file_hmac_start(f); + if (r < 0) + return r; + + /* All but state+reserved, boot_id, arena_size, + * tail_object_offset, n_objects, n_entries, + * tail_entry_seqnum, head_entry_seqnum, entry_array_offset, + * head_entry_realtime, tail_entry_realtime, + * tail_entry_monotonic, n_data, n_fields, n_tags, + * n_entry_arrays. */ + + gcry_md_write(f->hmac, f->header->signature, offsetof(Header, state) - offsetof(Header, signature)); + gcry_md_write(f->hmac, &f->header->file_id, offsetof(Header, boot_id) - offsetof(Header, file_id)); + gcry_md_write(f->hmac, &f->header->seqnum_id, offsetof(Header, arena_size) - offsetof(Header, seqnum_id)); + gcry_md_write(f->hmac, &f->header->data_hash_table_offset, offsetof(Header, tail_object_offset) - offsetof(Header, data_hash_table_offset)); + + return 0; +} + +int journal_file_fss_load(JournalFile *f) { + int r, fd = -1; + char *p = NULL; + struct stat st; + FSSHeader *m = NULL; + sd_id128_t machine; + + assert(f); + + /* This function is used to determine whether sealing should be enabled in the journal header so we + * can't check the header to check if sealing is enabled here. */ + + r = sd_id128_get_machine(&machine); + if (r < 0) + return r; + + if (asprintf(&p, "/var/log/journal/" SD_ID128_FORMAT_STR "/fss", + SD_ID128_FORMAT_VAL(machine)) < 0) + return -ENOMEM; + + fd = open(p, O_RDWR|O_CLOEXEC|O_NOCTTY, 0600); + if (fd < 0) { + if (errno != ENOENT) + log_error_errno(errno, "Failed to open %s: %m", p); + + r = -errno; + goto finish; + } + + if (fstat(fd, &st) < 0) { + r = -errno; + goto finish; + } + + if (st.st_size < (off_t) sizeof(FSSHeader)) { + r = -ENODATA; + goto finish; + } + + m = mmap(NULL, PAGE_ALIGN(sizeof(FSSHeader)), PROT_READ, MAP_SHARED, fd, 0); + if (m == MAP_FAILED) { + m = NULL; + r = -errno; + goto finish; + } + + if (memcmp(m->signature, FSS_HEADER_SIGNATURE, 8) != 0) { + r = -EBADMSG; + goto finish; + } + + if (m->incompatible_flags != 0) { + r = -EPROTONOSUPPORT; + goto finish; + } + + if (le64toh(m->header_size) < sizeof(FSSHeader)) { + r = -EBADMSG; + goto finish; + } + + if (le64toh(m->fsprg_state_size) != FSPRG_stateinbytes(le16toh(m->fsprg_secpar))) { + r = -EBADMSG; + goto finish; + } + + f->fss_file_size = le64toh(m->header_size) + le64toh(m->fsprg_state_size); + if ((uint64_t) st.st_size < f->fss_file_size) { + r = -ENODATA; + goto finish; + } + + if (!sd_id128_equal(machine, m->machine_id)) { + r = -EHOSTDOWN; + goto finish; + } + + if (le64toh(m->start_usec) <= 0 || + le64toh(m->interval_usec) <= 0) { + r = -EBADMSG; + goto finish; + } + + f->fss_file = mmap(NULL, PAGE_ALIGN(f->fss_file_size), PROT_READ|PROT_WRITE, MAP_SHARED, fd, 0); + if (f->fss_file == MAP_FAILED) { + f->fss_file = NULL; + r = -errno; + goto finish; + } + + f->fss_start_usec = le64toh(f->fss_file->start_usec); + f->fss_interval_usec = le64toh(f->fss_file->interval_usec); + + f->fsprg_state = (uint8_t*) f->fss_file + le64toh(f->fss_file->header_size); + f->fsprg_state_size = le64toh(f->fss_file->fsprg_state_size); + + r = 0; + +finish: + if (m) + munmap(m, PAGE_ALIGN(sizeof(FSSHeader))); + + safe_close(fd); + free(p); + + return r; +} + +int journal_file_hmac_setup(JournalFile *f) { + gcry_error_t e; + + if (!JOURNAL_HEADER_SEALED(f->header)) + return 0; + + initialize_libgcrypt(true); + + e = gcry_md_open(&f->hmac, GCRY_MD_SHA256, GCRY_MD_FLAG_HMAC); + if (e != 0) + return -EOPNOTSUPP; + + return 0; +} + +int journal_file_append_first_tag(JournalFile *f) { + int r; + uint64_t p; + + if (!JOURNAL_HEADER_SEALED(f->header)) + return 0; + + log_debug("Calculating first tag..."); + + r = journal_file_hmac_put_header(f); + if (r < 0) + return r; + + p = le64toh(f->header->field_hash_table_offset); + if (p < offsetof(Object, hash_table.items)) + return -EINVAL; + p -= offsetof(Object, hash_table.items); + + r = journal_file_hmac_put_object(f, OBJECT_FIELD_HASH_TABLE, NULL, p); + if (r < 0) + return r; + + p = le64toh(f->header->data_hash_table_offset); + if (p < offsetof(Object, hash_table.items)) + return -EINVAL; + p -= offsetof(Object, hash_table.items); + + r = journal_file_hmac_put_object(f, OBJECT_DATA_HASH_TABLE, NULL, p); + if (r < 0) + return r; + + r = journal_file_append_tag(f); + if (r < 0) + return r; + + return 0; +} + +int journal_file_parse_verification_key(JournalFile *f, const char *key) { + uint8_t *seed; + size_t seed_size, c; + const char *k; + int r; + unsigned long long start, interval; + + seed_size = FSPRG_RECOMMENDED_SEEDLEN; + seed = malloc(seed_size); + if (!seed) + return -ENOMEM; + + k = key; + for (c = 0; c < seed_size; c++) { + int x, y; + + while (*k == '-') + k++; + + x = unhexchar(*k); + if (x < 0) { + free(seed); + return -EINVAL; + } + k++; + y = unhexchar(*k); + if (y < 0) { + free(seed); + return -EINVAL; + } + k++; + + seed[c] = (uint8_t) (x * 16 + y); + } + + if (*k != '/') { + free(seed); + return -EINVAL; + } + k++; + + r = sscanf(k, "%llx-%llx", &start, &interval); + if (r != 2) { + free(seed); + return -EINVAL; + } + + f->fsprg_seed = seed; + f->fsprg_seed_size = seed_size; + + f->fss_start_usec = start * interval; + f->fss_interval_usec = interval; + + return 0; +} + +bool journal_file_next_evolve_usec(JournalFile *f, usec_t *u) { + uint64_t epoch; + + assert(f); + assert(u); + + if (!JOURNAL_HEADER_SEALED(f->header)) + return false; + + epoch = FSPRG_GetEpoch(f->fsprg_state); + + *u = (usec_t) (f->fss_start_usec + f->fss_interval_usec * epoch + f->fss_interval_usec); + + return true; +} diff --git a/src/libsystemd/sd-journal/journal-authenticate.h b/src/libsystemd/sd-journal/journal-authenticate.h new file mode 100644 index 0000000..e895722 --- /dev/null +++ b/src/libsystemd/sd-journal/journal-authenticate.h @@ -0,0 +1,23 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ +#pragma once + +#include <stdbool.h> + +#include "journal-file.h" + +int journal_file_append_tag(JournalFile *f); +int journal_file_maybe_append_tag(JournalFile *f, uint64_t realtime); +int journal_file_append_first_tag(JournalFile *f); + +int journal_file_hmac_setup(JournalFile *f); +int journal_file_hmac_start(JournalFile *f); +int journal_file_hmac_put_header(JournalFile *f); +int journal_file_hmac_put_object(JournalFile *f, ObjectType type, Object *o, uint64_t p); + +int journal_file_fss_load(JournalFile *f); +int journal_file_parse_verification_key(JournalFile *f, const char *key); + +int journal_file_fsprg_evolve(JournalFile *f, uint64_t realtime); +int journal_file_fsprg_seek(JournalFile *f, uint64_t epoch); + +bool journal_file_next_evolve_usec(JournalFile *f, usec_t *u); diff --git a/src/libsystemd/sd-journal/journal-def.h b/src/libsystemd/sd-journal/journal-def.h new file mode 100644 index 0000000..8f994b0 --- /dev/null +++ b/src/libsystemd/sd-journal/journal-def.h @@ -0,0 +1,259 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ +#pragma once + +#include "sd-id128.h" + +#include "macro.h" +#include "sparse-endian.h" + +/* + * If you change this file you probably should also change its documentation: + * + * https://systemd.io/JOURNAL_FILE_FORMAT + */ + +typedef struct Header Header; + +typedef struct ObjectHeader ObjectHeader; +typedef union Object Object; + +typedef struct DataObject DataObject; +typedef struct FieldObject FieldObject; +typedef struct EntryObject EntryObject; +typedef struct HashTableObject HashTableObject; +typedef struct EntryArrayObject EntryArrayObject; +typedef struct TagObject TagObject; + +typedef struct HashItem HashItem; + +typedef struct FSSHeader FSSHeader; + +/* Object types */ +typedef enum ObjectType { + OBJECT_UNUSED, /* also serves as "any type" or "additional context" */ + OBJECT_DATA, + OBJECT_FIELD, + OBJECT_ENTRY, + OBJECT_DATA_HASH_TABLE, + OBJECT_FIELD_HASH_TABLE, + OBJECT_ENTRY_ARRAY, + OBJECT_TAG, + _OBJECT_TYPE_MAX +} ObjectType; + +/* Object flags (note that src/basic/compress.h uses the same values for the compression types) */ +enum { + OBJECT_COMPRESSED_XZ = 1 << 0, + OBJECT_COMPRESSED_LZ4 = 1 << 1, + OBJECT_COMPRESSED_ZSTD = 1 << 2, + _OBJECT_COMPRESSED_MASK = OBJECT_COMPRESSED_XZ | OBJECT_COMPRESSED_LZ4 | OBJECT_COMPRESSED_ZSTD, +}; + +struct ObjectHeader { + uint8_t type; + uint8_t flags; + uint8_t reserved[6]; + le64_t size; + uint8_t payload[]; +} _packed_; + +#define DataObject__contents { \ + ObjectHeader object; \ + le64_t hash; \ + le64_t next_hash_offset; \ + le64_t next_field_offset; \ + le64_t entry_offset; /* the first array entry we store inline */ \ + le64_t entry_array_offset; \ + le64_t n_entries; \ + union { \ + struct { \ + uint8_t payload[0]; \ + } regular; \ + struct { \ + le32_t tail_entry_array_offset; \ + le32_t tail_entry_array_n_entries; \ + uint8_t payload[0]; \ + } compact; \ + }; \ +} + +struct DataObject DataObject__contents; +struct DataObject__packed DataObject__contents _packed_; +assert_cc(sizeof(struct DataObject) == sizeof(struct DataObject__packed)); + +#define FieldObject__contents { \ + ObjectHeader object; \ + le64_t hash; \ + le64_t next_hash_offset; \ + le64_t head_data_offset; \ + uint8_t payload[]; \ +} + +struct FieldObject FieldObject__contents; +struct FieldObject__packed FieldObject__contents _packed_; +assert_cc(sizeof(struct FieldObject) == sizeof(struct FieldObject__packed)); + +#define EntryObject__contents { \ + ObjectHeader object; \ + le64_t seqnum; \ + le64_t realtime; \ + le64_t monotonic; \ + sd_id128_t boot_id; \ + le64_t xor_hash; \ + union { \ + struct { \ + le64_t object_offset; \ + le64_t hash; \ + } regular[0]; \ + struct { \ + le32_t object_offset; \ + } compact[0]; \ + } items; \ +} + +struct EntryObject EntryObject__contents; +struct EntryObject__packed EntryObject__contents _packed_; +assert_cc(sizeof(struct EntryObject) == sizeof(struct EntryObject__packed)); + +struct HashItem { + le64_t head_hash_offset; + le64_t tail_hash_offset; +} _packed_; + +struct HashTableObject { + ObjectHeader object; + HashItem items[]; +} _packed_; + +struct EntryArrayObject { + ObjectHeader object; + le64_t next_entry_array_offset; + union { + le64_t regular[0]; + le32_t compact[0]; + } items; +} _packed_; + +#define TAG_LENGTH (256/8) + +struct TagObject { + ObjectHeader object; + le64_t seqnum; + le64_t epoch; + uint8_t tag[TAG_LENGTH]; /* SHA-256 HMAC */ +} _packed_; + +union Object { + ObjectHeader object; + DataObject data; + FieldObject field; + EntryObject entry; + HashTableObject hash_table; + EntryArrayObject entry_array; + TagObject tag; +}; + +enum { + STATE_OFFLINE = 0, + STATE_ONLINE = 1, + STATE_ARCHIVED = 2, + _STATE_MAX +}; + +/* Header flags */ +enum { + HEADER_INCOMPATIBLE_COMPRESSED_XZ = 1 << 0, + HEADER_INCOMPATIBLE_COMPRESSED_LZ4 = 1 << 1, + HEADER_INCOMPATIBLE_KEYED_HASH = 1 << 2, + HEADER_INCOMPATIBLE_COMPRESSED_ZSTD = 1 << 3, + HEADER_INCOMPATIBLE_COMPACT = 1 << 4, +}; + +#define HEADER_INCOMPATIBLE_ANY \ + (HEADER_INCOMPATIBLE_COMPRESSED_XZ | \ + HEADER_INCOMPATIBLE_COMPRESSED_LZ4 | \ + HEADER_INCOMPATIBLE_KEYED_HASH | \ + HEADER_INCOMPATIBLE_COMPRESSED_ZSTD | \ + HEADER_INCOMPATIBLE_COMPACT) + +#define HEADER_INCOMPATIBLE_SUPPORTED \ + ((HAVE_XZ ? HEADER_INCOMPATIBLE_COMPRESSED_XZ : 0) | \ + (HAVE_LZ4 ? HEADER_INCOMPATIBLE_COMPRESSED_LZ4 : 0) | \ + (HAVE_ZSTD ? HEADER_INCOMPATIBLE_COMPRESSED_ZSTD : 0) | \ + HEADER_INCOMPATIBLE_KEYED_HASH | \ + HEADER_INCOMPATIBLE_COMPACT) + +enum { + HEADER_COMPATIBLE_SEALED = 1 << 0, +}; + +#define HEADER_COMPATIBLE_ANY HEADER_COMPATIBLE_SEALED +#if HAVE_GCRYPT +# define HEADER_COMPATIBLE_SUPPORTED HEADER_COMPATIBLE_SEALED +#else +# define HEADER_COMPATIBLE_SUPPORTED 0 +#endif + +#define HEADER_SIGNATURE \ + ((const char[]) { 'L', 'P', 'K', 'S', 'H', 'H', 'R', 'H' }) + +#define struct_Header__contents { \ + uint8_t signature[8]; /* "LPKSHHRH" */ \ + le32_t compatible_flags; \ + le32_t incompatible_flags; \ + uint8_t state; \ + uint8_t reserved[7]; \ + sd_id128_t file_id; \ + sd_id128_t machine_id; \ + sd_id128_t boot_id; /* last writer */ \ + sd_id128_t seqnum_id; \ + le64_t header_size; \ + le64_t arena_size; \ + le64_t data_hash_table_offset; \ + le64_t data_hash_table_size; \ + le64_t field_hash_table_offset; \ + le64_t field_hash_table_size; \ + le64_t tail_object_offset; \ + le64_t n_objects; \ + le64_t n_entries; \ + le64_t tail_entry_seqnum; \ + le64_t head_entry_seqnum; \ + le64_t entry_array_offset; \ + le64_t head_entry_realtime; \ + le64_t tail_entry_realtime; \ + le64_t tail_entry_monotonic; \ + /* Added in 187 */ \ + le64_t n_data; \ + le64_t n_fields; \ + /* Added in 189 */ \ + le64_t n_tags; \ + le64_t n_entry_arrays; \ + /* Added in 246 */ \ + le64_t data_hash_chain_depth; \ + le64_t field_hash_chain_depth; \ + /* Added in 252 */ \ + le32_t tail_entry_array_offset; \ + le32_t tail_entry_array_n_entries; \ + } + +struct Header struct_Header__contents; +struct Header__packed struct_Header__contents _packed_; +assert_cc(sizeof(struct Header) == sizeof(struct Header__packed)); +assert_cc(sizeof(struct Header) == 264); + +#define FSS_HEADER_SIGNATURE \ + ((const char[]) { 'K', 'S', 'H', 'H', 'R', 'H', 'L', 'P' }) + +struct FSSHeader { + uint8_t signature[8]; /* "KSHHRHLP" */ + le32_t compatible_flags; + le32_t incompatible_flags; + sd_id128_t machine_id; + sd_id128_t boot_id; /* last writer */ + le64_t header_size; + le64_t start_usec; + le64_t interval_usec; + le16_t fsprg_secpar; + le16_t reserved[3]; + le64_t fsprg_state_size; +} _packed_; diff --git a/src/libsystemd/sd-journal/journal-file.c b/src/libsystemd/sd-journal/journal-file.c new file mode 100644 index 0000000..d3a9505 --- /dev/null +++ b/src/libsystemd/sd-journal/journal-file.c @@ -0,0 +1,4255 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ + +#include <errno.h> +#include <fcntl.h> +#include <linux/fs.h> +#include <linux/magic.h> +#include <pthread.h> +#include <stddef.h> +#include <sys/mman.h> +#include <sys/statvfs.h> +#include <sys/uio.h> +#include <unistd.h> + +#include "sd-event.h" + +#include "alloc-util.h" +#include "chattr-util.h" +#include "compress.h" +#include "env-util.h" +#include "fd-util.h" +#include "format-util.h" +#include "fs-util.h" +#include "journal-authenticate.h" +#include "journal-def.h" +#include "journal-file.h" +#include "lookup3.h" +#include "memory-util.h" +#include "path-util.h" +#include "random-util.h" +#include "set.h" +#include "sort-util.h" +#include "stat-util.h" +#include "string-table.h" +#include "string-util.h" +#include "strv.h" +#include "sync-util.h" +#include "xattr-util.h" + +#define DEFAULT_DATA_HASH_TABLE_SIZE (2047ULL*sizeof(HashItem)) +#define DEFAULT_FIELD_HASH_TABLE_SIZE (333ULL*sizeof(HashItem)) + +#define DEFAULT_COMPRESS_THRESHOLD (512ULL) +#define MIN_COMPRESS_THRESHOLD (8ULL) + +/* This is the minimum journal file size */ +#define JOURNAL_FILE_SIZE_MIN (512 * 1024ULL) /* 512 KiB */ +#define JOURNAL_COMPACT_SIZE_MAX UINT32_MAX /* 4 GiB */ + +/* These are the lower and upper bounds if we deduce the max_use value + * from the file system size */ +#define MAX_USE_LOWER (1 * 1024 * 1024ULL) /* 1 MiB */ +#define MAX_USE_UPPER (4 * 1024 * 1024 * 1024ULL) /* 4 GiB */ + +/* Those are the lower and upper bounds for the minimal use limit, + * i.e. how much we'll use even if keep_free suggests otherwise. */ +#define MIN_USE_LOW (1 * 1024 * 1024ULL) /* 1 MiB */ +#define MIN_USE_HIGH (16 * 1024 * 1024ULL) /* 16 MiB */ + +/* This is the upper bound if we deduce max_size from max_use */ +#define MAX_SIZE_UPPER (128 * 1024 * 1024ULL) /* 128 MiB */ + +/* This is the upper bound if we deduce the keep_free value from the + * file system size */ +#define KEEP_FREE_UPPER (4 * 1024 * 1024 * 1024ULL) /* 4 GiB */ + +/* This is the keep_free value when we can't determine the system + * size */ +#define DEFAULT_KEEP_FREE (1024 * 1024ULL) /* 1 MB */ + +/* This is the default maximum number of journal files to keep around. */ +#define DEFAULT_N_MAX_FILES 100 + +/* n_data was the first entry we added after the initial file format design */ +#define HEADER_SIZE_MIN ALIGN64(offsetof(Header, n_data)) + +/* How many entries to keep in the entry array chain cache at max */ +#define CHAIN_CACHE_MAX 20 + +/* How much to increase the journal file size at once each time we allocate something new. */ +#define FILE_SIZE_INCREASE (8 * 1024 * 1024ULL) /* 8MB */ + +/* Reread fstat() of the file for detecting deletions at least this often */ +#define LAST_STAT_REFRESH_USEC (5*USEC_PER_SEC) + +/* The mmap context to use for the header we pick as one above the last defined typed */ +#define CONTEXT_HEADER _OBJECT_TYPE_MAX + +/* Longest hash chain to rotate after */ +#define HASH_CHAIN_DEPTH_MAX 100 + +#ifdef __clang__ +# pragma GCC diagnostic ignored "-Waddress-of-packed-member" +#endif + +int journal_file_tail_end_by_pread(JournalFile *f, uint64_t *ret_offset) { + uint64_t p; + int r; + + assert(f); + assert(f->header); + assert(ret_offset); + + /* Same as journal_file_tail_end_by_mmap() below, but operates with pread() to avoid the mmap cache + * (and thus is thread safe) */ + + p = le64toh(f->header->tail_object_offset); + if (p == 0) + p = le64toh(f->header->header_size); + else { + Object tail; + uint64_t sz; + + r = journal_file_read_object_header(f, OBJECT_UNUSED, p, &tail); + if (r < 0) + return r; + + sz = le64toh(tail.object.size); + if (sz > UINT64_MAX - sizeof(uint64_t) + 1) + return -EBADMSG; + + sz = ALIGN64(sz); + if (p > UINT64_MAX - sz) + return -EBADMSG; + + p += sz; + } + + *ret_offset = p; + + return 0; +} + +int journal_file_tail_end_by_mmap(JournalFile *f, uint64_t *ret_offset) { + uint64_t p; + int r; + + assert(f); + assert(f->header); + assert(ret_offset); + + /* Same as journal_file_tail_end_by_pread() above, but operates with the usual mmap logic */ + + p = le64toh(f->header->tail_object_offset); + if (p == 0) + p = le64toh(f->header->header_size); + else { + Object *tail; + uint64_t sz; + + r = journal_file_move_to_object(f, OBJECT_UNUSED, p, &tail); + if (r < 0) + return r; + + sz = le64toh(READ_NOW(tail->object.size)); + if (sz > UINT64_MAX - sizeof(uint64_t) + 1) + return -EBADMSG; + + sz = ALIGN64(sz); + if (p > UINT64_MAX - sz) + return -EBADMSG; + + p += sz; + } + + *ret_offset = p; + + return 0; +} + +int journal_file_set_offline_thread_join(JournalFile *f) { + int r; + + assert(f); + + if (f->offline_state == OFFLINE_JOINED) + return 0; + + r = pthread_join(f->offline_thread, NULL); + if (r) + return -r; + + f->offline_state = OFFLINE_JOINED; + + if (mmap_cache_fd_got_sigbus(f->cache_fd)) + return -EIO; + + return 0; +} + +static int journal_file_set_online(JournalFile *f) { + bool wait = true; + + assert(f); + + if (!journal_file_writable(f)) + return -EPERM; + + if (f->fd < 0 || !f->header) + return -EINVAL; + + while (wait) { + switch (f->offline_state) { + case OFFLINE_JOINED: + /* No offline thread, no need to wait. */ + wait = false; + break; + + case OFFLINE_SYNCING: { + OfflineState tmp_state = OFFLINE_SYNCING; + if (!__atomic_compare_exchange_n(&f->offline_state, &tmp_state, OFFLINE_CANCEL, + false, __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST)) + continue; + } + /* Canceled syncing prior to offlining, no need to wait. */ + wait = false; + break; + + case OFFLINE_AGAIN_FROM_SYNCING: { + OfflineState tmp_state = OFFLINE_AGAIN_FROM_SYNCING; + if (!__atomic_compare_exchange_n(&f->offline_state, &tmp_state, OFFLINE_CANCEL, + false, __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST)) + continue; + } + /* Canceled restart from syncing, no need to wait. */ + wait = false; + break; + + case OFFLINE_AGAIN_FROM_OFFLINING: { + OfflineState tmp_state = OFFLINE_AGAIN_FROM_OFFLINING; + if (!__atomic_compare_exchange_n(&f->offline_state, &tmp_state, OFFLINE_CANCEL, + false, __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST)) + continue; + } + /* Canceled restart from offlining, must wait for offlining to complete however. */ + _fallthrough_; + default: { + int r; + + r = journal_file_set_offline_thread_join(f); + if (r < 0) + return r; + + wait = false; + break; + } + } + } + + if (mmap_cache_fd_got_sigbus(f->cache_fd)) + return -EIO; + + switch (f->header->state) { + case STATE_ONLINE: + return 0; + + case STATE_OFFLINE: + f->header->state = STATE_ONLINE; + (void) fsync(f->fd); + return 0; + + default: + return -EINVAL; + } +} + +JournalFile* journal_file_close(JournalFile *f) { + if (!f) + return NULL; + + if (f->cache_fd) + mmap_cache_fd_free(f->cache_fd); + + if (f->close_fd) + safe_close(f->fd); + free(f->path); + + ordered_hashmap_free_free(f->chain_cache); + +#if HAVE_COMPRESSION + free(f->compress_buffer); +#endif + +#if HAVE_GCRYPT + if (f->fss_file) + munmap(f->fss_file, PAGE_ALIGN(f->fss_file_size)); + else + free(f->fsprg_state); + + free(f->fsprg_seed); + + if (f->hmac) + gcry_md_close(f->hmac); +#endif + + return mfree(f); +} + +static bool keyed_hash_requested(void) { + int r; + + r = getenv_bool("SYSTEMD_JOURNAL_KEYED_HASH"); + if (r >= 0) + return r; + if (r != -ENXIO) + log_debug_errno(r, "Failed to parse $SYSTEMD_JOURNAL_KEYED_HASH environment variable, ignoring: %m"); + + return true; +} + +static bool compact_mode_requested(void) { + int r; + + r = getenv_bool("SYSTEMD_JOURNAL_COMPACT"); + if (r >= 0) + return r; + if (r != -ENXIO) + log_debug_errno(r, "Failed to parse $SYSTEMD_JOURNAL_COMPACT environment variable, ignoring: %m"); + + return true; +} + +static int journal_file_init_header(JournalFile *f, JournalFileFlags file_flags, JournalFile *template) { + Header h = {}; + ssize_t k; + bool seal = false; + int r; + + assert(f); + +#if HAVE_GCRYPT + /* Try to load the FSPRG state, and if we can't, then just don't do sealing */ + seal = FLAGS_SET(file_flags, JOURNAL_SEAL) && journal_file_fss_load(f) >= 0; +#endif + + memcpy(h.signature, HEADER_SIGNATURE, 8); + h.header_size = htole64(ALIGN64(sizeof(h))); + + h.incompatible_flags |= htole32( + FLAGS_SET(file_flags, JOURNAL_COMPRESS) * + COMPRESSION_TO_HEADER_INCOMPATIBLE_FLAG(DEFAULT_COMPRESSION) | + keyed_hash_requested() * HEADER_INCOMPATIBLE_KEYED_HASH | + compact_mode_requested() * HEADER_INCOMPATIBLE_COMPACT); + + h.compatible_flags = htole32(seal * HEADER_COMPATIBLE_SEALED); + + r = sd_id128_randomize(&h.file_id); + if (r < 0) + return r; + + if (template) { + h.seqnum_id = template->header->seqnum_id; + h.tail_entry_seqnum = template->header->tail_entry_seqnum; + } else + h.seqnum_id = h.file_id; + + k = pwrite(f->fd, &h, sizeof(h), 0); + if (k < 0) + return -errno; + + if (k != sizeof(h)) + return -EIO; + + return 0; +} + +static int journal_file_refresh_header(JournalFile *f) { + int r; + + assert(f); + assert(f->header); + + r = sd_id128_get_machine(&f->header->machine_id); + if (IN_SET(r, -ENOENT, -ENOMEDIUM, -ENOPKG)) + /* We don't have a machine-id, let's continue without */ + zero(f->header->machine_id); + else if (r < 0) + return r; + + r = sd_id128_get_boot(&f->header->boot_id); + if (r < 0) + return r; + + r = journal_file_set_online(f); + + /* Sync the online state to disk; likely just created a new file, also sync the directory this file + * is located in. */ + (void) fsync_full(f->fd); + + return r; +} + +static bool warn_wrong_flags(const JournalFile *f, bool compatible) { + const uint32_t any = compatible ? HEADER_COMPATIBLE_ANY : HEADER_INCOMPATIBLE_ANY, + supported = compatible ? HEADER_COMPATIBLE_SUPPORTED : HEADER_INCOMPATIBLE_SUPPORTED; + const char *type = compatible ? "compatible" : "incompatible"; + uint32_t flags; + + assert(f); + assert(f->header); + + flags = le32toh(compatible ? f->header->compatible_flags : f->header->incompatible_flags); + + if (flags & ~supported) { + if (flags & ~any) + log_debug("Journal file %s has unknown %s flags 0x%"PRIx32, + f->path, type, flags & ~any); + flags = (flags & any) & ~supported; + if (flags) { + const char* strv[6]; + size_t n = 0; + _cleanup_free_ char *t = NULL; + + if (compatible) { + if (flags & HEADER_COMPATIBLE_SEALED) + strv[n++] = "sealed"; + } else { + if (flags & HEADER_INCOMPATIBLE_COMPRESSED_XZ) + strv[n++] = "xz-compressed"; + if (flags & HEADER_INCOMPATIBLE_COMPRESSED_LZ4) + strv[n++] = "lz4-compressed"; + if (flags & HEADER_INCOMPATIBLE_COMPRESSED_ZSTD) + strv[n++] = "zstd-compressed"; + if (flags & HEADER_INCOMPATIBLE_KEYED_HASH) + strv[n++] = "keyed-hash"; + if (flags & HEADER_INCOMPATIBLE_COMPACT) + strv[n++] = "compact"; + } + strv[n] = NULL; + assert(n < ELEMENTSOF(strv)); + + t = strv_join((char**) strv, ", "); + log_debug("Journal file %s uses %s %s %s disabled at compilation time.", + f->path, type, n > 1 ? "flags" : "flag", strnull(t)); + } + return true; + } + + return false; +} + +static int journal_file_verify_header(JournalFile *f) { + uint64_t arena_size, header_size; + + assert(f); + assert(f->header); + + if (memcmp(f->header->signature, HEADER_SIGNATURE, 8)) + return -EBADMSG; + + /* In both read and write mode we refuse to open files with incompatible + * flags we don't know. */ + if (warn_wrong_flags(f, false)) + return -EPROTONOSUPPORT; + + /* When open for writing we refuse to open files with compatible flags, too. */ + if (journal_file_writable(f) && warn_wrong_flags(f, true)) + return -EPROTONOSUPPORT; + + if (f->header->state >= _STATE_MAX) + return -EBADMSG; + + header_size = le64toh(READ_NOW(f->header->header_size)); + + /* The first addition was n_data, so check that we are at least this large */ + if (header_size < HEADER_SIZE_MIN) + return -EBADMSG; + + if (JOURNAL_HEADER_SEALED(f->header) && !JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays)) + return -EBADMSG; + + arena_size = le64toh(READ_NOW(f->header->arena_size)); + + if (UINT64_MAX - header_size < arena_size || header_size + arena_size > (uint64_t) f->last_stat.st_size) + return -ENODATA; + + if (le64toh(f->header->tail_object_offset) > header_size + arena_size) + return -ENODATA; + + if (!VALID64(le64toh(f->header->data_hash_table_offset)) || + !VALID64(le64toh(f->header->field_hash_table_offset)) || + !VALID64(le64toh(f->header->tail_object_offset)) || + !VALID64(le64toh(f->header->entry_array_offset))) + return -ENODATA; + + if (journal_file_writable(f)) { + sd_id128_t machine_id; + uint8_t state; + int r; + + r = sd_id128_get_machine(&machine_id); + if (r < 0) + return r; + + if (!sd_id128_equal(machine_id, f->header->machine_id)) + return -EHOSTDOWN; + + state = f->header->state; + + if (state == STATE_ARCHIVED) + return -ESHUTDOWN; /* Already archived */ + else if (state == STATE_ONLINE) + return log_debug_errno(SYNTHETIC_ERRNO(EBUSY), + "Journal file %s is already online. Assuming unclean closing.", + f->path); + else if (state != STATE_OFFLINE) + return log_debug_errno(SYNTHETIC_ERRNO(EBUSY), + "Journal file %s has unknown state %i.", + f->path, state); + + if (f->header->field_hash_table_size == 0 || f->header->data_hash_table_size == 0) + return -EBADMSG; + + /* Don't permit appending to files from the future. Because otherwise the realtime timestamps wouldn't + * be strictly ordered in the entries in the file anymore, and we can't have that since it breaks + * bisection. */ + if (le64toh(f->header->tail_entry_realtime) > now(CLOCK_REALTIME)) + return log_debug_errno(SYNTHETIC_ERRNO(ETXTBSY), + "Journal file %s is from the future, refusing to append new data to it that'd be older.", + f->path); + } + + return 0; +} + +int journal_file_fstat(JournalFile *f) { + int r; + + assert(f); + assert(f->fd >= 0); + + if (fstat(f->fd, &f->last_stat) < 0) + return -errno; + + f->last_stat_usec = now(CLOCK_MONOTONIC); + + /* Refuse dealing with files that aren't regular */ + r = stat_verify_regular(&f->last_stat); + if (r < 0) + return r; + + /* Refuse appending to files that are already deleted */ + if (f->last_stat.st_nlink <= 0) + return -EIDRM; + + return 0; +} + +static int journal_file_allocate(JournalFile *f, uint64_t offset, uint64_t size) { + uint64_t old_size, new_size, old_header_size, old_arena_size; + int r; + + assert(f); + assert(f->header); + + /* We assume that this file is not sparse, and we know that for sure, since we always call + * posix_fallocate() ourselves */ + + if (size > PAGE_ALIGN_DOWN(UINT64_MAX) - offset) + return -EINVAL; + + if (mmap_cache_fd_got_sigbus(f->cache_fd)) + return -EIO; + + old_header_size = le64toh(READ_NOW(f->header->header_size)); + old_arena_size = le64toh(READ_NOW(f->header->arena_size)); + if (old_arena_size > PAGE_ALIGN_DOWN(UINT64_MAX) - old_header_size) + return -EBADMSG; + + old_size = old_header_size + old_arena_size; + + new_size = MAX(PAGE_ALIGN(offset + size), old_header_size); + + if (new_size <= old_size) { + + /* We already pre-allocated enough space, but before + * we write to it, let's check with fstat() if the + * file got deleted, in order make sure we don't throw + * away the data immediately. Don't check fstat() for + * all writes though, but only once ever 10s. */ + + if (f->last_stat_usec + LAST_STAT_REFRESH_USEC > now(CLOCK_MONOTONIC)) + return 0; + + return journal_file_fstat(f); + } + + /* Allocate more space. */ + + if (f->metrics.max_size > 0 && new_size > f->metrics.max_size) + return -E2BIG; + + /* Refuse to go over 4G in compact mode so offsets can be stored in 32-bit. */ + if (JOURNAL_HEADER_COMPACT(f->header) && new_size > UINT32_MAX) + return -E2BIG; + + if (new_size > f->metrics.min_size && f->metrics.keep_free > 0) { + struct statvfs svfs; + + if (fstatvfs(f->fd, &svfs) >= 0) { + uint64_t available; + + available = LESS_BY((uint64_t) svfs.f_bfree * (uint64_t) svfs.f_bsize, f->metrics.keep_free); + + if (new_size - old_size > available) + return -E2BIG; + } + } + + /* Increase by larger blocks at once */ + new_size = DIV_ROUND_UP(new_size, FILE_SIZE_INCREASE) * FILE_SIZE_INCREASE; + if (f->metrics.max_size > 0 && new_size > f->metrics.max_size) + new_size = f->metrics.max_size; + + /* Note that the glibc fallocate() fallback is very + inefficient, hence we try to minimize the allocation area + as we can. */ + r = posix_fallocate_loop(f->fd, old_size, new_size - old_size); + if (r < 0) + return r; + + f->header->arena_size = htole64(new_size - old_header_size); + + return journal_file_fstat(f); +} + +static unsigned type_to_context(ObjectType type) { + /* One context for each type, plus one catch-all for the rest */ + assert_cc(_OBJECT_TYPE_MAX <= MMAP_CACHE_MAX_CONTEXTS); + assert_cc(CONTEXT_HEADER < MMAP_CACHE_MAX_CONTEXTS); + return type > OBJECT_UNUSED && type < _OBJECT_TYPE_MAX ? type : 0; +} + +static int journal_file_move_to( + JournalFile *f, + ObjectType type, + bool keep_always, + uint64_t offset, + uint64_t size, + void **ret) { + + int r; + + assert(f); + assert(ret); + + /* This function may clear, overwrite, or alter previously cached entries. After this function has + * been called, all objects except for one obtained by this function are invalidated and must be + * re-read before use. */ + + if (size <= 0) + return -EINVAL; + + if (size > UINT64_MAX - offset) + return -EBADMSG; + + /* Avoid SIGBUS on invalid accesses */ + if (offset + size > (uint64_t) f->last_stat.st_size) { + /* Hmm, out of range? Let's refresh the fstat() data + * first, before we trust that check. */ + + r = journal_file_fstat(f); + if (r < 0) + return r; + + if (offset + size > (uint64_t) f->last_stat.st_size) + return -EADDRNOTAVAIL; + } + + return mmap_cache_fd_get(f->cache_fd, type_to_context(type), keep_always, offset, size, &f->last_stat, ret); +} + +static uint64_t minimum_header_size(JournalFile *f, Object *o) { + + static const uint64_t table[] = { + [OBJECT_DATA] = sizeof(DataObject), + [OBJECT_FIELD] = sizeof(FieldObject), + [OBJECT_ENTRY] = sizeof(EntryObject), + [OBJECT_DATA_HASH_TABLE] = sizeof(HashTableObject), + [OBJECT_FIELD_HASH_TABLE] = sizeof(HashTableObject), + [OBJECT_ENTRY_ARRAY] = sizeof(EntryArrayObject), + [OBJECT_TAG] = sizeof(TagObject), + }; + + assert(f); + assert(o); + + if (o->object.type == OBJECT_DATA) + return journal_file_data_payload_offset(f); + + if (o->object.type >= ELEMENTSOF(table) || table[o->object.type] <= 0) + return sizeof(ObjectHeader); + + return table[o->object.type]; +} + +static int check_object_header(JournalFile *f, Object *o, ObjectType type, uint64_t offset) { + uint64_t s; + + assert(f); + assert(o); + + s = le64toh(READ_NOW(o->object.size)); + if (s == 0) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Attempt to move to uninitialized object: %" PRIu64, + offset); + + if (s < sizeof(ObjectHeader)) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Attempt to move to overly short object: %" PRIu64, + offset); + + if (o->object.type <= OBJECT_UNUSED) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Attempt to move to object with invalid type: %" PRIu64, + offset); + + if (type > OBJECT_UNUSED && o->object.type != type) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Attempt to move to object of unexpected type: %" PRIu64, + offset); + + if (s < minimum_header_size(f, o)) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Attempt to move to truncated object: %" PRIu64, + offset); + + return 0; +} + +/* Lightweight object checks. We want this to be fast, so that we won't + * slowdown every journal_file_move_to_object() call too much. */ +static int check_object(JournalFile *f, Object *o, uint64_t offset) { + assert(f); + assert(o); + + switch (o->object.type) { + + case OBJECT_DATA: + if ((le64toh(o->data.entry_offset) == 0) ^ (le64toh(o->data.n_entries) == 0)) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Bad n_entries: %" PRIu64 ": %" PRIu64, + le64toh(o->data.n_entries), + offset); + + if (le64toh(o->object.size) <= journal_file_data_payload_offset(f)) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Bad object size (<= %zu): %" PRIu64 ": %" PRIu64, + journal_file_data_payload_offset(f), + le64toh(o->object.size), + offset); + + if (!VALID64(le64toh(o->data.next_hash_offset)) || + !VALID64(le64toh(o->data.next_field_offset)) || + !VALID64(le64toh(o->data.entry_offset)) || + !VALID64(le64toh(o->data.entry_array_offset))) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Invalid offset, next_hash_offset=" OFSfmt ", next_field_offset=" OFSfmt ", entry_offset=" OFSfmt ", entry_array_offset=" OFSfmt ": %" PRIu64, + le64toh(o->data.next_hash_offset), + le64toh(o->data.next_field_offset), + le64toh(o->data.entry_offset), + le64toh(o->data.entry_array_offset), + offset); + + break; + + case OBJECT_FIELD: + if (le64toh(o->object.size) <= offsetof(Object, field.payload)) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Bad field size (<= %zu): %" PRIu64 ": %" PRIu64, + offsetof(Object, field.payload), + le64toh(o->object.size), + offset); + + if (!VALID64(le64toh(o->field.next_hash_offset)) || + !VALID64(le64toh(o->field.head_data_offset))) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Invalid offset, next_hash_offset=" OFSfmt ", head_data_offset=" OFSfmt ": %" PRIu64, + le64toh(o->field.next_hash_offset), + le64toh(o->field.head_data_offset), + offset); + break; + + case OBJECT_ENTRY: { + uint64_t sz; + + sz = le64toh(READ_NOW(o->object.size)); + if (sz < offsetof(Object, entry.items) || + (sz - offsetof(Object, entry.items)) % journal_file_entry_item_size(f) != 0) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Bad entry size (<= %zu): %" PRIu64 ": %" PRIu64, + offsetof(Object, entry.items), + sz, + offset); + + if ((sz - offsetof(Object, entry.items)) / journal_file_entry_item_size(f) <= 0) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Invalid number items in entry: %" PRIu64 ": %" PRIu64, + (sz - offsetof(Object, entry.items)) / journal_file_entry_item_size(f), + offset); + + if (le64toh(o->entry.seqnum) <= 0) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Invalid entry seqnum: %" PRIx64 ": %" PRIu64, + le64toh(o->entry.seqnum), + offset); + + if (!VALID_REALTIME(le64toh(o->entry.realtime))) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Invalid entry realtime timestamp: %" PRIu64 ": %" PRIu64, + le64toh(o->entry.realtime), + offset); + + if (!VALID_MONOTONIC(le64toh(o->entry.monotonic))) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Invalid entry monotonic timestamp: %" PRIu64 ": %" PRIu64, + le64toh(o->entry.monotonic), + offset); + + break; + } + + case OBJECT_DATA_HASH_TABLE: + case OBJECT_FIELD_HASH_TABLE: { + uint64_t sz; + + sz = le64toh(READ_NOW(o->object.size)); + if (sz < offsetof(Object, hash_table.items) || + (sz - offsetof(Object, hash_table.items)) % sizeof(HashItem) != 0 || + (sz - offsetof(Object, hash_table.items)) / sizeof(HashItem) <= 0) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Invalid %s hash table size: %" PRIu64 ": %" PRIu64, + o->object.type == OBJECT_DATA_HASH_TABLE ? "data" : "field", + sz, + offset); + + break; + } + + case OBJECT_ENTRY_ARRAY: { + uint64_t sz, next; + + sz = le64toh(READ_NOW(o->object.size)); + if (sz < offsetof(Object, entry_array.items) || + (sz - offsetof(Object, entry_array.items)) % journal_file_entry_array_item_size(f) != 0 || + (sz - offsetof(Object, entry_array.items)) / journal_file_entry_array_item_size(f) <= 0) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Invalid object entry array size: %" PRIu64 ": %" PRIu64, + sz, + offset); + /* Here, we request that the offset of each entry array object is in strictly increasing order. */ + next = le64toh(o->entry_array.next_entry_array_offset); + if (!VALID64(next) || (next > 0 && next <= offset)) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Invalid object entry array next_entry_array_offset: %" PRIu64 ": %" PRIu64, + next, + offset); + + break; + } + + case OBJECT_TAG: + if (le64toh(o->object.size) != sizeof(TagObject)) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Invalid object tag size: %" PRIu64 ": %" PRIu64, + le64toh(o->object.size), + offset); + + if (!VALID_EPOCH(le64toh(o->tag.epoch))) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Invalid object tag epoch: %" PRIu64 ": %" PRIu64, + le64toh(o->tag.epoch), offset); + + break; + } + + return 0; +} + +int journal_file_move_to_object(JournalFile *f, ObjectType type, uint64_t offset, Object **ret) { + int r; + Object *o; + + assert(f); + + /* Even if this function fails, it may clear, overwrite, or alter previously cached entries. After + * this function has been called, all objects except for one obtained by this function are + * invalidated and must be re-read before use.. */ + + /* Objects may only be located at multiple of 64 bit */ + if (!VALID64(offset)) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Attempt to move to object at non-64bit boundary: %" PRIu64, + offset); + + /* Object may not be located in the file header */ + if (offset < le64toh(f->header->header_size)) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Attempt to move to object located in file header: %" PRIu64, + offset); + + r = journal_file_move_to(f, type, false, offset, sizeof(ObjectHeader), (void**) &o); + if (r < 0) + return r; + + r = check_object_header(f, o, type, offset); + if (r < 0) + return r; + + r = journal_file_move_to(f, type, false, offset, le64toh(READ_NOW(o->object.size)), (void**) &o); + if (r < 0) + return r; + + r = check_object_header(f, o, type, offset); + if (r < 0) + return r; + + r = check_object(f, o, offset); + if (r < 0) + return r; + + if (ret) + *ret = o; + + return 0; +} + +int journal_file_read_object_header(JournalFile *f, ObjectType type, uint64_t offset, Object *ret) { + ssize_t n; + Object o; + int r; + + assert(f); + + /* Objects may only be located at multiple of 64 bit */ + if (!VALID64(offset)) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Attempt to read object at non-64bit boundary: %" PRIu64, + offset); + + /* Object may not be located in the file header */ + if (offset < le64toh(f->header->header_size)) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Attempt to read object located in file header: %" PRIu64, + offset); + + /* This will likely read too much data but it avoids having to call pread() twice. */ + n = pread(f->fd, &o, sizeof(o), offset); + if (n < 0) + return log_debug_errno(errno, "Failed to read journal file at offset: %" PRIu64, + offset); + + if ((size_t) n < sizeof(o.object)) + return log_debug_errno(SYNTHETIC_ERRNO(EIO), + "Failed to read short object at offset: %" PRIu64, + offset); + + r = check_object_header(f, &o, type, offset); + if (r < 0) + return r; + + if ((size_t) n < minimum_header_size(f, &o)) + return log_debug_errno(SYNTHETIC_ERRNO(EIO), + "Short read while reading object: %" PRIu64, + offset); + + r = check_object(f, &o, offset); + if (r < 0) + return r; + + if (ret) + *ret = o; + + return 0; +} + +static uint64_t journal_file_entry_seqnum( + JournalFile *f, + uint64_t *seqnum) { + + uint64_t ret; + + assert(f); + assert(f->header); + + /* Picks a new sequence number for the entry we are about to add and returns it. */ + + ret = le64toh(f->header->tail_entry_seqnum) + 1; + + if (seqnum) { + /* If an external seqnum counter was passed, we update both the local and the external one, + * and set it to the maximum of both */ + + if (*seqnum + 1 > ret) + ret = *seqnum + 1; + + *seqnum = ret; + } + + f->header->tail_entry_seqnum = htole64(ret); + + if (f->header->head_entry_seqnum == 0) + f->header->head_entry_seqnum = htole64(ret); + + return ret; +} + +int journal_file_append_object( + JournalFile *f, + ObjectType type, + uint64_t size, + Object **ret_object, + uint64_t *ret_offset) { + + int r; + uint64_t p; + Object *o; + + assert(f); + assert(f->header); + assert(type > OBJECT_UNUSED && type < _OBJECT_TYPE_MAX); + assert(size >= sizeof(ObjectHeader)); + + r = journal_file_set_online(f); + if (r < 0) + return r; + + r = journal_file_tail_end_by_mmap(f, &p); + if (r < 0) + return r; + + r = journal_file_allocate(f, p, size); + if (r < 0) + return r; + + r = journal_file_move_to(f, type, false, p, size, (void**) &o); + if (r < 0) + return r; + + o->object = (ObjectHeader) { + .type = type, + .size = htole64(size), + }; + + f->header->tail_object_offset = htole64(p); + f->header->n_objects = htole64(le64toh(f->header->n_objects) + 1); + + if (ret_object) + *ret_object = o; + + if (ret_offset) + *ret_offset = p; + + return 0; +} + +static int journal_file_setup_data_hash_table(JournalFile *f) { + uint64_t s, p; + Object *o; + int r; + + assert(f); + assert(f->header); + + /* We estimate that we need 1 hash table entry per 768 bytes + of journal file and we want to make sure we never get + beyond 75% fill level. Calculate the hash table size for + the maximum file size based on these metrics. */ + + s = (f->metrics.max_size * 4 / 768 / 3) * sizeof(HashItem); + if (s < DEFAULT_DATA_HASH_TABLE_SIZE) + s = DEFAULT_DATA_HASH_TABLE_SIZE; + + log_debug("Reserving %"PRIu64" entries in data hash table.", s / sizeof(HashItem)); + + r = journal_file_append_object(f, + OBJECT_DATA_HASH_TABLE, + offsetof(Object, hash_table.items) + s, + &o, &p); + if (r < 0) + return r; + + memzero(o->hash_table.items, s); + + f->header->data_hash_table_offset = htole64(p + offsetof(Object, hash_table.items)); + f->header->data_hash_table_size = htole64(s); + + return 0; +} + +static int journal_file_setup_field_hash_table(JournalFile *f) { + uint64_t s, p; + Object *o; + int r; + + assert(f); + assert(f->header); + + /* We use a fixed size hash table for the fields as this + * number should grow very slowly only */ + + s = DEFAULT_FIELD_HASH_TABLE_SIZE; + log_debug("Reserving %"PRIu64" entries in field hash table.", s / sizeof(HashItem)); + + r = journal_file_append_object(f, + OBJECT_FIELD_HASH_TABLE, + offsetof(Object, hash_table.items) + s, + &o, &p); + if (r < 0) + return r; + + memzero(o->hash_table.items, s); + + f->header->field_hash_table_offset = htole64(p + offsetof(Object, hash_table.items)); + f->header->field_hash_table_size = htole64(s); + + return 0; +} + +int journal_file_map_data_hash_table(JournalFile *f) { + uint64_t s, p; + void *t; + int r; + + assert(f); + assert(f->header); + + if (f->data_hash_table) + return 0; + + p = le64toh(f->header->data_hash_table_offset); + s = le64toh(f->header->data_hash_table_size); + + r = journal_file_move_to(f, + OBJECT_DATA_HASH_TABLE, + true, + p, s, + &t); + if (r < 0) + return r; + + f->data_hash_table = t; + return 0; +} + +int journal_file_map_field_hash_table(JournalFile *f) { + uint64_t s, p; + void *t; + int r; + + assert(f); + assert(f->header); + + if (f->field_hash_table) + return 0; + + p = le64toh(f->header->field_hash_table_offset); + s = le64toh(f->header->field_hash_table_size); + + r = journal_file_move_to(f, + OBJECT_FIELD_HASH_TABLE, + true, + p, s, + &t); + if (r < 0) + return r; + + f->field_hash_table = t; + return 0; +} + +static int journal_file_link_field( + JournalFile *f, + Object *o, + uint64_t offset, + uint64_t hash) { + + uint64_t p, h, m; + int r; + + assert(f); + assert(f->header); + assert(f->field_hash_table); + assert(o); + assert(offset > 0); + + if (o->object.type != OBJECT_FIELD) + return -EINVAL; + + m = le64toh(READ_NOW(f->header->field_hash_table_size)) / sizeof(HashItem); + if (m <= 0) + return -EBADMSG; + + /* This might alter the window we are looking at */ + o->field.next_hash_offset = o->field.head_data_offset = 0; + + h = hash % m; + p = le64toh(f->field_hash_table[h].tail_hash_offset); + if (p == 0) + f->field_hash_table[h].head_hash_offset = htole64(offset); + else { + r = journal_file_move_to_object(f, OBJECT_FIELD, p, &o); + if (r < 0) + return r; + + o->field.next_hash_offset = htole64(offset); + } + + f->field_hash_table[h].tail_hash_offset = htole64(offset); + + if (JOURNAL_HEADER_CONTAINS(f->header, n_fields)) + f->header->n_fields = htole64(le64toh(f->header->n_fields) + 1); + + return 0; +} + +static int journal_file_link_data( + JournalFile *f, + Object *o, + uint64_t offset, + uint64_t hash) { + + uint64_t p, h, m; + int r; + + assert(f); + assert(f->header); + assert(f->data_hash_table); + assert(o); + assert(offset > 0); + + if (o->object.type != OBJECT_DATA) + return -EINVAL; + + m = le64toh(READ_NOW(f->header->data_hash_table_size)) / sizeof(HashItem); + if (m <= 0) + return -EBADMSG; + + /* This might alter the window we are looking at */ + o->data.next_hash_offset = o->data.next_field_offset = 0; + o->data.entry_offset = o->data.entry_array_offset = 0; + o->data.n_entries = 0; + + h = hash % m; + p = le64toh(f->data_hash_table[h].tail_hash_offset); + if (p == 0) + /* Only entry in the hash table is easy */ + f->data_hash_table[h].head_hash_offset = htole64(offset); + else { + /* Move back to the previous data object, to patch in + * pointer */ + + r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); + if (r < 0) + return r; + + o->data.next_hash_offset = htole64(offset); + } + + f->data_hash_table[h].tail_hash_offset = htole64(offset); + + if (JOURNAL_HEADER_CONTAINS(f->header, n_data)) + f->header->n_data = htole64(le64toh(f->header->n_data) + 1); + + return 0; +} + +static int get_next_hash_offset( + JournalFile *f, + uint64_t *p, + le64_t *next_hash_offset, + uint64_t *depth, + le64_t *header_max_depth) { + + uint64_t nextp; + + assert(f); + assert(p); + assert(next_hash_offset); + assert(depth); + + nextp = le64toh(READ_NOW(*next_hash_offset)); + if (nextp > 0) { + if (nextp <= *p) /* Refuse going in loops */ + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Detected hash item loop in %s, refusing.", f->path); + + (*depth)++; + + /* If the depth of this hash chain is larger than all others we have seen so far, record it */ + if (header_max_depth && journal_file_writable(f)) + *header_max_depth = htole64(MAX(*depth, le64toh(*header_max_depth))); + } + + *p = nextp; + return 0; +} + +int journal_file_find_field_object_with_hash( + JournalFile *f, + const void *field, + uint64_t size, + uint64_t hash, + Object **ret_object, + uint64_t *ret_offset) { + + uint64_t p, osize, h, m, depth = 0; + int r; + + assert(f); + assert(f->header); + assert(field); + assert(size > 0); + + /* If the field hash table is empty, we can't find anything */ + if (le64toh(f->header->field_hash_table_size) <= 0) + return 0; + + /* Map the field hash table, if it isn't mapped yet. */ + r = journal_file_map_field_hash_table(f); + if (r < 0) + return r; + + osize = offsetof(Object, field.payload) + size; + + m = le64toh(READ_NOW(f->header->field_hash_table_size)) / sizeof(HashItem); + if (m <= 0) + return -EBADMSG; + + h = hash % m; + p = le64toh(f->field_hash_table[h].head_hash_offset); + while (p > 0) { + Object *o; + + r = journal_file_move_to_object(f, OBJECT_FIELD, p, &o); + if (r < 0) + return r; + + if (le64toh(o->field.hash) == hash && + le64toh(o->object.size) == osize && + memcmp(o->field.payload, field, size) == 0) { + + if (ret_object) + *ret_object = o; + if (ret_offset) + *ret_offset = p; + + return 1; + } + + r = get_next_hash_offset( + f, + &p, + &o->field.next_hash_offset, + &depth, + JOURNAL_HEADER_CONTAINS(f->header, field_hash_chain_depth) ? &f->header->field_hash_chain_depth : NULL); + if (r < 0) + return r; + } + + return 0; +} + +uint64_t journal_file_hash_data( + JournalFile *f, + const void *data, + size_t sz) { + + assert(f); + assert(f->header); + assert(data || sz == 0); + + /* We try to unify our codebase on siphash, hence new-styled journal files utilizing the keyed hash + * function use siphash. Old journal files use the Jenkins hash. */ + + if (JOURNAL_HEADER_KEYED_HASH(f->header)) + return siphash24(data, sz, f->header->file_id.bytes); + + return jenkins_hash64(data, sz); +} + +int journal_file_find_field_object( + JournalFile *f, + const void *field, + uint64_t size, + Object **ret_object, + uint64_t *ret_offset) { + + assert(f); + assert(field); + assert(size > 0); + + return journal_file_find_field_object_with_hash( + f, + field, size, + journal_file_hash_data(f, field, size), + ret_object, ret_offset); +} + +int journal_file_find_data_object_with_hash( + JournalFile *f, + const void *data, + uint64_t size, + uint64_t hash, + Object **ret_object, + uint64_t *ret_offset) { + + uint64_t p, h, m, depth = 0; + int r; + + assert(f); + assert(f->header); + assert(data || size == 0); + + /* If there's no data hash table, then there's no entry. */ + if (le64toh(f->header->data_hash_table_size) <= 0) + return 0; + + /* Map the data hash table, if it isn't mapped yet. */ + r = journal_file_map_data_hash_table(f); + if (r < 0) + return r; + + m = le64toh(READ_NOW(f->header->data_hash_table_size)) / sizeof(HashItem); + if (m <= 0) + return -EBADMSG; + + h = hash % m; + p = le64toh(f->data_hash_table[h].head_hash_offset); + + while (p > 0) { + Object *o; + void *d; + size_t rsize; + + r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); + if (r < 0) + return r; + + if (le64toh(o->data.hash) != hash) + goto next; + + r = journal_file_data_payload(f, o, p, NULL, 0, 0, &d, &rsize); + if (r < 0) + return r; + assert(r > 0); /* journal_file_data_payload() always returns > 0 if no field is provided. */ + + if (memcmp_nn(data, size, d, rsize) == 0) { + if (ret_object) + *ret_object = o; + + if (ret_offset) + *ret_offset = p; + + return 1; + } + + next: + r = get_next_hash_offset( + f, + &p, + &o->data.next_hash_offset, + &depth, + JOURNAL_HEADER_CONTAINS(f->header, data_hash_chain_depth) ? &f->header->data_hash_chain_depth : NULL); + if (r < 0) + return r; + } + + return 0; +} + +int journal_file_find_data_object( + JournalFile *f, + const void *data, + uint64_t size, + Object **ret_object, + uint64_t *ret_offset) { + + assert(f); + assert(data || size == 0); + + return journal_file_find_data_object_with_hash( + f, + data, size, + journal_file_hash_data(f, data, size), + ret_object, ret_offset); +} + +bool journal_field_valid(const char *p, size_t l, bool allow_protected) { + /* We kinda enforce POSIX syntax recommendations for + environment variables here, but make a couple of additional + requirements. + + http://pubs.opengroup.org/onlinepubs/000095399/basedefs/xbd_chap08.html */ + + assert(p); + + if (l == SIZE_MAX) + l = strlen(p); + + /* No empty field names */ + if (l <= 0) + return false; + + /* Don't allow names longer than 64 chars */ + if (l > 64) + return false; + + /* Variables starting with an underscore are protected */ + if (!allow_protected && p[0] == '_') + return false; + + /* Don't allow digits as first character */ + if (ascii_isdigit(p[0])) + return false; + + /* Only allow A-Z0-9 and '_' */ + for (const char *a = p; a < p + l; a++) + if ((*a < 'A' || *a > 'Z') && + !ascii_isdigit(*a) && + *a != '_') + return false; + + return true; +} + +static int journal_file_append_field( + JournalFile *f, + const void *field, + uint64_t size, + Object **ret_object, + uint64_t *ret_offset) { + + uint64_t hash, p; + uint64_t osize; + Object *o; + int r; + + assert(f); + assert(field); + assert(size > 0); + + if (!journal_field_valid(field, size, true)) + return -EBADMSG; + + hash = journal_file_hash_data(f, field, size); + + r = journal_file_find_field_object_with_hash(f, field, size, hash, ret_object, ret_offset); + if (r < 0) + return r; + if (r > 0) + return 0; + + osize = offsetof(Object, field.payload) + size; + r = journal_file_append_object(f, OBJECT_FIELD, osize, &o, &p); + if (r < 0) + return r; + + o->field.hash = htole64(hash); + memcpy(o->field.payload, field, size); + + r = journal_file_link_field(f, o, p, hash); + if (r < 0) + return r; + + /* The linking might have altered the window, so let's only pass the offset to hmac which will + * move to the object again if needed. */ + +#if HAVE_GCRYPT + r = journal_file_hmac_put_object(f, OBJECT_FIELD, NULL, p); + if (r < 0) + return r; +#endif + + if (ret_object) { + r = journal_file_move_to_object(f, OBJECT_FIELD, p, ret_object); + if (r < 0) + return r; + } + + if (ret_offset) + *ret_offset = p; + + return 0; +} + +static Compression maybe_compress_payload(JournalFile *f, uint8_t *dst, const uint8_t *src, uint64_t size, size_t *rsize) { + Compression compression = COMPRESSION_NONE; + + assert(f); + assert(f->header); + +#if HAVE_COMPRESSION + if (JOURNAL_FILE_COMPRESS(f) && size >= f->compress_threshold_bytes) { + compression = compress_blob(src, size, dst, size - 1, rsize); + if (compression > 0) + log_debug("Compressed data object %"PRIu64" -> %zu using %s", + size, *rsize, compression_to_string(compression)); + else + /* Compression didn't work, we don't really care why, let's continue without compression */ + compression = COMPRESSION_NONE; + } +#endif + + return compression; +} + +static int journal_file_append_data( + JournalFile *f, + const void *data, + uint64_t size, + Object **ret_object, + uint64_t *ret_offset) { + + uint64_t hash, p, osize; + Object *o, *fo; + size_t rsize = 0; + Compression c; + const void *eq; + int r; + + assert(f); + + if (!data || size == 0) + return -EINVAL; + + hash = journal_file_hash_data(f, data, size); + + r = journal_file_find_data_object_with_hash(f, data, size, hash, ret_object, ret_offset); + if (r < 0) + return r; + if (r > 0) + return 0; + + eq = memchr(data, '=', size); + if (!eq) + return -EINVAL; + + osize = journal_file_data_payload_offset(f) + size; + r = journal_file_append_object(f, OBJECT_DATA, osize, &o, &p); + if (r < 0) + return r; + + o->data.hash = htole64(hash); + + c = maybe_compress_payload(f, journal_file_data_payload_field(f, o), data, size, &rsize); + + if (c != COMPRESSION_NONE) { + o->object.size = htole64(journal_file_data_payload_offset(f) + rsize); + o->object.flags |= COMPRESSION_TO_OBJECT_FLAG(c); + } else + memcpy_safe(journal_file_data_payload_field(f, o), data, size); + + r = journal_file_link_data(f, o, p, hash); + if (r < 0) + return r; + + /* The linking might have altered the window, so let's refresh our pointer. */ + r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); + if (r < 0) + return r; + +#if HAVE_GCRYPT + r = journal_file_hmac_put_object(f, OBJECT_DATA, o, p); + if (r < 0) + return r; +#endif + + /* Create field object ... */ + r = journal_file_append_field(f, data, (uint8_t*) eq - (uint8_t*) data, &fo, NULL); + if (r < 0) + return r; + + /* ... and link it in. */ + o->data.next_field_offset = fo->field.head_data_offset; + fo->field.head_data_offset = le64toh(p); + + if (ret_object) + *ret_object = o; + + if (ret_offset) + *ret_offset = p; + + return 0; +} + +static int maybe_decompress_payload( + JournalFile *f, + uint8_t *payload, + uint64_t size, + Compression compression, + const char *field, + size_t field_length, + size_t data_threshold, + void **ret_data, + size_t *ret_size) { + + assert(f); + + /* We can't read objects larger than 4G on a 32bit machine */ + if ((uint64_t) (size_t) size != size) + return -E2BIG; + + if (compression != COMPRESSION_NONE) { +#if HAVE_COMPRESSION + size_t rsize; + int r; + + if (field) { + r = decompress_startswith(compression, payload, size, &f->compress_buffer, field, + field_length, '='); + if (r < 0) + return log_debug_errno(r, + "Cannot decompress %s object of length %" PRIu64 ": %m", + compression_to_string(compression), + size); + if (r == 0) { + if (ret_data) + *ret_data = NULL; + if (ret_size) + *ret_size = 0; + return 0; + } + } + + r = decompress_blob(compression, payload, size, &f->compress_buffer, &rsize, 0); + if (r < 0) + return r; + + if (ret_data) + *ret_data = f->compress_buffer; + if (ret_size) + *ret_size = rsize; +#else + return -EPROTONOSUPPORT; +#endif + } else { + if (field && (size < field_length + 1 || memcmp(payload, field, field_length) != 0 || payload[field_length] != '=')) { + if (ret_data) + *ret_data = NULL; + if (ret_size) + *ret_size = 0; + return 0; + } + + if (ret_data) + *ret_data = payload; + if (ret_size) + *ret_size = (size_t) size; + } + + return 1; +} + +int journal_file_data_payload( + JournalFile *f, + Object *o, + uint64_t offset, + const char *field, + size_t field_length, + size_t data_threshold, + void **ret_data, + size_t *ret_size) { + + uint64_t size; + Compression c; + int r; + + assert(f); + assert(!field == (field_length == 0)); /* These must be specified together. */ + + if (!o) { + r = journal_file_move_to_object(f, OBJECT_DATA, offset, &o); + if (r < 0) + return r; + } + + size = le64toh(READ_NOW(o->object.size)); + if (size < journal_file_data_payload_offset(f)) + return -EBADMSG; + + size -= journal_file_data_payload_offset(f); + + c = COMPRESSION_FROM_OBJECT(o); + if (c < 0) + return -EPROTONOSUPPORT; + + return maybe_decompress_payload(f, journal_file_data_payload_field(f, o), size, c, field, + field_length, data_threshold, ret_data, ret_size); +} + +uint64_t journal_file_entry_n_items(JournalFile *f, Object *o) { + uint64_t sz; + + assert(f); + assert(o); + + if (o->object.type != OBJECT_ENTRY) + return 0; + + sz = le64toh(READ_NOW(o->object.size)); + if (sz < offsetof(Object, entry.items)) + return 0; + + return (sz - offsetof(Object, entry.items)) / journal_file_entry_item_size(f); +} + +uint64_t journal_file_entry_array_n_items(JournalFile *f, Object *o) { + uint64_t sz; + + assert(f); + assert(o); + + if (o->object.type != OBJECT_ENTRY_ARRAY) + return 0; + + sz = le64toh(READ_NOW(o->object.size)); + if (sz < offsetof(Object, entry_array.items)) + return 0; + + return (sz - offsetof(Object, entry_array.items)) / journal_file_entry_array_item_size(f); +} + +uint64_t journal_file_hash_table_n_items(Object *o) { + uint64_t sz; + + assert(o); + + if (!IN_SET(o->object.type, OBJECT_DATA_HASH_TABLE, OBJECT_FIELD_HASH_TABLE)) + return 0; + + sz = le64toh(READ_NOW(o->object.size)); + if (sz < offsetof(Object, hash_table.items)) + return 0; + + return (sz - offsetof(Object, hash_table.items)) / sizeof(HashItem); +} + +static void write_entry_array_item(JournalFile *f, Object *o, uint64_t i, uint64_t p) { + assert(f); + assert(o); + + if (JOURNAL_HEADER_COMPACT(f->header)) { + assert(p <= UINT32_MAX); + o->entry_array.items.compact[i] = htole32(p); + } else + o->entry_array.items.regular[i] = htole64(p); +} + +static int link_entry_into_array( + JournalFile *f, + le64_t *first, + le64_t *idx, + le32_t *tail, + le32_t *tidx, + uint64_t p) { + + uint64_t n = 0, ap = 0, q, i, a, hidx; + Object *o; + int r; + + assert(f); + assert(f->header); + assert(first); + assert(idx); + assert(p > 0); + + a = tail ? le32toh(*tail) : le64toh(*first); + hidx = le64toh(READ_NOW(*idx)); + i = tidx ? le32toh(READ_NOW(*tidx)) : hidx; + + while (a > 0) { + r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o); + if (r < 0) + return r; + + n = journal_file_entry_array_n_items(f, o); + if (i < n) { + write_entry_array_item(f, o, i, p); + *idx = htole64(hidx + 1); + if (tidx) + *tidx = htole32(le32toh(*tidx) + 1); + return 0; + } + + i -= n; + ap = a; + a = le64toh(o->entry_array.next_entry_array_offset); + } + + if (hidx > n) + n = (hidx+1) * 2; + else + n = n * 2; + + if (n < 4) + n = 4; + + r = journal_file_append_object(f, OBJECT_ENTRY_ARRAY, + offsetof(Object, entry_array.items) + n * journal_file_entry_array_item_size(f), + &o, &q); + if (r < 0) + return r; + +#if HAVE_GCRYPT + r = journal_file_hmac_put_object(f, OBJECT_ENTRY_ARRAY, o, q); + if (r < 0) + return r; +#endif + + write_entry_array_item(f, o, i, p); + + if (ap == 0) + *first = htole64(q); + else { + r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, ap, &o); + if (r < 0) + return r; + + o->entry_array.next_entry_array_offset = htole64(q); + } + + if (tail) + *tail = htole32(q); + + if (JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays)) + f->header->n_entry_arrays = htole64(le64toh(f->header->n_entry_arrays) + 1); + + *idx = htole64(hidx + 1); + if (tidx) + *tidx = htole32(1); + + return 0; +} + +static int link_entry_into_array_plus_one( + JournalFile *f, + le64_t *extra, + le64_t *first, + le64_t *idx, + le32_t *tail, + le32_t *tidx, + uint64_t p) { + + uint64_t hidx; + int r; + + assert(f); + assert(extra); + assert(first); + assert(idx); + assert(p > 0); + + hidx = le64toh(READ_NOW(*idx)); + if (hidx == UINT64_MAX) + return -EBADMSG; + if (hidx == 0) + *extra = htole64(p); + else { + le64_t i; + + i = htole64(hidx - 1); + r = link_entry_into_array(f, first, &i, tail, tidx, p); + if (r < 0) + return r; + } + + *idx = htole64(hidx + 1); + return 0; +} + +static int journal_file_link_entry_item(JournalFile *f, uint64_t offset, uint64_t p) { + Object *o; + int r; + + assert(f); + assert(offset > 0); + + r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); + if (r < 0) + return r; + + return link_entry_into_array_plus_one(f, + &o->data.entry_offset, + &o->data.entry_array_offset, + &o->data.n_entries, + JOURNAL_HEADER_COMPACT(f->header) ? &o->data.compact.tail_entry_array_offset : NULL, + JOURNAL_HEADER_COMPACT(f->header) ? &o->data.compact.tail_entry_array_n_entries : NULL, + offset); +} + +static int journal_file_link_entry( + JournalFile *f, + Object *o, + uint64_t offset, + const EntryItem items[], + size_t n_items) { + + int r; + + assert(f); + assert(f->header); + assert(o); + assert(offset > 0); + + if (o->object.type != OBJECT_ENTRY) + return -EINVAL; + + __atomic_thread_fence(__ATOMIC_SEQ_CST); + + /* Link up the entry itself */ + r = link_entry_into_array(f, + &f->header->entry_array_offset, + &f->header->n_entries, + JOURNAL_HEADER_CONTAINS(f->header, tail_entry_array_offset) ? &f->header->tail_entry_array_offset : NULL, + JOURNAL_HEADER_CONTAINS(f->header, tail_entry_array_n_entries) ? &f->header->tail_entry_array_n_entries : NULL, + offset); + if (r < 0) + return r; + + /* log_debug("=> %s seqnr=%"PRIu64" n_entries=%"PRIu64, f->path, o->entry.seqnum, f->header->n_entries); */ + + if (f->header->head_entry_realtime == 0) + f->header->head_entry_realtime = o->entry.realtime; + + f->header->tail_entry_realtime = o->entry.realtime; + f->header->tail_entry_monotonic = o->entry.monotonic; + + /* Link up the items */ + for (uint64_t i = 0; i < n_items; i++) { + int k; + + /* If we fail to link an entry item because we can't allocate a new entry array, don't fail + * immediately but try to link the other entry items since it might still be possible to link + * those if they don't require a new entry array to be allocated. */ + + k = journal_file_link_entry_item(f, offset, items[i].object_offset); + if (k == -E2BIG) + r = k; + else if (k < 0) + return k; + } + + return r; +} + +static void write_entry_item(JournalFile *f, Object *o, uint64_t i, const EntryItem *item) { + assert(f); + assert(o); + assert(item); + + if (JOURNAL_HEADER_COMPACT(f->header)) { + assert(item->object_offset <= UINT32_MAX); + o->entry.items.compact[i].object_offset = htole32(item->object_offset); + } else { + o->entry.items.regular[i].object_offset = htole64(item->object_offset); + o->entry.items.regular[i].hash = htole64(item->hash); + } +} + +static int journal_file_append_entry_internal( + JournalFile *f, + const dual_timestamp *ts, + const sd_id128_t *boot_id, + uint64_t xor_hash, + const EntryItem items[], + size_t n_items, + uint64_t *seqnum, + Object **ret_object, + uint64_t *ret_offset) { + + uint64_t np; + uint64_t osize; + Object *o; + int r; + + assert(f); + assert(f->header); + assert(ts); + assert(items || n_items == 0); + + osize = offsetof(Object, entry.items) + (n_items * journal_file_entry_item_size(f)); + + r = journal_file_append_object(f, OBJECT_ENTRY, osize, &o, &np); + if (r < 0) + return r; + + o->entry.seqnum = htole64(journal_file_entry_seqnum(f, seqnum)); + o->entry.realtime = htole64(ts->realtime); + o->entry.monotonic = htole64(ts->monotonic); + o->entry.xor_hash = htole64(xor_hash); + if (boot_id) + f->header->boot_id = *boot_id; + o->entry.boot_id = f->header->boot_id; + + for (size_t i = 0; i < n_items; i++) + write_entry_item(f, o, i, &items[i]); + +#if HAVE_GCRYPT + r = journal_file_hmac_put_object(f, OBJECT_ENTRY, o, np); + if (r < 0) + return r; +#endif + + r = journal_file_link_entry(f, o, np, items, n_items); + if (r < 0) + return r; + + if (ret_object) + *ret_object = o; + + if (ret_offset) + *ret_offset = np; + + return r; +} + +void journal_file_post_change(JournalFile *f) { + assert(f); + + if (f->fd < 0) + return; + + /* inotify() does not receive IN_MODIFY events from file + * accesses done via mmap(). After each access we hence + * trigger IN_MODIFY by truncating the journal file to its + * current size which triggers IN_MODIFY. */ + + __atomic_thread_fence(__ATOMIC_SEQ_CST); + + if (ftruncate(f->fd, f->last_stat.st_size) < 0) + log_debug_errno(errno, "Failed to truncate file to its own size: %m"); +} + +static int post_change_thunk(sd_event_source *timer, uint64_t usec, void *userdata) { + assert(userdata); + + journal_file_post_change(userdata); + + return 1; +} + +static void schedule_post_change(JournalFile *f) { + sd_event *e; + int r; + + assert(f); + assert(f->post_change_timer); + + assert_se(e = sd_event_source_get_event(f->post_change_timer)); + + /* If we are already going down, post the change immediately. */ + if (IN_SET(sd_event_get_state(e), SD_EVENT_EXITING, SD_EVENT_FINISHED)) + goto fail; + + r = sd_event_source_get_enabled(f->post_change_timer, NULL); + if (r < 0) { + log_debug_errno(r, "Failed to get ftruncate timer state: %m"); + goto fail; + } + if (r > 0) + return; + + r = sd_event_source_set_time_relative(f->post_change_timer, f->post_change_timer_period); + if (r < 0) { + log_debug_errno(r, "Failed to set time for scheduling ftruncate: %m"); + goto fail; + } + + r = sd_event_source_set_enabled(f->post_change_timer, SD_EVENT_ONESHOT); + if (r < 0) { + log_debug_errno(r, "Failed to enable scheduled ftruncate: %m"); + goto fail; + } + + return; + +fail: + /* On failure, let's simply post the change immediately. */ + journal_file_post_change(f); +} + +/* Enable coalesced change posting in a timer on the provided sd_event instance */ +int journal_file_enable_post_change_timer(JournalFile *f, sd_event *e, usec_t t) { + _cleanup_(sd_event_source_unrefp) sd_event_source *timer = NULL; + int r; + + assert(f); + assert_return(!f->post_change_timer, -EINVAL); + assert(e); + assert(t); + + /* If we are already going down, we cannot install the timer. + * In such case, the caller needs to call journal_file_post_change() explicitly. */ + if (IN_SET(sd_event_get_state(e), SD_EVENT_EXITING, SD_EVENT_FINISHED)) + return 0; + + r = sd_event_add_time(e, &timer, CLOCK_MONOTONIC, 0, 0, post_change_thunk, f); + if (r < 0) + return r; + + r = sd_event_source_set_enabled(timer, SD_EVENT_OFF); + if (r < 0) + return r; + + f->post_change_timer = TAKE_PTR(timer); + f->post_change_timer_period = t; + + return 1; +} + +static int entry_item_cmp(const EntryItem *a, const EntryItem *b) { + return CMP(ASSERT_PTR(a)->object_offset, ASSERT_PTR(b)->object_offset); +} + +static size_t remove_duplicate_entry_items(EntryItem items[], size_t n) { + size_t j = 1; + + assert(items || n == 0); + + if (n <= 1) + return n; + + for (size_t i = 1; i < n; i++) + if (items[i].object_offset != items[j - 1].object_offset) + items[j++] = items[i]; + + return j; +} + +int journal_file_append_entry( + JournalFile *f, + const dual_timestamp *ts, + const sd_id128_t *boot_id, + const struct iovec iovec[], + unsigned n_iovec, + uint64_t *seqnum, + Object **ret_object, + uint64_t *ret_offset) { + + _cleanup_free_ EntryItem *items_alloc = NULL; + EntryItem *items; + uint64_t xor_hash = 0; + struct dual_timestamp _ts; + int r; + + assert(f); + assert(f->header); + assert(iovec); + assert(n_iovec > 0); + + if (ts) { + if (!VALID_REALTIME(ts->realtime)) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Invalid realtime timestamp %" PRIu64 ", refusing entry.", + ts->realtime); + if (!VALID_MONOTONIC(ts->monotonic)) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "Invalid monotomic timestamp %" PRIu64 ", refusing entry.", + ts->monotonic); + } else { + dual_timestamp_get(&_ts); + ts = &_ts; + } + +#if HAVE_GCRYPT + r = journal_file_maybe_append_tag(f, ts->realtime); + if (r < 0) + return r; +#endif + + if (n_iovec < ALLOCA_MAX / sizeof(EntryItem) / 2) + items = newa(EntryItem, n_iovec); + else { + items_alloc = new(EntryItem, n_iovec); + if (!items_alloc) + return -ENOMEM; + + items = items_alloc; + } + + for (size_t i = 0; i < n_iovec; i++) { + uint64_t p; + Object *o; + + r = journal_file_append_data(f, iovec[i].iov_base, iovec[i].iov_len, &o, &p); + if (r < 0) + return r; + + /* When calculating the XOR hash field, we need to take special care if the "keyed-hash" + * journal file flag is on. We use the XOR hash field to quickly determine the identity of a + * specific record, and give records with otherwise identical position (i.e. match in seqno, + * timestamp, …) a stable ordering. But for that we can't have it that the hash of the + * objects in each file is different since they are keyed. Hence let's calculate the Jenkins + * hash here for that. This also has the benefit that cursors for old and new journal files + * are completely identical (they include the XOR hash after all). For classic Jenkins-hash + * files things are easier, we can just take the value from the stored record directly. */ + + if (JOURNAL_HEADER_KEYED_HASH(f->header)) + xor_hash ^= jenkins_hash64(iovec[i].iov_base, iovec[i].iov_len); + else + xor_hash ^= le64toh(o->data.hash); + + items[i] = (EntryItem) { + .object_offset = p, + .hash = le64toh(o->data.hash), + }; + } + + /* Order by the position on disk, in order to improve seek + * times for rotating media. */ + typesafe_qsort(items, n_iovec, entry_item_cmp); + n_iovec = remove_duplicate_entry_items(items, n_iovec); + + r = journal_file_append_entry_internal(f, ts, boot_id, xor_hash, items, n_iovec, seqnum, ret_object, ret_offset); + + /* If the memory mapping triggered a SIGBUS then we return an + * IO error and ignore the error code passed down to us, since + * it is very likely just an effect of a nullified replacement + * mapping page */ + + if (mmap_cache_fd_got_sigbus(f->cache_fd)) + r = -EIO; + + if (f->post_change_timer) + schedule_post_change(f); + else + journal_file_post_change(f); + + return r; +} + +typedef struct ChainCacheItem { + uint64_t first; /* the array at the beginning of the chain */ + uint64_t array; /* the cached array */ + uint64_t begin; /* the first item in the cached array */ + uint64_t total; /* the total number of items in all arrays before this one in the chain */ + uint64_t last_index; /* the last index we looked at, to optimize locality when bisecting */ +} ChainCacheItem; + +static void chain_cache_put( + OrderedHashmap *h, + ChainCacheItem *ci, + uint64_t first, + uint64_t array, + uint64_t begin, + uint64_t total, + uint64_t last_index) { + + assert(h); + + if (!ci) { + /* If the chain item to cache for this chain is the + * first one it's not worth caching anything */ + if (array == first) + return; + + if (ordered_hashmap_size(h) >= CHAIN_CACHE_MAX) { + ci = ordered_hashmap_steal_first(h); + assert(ci); + } else { + ci = new(ChainCacheItem, 1); + if (!ci) + return; + } + + ci->first = first; + + if (ordered_hashmap_put(h, &ci->first, ci) < 0) { + free(ci); + return; + } + } else + assert(ci->first == first); + + ci->array = array; + ci->begin = begin; + ci->total = total; + ci->last_index = last_index; +} + +static int bump_array_index(uint64_t *i, direction_t direction, uint64_t n) { + assert(i); + + /* Increase or decrease the specified index, in the right direction. */ + + if (direction == DIRECTION_DOWN) { + if (*i >= n - 1) + return 0; + + (*i)++; + } else { + if (*i <= 0) + return 0; + + (*i)--; + } + + return 1; +} + +static int bump_entry_array( + JournalFile *f, + Object *o, + uint64_t offset, + uint64_t first, + direction_t direction, + uint64_t *ret) { + + uint64_t p, q = 0; + int r; + + assert(f); + assert(offset); + assert(ret); + + if (direction == DIRECTION_DOWN) { + assert(o); + *ret = le64toh(o->entry_array.next_entry_array_offset); + return 0; + } + + /* Entry array chains are a singly linked list, so to find the previous array in the chain, we have + * to start iterating from the top. */ + + p = first; + + while (p > 0 && p != offset) { + r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, p, &o); + if (r < 0) + return r; + + q = p; + p = le64toh(o->entry_array.next_entry_array_offset); + } + + /* If we can't find the previous entry array in the entry array chain, we're likely dealing with a + * corrupted journal file. */ + if (p == 0) + return -EBADMSG; + + *ret = q; + + return 0; +} + +static int generic_array_get( + JournalFile *f, + uint64_t first, + uint64_t i, + direction_t direction, + Object **ret_object, + uint64_t *ret_offset) { + + uint64_t p = 0, a, t = 0, k; + ChainCacheItem *ci; + Object *o; + int r; + + assert(f); + + /* FIXME: fix return value assignment on success. */ + + a = first; + + /* Try the chain cache first */ + ci = ordered_hashmap_get(f->chain_cache, &first); + if (ci && i > ci->total) { + a = ci->array; + i -= ci->total; + t = ci->total; + } + + while (a > 0) { + r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o); + if (IN_SET(r, -EBADMSG, -EADDRNOTAVAIL)) { + /* If there's corruption and we're going downwards, let's pretend we reached the + * final entry in the entry array chain. */ + + if (direction == DIRECTION_DOWN) + return 0; + + /* If there's corruption and we're going upwards, move back to the previous entry + * array and start iterating entries from there. */ + + r = bump_entry_array(f, NULL, a, first, DIRECTION_UP, &a); + if (r < 0) + return r; + + i = UINT64_MAX; + + break; + } + if (r < 0) + return r; + + k = journal_file_entry_array_n_items(f, o); + if (i < k) + break; + + i -= k; + t += k; + a = le64toh(o->entry_array.next_entry_array_offset); + } + + /* If we've found the right location, now look for the first non-corrupt entry object (in the right + * direction). */ + + while (a > 0) { + /* In the first iteration of the while loop, we reuse i, k and o from the previous while + * loop. */ + if (i == UINT64_MAX) { + r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o); + if (r < 0) + return r; + + k = journal_file_entry_array_n_items(f, o); + if (k == 0) + break; + + i = direction == DIRECTION_DOWN ? 0 : k - 1; + } + + do { + p = journal_file_entry_array_item(f, o, i); + + r = journal_file_move_to_object(f, OBJECT_ENTRY, p, ret_object); + if (r >= 0) { + /* Let's cache this item for the next invocation */ + chain_cache_put(f->chain_cache, ci, first, a, journal_file_entry_array_item(f, o, 0), t, i); + + if (ret_offset) + *ret_offset = p; + + return 1; + } + if (!IN_SET(r, -EADDRNOTAVAIL, -EBADMSG)) + return r; + + /* OK, so this entry is borked. Most likely some entry didn't get synced to + * disk properly, let's see if the next one might work for us instead. */ + log_debug_errno(r, "Entry item %" PRIu64 " is bad, skipping over it.", i); + } while (bump_array_index(&i, direction, k) > 0); + + r = bump_entry_array(f, o, a, first, direction, &a); + if (r < 0) + return r; + + t += k; + i = UINT64_MAX; + } + + return 0; +} + +static int generic_array_get_plus_one( + JournalFile *f, + uint64_t extra, + uint64_t first, + uint64_t i, + direction_t direction, + Object **ret_object, + uint64_t *ret_offset) { + + int r; + + assert(f); + + /* FIXME: fix return value assignment on success. */ + + if (i == 0) { + r = journal_file_move_to_object(f, OBJECT_ENTRY, extra, ret_object); + if (IN_SET(r, -EADDRNOTAVAIL, -EBADMSG)) + return generic_array_get(f, first, 0, direction, ret_object, ret_offset); + if (r < 0) + return r; + + if (ret_offset) + *ret_offset = extra; + + return 1; + } + + return generic_array_get(f, first, i - 1, direction, ret_object, ret_offset); +} + +enum { + TEST_FOUND, + TEST_LEFT, + TEST_RIGHT +}; + +static int generic_array_bisect( + JournalFile *f, + uint64_t first, + uint64_t n, + uint64_t needle, + int (*test_object)(JournalFile *f, uint64_t p, uint64_t needle), + direction_t direction, + Object **ret_object, + uint64_t *ret_offset, + uint64_t *ret_idx) { + + /* Given an entry array chain, this function finds the object "closest" to the given needle in the + * chain, taking into account the provided direction. A function can be provided to determine how + * an object is matched against the given needle. + * + * Given a journal file, the offset of an object and the needle, the test_object() function should + * return TEST_LEFT if the needle is located earlier in the entry array chain, TEST_RIGHT if the + * needle is located later in the entry array chain and TEST_FOUND if the object matches the needle. + * If test_object() returns TEST_FOUND for a specific object, that object's information will be used + * to populate the return values of this function. If test_object() never returns TEST_FOUND, the + * return values are populated with the details of one of the objects closest to the needle. If the + * direction is DIRECTION_UP, the earlier object is used. Otherwise, the later object is used. + */ + + uint64_t a, p, t = 0, i = 0, last_p = 0, last_index = UINT64_MAX; + bool subtract_one = false; + Object *array = NULL; + ChainCacheItem *ci; + int r; + + assert(f); + assert(test_object); + + /* Start with the first array in the chain */ + a = first; + + ci = ordered_hashmap_get(f->chain_cache, &first); + if (ci && n > ci->total && ci->begin != 0) { + /* Ah, we have iterated this bisection array chain + * previously! Let's see if we can skip ahead in the + * chain, as far as the last time. But we can't jump + * backwards in the chain, so let's check that + * first. */ + + r = test_object(f, ci->begin, needle); + if (r < 0) + return r; + + if (r == TEST_LEFT) { + /* OK, what we are looking for is right of the + * begin of this EntryArray, so let's jump + * straight to previously cached array in the + * chain */ + + a = ci->array; + n -= ci->total; + t = ci->total; + last_index = ci->last_index; + } + } + + while (a > 0) { + uint64_t left, right, k, lp; + + r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &array); + if (r < 0) + return r; + + k = journal_file_entry_array_n_items(f, array); + right = MIN(k, n); + if (right <= 0) + return 0; + + i = right - 1; + lp = p = journal_file_entry_array_item(f, array, i); + if (p <= 0) + r = -EBADMSG; + else + r = test_object(f, p, needle); + if (r == -EBADMSG) { + log_debug_errno(r, "Encountered invalid entry while bisecting, cutting algorithm short. (1)"); + n = i; + continue; + } + if (r < 0) + return r; + + if (r == TEST_FOUND) + r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; + + if (r == TEST_RIGHT) { + left = 0; + right -= 1; + + if (last_index != UINT64_MAX) { + assert(last_index <= right); + + /* If we cached the last index we + * looked at, let's try to not to jump + * too wildly around and see if we can + * limit the range to look at early to + * the immediate neighbors of the last + * index we looked at. */ + + if (last_index > 0) { + uint64_t x = last_index - 1; + + p = journal_file_entry_array_item(f, array, x); + if (p <= 0) + return -EBADMSG; + + r = test_object(f, p, needle); + if (r < 0) + return r; + + if (r == TEST_FOUND) + r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; + + if (r == TEST_RIGHT) + right = x; + else + left = x + 1; + } + + if (last_index < right) { + uint64_t y = last_index + 1; + + p = journal_file_entry_array_item(f, array, y); + if (p <= 0) + return -EBADMSG; + + r = test_object(f, p, needle); + if (r < 0) + return r; + + if (r == TEST_FOUND) + r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; + + if (r == TEST_RIGHT) + right = y; + else + left = y + 1; + } + } + + for (;;) { + if (left == right) { + if (direction == DIRECTION_UP) + subtract_one = true; + + i = left; + goto found; + } + + assert(left < right); + i = (left + right) / 2; + + p = journal_file_entry_array_item(f, array, i); + if (p <= 0) + r = -EBADMSG; + else + r = test_object(f, p, needle); + if (r == -EBADMSG) { + log_debug_errno(r, "Encountered invalid entry while bisecting, cutting algorithm short. (2)"); + right = n = i; + continue; + } + if (r < 0) + return r; + + if (r == TEST_FOUND) + r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; + + if (r == TEST_RIGHT) + right = i; + else + left = i + 1; + } + } + + if (k >= n) { + if (direction == DIRECTION_UP) { + i = n; + subtract_one = true; + goto found; + } + + return 0; + } + + last_p = lp; + + n -= k; + t += k; + last_index = UINT64_MAX; + a = le64toh(array->entry_array.next_entry_array_offset); + } + + return 0; + +found: + if (subtract_one && t == 0 && i == 0) + return 0; + + /* Let's cache this item for the next invocation */ + chain_cache_put(f->chain_cache, ci, first, a, journal_file_entry_array_item(f, array, 0), t, subtract_one ? (i > 0 ? i-1 : UINT64_MAX) : i); + + if (subtract_one && i == 0) + p = last_p; + else if (subtract_one) + p = journal_file_entry_array_item(f, array, i - 1); + else + p = journal_file_entry_array_item(f, array, i); + + if (ret_object) { + r = journal_file_move_to_object(f, OBJECT_ENTRY, p, ret_object); + if (r < 0) + return r; + } + + if (ret_offset) + *ret_offset = p; + + if (ret_idx) + *ret_idx = t + i + (subtract_one ? -1 : 0); + + return 1; +} + +static int generic_array_bisect_plus_one( + JournalFile *f, + uint64_t extra, + uint64_t first, + uint64_t n, + uint64_t needle, + int (*test_object)(JournalFile *f, uint64_t p, uint64_t needle), + direction_t direction, + Object **ret_object, + uint64_t *ret_offset, + uint64_t *ret_idx) { + + int r; + bool step_back = false; + + assert(f); + assert(test_object); + + if (n <= 0) + return 0; + + /* This bisects the array in object 'first', but first checks + * an extra */ + r = test_object(f, extra, needle); + if (r < 0) + return r; + + if (r == TEST_FOUND) + r = direction == DIRECTION_DOWN ? TEST_RIGHT : TEST_LEFT; + + /* if we are looking with DIRECTION_UP then we need to first + see if in the actual array there is a matching entry, and + return the last one of that. But if there isn't any we need + to return this one. Hence remember this, and return it + below. */ + if (r == TEST_LEFT) + step_back = direction == DIRECTION_UP; + + if (r == TEST_RIGHT) { + if (direction == DIRECTION_DOWN) + goto found; + else + return 0; + } + + r = generic_array_bisect(f, first, n-1, needle, test_object, direction, ret_object, ret_offset, ret_idx); + + if (r == 0 && step_back) + goto found; + + if (r > 0 && ret_idx) + (*ret_idx)++; + + return r; + +found: + if (ret_object) { + r = journal_file_move_to_object(f, OBJECT_ENTRY, extra, ret_object); + if (r < 0) + return r; + } + + if (ret_offset) + *ret_offset = extra; + + if (ret_idx) + *ret_idx = 0; + + return 1; +} + +_pure_ static int test_object_offset(JournalFile *f, uint64_t p, uint64_t needle) { + assert(f); + assert(p > 0); + + if (p == needle) + return TEST_FOUND; + else if (p < needle) + return TEST_LEFT; + else + return TEST_RIGHT; +} + +int journal_file_move_to_entry_by_offset( + JournalFile *f, + uint64_t p, + direction_t direction, + Object **ret_object, + uint64_t *ret_offset) { + + assert(f); + assert(f->header); + + return generic_array_bisect( + f, + le64toh(f->header->entry_array_offset), + le64toh(f->header->n_entries), + p, + test_object_offset, + direction, + ret_object, ret_offset, NULL); +} + +static int test_object_seqnum(JournalFile *f, uint64_t p, uint64_t needle) { + uint64_t sq; + Object *o; + int r; + + assert(f); + assert(p > 0); + + r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); + if (r < 0) + return r; + + sq = le64toh(READ_NOW(o->entry.seqnum)); + if (sq == needle) + return TEST_FOUND; + else if (sq < needle) + return TEST_LEFT; + else + return TEST_RIGHT; +} + +int journal_file_move_to_entry_by_seqnum( + JournalFile *f, + uint64_t seqnum, + direction_t direction, + Object **ret_object, + uint64_t *ret_offset) { + + assert(f); + assert(f->header); + + return generic_array_bisect( + f, + le64toh(f->header->entry_array_offset), + le64toh(f->header->n_entries), + seqnum, + test_object_seqnum, + direction, + ret_object, ret_offset, NULL); +} + +static int test_object_realtime(JournalFile *f, uint64_t p, uint64_t needle) { + Object *o; + uint64_t rt; + int r; + + assert(f); + assert(p > 0); + + r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); + if (r < 0) + return r; + + rt = le64toh(READ_NOW(o->entry.realtime)); + if (rt == needle) + return TEST_FOUND; + else if (rt < needle) + return TEST_LEFT; + else + return TEST_RIGHT; +} + +int journal_file_move_to_entry_by_realtime( + JournalFile *f, + uint64_t realtime, + direction_t direction, + Object **ret_object, + uint64_t *ret_offset) { + + assert(f); + assert(f->header); + + return generic_array_bisect( + f, + le64toh(f->header->entry_array_offset), + le64toh(f->header->n_entries), + realtime, + test_object_realtime, + direction, + ret_object, ret_offset, NULL); +} + +static int test_object_monotonic(JournalFile *f, uint64_t p, uint64_t needle) { + Object *o; + uint64_t m; + int r; + + assert(f); + assert(p > 0); + + r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); + if (r < 0) + return r; + + m = le64toh(READ_NOW(o->entry.monotonic)); + if (m == needle) + return TEST_FOUND; + else if (m < needle) + return TEST_LEFT; + else + return TEST_RIGHT; +} + +static int find_data_object_by_boot_id( + JournalFile *f, + sd_id128_t boot_id, + Object **ret_object, + uint64_t *ret_offset) { + + char t[STRLEN("_BOOT_ID=") + 32 + 1] = "_BOOT_ID="; + + assert(f); + + sd_id128_to_string(boot_id, t + 9); + return journal_file_find_data_object(f, t, sizeof(t) - 1, ret_object, ret_offset); +} + +int journal_file_move_to_entry_by_monotonic( + JournalFile *f, + sd_id128_t boot_id, + uint64_t monotonic, + direction_t direction, + Object **ret_object, + uint64_t *ret_offset) { + + Object *o; + int r; + + assert(f); + + r = find_data_object_by_boot_id(f, boot_id, &o, NULL); + if (r <= 0) + return r; + + return generic_array_bisect_plus_one( + f, + le64toh(o->data.entry_offset), + le64toh(o->data.entry_array_offset), + le64toh(o->data.n_entries), + monotonic, + test_object_monotonic, + direction, + ret_object, ret_offset, NULL); +} + +void journal_file_reset_location(JournalFile *f) { + assert(f); + + f->location_type = LOCATION_HEAD; + f->current_offset = 0; + f->current_seqnum = 0; + f->current_realtime = 0; + f->current_monotonic = 0; + zero(f->current_boot_id); + f->current_xor_hash = 0; +} + +void journal_file_save_location(JournalFile *f, Object *o, uint64_t offset) { + assert(f); + assert(o); + + f->location_type = LOCATION_SEEK; + f->current_offset = offset; + f->current_seqnum = le64toh(o->entry.seqnum); + f->current_realtime = le64toh(o->entry.realtime); + f->current_monotonic = le64toh(o->entry.monotonic); + f->current_boot_id = o->entry.boot_id; + f->current_xor_hash = le64toh(o->entry.xor_hash); +} + +int journal_file_compare_locations(JournalFile *af, JournalFile *bf) { + int r; + + assert(af); + assert(af->header); + assert(bf); + assert(bf->header); + assert(af->location_type == LOCATION_SEEK); + assert(bf->location_type == LOCATION_SEEK); + + /* If contents, timestamps and seqnum match, these entries are + * identical. */ + if (sd_id128_equal(af->current_boot_id, bf->current_boot_id) && + af->current_monotonic == bf->current_monotonic && + af->current_realtime == bf->current_realtime && + af->current_xor_hash == bf->current_xor_hash && + sd_id128_equal(af->header->seqnum_id, bf->header->seqnum_id) && + af->current_seqnum == bf->current_seqnum) + return 0; + + if (sd_id128_equal(af->header->seqnum_id, bf->header->seqnum_id)) { + + /* If this is from the same seqnum source, compare + * seqnums */ + r = CMP(af->current_seqnum, bf->current_seqnum); + if (r != 0) + return r; + + /* Wow! This is weird, different data but the same + * seqnums? Something is borked, but let's make the + * best of it and compare by time. */ + } + + if (sd_id128_equal(af->current_boot_id, bf->current_boot_id)) { + + /* If the boot id matches, compare monotonic time */ + r = CMP(af->current_monotonic, bf->current_monotonic); + if (r != 0) + return r; + } + + /* Otherwise, compare UTC time */ + r = CMP(af->current_realtime, bf->current_realtime); + if (r != 0) + return r; + + /* Finally, compare by contents */ + return CMP(af->current_xor_hash, bf->current_xor_hash); +} + +static bool check_properly_ordered(uint64_t new_offset, uint64_t old_offset, direction_t direction) { + + /* Consider it an error if any of the two offsets is uninitialized */ + if (old_offset == 0 || new_offset == 0) + return false; + + /* If we go down, the new offset must be larger than the old one. */ + return direction == DIRECTION_DOWN ? + new_offset > old_offset : + new_offset < old_offset; +} + +int journal_file_next_entry( + JournalFile *f, + uint64_t p, + direction_t direction, + Object **ret_object, + uint64_t *ret_offset) { + + uint64_t i, n, ofs; + int r; + + assert(f); + assert(f->header); + + /* FIXME: fix return value assignment. */ + + n = le64toh(READ_NOW(f->header->n_entries)); + if (n <= 0) + return 0; + + if (p == 0) + i = direction == DIRECTION_DOWN ? 0 : n - 1; + else { + r = generic_array_bisect(f, + le64toh(f->header->entry_array_offset), + le64toh(f->header->n_entries), + p, + test_object_offset, + DIRECTION_DOWN, + NULL, NULL, + &i); + if (r <= 0) + return r; + + r = bump_array_index(&i, direction, n); + if (r <= 0) + return r; + } + + /* And jump to it */ + r = generic_array_get(f, le64toh(f->header->entry_array_offset), i, direction, ret_object, &ofs); + if (r <= 0) + return r; + + /* Ensure our array is properly ordered. */ + if (p > 0 && !check_properly_ordered(ofs, p, direction)) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "%s: entry array not properly ordered at entry %" PRIu64, + f->path, i); + + if (ret_offset) + *ret_offset = ofs; + + return 1; +} + +int journal_file_next_entry_for_data( + JournalFile *f, + Object *d, + direction_t direction, + Object **ret_object, + uint64_t *ret_offset) { + + uint64_t i, n, ofs; + int r; + + assert(f); + assert(d); + assert(d->object.type == OBJECT_DATA); + + /* FIXME: fix return value assignment. */ + + n = le64toh(READ_NOW(d->data.n_entries)); + if (n <= 0) + return n; + + i = direction == DIRECTION_DOWN ? 0 : n - 1; + + r = generic_array_get_plus_one(f, + le64toh(d->data.entry_offset), + le64toh(d->data.entry_array_offset), + i, + direction, + ret_object, &ofs); + if (r <= 0) + return r; + + if (ret_offset) + *ret_offset = ofs; + + return 1; +} + +int journal_file_move_to_entry_by_offset_for_data( + JournalFile *f, + Object *d, + uint64_t p, + direction_t direction, + Object **ret, uint64_t *ret_offset) { + + assert(f); + assert(d); + assert(d->object.type == OBJECT_DATA); + + return generic_array_bisect_plus_one( + f, + le64toh(d->data.entry_offset), + le64toh(d->data.entry_array_offset), + le64toh(d->data.n_entries), + p, + test_object_offset, + direction, + ret, ret_offset, NULL); +} + +int journal_file_move_to_entry_by_monotonic_for_data( + JournalFile *f, + Object *d, + sd_id128_t boot_id, + uint64_t monotonic, + direction_t direction, + Object **ret_object, + uint64_t *ret_offset) { + + uint64_t b, z, entry_offset, entry_array_offset, n_entries; + Object *o; + int r; + + assert(f); + assert(d); + assert(d->object.type == OBJECT_DATA); + + /* Save all the required data before the data object gets invalidated. */ + entry_offset = le64toh(READ_NOW(d->data.entry_offset)); + entry_array_offset = le64toh(READ_NOW(d->data.entry_array_offset)); + n_entries = le64toh(READ_NOW(d->data.n_entries)); + + /* First, seek by time */ + r = find_data_object_by_boot_id(f, boot_id, &o, &b); + if (r <= 0) + return r; + + r = generic_array_bisect_plus_one(f, + le64toh(o->data.entry_offset), + le64toh(o->data.entry_array_offset), + le64toh(o->data.n_entries), + monotonic, + test_object_monotonic, + direction, + NULL, &z, NULL); + if (r <= 0) + return r; + + /* And now, continue seeking until we find an entry that + * exists in both bisection arrays */ + + r = journal_file_move_to_object(f, OBJECT_DATA, b, &o); + if (r < 0) + return r; + + for (;;) { + uint64_t p, q; + + r = generic_array_bisect_plus_one(f, + entry_offset, + entry_array_offset, + n_entries, + z, + test_object_offset, + direction, + NULL, &p, NULL); + if (r <= 0) + return r; + + r = generic_array_bisect_plus_one(f, + le64toh(o->data.entry_offset), + le64toh(o->data.entry_array_offset), + le64toh(o->data.n_entries), + p, + test_object_offset, + direction, + NULL, &q, NULL); + + if (r <= 0) + return r; + + if (p == q) { + if (ret_object) { + r = journal_file_move_to_object(f, OBJECT_ENTRY, q, ret_object); + if (r < 0) + return r; + } + + if (ret_offset) + *ret_offset = q; + + return 1; + } + + z = q; + } +} + +int journal_file_move_to_entry_by_seqnum_for_data( + JournalFile *f, + Object *d, + uint64_t seqnum, + direction_t direction, + Object **ret_object, + uint64_t *ret_offset) { + + assert(f); + assert(d); + assert(d->object.type == OBJECT_DATA); + + return generic_array_bisect_plus_one( + f, + le64toh(d->data.entry_offset), + le64toh(d->data.entry_array_offset), + le64toh(d->data.n_entries), + seqnum, + test_object_seqnum, + direction, + ret_object, ret_offset, NULL); +} + +int journal_file_move_to_entry_by_realtime_for_data( + JournalFile *f, + Object *d, + uint64_t realtime, + direction_t direction, + Object **ret, uint64_t *ret_offset) { + + assert(f); + assert(d); + assert(d->object.type == OBJECT_DATA); + + return generic_array_bisect_plus_one( + f, + le64toh(d->data.entry_offset), + le64toh(d->data.entry_array_offset), + le64toh(d->data.n_entries), + realtime, + test_object_realtime, + direction, + ret, ret_offset, NULL); +} + +void journal_file_dump(JournalFile *f) { + Object *o; + uint64_t p; + int r; + + assert(f); + assert(f->header); + + journal_file_print_header(f); + + p = le64toh(READ_NOW(f->header->header_size)); + while (p != 0) { + const char *s; + Compression c; + + r = journal_file_move_to_object(f, OBJECT_UNUSED, p, &o); + if (r < 0) + goto fail; + + s = journal_object_type_to_string(o->object.type); + + switch (o->object.type) { + + case OBJECT_ENTRY: + assert(s); + + printf("Type: %s seqnum=%"PRIu64" monotonic=%"PRIu64" realtime=%"PRIu64"\n", + s, + le64toh(o->entry.seqnum), + le64toh(o->entry.monotonic), + le64toh(o->entry.realtime)); + break; + + case OBJECT_TAG: + assert(s); + + printf("Type: %s seqnum=%"PRIu64" epoch=%"PRIu64"\n", + s, + le64toh(o->tag.seqnum), + le64toh(o->tag.epoch)); + break; + + default: + if (s) + printf("Type: %s \n", s); + else + printf("Type: unknown (%i)", o->object.type); + + break; + } + + c = COMPRESSION_FROM_OBJECT(o); + if (c > COMPRESSION_NONE) + printf("Flags: %s\n", + compression_to_string(c)); + + if (p == le64toh(f->header->tail_object_offset)) + p = 0; + else + p += ALIGN64(le64toh(o->object.size)); + } + + return; +fail: + log_error("File corrupt"); +} + +/* Note: the lifetime of the compound literal is the immediately surrounding block. */ +#define FORMAT_TIMESTAMP_SAFE(t) (FORMAT_TIMESTAMP(t) ?: " --- ") + +void journal_file_print_header(JournalFile *f) { + struct stat st; + + assert(f); + assert(f->header); + + printf("File path: %s\n" + "File ID: %s\n" + "Machine ID: %s\n" + "Boot ID: %s\n" + "Sequential number ID: %s\n" + "State: %s\n" + "Compatible flags:%s%s\n" + "Incompatible flags:%s%s%s%s%s%s\n" + "Header size: %"PRIu64"\n" + "Arena size: %"PRIu64"\n" + "Data hash table size: %"PRIu64"\n" + "Field hash table size: %"PRIu64"\n" + "Rotate suggested: %s\n" + "Head sequential number: %"PRIu64" (%"PRIx64")\n" + "Tail sequential number: %"PRIu64" (%"PRIx64")\n" + "Head realtime timestamp: %s (%"PRIx64")\n" + "Tail realtime timestamp: %s (%"PRIx64")\n" + "Tail monotonic timestamp: %s (%"PRIx64")\n" + "Objects: %"PRIu64"\n" + "Entry objects: %"PRIu64"\n", + f->path, + SD_ID128_TO_STRING(f->header->file_id), + SD_ID128_TO_STRING(f->header->machine_id), + SD_ID128_TO_STRING(f->header->boot_id), + SD_ID128_TO_STRING(f->header->seqnum_id), + f->header->state == STATE_OFFLINE ? "OFFLINE" : + f->header->state == STATE_ONLINE ? "ONLINE" : + f->header->state == STATE_ARCHIVED ? "ARCHIVED" : "UNKNOWN", + JOURNAL_HEADER_SEALED(f->header) ? " SEALED" : "", + (le32toh(f->header->compatible_flags) & ~HEADER_COMPATIBLE_ANY) ? " ???" : "", + JOURNAL_HEADER_COMPRESSED_XZ(f->header) ? " COMPRESSED-XZ" : "", + JOURNAL_HEADER_COMPRESSED_LZ4(f->header) ? " COMPRESSED-LZ4" : "", + JOURNAL_HEADER_COMPRESSED_ZSTD(f->header) ? " COMPRESSED-ZSTD" : "", + JOURNAL_HEADER_KEYED_HASH(f->header) ? " KEYED-HASH" : "", + JOURNAL_HEADER_COMPACT(f->header) ? " COMPACT" : "", + (le32toh(f->header->incompatible_flags) & ~HEADER_INCOMPATIBLE_ANY) ? " ???" : "", + le64toh(f->header->header_size), + le64toh(f->header->arena_size), + le64toh(f->header->data_hash_table_size) / sizeof(HashItem), + le64toh(f->header->field_hash_table_size) / sizeof(HashItem), + yes_no(journal_file_rotate_suggested(f, 0, LOG_DEBUG)), + le64toh(f->header->head_entry_seqnum), le64toh(f->header->head_entry_seqnum), + le64toh(f->header->tail_entry_seqnum), le64toh(f->header->tail_entry_seqnum), + FORMAT_TIMESTAMP_SAFE(le64toh(f->header->head_entry_realtime)), le64toh(f->header->head_entry_realtime), + FORMAT_TIMESTAMP_SAFE(le64toh(f->header->tail_entry_realtime)), le64toh(f->header->tail_entry_realtime), + FORMAT_TIMESPAN(le64toh(f->header->tail_entry_monotonic), USEC_PER_MSEC), le64toh(f->header->tail_entry_monotonic), + le64toh(f->header->n_objects), + le64toh(f->header->n_entries)); + + if (JOURNAL_HEADER_CONTAINS(f->header, n_data)) + printf("Data objects: %"PRIu64"\n" + "Data hash table fill: %.1f%%\n", + le64toh(f->header->n_data), + 100.0 * (double) le64toh(f->header->n_data) / ((double) (le64toh(f->header->data_hash_table_size) / sizeof(HashItem)))); + + if (JOURNAL_HEADER_CONTAINS(f->header, n_fields)) + printf("Field objects: %"PRIu64"\n" + "Field hash table fill: %.1f%%\n", + le64toh(f->header->n_fields), + 100.0 * (double) le64toh(f->header->n_fields) / ((double) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem)))); + + if (JOURNAL_HEADER_CONTAINS(f->header, n_tags)) + printf("Tag objects: %"PRIu64"\n", + le64toh(f->header->n_tags)); + if (JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays)) + printf("Entry array objects: %"PRIu64"\n", + le64toh(f->header->n_entry_arrays)); + + if (JOURNAL_HEADER_CONTAINS(f->header, field_hash_chain_depth)) + printf("Deepest field hash chain: %" PRIu64"\n", + f->header->field_hash_chain_depth); + + if (JOURNAL_HEADER_CONTAINS(f->header, data_hash_chain_depth)) + printf("Deepest data hash chain: %" PRIu64"\n", + f->header->data_hash_chain_depth); + + if (fstat(f->fd, &st) >= 0) + printf("Disk usage: %s\n", FORMAT_BYTES((uint64_t) st.st_blocks * 512ULL)); +} + +static int journal_file_warn_btrfs(JournalFile *f) { + unsigned attrs; + int r; + + assert(f); + + /* Before we write anything, check if the COW logic is turned + * off on btrfs. Given our write pattern that is quite + * unfriendly to COW file systems this should greatly improve + * performance on COW file systems, such as btrfs, at the + * expense of data integrity features (which shouldn't be too + * bad, given that we do our own checksumming). */ + + r = fd_is_fs_type(f->fd, BTRFS_SUPER_MAGIC); + if (r < 0) + return log_warning_errno(r, "Failed to determine if journal is on btrfs: %m"); + if (!r) + return 0; + + r = read_attr_fd(f->fd, &attrs); + if (r < 0) + return log_warning_errno(r, "Failed to read file attributes: %m"); + + if (attrs & FS_NOCOW_FL) { + log_debug("Detected btrfs file system with copy-on-write disabled, all is good."); + return 0; + } + + log_notice("Creating journal file %s on a btrfs file system, and copy-on-write is enabled. " + "This is likely to slow down journal access substantially, please consider turning " + "off the copy-on-write file attribute on the journal directory, using chattr +C.", f->path); + + return 1; +} + +static void journal_default_metrics(JournalMetrics *m, int fd, bool compact) { + struct statvfs ss; + uint64_t fs_size = 0; + + assert(m); + assert(fd >= 0); + + if (fstatvfs(fd, &ss) >= 0) + fs_size = ss.f_frsize * ss.f_blocks; + else + log_debug_errno(errno, "Failed to determine disk size: %m"); + + if (m->max_use == UINT64_MAX) { + + if (fs_size > 0) + m->max_use = CLAMP(PAGE_ALIGN(fs_size / 10), /* 10% of file system size */ + MAX_USE_LOWER, MAX_USE_UPPER); + else + m->max_use = MAX_USE_LOWER; + } else { + m->max_use = PAGE_ALIGN(m->max_use); + + if (m->max_use != 0 && m->max_use < JOURNAL_FILE_SIZE_MIN*2) + m->max_use = JOURNAL_FILE_SIZE_MIN*2; + } + + if (m->min_use == UINT64_MAX) { + if (fs_size > 0) + m->min_use = CLAMP(PAGE_ALIGN(fs_size / 50), /* 2% of file system size */ + MIN_USE_LOW, MIN_USE_HIGH); + else + m->min_use = MIN_USE_LOW; + } + + if (m->min_use > m->max_use) + m->min_use = m->max_use; + + if (m->max_size == UINT64_MAX) + m->max_size = MIN(PAGE_ALIGN(m->max_use / 8), /* 8 chunks */ + MAX_SIZE_UPPER); + else + m->max_size = PAGE_ALIGN(m->max_size); + + if (compact && m->max_size > JOURNAL_COMPACT_SIZE_MAX) + m->max_size = JOURNAL_COMPACT_SIZE_MAX; + + if (m->max_size != 0) { + if (m->max_size < JOURNAL_FILE_SIZE_MIN) + m->max_size = JOURNAL_FILE_SIZE_MIN; + + if (m->max_use != 0 && m->max_size*2 > m->max_use) + m->max_use = m->max_size*2; + } + + if (m->min_size == UINT64_MAX) + m->min_size = JOURNAL_FILE_SIZE_MIN; + else + m->min_size = CLAMP(PAGE_ALIGN(m->min_size), + JOURNAL_FILE_SIZE_MIN, + m->max_size ?: UINT64_MAX); + + if (m->keep_free == UINT64_MAX) { + if (fs_size > 0) + m->keep_free = MIN(PAGE_ALIGN(fs_size / 20), /* 5% of file system size */ + KEEP_FREE_UPPER); + else + m->keep_free = DEFAULT_KEEP_FREE; + } + + if (m->n_max_files == UINT64_MAX) + m->n_max_files = DEFAULT_N_MAX_FILES; + + log_debug("Fixed min_use=%s max_use=%s max_size=%s min_size=%s keep_free=%s n_max_files=%" PRIu64, + FORMAT_BYTES(m->min_use), + FORMAT_BYTES(m->max_use), + FORMAT_BYTES(m->max_size), + FORMAT_BYTES(m->min_size), + FORMAT_BYTES(m->keep_free), + m->n_max_files); +} + +int journal_file_open( + int fd, + const char *fname, + int open_flags, + JournalFileFlags file_flags, + mode_t mode, + uint64_t compress_threshold_bytes, + JournalMetrics *metrics, + MMapCache *mmap_cache, + JournalFile *template, + JournalFile **ret) { + + bool newly_created = false; + JournalFile *f; + void *h; + int r; + + assert(fd >= 0 || fname); + assert(mmap_cache); + assert(ret); + + if (!IN_SET((open_flags & O_ACCMODE), O_RDONLY, O_RDWR)) + return -EINVAL; + + if ((open_flags & O_ACCMODE) == O_RDONLY && FLAGS_SET(open_flags, O_CREAT)) + return -EINVAL; + + if (fname && (open_flags & O_CREAT) && !endswith(fname, ".journal")) + return -EINVAL; + + f = new(JournalFile, 1); + if (!f) + return -ENOMEM; + + *f = (JournalFile) { + .fd = fd, + .mode = mode, + .open_flags = open_flags, + .compress_threshold_bytes = compress_threshold_bytes == UINT64_MAX ? + DEFAULT_COMPRESS_THRESHOLD : + MAX(MIN_COMPRESS_THRESHOLD, compress_threshold_bytes), + }; + + if (fname) { + f->path = strdup(fname); + if (!f->path) { + r = -ENOMEM; + goto fail; + } + } else { + assert(fd >= 0); + + /* If we don't know the path, fill in something explanatory and vaguely useful */ + if (asprintf(&f->path, "/proc/self/%i", fd) < 0) { + r = -ENOMEM; + goto fail; + } + } + + f->chain_cache = ordered_hashmap_new(&uint64_hash_ops); + if (!f->chain_cache) { + r = -ENOMEM; + goto fail; + } + + if (f->fd < 0) { + /* We pass O_NONBLOCK here, so that in case somebody pointed us to some character device node or FIFO + * or so, we likely fail quickly than block for long. For regular files O_NONBLOCK has no effect, hence + * it doesn't hurt in that case. */ + + f->fd = openat_report_new(AT_FDCWD, f->path, f->open_flags|O_CLOEXEC|O_NONBLOCK, f->mode, &newly_created); + if (f->fd < 0) { + r = f->fd; + goto fail; + } + + /* fds we opened here by us should also be closed by us. */ + f->close_fd = true; + + r = fd_nonblock(f->fd, false); + if (r < 0) + goto fail; + + if (!newly_created) { + r = journal_file_fstat(f); + if (r < 0) + goto fail; + } + } else { + r = journal_file_fstat(f); + if (r < 0) + goto fail; + + /* If we just got the fd passed in, we don't really know if we created the file anew */ + newly_created = f->last_stat.st_size == 0 && journal_file_writable(f); + } + + f->cache_fd = mmap_cache_add_fd(mmap_cache, f->fd, prot_from_flags(open_flags)); + if (!f->cache_fd) { + r = -ENOMEM; + goto fail; + } + + if (newly_created) { + (void) journal_file_warn_btrfs(f); + + /* Let's attach the creation time to the journal file, so that the vacuuming code knows the age of this + * file even if the file might end up corrupted one day... Ideally we'd just use the creation time many + * file systems maintain for each file, but the API to query this is very new, hence let's emulate this + * via extended attributes. If extended attributes are not supported we'll just skip this, and rely + * solely on mtime/atime/ctime of the file. */ + (void) fd_setcrtime(f->fd, 0); + + r = journal_file_init_header(f, file_flags, template); + if (r < 0) + goto fail; + + r = journal_file_fstat(f); + if (r < 0) + goto fail; + } + + if (f->last_stat.st_size < (off_t) HEADER_SIZE_MIN) { + r = -ENODATA; + goto fail; + } + + r = mmap_cache_fd_get(f->cache_fd, CONTEXT_HEADER, true, 0, PAGE_ALIGN(sizeof(Header)), &f->last_stat, &h); + if (r == -EINVAL) { + /* Some file systems (jffs2 or p9fs) don't support mmap() properly (or only read-only + * mmap()), and return EINVAL in that case. Let's propagate that as a more recognizable error + * code. */ + r = -EAFNOSUPPORT; + goto fail; + } + if (r < 0) + goto fail; + + f->header = h; + + if (!newly_created) { + r = journal_file_verify_header(f); + if (r < 0) + goto fail; + } + +#if HAVE_GCRYPT + if (!newly_created && journal_file_writable(f) && JOURNAL_HEADER_SEALED(f->header)) { + r = journal_file_fss_load(f); + if (r < 0) + goto fail; + } +#endif + + if (journal_file_writable(f)) { + if (metrics) { + journal_default_metrics(metrics, f->fd, JOURNAL_HEADER_COMPACT(f->header)); + f->metrics = *metrics; + } else if (template) + f->metrics = template->metrics; + + r = journal_file_refresh_header(f); + if (r < 0) + goto fail; + } + +#if HAVE_GCRYPT + r = journal_file_hmac_setup(f); + if (r < 0) + goto fail; +#endif + + if (newly_created) { + r = journal_file_setup_field_hash_table(f); + if (r < 0) + goto fail; + + r = journal_file_setup_data_hash_table(f); + if (r < 0) + goto fail; + +#if HAVE_GCRYPT + r = journal_file_append_first_tag(f); + if (r < 0) + goto fail; +#endif + } + + if (mmap_cache_fd_got_sigbus(f->cache_fd)) { + r = -EIO; + goto fail; + } + + if (template && template->post_change_timer) { + r = journal_file_enable_post_change_timer( + f, + sd_event_source_get_event(template->post_change_timer), + template->post_change_timer_period); + + if (r < 0) + goto fail; + } + + /* The file is opened now successfully, thus we take possession of any passed in fd. */ + f->close_fd = true; + + if (DEBUG_LOGGING) { + static int last_seal = -1, last_compress = -1, last_keyed_hash = -1; + static uint64_t last_bytes = UINT64_MAX; + + if (last_seal != JOURNAL_HEADER_SEALED(f->header) || + last_keyed_hash != JOURNAL_HEADER_KEYED_HASH(f->header) || + last_compress != JOURNAL_FILE_COMPRESS(f) || + last_bytes != f->compress_threshold_bytes) { + + log_debug("Journal effective settings seal=%s keyed_hash=%s compress=%s compress_threshold_bytes=%s", + yes_no(JOURNAL_HEADER_SEALED(f->header)), yes_no(JOURNAL_HEADER_KEYED_HASH(f->header)), + yes_no(JOURNAL_FILE_COMPRESS(f)), FORMAT_BYTES(f->compress_threshold_bytes)); + last_seal = JOURNAL_HEADER_SEALED(f->header); + last_keyed_hash = JOURNAL_HEADER_KEYED_HASH(f->header); + last_compress = JOURNAL_FILE_COMPRESS(f); + last_bytes = f->compress_threshold_bytes; + } + } + + *ret = f; + return 0; + +fail: + if (f->cache_fd && mmap_cache_fd_got_sigbus(f->cache_fd)) + r = -EIO; + + (void) journal_file_close(f); + + if (newly_created && fd < 0) + (void) unlink(fname); + + return r; +} + +int journal_file_archive(JournalFile *f, char **ret_previous_path) { + _cleanup_free_ char *p = NULL; + + assert(f); + + if (!journal_file_writable(f)) + return -EINVAL; + + /* Is this a journal file that was passed to us as fd? If so, we synthesized a path name for it, and we refuse + * rotation, since we don't know the actual path, and couldn't rename the file hence. */ + if (path_startswith(f->path, "/proc/self/fd")) + return -EINVAL; + + if (!endswith(f->path, ".journal")) + return -EINVAL; + + if (asprintf(&p, "%.*s@" SD_ID128_FORMAT_STR "-%016"PRIx64"-%016"PRIx64".journal", + (int) strlen(f->path) - 8, f->path, + SD_ID128_FORMAT_VAL(f->header->seqnum_id), + le64toh(f->header->head_entry_seqnum), + le64toh(f->header->head_entry_realtime)) < 0) + return -ENOMEM; + + /* Try to rename the file to the archived version. If the file already was deleted, we'll get ENOENT, let's + * ignore that case. */ + if (rename(f->path, p) < 0 && errno != ENOENT) + return -errno; + + /* Sync the rename to disk */ + (void) fsync_directory_of_file(f->fd); + + if (ret_previous_path) + *ret_previous_path = f->path; + else + free(f->path); + + f->path = TAKE_PTR(p); + + /* Set as archive so offlining commits w/state=STATE_ARCHIVED. Previously we would set old_file->header->state + * to STATE_ARCHIVED directly here, but journal_file_set_offline() short-circuits when state != STATE_ONLINE, + * which would result in the rotated journal never getting fsync() called before closing. Now we simply queue + * the archive state by setting an archive bit, leaving the state as STATE_ONLINE so proper offlining + * occurs. */ + f->archive = true; + + return 0; +} + +int journal_file_dispose(int dir_fd, const char *fname) { + _cleanup_free_ char *p = NULL; + + assert(fname); + + /* Renames a journal file to *.journal~, i.e. to mark it as corrupted or otherwise uncleanly shutdown. Note that + * this is done without looking into the file or changing any of its contents. The idea is that this is called + * whenever something is suspicious and we want to move the file away and make clear that it is not accessed + * for writing anymore. */ + + if (!endswith(fname, ".journal")) + return -EINVAL; + + if (asprintf(&p, "%.*s@%016" PRIx64 "-%016" PRIx64 ".journal~", + (int) strlen(fname) - 8, fname, + now(CLOCK_REALTIME), + random_u64()) < 0) + return -ENOMEM; + + if (renameat(dir_fd, fname, dir_fd, p) < 0) + return -errno; + + return 0; +} + +int journal_file_copy_entry(JournalFile *from, JournalFile *to, Object *o, uint64_t p) { + _cleanup_free_ EntryItem *items_alloc = NULL; + EntryItem *items; + uint64_t n, m = 0, xor_hash = 0; + sd_id128_t boot_id; + dual_timestamp ts; + int r; + + assert(from); + assert(to); + assert(o); + assert(p > 0); + + if (!journal_file_writable(to)) + return -EPERM; + + ts = (dual_timestamp) { + .monotonic = le64toh(o->entry.monotonic), + .realtime = le64toh(o->entry.realtime), + }; + boot_id = o->entry.boot_id; + + n = journal_file_entry_n_items(from, o); + if (n == 0) + return 0; + + if (n < ALLOCA_MAX / sizeof(EntryItem) / 2) + items = newa(EntryItem, n); + else { + items_alloc = new(EntryItem, n); + if (!items_alloc) + return -ENOMEM; + + items = items_alloc; + } + + for (uint64_t i = 0; i < n; i++) { + uint64_t h, q; + void *data; + size_t l; + Object *u; + + q = journal_file_entry_item_object_offset(from, o, i); + r = journal_file_data_payload(from, NULL, q, NULL, 0, 0, &data, &l); + if (IN_SET(r, -EADDRNOTAVAIL, -EBADMSG)) { + log_debug_errno(r, "Entry item %"PRIu64" data object is bad, skipping over it: %m", i); + goto next; + } + if (r < 0) + return r; + assert(r > 0); + + if (l == 0) + return -EBADMSG; + + r = journal_file_append_data(to, data, l, &u, &h); + if (r < 0) + return r; + + if (JOURNAL_HEADER_KEYED_HASH(to->header)) + xor_hash ^= jenkins_hash64(data, l); + else + xor_hash ^= le64toh(u->data.hash); + + items[m++] = (EntryItem) { + .object_offset = h, + .hash = le64toh(u->data.hash), + }; + + next: + /* The above journal_file_data_payload() may clear or overwrite cached object. Hence, we need + * to re-read the object from the cache. */ + r = journal_file_move_to_object(from, OBJECT_ENTRY, p, &o); + if (r < 0) + return r; + } + + if (m == 0) + return 0; + + r = journal_file_append_entry_internal(to, &ts, &boot_id, xor_hash, items, m, NULL, NULL, NULL); + + if (mmap_cache_fd_got_sigbus(to->cache_fd)) + return -EIO; + + return r; +} + +void journal_reset_metrics(JournalMetrics *m) { + assert(m); + + /* Set everything to "pick automatic values". */ + + *m = (JournalMetrics) { + .min_use = UINT64_MAX, + .max_use = UINT64_MAX, + .min_size = UINT64_MAX, + .max_size = UINT64_MAX, + .keep_free = UINT64_MAX, + .n_max_files = UINT64_MAX, + }; +} + +int journal_file_get_cutoff_realtime_usec(JournalFile *f, usec_t *ret_from, usec_t *ret_to) { + assert(f); + assert(f->header); + assert(ret_from || ret_to); + + if (ret_from) { + if (f->header->head_entry_realtime == 0) + return -ENOENT; + + *ret_from = le64toh(f->header->head_entry_realtime); + } + + if (ret_to) { + if (f->header->tail_entry_realtime == 0) + return -ENOENT; + + *ret_to = le64toh(f->header->tail_entry_realtime); + } + + return 1; +} + +int journal_file_get_cutoff_monotonic_usec(JournalFile *f, sd_id128_t boot_id, usec_t *ret_from, usec_t *ret_to) { + Object *o; + uint64_t p; + int r; + + assert(f); + assert(ret_from || ret_to); + + /* FIXME: fix return value assignment on success with 0. */ + + r = find_data_object_by_boot_id(f, boot_id, &o, &p); + if (r <= 0) + return r; + + if (le64toh(o->data.n_entries) <= 0) + return 0; + + if (ret_from) { + r = journal_file_move_to_object(f, OBJECT_ENTRY, le64toh(o->data.entry_offset), &o); + if (r < 0) + return r; + + *ret_from = le64toh(o->entry.monotonic); + } + + if (ret_to) { + r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); + if (r < 0) + return r; + + r = generic_array_get_plus_one(f, + le64toh(o->data.entry_offset), + le64toh(o->data.entry_array_offset), + le64toh(o->data.n_entries) - 1, + DIRECTION_UP, + &o, NULL); + if (r <= 0) + return r; + + *ret_to = le64toh(o->entry.monotonic); + } + + return 1; +} + +bool journal_file_rotate_suggested(JournalFile *f, usec_t max_file_usec, int log_level) { + assert(f); + assert(f->header); + + /* If we gained new header fields we gained new features, + * hence suggest a rotation */ + if (le64toh(f->header->header_size) < sizeof(Header)) { + log_full(log_level, "%s uses an outdated header, suggesting rotation.", f->path); + return true; + } + + /* Let's check if the hash tables grew over a certain fill level (75%, borrowing this value from + * Java's hash table implementation), and if so suggest a rotation. To calculate the fill level we + * need the n_data field, which only exists in newer versions. */ + + if (JOURNAL_HEADER_CONTAINS(f->header, n_data)) + if (le64toh(f->header->n_data) * 4ULL > (le64toh(f->header->data_hash_table_size) / sizeof(HashItem)) * 3ULL) { + log_full(log_level, + "Data hash table of %s has a fill level at %.1f (%"PRIu64" of %"PRIu64" items, %llu file size, %"PRIu64" bytes per hash table item), suggesting rotation.", + f->path, + 100.0 * (double) le64toh(f->header->n_data) / ((double) (le64toh(f->header->data_hash_table_size) / sizeof(HashItem))), + le64toh(f->header->n_data), + le64toh(f->header->data_hash_table_size) / sizeof(HashItem), + (unsigned long long) f->last_stat.st_size, + f->last_stat.st_size / le64toh(f->header->n_data)); + return true; + } + + if (JOURNAL_HEADER_CONTAINS(f->header, n_fields)) + if (le64toh(f->header->n_fields) * 4ULL > (le64toh(f->header->field_hash_table_size) / sizeof(HashItem)) * 3ULL) { + log_full(log_level, + "Field hash table of %s has a fill level at %.1f (%"PRIu64" of %"PRIu64" items), suggesting rotation.", + f->path, + 100.0 * (double) le64toh(f->header->n_fields) / ((double) (le64toh(f->header->field_hash_table_size) / sizeof(HashItem))), + le64toh(f->header->n_fields), + le64toh(f->header->field_hash_table_size) / sizeof(HashItem)); + return true; + } + + /* If there are too many hash collisions somebody is most likely playing games with us. Hence, if our + * longest chain is longer than some threshold, let's suggest rotation. */ + if (JOURNAL_HEADER_CONTAINS(f->header, data_hash_chain_depth) && + le64toh(f->header->data_hash_chain_depth) > HASH_CHAIN_DEPTH_MAX) { + log_full(log_level, + "Data hash table of %s has deepest hash chain of length %" PRIu64 ", suggesting rotation.", + f->path, le64toh(f->header->data_hash_chain_depth)); + return true; + } + + if (JOURNAL_HEADER_CONTAINS(f->header, field_hash_chain_depth) && + le64toh(f->header->field_hash_chain_depth) > HASH_CHAIN_DEPTH_MAX) { + log_full(log_level, + "Field hash table of %s has deepest hash chain of length at %" PRIu64 ", suggesting rotation.", + f->path, le64toh(f->header->field_hash_chain_depth)); + return true; + } + + /* Are the data objects properly indexed by field objects? */ + if (JOURNAL_HEADER_CONTAINS(f->header, n_data) && + JOURNAL_HEADER_CONTAINS(f->header, n_fields) && + le64toh(f->header->n_data) > 0 && + le64toh(f->header->n_fields) == 0) { + log_full(log_level, + "Data objects of %s are not indexed by field objects, suggesting rotation.", + f->path); + return true; + } + + if (max_file_usec > 0) { + usec_t t, h; + + h = le64toh(f->header->head_entry_realtime); + t = now(CLOCK_REALTIME); + + if (h > 0 && t > h + max_file_usec) { + log_full(log_level, + "Oldest entry in %s is older than the configured file retention duration (%s), suggesting rotation.", + f->path, FORMAT_TIMESPAN(max_file_usec, USEC_PER_SEC)); + return true; + } + } + + return false; +} + +static const char * const journal_object_type_table[] = { + [OBJECT_UNUSED] = "unused", + [OBJECT_DATA] = "data", + [OBJECT_FIELD] = "field", + [OBJECT_ENTRY] = "entry", + [OBJECT_DATA_HASH_TABLE] = "data hash table", + [OBJECT_FIELD_HASH_TABLE] = "field hash table", + [OBJECT_ENTRY_ARRAY] = "entry array", + [OBJECT_TAG] = "tag", +}; + +DEFINE_STRING_TABLE_LOOKUP_TO_STRING(journal_object_type, ObjectType); diff --git a/src/libsystemd/sd-journal/journal-file.h b/src/libsystemd/sd-journal/journal-file.h new file mode 100644 index 0000000..1f3c80c --- /dev/null +++ b/src/libsystemd/sd-journal/journal-file.h @@ -0,0 +1,366 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ +#pragma once + +#include <fcntl.h> +#include <inttypes.h> +#include <sys/uio.h> + +#if HAVE_GCRYPT +# include <gcrypt.h> +#endif + +#include "sd-event.h" +#include "sd-id128.h" + +#include "compress.h" +#include "hashmap.h" +#include "journal-def.h" +#include "mmap-cache.h" +#include "sparse-endian.h" +#include "time-util.h" + +typedef struct JournalMetrics { + /* For all these: UINT64_MAX means "pick automatically", and 0 means "no limit enforced" */ + uint64_t max_size; /* how large journal files grow at max */ + uint64_t min_size; /* how large journal files grow at least */ + uint64_t max_use; /* how much disk space to use in total at max, keep_free permitting */ + uint64_t min_use; /* how much disk space to use in total at least, even if keep_free says not to */ + uint64_t keep_free; /* how much to keep free on disk */ + uint64_t n_max_files; /* how many files to keep around at max */ +} JournalMetrics; + +typedef enum direction { + DIRECTION_UP, + DIRECTION_DOWN +} direction_t; + +typedef enum LocationType { + /* The first and last entries, resp. */ + LOCATION_HEAD, + LOCATION_TAIL, + + /* We already read the entry we currently point to, and the + * next one to read should probably not be this one again. */ + LOCATION_DISCRETE, + + /* We should seek to the precise location specified, and + * return it, as we haven't read it yet. */ + LOCATION_SEEK +} LocationType; + +typedef enum OfflineState { + OFFLINE_JOINED, + OFFLINE_SYNCING, + OFFLINE_OFFLINING, + OFFLINE_CANCEL, + OFFLINE_AGAIN_FROM_SYNCING, + OFFLINE_AGAIN_FROM_OFFLINING, + OFFLINE_DONE +} OfflineState; + +typedef struct JournalFile { + int fd; + MMapFileDescriptor *cache_fd; + + mode_t mode; + + int open_flags; + bool close_fd:1; + bool archive:1; + + direction_t last_direction; + LocationType location_type; + uint64_t last_n_entries; + + char *path; + struct stat last_stat; + usec_t last_stat_usec; + + Header *header; + HashItem *data_hash_table; + HashItem *field_hash_table; + + uint64_t current_offset; + uint64_t current_seqnum; + uint64_t current_realtime; + uint64_t current_monotonic; + sd_id128_t current_boot_id; + uint64_t current_xor_hash; + + JournalMetrics metrics; + + sd_event_source *post_change_timer; + usec_t post_change_timer_period; + + OrderedHashmap *chain_cache; + + pthread_t offline_thread; + volatile OfflineState offline_state; + + unsigned last_seen_generation; + + uint64_t compress_threshold_bytes; +#if HAVE_COMPRESSION + void *compress_buffer; +#endif + +#if HAVE_GCRYPT + gcry_md_hd_t hmac; + bool hmac_running; + + FSSHeader *fss_file; + size_t fss_file_size; + + uint64_t fss_start_usec; + uint64_t fss_interval_usec; + + void *fsprg_state; + size_t fsprg_state_size; + + void *fsprg_seed; + size_t fsprg_seed_size; +#endif +} JournalFile; + +typedef enum JournalFileFlags { + JOURNAL_COMPRESS = 1 << 0, + JOURNAL_SEAL = 1 << 1, +} JournalFileFlags; + +typedef struct { + uint64_t object_offset; + uint64_t hash; +} EntryItem; + +int journal_file_open( + int fd, + const char *fname, + int open_flags, + JournalFileFlags file_flags, + mode_t mode, + uint64_t compress_threshold_bytes, + JournalMetrics *metrics, + MMapCache *mmap_cache, + JournalFile *template, + JournalFile **ret); + +int journal_file_set_offline_thread_join(JournalFile *f); +JournalFile* journal_file_close(JournalFile *j); +int journal_file_fstat(JournalFile *f); +DEFINE_TRIVIAL_CLEANUP_FUNC(JournalFile*, journal_file_close); + +#define ALIGN64(x) (((x) + 7ULL) & ~7ULL) +#define VALID64(x) (((x) & 7ULL) == 0ULL) + +/* Use six characters to cover the offsets common in smallish journal + * files without adding too many zeros. */ +#define OFSfmt "%06"PRIx64 + +static inline bool VALID_REALTIME(uint64_t u) { + /* This considers timestamps until the year 3112 valid. That should be plenty room... */ + return u > 0 && u < (1ULL << 55); +} + +static inline bool VALID_MONOTONIC(uint64_t u) { + /* This considers timestamps until 1142 years of runtime valid. */ + return u < (1ULL << 55); +} + +static inline bool VALID_EPOCH(uint64_t u) { + /* This allows changing the key for 1142 years, every usec. */ + return u < (1ULL << 55); +} + +#define JOURNAL_HEADER_CONTAINS(h, field) \ + (le64toh((h)->header_size) >= offsetof(Header, field) + sizeof((h)->field)) + +#define JOURNAL_HEADER_SEALED(h) \ + FLAGS_SET(le32toh((h)->compatible_flags), HEADER_COMPATIBLE_SEALED) + +#define JOURNAL_HEADER_COMPRESSED_XZ(h) \ + FLAGS_SET(le32toh((h)->incompatible_flags), HEADER_INCOMPATIBLE_COMPRESSED_XZ) + +#define JOURNAL_HEADER_COMPRESSED_LZ4(h) \ + FLAGS_SET(le32toh((h)->incompatible_flags), HEADER_INCOMPATIBLE_COMPRESSED_LZ4) + +#define JOURNAL_HEADER_COMPRESSED_ZSTD(h) \ + FLAGS_SET(le32toh((h)->incompatible_flags), HEADER_INCOMPATIBLE_COMPRESSED_ZSTD) + +#define JOURNAL_HEADER_KEYED_HASH(h) \ + FLAGS_SET(le32toh((h)->incompatible_flags), HEADER_INCOMPATIBLE_KEYED_HASH) + +#define JOURNAL_HEADER_COMPACT(h) \ + FLAGS_SET(le32toh((h)->incompatible_flags), HEADER_INCOMPATIBLE_COMPACT) + +int journal_file_move_to_object(JournalFile *f, ObjectType type, uint64_t offset, Object **ret); +int journal_file_read_object_header(JournalFile *f, ObjectType type, uint64_t offset, Object *ret); + +int journal_file_tail_end_by_pread(JournalFile *f, uint64_t *ret_offset); +int journal_file_tail_end_by_mmap(JournalFile *f, uint64_t *ret_offset); + +static inline uint64_t journal_file_entry_item_object_offset(JournalFile *f, Object *o, size_t i) { + assert(f); + assert(o); + return JOURNAL_HEADER_COMPACT(f->header) ? le32toh(o->entry.items.compact[i].object_offset) : + le64toh(o->entry.items.regular[i].object_offset); +} + +static inline size_t journal_file_entry_item_size(JournalFile *f) { + assert(f); + return JOURNAL_HEADER_COMPACT(f->header) ? sizeof_field(Object, entry.items.compact[0]) : + sizeof_field(Object, entry.items.regular[0]); +} + +uint64_t journal_file_entry_n_items(JournalFile *f, Object *o) _pure_; + +int journal_file_data_payload( + JournalFile *f, + Object *o, + uint64_t offset, + const char *field, + size_t field_length, + size_t data_threshold, + void **ret_data, + size_t *ret_size); + +static inline size_t journal_file_data_payload_offset(JournalFile *f) { + return JOURNAL_HEADER_COMPACT(f->header) + ? offsetof(Object, data.compact.payload) + : offsetof(Object, data.regular.payload); +} + +static inline uint8_t* journal_file_data_payload_field(JournalFile *f, Object *o) { + return JOURNAL_HEADER_COMPACT(f->header) ? o->data.compact.payload : o->data.regular.payload; +} + +uint64_t journal_file_entry_array_n_items(JournalFile *f, Object *o) _pure_; + +static inline uint64_t journal_file_entry_array_item(JournalFile *f, Object *o, size_t i) { + assert(f); + assert(o); + return JOURNAL_HEADER_COMPACT(f->header) ? le32toh(o->entry_array.items.compact[i]) : + le64toh(o->entry_array.items.regular[i]); +} + +static inline size_t journal_file_entry_array_item_size(JournalFile *f) { + assert(f); + return JOURNAL_HEADER_COMPACT(f->header) ? sizeof(le32_t) : sizeof(le64_t); +} + +uint64_t journal_file_hash_table_n_items(Object *o) _pure_; + +int journal_file_append_object(JournalFile *f, ObjectType type, uint64_t size, Object **ret_object, uint64_t *ret_offset); +int journal_file_append_entry( + JournalFile *f, + const dual_timestamp *ts, + const sd_id128_t *boot_id, + const struct iovec iovec[], unsigned n_iovec, + uint64_t *seqno, + Object **ret_object, + uint64_t *ret_offset); + +int journal_file_find_data_object(JournalFile *f, const void *data, uint64_t size, Object **ret_object, uint64_t *ret_offset); +int journal_file_find_data_object_with_hash(JournalFile *f, const void *data, uint64_t size, uint64_t hash, Object **ret_object, uint64_t *ret_offset); + +int journal_file_find_field_object(JournalFile *f, const void *field, uint64_t size, Object **ret_object, uint64_t *ret_offset); +int journal_file_find_field_object_with_hash(JournalFile *f, const void *field, uint64_t size, uint64_t hash, Object **ret_object, uint64_t *ret_offset); + +void journal_file_reset_location(JournalFile *f); +void journal_file_save_location(JournalFile *f, Object *o, uint64_t offset); +int journal_file_compare_locations(JournalFile *af, JournalFile *bf); +int journal_file_next_entry(JournalFile *f, uint64_t p, direction_t direction, Object **ret_object, uint64_t *ret_offset); + +int journal_file_next_entry_for_data(JournalFile *f, Object *d, direction_t direction, Object **ret_object, uint64_t *ret_offset); + +int journal_file_move_to_entry_by_offset(JournalFile *f, uint64_t p, direction_t direction, Object **ret_object, uint64_t *ret_offset); +int journal_file_move_to_entry_by_seqnum(JournalFile *f, uint64_t seqnum, direction_t direction, Object **ret_object, uint64_t *ret_offset); +int journal_file_move_to_entry_by_realtime(JournalFile *f, uint64_t realtime, direction_t direction, Object **ret_object, uint64_t *ret_offset); +int journal_file_move_to_entry_by_monotonic(JournalFile *f, sd_id128_t boot_id, uint64_t monotonic, direction_t direction, Object **ret_object, uint64_t *ret_offset); + +int journal_file_move_to_entry_by_offset_for_data(JournalFile *f, Object *d, uint64_t p, direction_t direction, Object **ret_object, uint64_t *ret_offset); +int journal_file_move_to_entry_by_seqnum_for_data(JournalFile *f, Object *d, uint64_t seqnum, direction_t direction, Object **ret_object, uint64_t *ret_offset); +int journal_file_move_to_entry_by_realtime_for_data(JournalFile *f, Object *d, uint64_t realtime, direction_t direction, Object **ret_object, uint64_t *ret_offset); +int journal_file_move_to_entry_by_monotonic_for_data(JournalFile *f, Object *d, sd_id128_t boot_id, uint64_t monotonic, direction_t direction, Object **ret_object, uint64_t *ret_offset); + +int journal_file_copy_entry(JournalFile *from, JournalFile *to, Object *o, uint64_t p); + +void journal_file_dump(JournalFile *f); +void journal_file_print_header(JournalFile *f); + +int journal_file_archive(JournalFile *f, char **ret_previous_path); +JournalFile* journal_initiate_close(JournalFile *f, Set *deferred_closes); + +int journal_file_dispose(int dir_fd, const char *fname); + +void journal_file_post_change(JournalFile *f); +int journal_file_enable_post_change_timer(JournalFile *f, sd_event *e, usec_t t); + +void journal_reset_metrics(JournalMetrics *m); + +int journal_file_get_cutoff_realtime_usec(JournalFile *f, usec_t *ret_from, usec_t *ret_to); +int journal_file_get_cutoff_monotonic_usec(JournalFile *f, sd_id128_t boot, usec_t *ret_from, usec_t *ret_to); + +bool journal_file_rotate_suggested(JournalFile *f, usec_t max_file_usec, int log_level); + +int journal_file_map_data_hash_table(JournalFile *f); +int journal_file_map_field_hash_table(JournalFile *f); + +static inline bool JOURNAL_FILE_COMPRESS(JournalFile *f) { + assert(f); + return JOURNAL_HEADER_COMPRESSED_XZ(f->header) || JOURNAL_HEADER_COMPRESSED_LZ4(f->header) || + JOURNAL_HEADER_COMPRESSED_ZSTD(f->header); +} + +uint64_t journal_file_hash_data(JournalFile *f, const void *data, size_t sz); + +bool journal_field_valid(const char *p, size_t l, bool allow_protected); + +const char* journal_object_type_to_string(ObjectType type) _const_; + +static inline Compression COMPRESSION_FROM_OBJECT(const Object *o) { + assert(o); + + switch (o->object.flags & _OBJECT_COMPRESSED_MASK) { + case 0: + return COMPRESSION_NONE; + case OBJECT_COMPRESSED_XZ: + return COMPRESSION_XZ; + case OBJECT_COMPRESSED_LZ4: + return COMPRESSION_LZ4; + case OBJECT_COMPRESSED_ZSTD: + return COMPRESSION_ZSTD; + default: + return _COMPRESSION_INVALID; + } +} + +static inline uint8_t COMPRESSION_TO_OBJECT_FLAG(Compression c) { + switch (c) { + case COMPRESSION_XZ: + return OBJECT_COMPRESSED_XZ; + case COMPRESSION_LZ4: + return OBJECT_COMPRESSED_LZ4; + case COMPRESSION_ZSTD: + return OBJECT_COMPRESSED_ZSTD; + default: + return 0; + } +} + +static inline uint32_t COMPRESSION_TO_HEADER_INCOMPATIBLE_FLAG(Compression c) { + switch (c) { + case COMPRESSION_XZ: + return HEADER_INCOMPATIBLE_COMPRESSED_XZ; + case COMPRESSION_LZ4: + return HEADER_INCOMPATIBLE_COMPRESSED_LZ4; + case COMPRESSION_ZSTD: + return HEADER_INCOMPATIBLE_COMPRESSED_ZSTD; + default: + return 0; + } +} + +static inline bool journal_file_writable(JournalFile *f) { + assert(f); + return (f->open_flags & O_ACCMODE) != O_RDONLY; +} diff --git a/src/libsystemd/sd-journal/journal-internal.h b/src/libsystemd/sd-journal/journal-internal.h new file mode 100644 index 0000000..388c5b0 --- /dev/null +++ b/src/libsystemd/sd-journal/journal-internal.h @@ -0,0 +1,139 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ +#pragma once + +#include <inttypes.h> +#include <stdbool.h> +#include <sys/types.h> + +#include "sd-id128.h" +#include "sd-journal.h" + +#include "hashmap.h" +#include "journal-def.h" +#include "journal-file.h" +#include "list.h" +#include "set.h" + +#define JOURNAL_FILES_MAX 7168u + +typedef struct Match Match; +typedef struct Location Location; +typedef struct Directory Directory; + +typedef enum MatchType { + MATCH_DISCRETE, + MATCH_OR_TERM, + MATCH_AND_TERM +} MatchType; + +struct Match { + MatchType type; + Match *parent; + LIST_FIELDS(Match, matches); + + /* For concrete matches */ + char *data; + size_t size; + uint64_t hash; /* old-style jenkins hash. New-style siphash is different per file, hence won't be cached here */ + + /* For terms */ + LIST_HEAD(Match, matches); +}; + +struct Location { + LocationType type; + + bool seqnum_set:1; + bool realtime_set:1; + bool monotonic_set:1; + bool xor_hash_set:1; + + uint64_t seqnum; + sd_id128_t seqnum_id; + + uint64_t realtime; + + uint64_t monotonic; + sd_id128_t boot_id; + + uint64_t xor_hash; +}; + +struct Directory { + char *path; + int wd; + bool is_root; + unsigned last_seen_generation; +}; + +struct sd_journal { + int toplevel_fd; + + char *path; + char *prefix; + char *namespace; + + OrderedHashmap *files; + IteratedCache *files_cache; + MMapCache *mmap; + + Location current_location; + + JournalFile *current_file; + uint64_t current_field; + + Match *level0, *level1, *level2; + + pid_t original_pid; + + int inotify_fd; + unsigned current_invalidate_counter, last_invalidate_counter; + usec_t last_process_usec; + unsigned generation; + + /* Iterating through unique fields and their data values */ + char *unique_field; + JournalFile *unique_file; + uint64_t unique_offset; + + /* Iterating through known fields */ + JournalFile *fields_file; + uint64_t fields_offset; + uint64_t fields_hash_table_index; + char *fields_buffer; + + int flags; + + bool on_network:1; + bool no_new_files:1; + bool no_inotify:1; + bool unique_file_lost:1; /* File we were iterating over got + removed, and there were no more + files, so sd_j_enumerate_unique + will return a value equal to 0. */ + bool fields_file_lost:1; + bool has_runtime_files:1; + bool has_persistent_files:1; + + size_t data_threshold; + + Hashmap *directories_by_path; + Hashmap *directories_by_wd; + + Hashmap *errors; +}; + +char *journal_make_match_string(sd_journal *j); +void journal_print_header(sd_journal *j); + +#define JOURNAL_FOREACH_DATA_RETVAL(j, data, l, retval) \ + for (sd_journal_restart_data(j); ((retval) = sd_journal_enumerate_data((j), &(data), &(l))) > 0; ) + +/* All errors that we might encounter while extracting a field that are not real errors, + * but only mean that the field is too large or we don't support the compression. */ +static inline bool JOURNAL_ERRNO_IS_UNAVAILABLE_FIELD(int r) { + return IN_SET(abs(r), + ENOBUFS, /* Field or decompressed field too large */ + E2BIG, /* Field too large for pointer width */ + EPROTONOSUPPORT); /* Unsupported compression */ +} diff --git a/src/libsystemd/sd-journal/journal-send.c b/src/libsystemd/sd-journal/journal-send.c new file mode 100644 index 0000000..db53dd7 --- /dev/null +++ b/src/libsystemd/sd-journal/journal-send.c @@ -0,0 +1,598 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ + +#include <errno.h> +#include <fcntl.h> +#include <printf.h> +#include <stddef.h> +#include <sys/un.h> +#include <unistd.h> +#if HAVE_VALGRIND_VALGRIND_H +#include <valgrind/valgrind.h> +#endif + +#define SD_JOURNAL_SUPPRESS_LOCATION + +#include "sd-journal.h" + +#include "alloc-util.h" +#include "errno-util.h" +#include "fd-util.h" +#include "fileio.h" +#include "io-util.h" +#include "journal-send.h" +#include "memfd-util.h" +#include "socket-util.h" +#include "stdio-util.h" +#include "string-util.h" +#include "tmpfile-util.h" + +#define SNDBUF_SIZE (8*1024*1024) + +#define ALLOCA_CODE_FUNC(f, func) \ + do { \ + size_t _fl; \ + const char *_func = (func); \ + char **_f = &(f); \ + _fl = strlen(_func) + 1; \ + *_f = newa(char, _fl + 10); \ + memcpy(*_f, "CODE_FUNC=", 10); \ + memcpy(*_f + 10, _func, _fl); \ + } while (false) + +/* We open a single fd, and we'll share it with the current process, + * all its threads, and all its subprocesses. This means we need to + * initialize it atomically, and need to operate on it atomically + * never assuming we are the only user */ +static int fd_plus_one = 0; + +static int journal_fd(void) { + int fd; + +retry: + if (fd_plus_one > 0) + return fd_plus_one - 1; + + fd = socket(AF_UNIX, SOCK_DGRAM|SOCK_CLOEXEC, 0); + if (fd < 0) + return -errno; + + fd_inc_sndbuf(fd, SNDBUF_SIZE); + + if (!__atomic_compare_exchange_n(&fd_plus_one, &(int){0}, fd+1, + false, __ATOMIC_SEQ_CST, __ATOMIC_SEQ_CST)) { + safe_close(fd); + goto retry; + } + + return fd; +} + +int journal_fd_nonblock(bool nonblock) { + int r; + + r = journal_fd(); + if (r < 0) + return r; + + return fd_nonblock(r, nonblock); +} + +#if VALGRIND +void close_journal_fd(void) { + /* Be nice to valgrind. This is not atomic. This must be used only in tests. */ + + if (!RUNNING_ON_VALGRIND) + return; + + if (getpid() != gettid()) + return; + + if (fd_plus_one <= 0) + return; + + safe_close(fd_plus_one - 1); + fd_plus_one = 0; +} +#endif + +_public_ int sd_journal_print(int priority, const char *format, ...) { + int r; + va_list ap; + + va_start(ap, format); + r = sd_journal_printv(priority, format, ap); + va_end(ap); + + return r; +} + +_public_ int sd_journal_printv(int priority, const char *format, va_list ap) { + char p[STRLEN("PRIORITY=") + DECIMAL_STR_MAX(int) + 1]; + char sbuf[LINE_MAX + 8] = "MESSAGE="; + struct iovec iov[2]; + int len; + va_list aq; + char *buffer = sbuf; + + assert_return(priority >= 0, -EINVAL); + assert_return(priority <= 7, -EINVAL); + assert_return(format, -EINVAL); + + xsprintf(p, "PRIORITY=%i", priority & LOG_PRIMASK); + + va_copy(aq, ap); + len = vsnprintf(buffer + 8, LINE_MAX, format, aq); + va_end(aq); + + if (len >= (int)LONG_LINE_MAX - 8) + return -ENOBUFS; + + /* Allocate large buffer to accommodate big message */ + if (len >= LINE_MAX) { + buffer = alloca_safe(len + 9); + memcpy(buffer, "MESSAGE=", 8); + assert_se(vsnprintf(buffer + 8, len + 1, format, ap) == len); + } + + /* Strip trailing whitespace, keep prefix whitespace. */ + (void) strstrip(buffer); + + /* Suppress empty lines */ + if (isempty(buffer + 8)) + return 0; + + iov[0] = IOVEC_MAKE_STRING(buffer); + iov[1] = IOVEC_MAKE_STRING(p); + + return sd_journal_sendv(iov, 2); +} + +_printf_(1, 0) static int fill_iovec_sprintf(const char *format, va_list ap, int extra, struct iovec **_iov) { + PROTECT_ERRNO; + int r, n = 0, i = 0, j; + struct iovec *iov = NULL; + + assert(_iov); + + if (extra > 0) { + n = MAX(extra * 2, extra + 4); + iov = malloc0(n * sizeof(struct iovec)); + if (!iov) { + r = -ENOMEM; + goto fail; + } + + i = extra; + } + + while (format) { + struct iovec *c; + char *buffer; + va_list aq; + + if (i >= n) { + n = MAX(i*2, 4); + c = reallocarray(iov, n, sizeof(struct iovec)); + if (!c) { + r = -ENOMEM; + goto fail; + } + + iov = c; + } + + va_copy(aq, ap); + if (vasprintf(&buffer, format, aq) < 0) { + va_end(aq); + r = -ENOMEM; + goto fail; + } + va_end(aq); + + VA_FORMAT_ADVANCE(format, ap); + + (void) strstrip(buffer); /* strip trailing whitespace, keep prefixing whitespace */ + + iov[i++] = IOVEC_MAKE_STRING(buffer); + + format = va_arg(ap, char *); + } + + *_iov = iov; + + return i; + +fail: + for (j = 0; j < i; j++) + free(iov[j].iov_base); + + free(iov); + + return r; +} + +_public_ int sd_journal_send(const char *format, ...) { + int r, i, j; + va_list ap; + struct iovec *iov = NULL; + + va_start(ap, format); + i = fill_iovec_sprintf(format, ap, 0, &iov); + va_end(ap); + + if (_unlikely_(i < 0)) { + r = i; + goto finish; + } + + r = sd_journal_sendv(iov, i); + +finish: + for (j = 0; j < i; j++) + free(iov[j].iov_base); + + free(iov); + + return r; +} + +_public_ int sd_journal_sendv(const struct iovec *iov, int n) { + PROTECT_ERRNO; + int fd, r; + _cleanup_close_ int buffer_fd = -1; + struct iovec *w; + uint64_t *l; + int i, j = 0; + static const union sockaddr_union sa = { + .un.sun_family = AF_UNIX, + .un.sun_path = "/run/systemd/journal/socket", + }; + struct msghdr mh = { + .msg_name = (struct sockaddr*) &sa.sa, + .msg_namelen = SOCKADDR_UN_LEN(sa.un), + }; + ssize_t k; + bool have_syslog_identifier = false; + bool seal = true; + + assert_return(iov, -EINVAL); + assert_return(n > 0, -EINVAL); + + w = newa(struct iovec, n * 5 + 3); + l = newa(uint64_t, n); + + for (i = 0; i < n; i++) { + char *c, *nl; + + if (_unlikely_(!iov[i].iov_base || iov[i].iov_len <= 1)) + return -EINVAL; + + c = memchr(iov[i].iov_base, '=', iov[i].iov_len); + if (_unlikely_(!c || c == iov[i].iov_base)) + return -EINVAL; + + have_syslog_identifier = have_syslog_identifier || + (c == (char *) iov[i].iov_base + 17 && + startswith(iov[i].iov_base, "SYSLOG_IDENTIFIER")); + + nl = memchr(iov[i].iov_base, '\n', iov[i].iov_len); + if (nl) { + if (_unlikely_(nl < c)) + return -EINVAL; + + /* Already includes a newline? Bummer, then + * let's write the variable name, then a + * newline, then the size (64bit LE), followed + * by the data and a final newline */ + + w[j++] = IOVEC_MAKE(iov[i].iov_base, c - (char*) iov[i].iov_base); + w[j++] = IOVEC_MAKE_STRING("\n"); + + l[i] = htole64(iov[i].iov_len - (c - (char*) iov[i].iov_base) - 1); + w[j++] = IOVEC_MAKE(&l[i], sizeof(uint64_t)); + + w[j++] = IOVEC_MAKE(c + 1, iov[i].iov_len - (c - (char*) iov[i].iov_base) - 1); + } else + /* Nothing special? Then just add the line and + * append a newline */ + w[j++] = iov[i]; + + w[j++] = IOVEC_MAKE_STRING("\n"); + } + + if (!have_syslog_identifier && + string_is_safe(program_invocation_short_name)) { + + /* Implicitly add program_invocation_short_name, if it + * is not set explicitly. We only do this for + * program_invocation_short_name, and nothing else + * since everything else is much nicer to retrieve + * from the outside. */ + + w[j++] = IOVEC_MAKE_STRING("SYSLOG_IDENTIFIER="); + w[j++] = IOVEC_MAKE_STRING(program_invocation_short_name); + w[j++] = IOVEC_MAKE_STRING("\n"); + } + + fd = journal_fd(); + if (_unlikely_(fd < 0)) + return fd; + + mh.msg_iov = w; + mh.msg_iovlen = j; + + k = sendmsg(fd, &mh, MSG_NOSIGNAL); + if (k >= 0) + return 0; + + /* Fail silently if the journal is not available */ + if (errno == ENOENT) + return 0; + + if (!IN_SET(errno, EMSGSIZE, ENOBUFS, EAGAIN)) + return -errno; + + /* Message doesn't fit... Let's dump the data in a memfd or + * temporary file and just pass a file descriptor of it to the + * other side. + * + * For the temporary files we use /dev/shm instead of /tmp + * here, since we want this to be a tmpfs, and one that is + * available from early boot on and where unprivileged users + * can create files. */ + buffer_fd = memfd_new(NULL); + if (buffer_fd < 0) { + if (buffer_fd == -ENOSYS) { + buffer_fd = open_tmpfile_unlinkable("/dev/shm", O_RDWR | O_CLOEXEC); + if (buffer_fd < 0) + return buffer_fd; + + seal = false; + } else + return buffer_fd; + } + + n = writev(buffer_fd, w, j); + if (n < 0) + return -errno; + + if (seal) { + r = memfd_set_sealed(buffer_fd); + if (r < 0) + return r; + } + + r = send_one_fd_sa(fd, buffer_fd, mh.msg_name, mh.msg_namelen, 0); + if (r == -ENOENT) + /* Fail silently if the journal is not available */ + return 0; + return r; +} + +static int fill_iovec_perror_and_send(const char *message, int skip, struct iovec iov[]) { + PROTECT_ERRNO; + size_t n, k; + + k = isempty(message) ? 0 : strlen(message) + 2; + n = 8 + k + 256 + 1; + + for (;;) { + char buffer[n]; + char* j; + + errno = 0; + j = strerror_r(_saved_errno_, buffer + 8 + k, n - 8 - k); + if (errno == 0) { + char error[STRLEN("ERRNO=") + DECIMAL_STR_MAX(int) + 1]; + + if (j != buffer + 8 + k) + memmove(buffer + 8 + k, j, strlen(j)+1); + + memcpy(buffer, "MESSAGE=", 8); + + if (k > 0) { + memcpy(buffer + 8, message, k - 2); + memcpy(buffer + 8 + k - 2, ": ", 2); + } + + xsprintf(error, "ERRNO=%i", _saved_errno_); + + assert_cc(3 == LOG_ERR); + iov[skip+0] = IOVEC_MAKE_STRING("PRIORITY=3"); + iov[skip+1] = IOVEC_MAKE_STRING(buffer); + iov[skip+2] = IOVEC_MAKE_STRING(error); + + return sd_journal_sendv(iov, skip + 3); + } + + if (errno != ERANGE) + return -errno; + + n *= 2; + } +} + +_public_ int sd_journal_perror(const char *message) { + struct iovec iovec[3]; + + return fill_iovec_perror_and_send(message, 0, iovec); +} + +_public_ int sd_journal_stream_fd(const char *identifier, int priority, int level_prefix) { + _cleanup_close_ int fd = -1; + char *header; + size_t l; + int r; + + assert_return(priority >= 0, -EINVAL); + assert_return(priority <= 7, -EINVAL); + + fd = socket(AF_UNIX, SOCK_STREAM|SOCK_CLOEXEC, 0); + if (fd < 0) + return -errno; + + r = connect_unix_path(fd, AT_FDCWD, "/run/systemd/journal/stdout"); + if (r < 0) + return r; + + if (shutdown(fd, SHUT_RD) < 0) + return -errno; + + (void) fd_inc_sndbuf(fd, SNDBUF_SIZE); + + identifier = strempty(identifier); + + l = strlen(identifier); + header = newa(char, l + 1 + 1 + 2 + 2 + 2 + 2 + 2); + + memcpy(header, identifier, l); + header[l++] = '\n'; + header[l++] = '\n'; /* unit id */ + header[l++] = '0' + priority; + header[l++] = '\n'; + header[l++] = '0' + !!level_prefix; + header[l++] = '\n'; + header[l++] = '0'; + header[l++] = '\n'; + header[l++] = '0'; + header[l++] = '\n'; + header[l++] = '0'; + header[l++] = '\n'; + + r = loop_write(fd, header, l, false); + if (r < 0) + return r; + + return TAKE_FD(fd); +} + +_public_ int sd_journal_print_with_location(int priority, const char *file, const char *line, const char *func, const char *format, ...) { + int r; + va_list ap; + + va_start(ap, format); + r = sd_journal_printv_with_location(priority, file, line, func, format, ap); + va_end(ap); + + return r; +} + +_public_ int sd_journal_printv_with_location(int priority, const char *file, const char *line, const char *func, const char *format, va_list ap) { + char p[STRLEN("PRIORITY=") + DECIMAL_STR_MAX(int) + 1]; + char sbuf[LINE_MAX + 8] = "MESSAGE="; + struct iovec iov[5]; + char *f; + int len; + char *buffer = sbuf; + va_list aq; + + assert_return(priority >= 0, -EINVAL); + assert_return(priority <= 7, -EINVAL); + assert_return(format, -EINVAL); + + xsprintf(p, "PRIORITY=%i", priority & LOG_PRIMASK); + + va_copy(aq, ap); + len = vsnprintf(buffer + 8, LINE_MAX, format, aq); + va_end(aq); + + if (len >= (int)LONG_LINE_MAX - 8) + return -ENOBUFS; + + /* Allocate large buffer to accommodate big message */ + if (len >= LINE_MAX) { + buffer = alloca_safe(len + 9); + memcpy(buffer, "MESSAGE=", 8); + assert_se(vsnprintf(buffer + 8, len + 1, format, ap) == len); + } + + /* Strip trailing whitespace, keep prefixing whitespace */ + (void) strstrip(buffer); + + /* Suppress empty lines */ + if (isempty(buffer + 8)) + return 0; + + /* func is initialized from __func__ which is not a macro, but + * a static const char[], hence cannot easily be prefixed with + * CODE_FUNC=, hence let's do it manually here. */ + ALLOCA_CODE_FUNC(f, func); + + iov[0] = IOVEC_MAKE_STRING(buffer); + iov[1] = IOVEC_MAKE_STRING(p); + iov[2] = IOVEC_MAKE_STRING(file); + iov[3] = IOVEC_MAKE_STRING(line); + iov[4] = IOVEC_MAKE_STRING(f); + + return sd_journal_sendv(iov, ELEMENTSOF(iov)); +} + +_public_ int sd_journal_send_with_location(const char *file, const char *line, const char *func, const char *format, ...) { + _cleanup_free_ struct iovec *iov = NULL; + int r, i, j; + va_list ap; + char *f; + + va_start(ap, format); + i = fill_iovec_sprintf(format, ap, 3, &iov); + va_end(ap); + + if (_unlikely_(i < 0)) { + r = i; + goto finish; + } + + ALLOCA_CODE_FUNC(f, func); + + iov[0] = IOVEC_MAKE_STRING(file); + iov[1] = IOVEC_MAKE_STRING(line); + iov[2] = IOVEC_MAKE_STRING(f); + + r = sd_journal_sendv(iov, i); + +finish: + for (j = 3; j < i; j++) + free(iov[j].iov_base); + + return r; +} + +_public_ int sd_journal_sendv_with_location( + const char *file, const char *line, + const char *func, + const struct iovec *iov, int n) { + + struct iovec *niov; + char *f; + + assert_return(iov, -EINVAL); + assert_return(n > 0, -EINVAL); + + niov = newa(struct iovec, n + 3); + memcpy(niov, iov, sizeof(struct iovec) * n); + + ALLOCA_CODE_FUNC(f, func); + + niov[n++] = IOVEC_MAKE_STRING(file); + niov[n++] = IOVEC_MAKE_STRING(line); + niov[n++] = IOVEC_MAKE_STRING(f); + + return sd_journal_sendv(niov, n); +} + +_public_ int sd_journal_perror_with_location( + const char *file, const char *line, + const char *func, + const char *message) { + + struct iovec iov[6]; + char *f; + + ALLOCA_CODE_FUNC(f, func); + + iov[0] = IOVEC_MAKE_STRING(file); + iov[1] = IOVEC_MAKE_STRING(line); + iov[2] = IOVEC_MAKE_STRING(f); + + return fill_iovec_perror_and_send(message, 3, iov); +} diff --git a/src/libsystemd/sd-journal/journal-send.h b/src/libsystemd/sd-journal/journal-send.h new file mode 100644 index 0000000..558d39a --- /dev/null +++ b/src/libsystemd/sd-journal/journal-send.h @@ -0,0 +1,12 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ +#pragma once + +#include <stdbool.h> + +int journal_fd_nonblock(bool nonblock); + +#if VALGRIND +void close_journal_fd(void); +#else +static inline void close_journal_fd(void) {} +#endif diff --git a/src/libsystemd/sd-journal/journal-vacuum.c b/src/libsystemd/sd-journal/journal-vacuum.c new file mode 100644 index 0000000..23497af --- /dev/null +++ b/src/libsystemd/sd-journal/journal-vacuum.c @@ -0,0 +1,326 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ + +#include <fcntl.h> +#include <sys/stat.h> +#include <unistd.h> + +#include "sd-id128.h" + +#include "alloc-util.h" +#include "dirent-util.h" +#include "fd-util.h" +#include "format-util.h" +#include "fs-util.h" +#include "journal-def.h" +#include "journal-file.h" +#include "journal-vacuum.h" +#include "sort-util.h" +#include "string-util.h" +#include "time-util.h" +#include "xattr-util.h" + +struct vacuum_info { + uint64_t usage; + char *filename; + + uint64_t realtime; + + sd_id128_t seqnum_id; + uint64_t seqnum; + bool have_seqnum; +}; + +static int vacuum_compare(const struct vacuum_info *a, const struct vacuum_info *b) { + int r; + + if (a->have_seqnum && b->have_seqnum && + sd_id128_equal(a->seqnum_id, b->seqnum_id)) + return CMP(a->seqnum, b->seqnum); + + r = CMP(a->realtime, b->realtime); + if (r != 0) + return r; + + if (a->have_seqnum && b->have_seqnum) + return memcmp(&a->seqnum_id, &b->seqnum_id, 16); + + return strcmp(a->filename, b->filename); +} + +static void patch_realtime( + int fd, + const char *fn, + const struct stat *st, + unsigned long long *realtime) { + + usec_t x; + + /* The timestamp was determined by the file name, but let's see if the file might actually be older + * than the file name suggested... */ + + assert(fd >= 0); + assert(fn); + assert(st); + assert(realtime); + + x = timespec_load(&st->st_ctim); + if (timestamp_is_set(x) && x < *realtime) + *realtime = x; + + x = timespec_load(&st->st_atim); + if (timestamp_is_set(x) && x < *realtime) + *realtime = x; + + x = timespec_load(&st->st_mtim); + if (timestamp_is_set(x) && x < *realtime) + *realtime = x; + + /* Let's read the original creation time, if possible. Ideally we'd just query the creation time the + * FS might provide, but unfortunately there's currently no sane API to query it. Hence let's + * implement this manually... */ + + if (fd_getcrtime_at(fd, fn, AT_SYMLINK_FOLLOW, &x) >= 0 && x < *realtime) + *realtime = x; +} + +static int journal_file_empty(int dir_fd, const char *name) { + _cleanup_close_ int fd = -1; + struct stat st; + le64_t n_entries; + ssize_t n; + + fd = openat(dir_fd, name, O_RDONLY|O_CLOEXEC|O_NOFOLLOW|O_NONBLOCK|O_NOATIME); + if (fd < 0) { + /* Maybe failed due to O_NOATIME and lack of privileges? */ + fd = openat(dir_fd, name, O_RDONLY|O_CLOEXEC|O_NOFOLLOW|O_NONBLOCK); + if (fd < 0) + return -errno; + } + + if (fstat(fd, &st) < 0) + return -errno; + + /* If an offline file doesn't even have a header we consider it empty */ + if (st.st_size < (off_t) sizeof(Header)) + return 1; + + /* If the number of entries is empty, we consider it empty, too */ + n = pread(fd, &n_entries, sizeof(n_entries), offsetof(Header, n_entries)); + if (n < 0) + return -errno; + if (n != sizeof(n_entries)) + return -EIO; + + return le64toh(n_entries) <= 0; +} + +int journal_directory_vacuum( + const char *directory, + uint64_t max_use, + uint64_t n_max_files, + usec_t max_retention_usec, + usec_t *oldest_usec, + bool verbose) { + + uint64_t sum = 0, freed = 0, n_active_files = 0; + size_t n_list = 0, i; + _cleanup_closedir_ DIR *d = NULL; + struct vacuum_info *list = NULL; + usec_t retention_limit = 0; + int r; + + assert(directory); + + if (max_use <= 0 && max_retention_usec <= 0 && n_max_files <= 0) + return 0; + + if (max_retention_usec > 0) + retention_limit = usec_sub_unsigned(now(CLOCK_REALTIME), max_retention_usec); + + d = opendir(directory); + if (!d) + return -errno; + + FOREACH_DIRENT_ALL(de, d, r = -errno; goto finish) { + unsigned long long seqnum = 0, realtime; + _cleanup_free_ char *p = NULL; + sd_id128_t seqnum_id; + bool have_seqnum; + uint64_t size; + struct stat st; + size_t q; + + if (fstatat(dirfd(d), de->d_name, &st, AT_SYMLINK_NOFOLLOW) < 0) { + log_debug_errno(errno, "Failed to stat file %s while vacuuming, ignoring: %m", de->d_name); + continue; + } + + if (!S_ISREG(st.st_mode)) + continue; + + size = 512UL * (uint64_t) st.st_blocks; + + q = strlen(de->d_name); + + if (endswith(de->d_name, ".journal")) { + + /* Vacuum archived files. Active files are + * left around */ + + if (q < 1 + 32 + 1 + 16 + 1 + 16 + 8) { + n_active_files++; + sum += size; + continue; + } + + if (de->d_name[q-8-16-1] != '-' || + de->d_name[q-8-16-1-16-1] != '-' || + de->d_name[q-8-16-1-16-1-32-1] != '@') { + n_active_files++; + sum += size; + continue; + } + + p = strdup(de->d_name); + if (!p) { + r = -ENOMEM; + goto finish; + } + + de->d_name[q-8-16-1-16-1] = 0; + if (sd_id128_from_string(de->d_name + q-8-16-1-16-1-32, &seqnum_id) < 0) { + n_active_files++; + sum += size; + continue; + } + + if (sscanf(de->d_name + q-8-16-1-16, "%16llx-%16llx.journal", &seqnum, &realtime) != 2) { + n_active_files++; + sum += size; + continue; + } + + have_seqnum = true; + + } else if (endswith(de->d_name, ".journal~")) { + unsigned long long tmp; + + /* seqnum_id won't be initialised before use below, so set to 0 */ + seqnum_id = SD_ID128_NULL; + + /* Vacuum corrupted files */ + + if (q < 1 + 16 + 1 + 16 + 8 + 1) { + n_active_files++; + sum += size; + continue; + } + + if (de->d_name[q-1-8-16-1] != '-' || + de->d_name[q-1-8-16-1-16-1] != '@') { + n_active_files++; + sum += size; + continue; + } + + p = strdup(de->d_name); + if (!p) { + r = -ENOMEM; + goto finish; + } + + if (sscanf(de->d_name + q-1-8-16-1-16, "%16llx-%16llx.journal~", &realtime, &tmp) != 2) { + n_active_files++; + sum += size; + continue; + } + + have_seqnum = false; + } else { + /* We do not vacuum unknown files! */ + log_debug("Not vacuuming unknown file %s.", de->d_name); + continue; + } + + r = journal_file_empty(dirfd(d), p); + if (r < 0) { + log_debug_errno(r, "Failed check if %s is empty, ignoring: %m", p); + continue; + } + if (r > 0) { + /* Always vacuum empty non-online files. */ + + r = unlinkat_deallocate(dirfd(d), p, 0); + if (r >= 0) { + + log_full(verbose ? LOG_INFO : LOG_DEBUG, + "Deleted empty archived journal %s/%s (%s).", directory, p, FORMAT_BYTES(size)); + + freed += size; + } else if (r != -ENOENT) + log_warning_errno(r, "Failed to delete empty archived journal %s/%s: %m", directory, p); + + continue; + } + + patch_realtime(dirfd(d), p, &st, &realtime); + + if (!GREEDY_REALLOC(list, n_list + 1)) { + r = -ENOMEM; + goto finish; + } + + list[n_list++] = (struct vacuum_info) { + .filename = TAKE_PTR(p), + .usage = size, + .seqnum = seqnum, + .realtime = realtime, + .seqnum_id = seqnum_id, + .have_seqnum = have_seqnum, + }; + + sum += size; + } + + typesafe_qsort(list, n_list, vacuum_compare); + + for (i = 0; i < n_list; i++) { + uint64_t left; + + left = n_active_files + n_list - i; + + if ((max_retention_usec <= 0 || list[i].realtime >= retention_limit) && + (max_use <= 0 || sum <= max_use) && + (n_max_files <= 0 || left <= n_max_files)) + break; + + r = unlinkat_deallocate(dirfd(d), list[i].filename, 0); + if (r >= 0) { + log_full(verbose ? LOG_INFO : LOG_DEBUG, "Deleted archived journal %s/%s (%s).", + directory, list[i].filename, FORMAT_BYTES(list[i].usage)); + freed += list[i].usage; + + if (list[i].usage < sum) + sum -= list[i].usage; + else + sum = 0; + + } else if (r != -ENOENT) + log_warning_errno(r, "Failed to delete archived journal %s/%s: %m", directory, list[i].filename); + } + + if (oldest_usec && i < n_list && (*oldest_usec == 0 || list[i].realtime < *oldest_usec)) + *oldest_usec = list[i].realtime; + + r = 0; + +finish: + for (i = 0; i < n_list; i++) + free(list[i].filename); + free(list); + + log_full(verbose ? LOG_INFO : LOG_DEBUG, "Vacuuming done, freed %s of archived journals from %s.", + FORMAT_BYTES(freed), directory); + + return r; +} diff --git a/src/libsystemd/sd-journal/journal-vacuum.h b/src/libsystemd/sd-journal/journal-vacuum.h new file mode 100644 index 0000000..d87c847 --- /dev/null +++ b/src/libsystemd/sd-journal/journal-vacuum.h @@ -0,0 +1,9 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ +#pragma once + +#include <inttypes.h> +#include <stdbool.h> + +#include "time-util.h" + +int journal_directory_vacuum(const char *directory, uint64_t max_use, uint64_t n_max_files, usec_t max_retention_usec, usec_t *oldest_usec, bool verbose); diff --git a/src/libsystemd/sd-journal/journal-verify.c b/src/libsystemd/sd-journal/journal-verify.c new file mode 100644 index 0000000..415fec9 --- /dev/null +++ b/src/libsystemd/sd-journal/journal-verify.c @@ -0,0 +1,1419 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ + +#include <fcntl.h> +#include <stddef.h> +#include <sys/mman.h> +#include <unistd.h> + +#include "alloc-util.h" +#include "compress.h" +#include "fd-util.h" +#include "fileio.h" +#include "fs-util.h" +#include "journal-authenticate.h" +#include "journal-def.h" +#include "journal-file.h" +#include "journal-verify.h" +#include "lookup3.h" +#include "macro.h" +#include "terminal-util.h" +#include "tmpfile-util.h" +#include "util.h" + +static void draw_progress(uint64_t p, usec_t *last_usec) { + unsigned n, i, j, k; + usec_t z, x; + + if (!on_tty()) + return; + + z = now(CLOCK_MONOTONIC); + x = *last_usec; + + if (x != 0 && x + 40 * USEC_PER_MSEC > z) + return; + + *last_usec = z; + + n = (3 * columns()) / 4; + j = (n * (unsigned) p) / 65535ULL; + k = n - j; + + fputs("\r", stdout); + if (colors_enabled()) + fputs("\x1B[?25l", stdout); + + fputs(ansi_highlight_green(), stdout); + + for (i = 0; i < j; i++) + fputs("\xe2\x96\x88", stdout); + + fputs(ansi_normal(), stdout); + + for (i = 0; i < k; i++) + fputs("\xe2\x96\x91", stdout); + + printf(" %3"PRIu64"%%", 100U * p / 65535U); + + fputs("\r", stdout); + if (colors_enabled()) + fputs("\x1B[?25h", stdout); + + fflush(stdout); +} + +static uint64_t scale_progress(uint64_t scale, uint64_t p, uint64_t m) { + /* Calculates scale * p / m, but handles m == 0 safely, and saturates. + * Currently all callers use m >= 1, but we keep the check to be defensive. + */ + + if (p >= m || m == 0) + return scale; + + return scale * p / m; +} + +static void flush_progress(void) { + unsigned n, i; + + if (!on_tty()) + return; + + n = (3 * columns()) / 4; + + putchar('\r'); + + for (i = 0; i < n + 5; i++) + putchar(' '); + + putchar('\r'); + fflush(stdout); +} + +#define debug(_offset, _fmt, ...) do { \ + flush_progress(); \ + log_debug(OFSfmt": " _fmt, _offset, ##__VA_ARGS__); \ + } while (0) + +#define warning(_offset, _fmt, ...) do { \ + flush_progress(); \ + log_warning(OFSfmt": " _fmt, _offset, ##__VA_ARGS__); \ + } while (0) + +#define error(_offset, _fmt, ...) do { \ + flush_progress(); \ + log_error(OFSfmt": " _fmt, (uint64_t)_offset, ##__VA_ARGS__); \ + } while (0) + +#define error_errno(_offset, error, _fmt, ...) do { \ + flush_progress(); \ + log_error_errno(error, OFSfmt": " _fmt, (uint64_t)_offset, ##__VA_ARGS__); \ + } while (0) + +static int hash_payload(JournalFile *f, Object *o, uint64_t offset, const uint8_t *src, uint64_t size, uint64_t *res_hash) { + Compression c; + int r; + + assert(o); + assert(src); + assert(res_hash); + + c = COMPRESSION_FROM_OBJECT(o); + if (c < 0) + return -EBADMSG; + if (c != COMPRESSION_NONE) { + _cleanup_free_ void *b = NULL; + size_t b_size; + + r = decompress_blob(c, src, size, &b, &b_size, 0); + if (r < 0) { + error_errno(offset, r, "%s decompression failed: %m", + compression_to_string(c)); + return r; + } + + *res_hash = journal_file_hash_data(f, b, b_size); + } else + *res_hash = journal_file_hash_data(f, src, size); + + return 0; +} + +static int journal_file_object_verify(JournalFile *f, uint64_t offset, Object *o) { + assert(f); + assert(offset); + assert(o); + + /* This does various superficial tests about the length an + * possible field values. It does not follow any references to + * other objects. */ + + if ((o->object.flags & _OBJECT_COMPRESSED_MASK) != 0 && + o->object.type != OBJECT_DATA) { + error(offset, + "Found compressed object of type %s that isn't of type data, which is not allowed.", + journal_object_type_to_string(o->object.type)); + return -EBADMSG; + } + + switch (o->object.type) { + + case OBJECT_DATA: { + uint64_t h1, h2; + int r; + + if (le64toh(o->data.entry_offset) == 0) + warning(offset, "Unused data (entry_offset==0)"); + + if ((le64toh(o->data.entry_offset) == 0) ^ (le64toh(o->data.n_entries) == 0)) { + error(offset, "Bad n_entries: %"PRIu64, le64toh(o->data.n_entries)); + return -EBADMSG; + } + + if (le64toh(o->object.size) - journal_file_data_payload_offset(f) <= 0) { + error(offset, "Bad object size (<= %zu): %"PRIu64, + journal_file_data_payload_offset(f), + le64toh(o->object.size)); + return -EBADMSG; + } + + h1 = le64toh(o->data.hash); + r = hash_payload(f, o, offset, journal_file_data_payload_field(f, o), + le64toh(o->object.size) - journal_file_data_payload_offset(f), + &h2); + if (r < 0) + return r; + + if (h1 != h2) { + error(offset, "Invalid hash (%08" PRIx64 " vs. %08" PRIx64 ")", h1, h2); + return -EBADMSG; + } + + if (!VALID64(le64toh(o->data.next_hash_offset)) || + !VALID64(le64toh(o->data.next_field_offset)) || + !VALID64(le64toh(o->data.entry_offset)) || + !VALID64(le64toh(o->data.entry_array_offset))) { + error(offset, "Invalid offset (next_hash_offset="OFSfmt", next_field_offset="OFSfmt", entry_offset="OFSfmt", entry_array_offset="OFSfmt, + le64toh(o->data.next_hash_offset), + le64toh(o->data.next_field_offset), + le64toh(o->data.entry_offset), + le64toh(o->data.entry_array_offset)); + return -EBADMSG; + } + + break; + } + + case OBJECT_FIELD: { + uint64_t h1, h2; + int r; + + if (le64toh(o->object.size) - offsetof(Object, field.payload) <= 0) { + error(offset, + "Bad field size (<= %zu): %"PRIu64, + offsetof(Object, field.payload), + le64toh(o->object.size)); + return -EBADMSG; + } + + h1 = le64toh(o->field.hash); + r = hash_payload(f, o, offset, o->field.payload, + le64toh(o->object.size) - offsetof(Object, field.payload), + &h2); + if (r < 0) + return r; + + if (h1 != h2) { + error(offset, "Invalid hash (%08" PRIx64 " vs. %08" PRIx64 ")", h1, h2); + return -EBADMSG; + } + + if (!VALID64(le64toh(o->field.next_hash_offset)) || + !VALID64(le64toh(o->field.head_data_offset))) { + error(offset, + "Invalid offset (next_hash_offset="OFSfmt", head_data_offset="OFSfmt, + le64toh(o->field.next_hash_offset), + le64toh(o->field.head_data_offset)); + return -EBADMSG; + } + break; + } + + case OBJECT_ENTRY: + if ((le64toh(o->object.size) - offsetof(Object, entry.items)) % journal_file_entry_item_size(f) != 0) { + error(offset, + "Bad entry size (<= %zu): %"PRIu64, + offsetof(Object, entry.items), + le64toh(o->object.size)); + return -EBADMSG; + } + + if ((le64toh(o->object.size) - offsetof(Object, entry.items)) / journal_file_entry_item_size(f) <= 0) { + error(offset, + "Invalid number items in entry: %"PRIu64, + (le64toh(o->object.size) - offsetof(Object, entry.items)) / journal_file_entry_item_size(f)); + return -EBADMSG; + } + + if (le64toh(o->entry.seqnum) <= 0) { + error(offset, + "Invalid entry seqnum: %"PRIx64, + le64toh(o->entry.seqnum)); + return -EBADMSG; + } + + if (!VALID_REALTIME(le64toh(o->entry.realtime))) { + error(offset, + "Invalid entry realtime timestamp: %"PRIu64, + le64toh(o->entry.realtime)); + return -EBADMSG; + } + + if (!VALID_MONOTONIC(le64toh(o->entry.monotonic))) { + error(offset, + "Invalid entry monotonic timestamp: %"PRIu64, + le64toh(o->entry.monotonic)); + return -EBADMSG; + } + + for (uint64_t i = 0; i < journal_file_entry_n_items(f, o); i++) { + if (journal_file_entry_item_object_offset(f, o, i) == 0 || + !VALID64(journal_file_entry_item_object_offset(f, o, i))) { + error(offset, + "Invalid entry item (%"PRIu64"/%"PRIu64") offset: "OFSfmt, + i, journal_file_entry_n_items(f, o), + journal_file_entry_item_object_offset(f, o, i)); + return -EBADMSG; + } + } + + break; + + case OBJECT_DATA_HASH_TABLE: + case OBJECT_FIELD_HASH_TABLE: + if ((le64toh(o->object.size) - offsetof(Object, hash_table.items)) % sizeof(HashItem) != 0 || + (le64toh(o->object.size) - offsetof(Object, hash_table.items)) / sizeof(HashItem) <= 0) { + error(offset, + "Invalid %s size: %"PRIu64, + journal_object_type_to_string(o->object.type), + le64toh(o->object.size)); + return -EBADMSG; + } + + for (uint64_t i = 0; i < journal_file_hash_table_n_items(o); i++) { + if (o->hash_table.items[i].head_hash_offset != 0 && + !VALID64(le64toh(o->hash_table.items[i].head_hash_offset))) { + error(offset, + "Invalid %s hash table item (%"PRIu64"/%"PRIu64") head_hash_offset: "OFSfmt, + journal_object_type_to_string(o->object.type), + i, journal_file_hash_table_n_items(o), + le64toh(o->hash_table.items[i].head_hash_offset)); + return -EBADMSG; + } + if (o->hash_table.items[i].tail_hash_offset != 0 && + !VALID64(le64toh(o->hash_table.items[i].tail_hash_offset))) { + error(offset, + "Invalid %s hash table item (%"PRIu64"/%"PRIu64") tail_hash_offset: "OFSfmt, + journal_object_type_to_string(o->object.type), + i, journal_file_hash_table_n_items(o), + le64toh(o->hash_table.items[i].tail_hash_offset)); + return -EBADMSG; + } + + if ((o->hash_table.items[i].head_hash_offset != 0) != + (o->hash_table.items[i].tail_hash_offset != 0)) { + error(offset, + "Invalid %s hash table item (%"PRIu64"/%"PRIu64"): head_hash_offset="OFSfmt" tail_hash_offset="OFSfmt, + journal_object_type_to_string(o->object.type), + i, journal_file_hash_table_n_items(o), + le64toh(o->hash_table.items[i].head_hash_offset), + le64toh(o->hash_table.items[i].tail_hash_offset)); + return -EBADMSG; + } + } + + break; + + case OBJECT_ENTRY_ARRAY: + if ((le64toh(o->object.size) - offsetof(Object, entry_array.items)) % journal_file_entry_array_item_size(f) != 0 || + (le64toh(o->object.size) - offsetof(Object, entry_array.items)) / journal_file_entry_array_item_size(f) <= 0) { + error(offset, + "Invalid object entry array size: %"PRIu64, + le64toh(o->object.size)); + return -EBADMSG; + } + + if (!VALID64(le64toh(o->entry_array.next_entry_array_offset))) { + error(offset, + "Invalid object entry array next_entry_array_offset: "OFSfmt, + le64toh(o->entry_array.next_entry_array_offset)); + return -EBADMSG; + } + + for (uint64_t i = 0; i < journal_file_entry_array_n_items(f, o); i++) { + uint64_t q = journal_file_entry_array_item(f, o, i); + if (q != 0 && !VALID64(q)) { + error(offset, + "Invalid object entry array item (%"PRIu64"/%"PRIu64"): "OFSfmt, + i, journal_file_entry_array_n_items(f, o), q); + return -EBADMSG; + } + } + + break; + + case OBJECT_TAG: + if (le64toh(o->object.size) != sizeof(TagObject)) { + error(offset, + "Invalid object tag size: %"PRIu64, + le64toh(o->object.size)); + return -EBADMSG; + } + + if (!VALID_EPOCH(le64toh(o->tag.epoch))) { + error(offset, + "Invalid object tag epoch: %"PRIu64, + le64toh(o->tag.epoch)); + return -EBADMSG; + } + + break; + } + + return 0; +} + +static int write_uint64(FILE *fp, uint64_t p) { + if (fwrite(&p, sizeof(p), 1, fp) != 1) + return -EIO; + + return 0; +} + +static int contains_uint64(MMapFileDescriptor *f, uint64_t n, uint64_t p) { + uint64_t a, b; + int r; + + assert(f); + + /* Bisection ... */ + + a = 0; b = n; + while (a < b) { + uint64_t c, *z; + + c = (a + b) / 2; + + r = mmap_cache_fd_get(f, 0, false, c * sizeof(uint64_t), sizeof(uint64_t), NULL, (void **) &z); + if (r < 0) + return r; + + if (*z == p) + return 1; + + if (a + 1 >= b) + return 0; + + if (p < *z) + b = c; + else + a = c; + } + + return 0; +} + +static int verify_data( + JournalFile *f, + Object *o, uint64_t p, + MMapFileDescriptor *cache_entry_fd, uint64_t n_entries, + MMapFileDescriptor *cache_entry_array_fd, uint64_t n_entry_arrays) { + + uint64_t i, n, a, last, q; + int r; + + assert(f); + assert(o); + assert(cache_entry_fd); + assert(cache_entry_array_fd); + + n = le64toh(o->data.n_entries); + a = le64toh(o->data.entry_array_offset); + + /* Entry array means at least two objects */ + if (a && n < 2) { + error(p, "Entry array present (entry_array_offset="OFSfmt", but n_entries=%"PRIu64")", a, n); + return -EBADMSG; + } + + if (n == 0) + return 0; + + /* We already checked that earlier */ + assert(o->data.entry_offset); + + last = q = le64toh(o->data.entry_offset); + if (!contains_uint64(cache_entry_fd, n_entries, q)) { + error(p, "Data object references invalid entry at "OFSfmt, q); + return -EBADMSG; + } + + r = journal_file_move_to_entry_by_offset(f, q, DIRECTION_DOWN, NULL, NULL); + if (r < 0) + return r; + if (r == 0) { + error(q, "Entry object doesn't exist in the main entry array"); + return -EBADMSG; + } + + i = 1; + while (i < n) { + uint64_t next, m, j; + + if (a == 0) { + error(p, "Array chain too short"); + return -EBADMSG; + } + + if (!contains_uint64(cache_entry_array_fd, n_entry_arrays, a)) { + error(p, "Invalid array offset "OFSfmt, a); + return -EBADMSG; + } + + r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o); + if (r < 0) + return r; + + next = le64toh(o->entry_array.next_entry_array_offset); + if (next != 0 && next <= a) { + error(p, "Array chain has cycle (jumps back from "OFSfmt" to "OFSfmt")", a, next); + return -EBADMSG; + } + + m = journal_file_entry_array_n_items(f, o); + for (j = 0; i < n && j < m; i++, j++) { + + q = journal_file_entry_array_item(f, o, j); + if (q <= last) { + error(p, "Data object's entry array not sorted (%"PRIu64" <= %"PRIu64")", q, last); + return -EBADMSG; + } + last = q; + + if (!contains_uint64(cache_entry_fd, n_entries, q)) { + error(p, "Data object references invalid entry at "OFSfmt, q); + return -EBADMSG; + } + + r = journal_file_move_to_entry_by_offset(f, q, DIRECTION_DOWN, NULL, NULL); + if (r < 0) + return r; + if (r == 0) { + error(q, "Entry object doesn't exist in the main entry array"); + return -EBADMSG; + } + + /* Pointer might have moved, reposition */ + r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o); + if (r < 0) + return r; + } + + a = next; + } + + return 0; +} + +static int verify_data_hash_table( + JournalFile *f, + MMapFileDescriptor *cache_data_fd, uint64_t n_data, + MMapFileDescriptor *cache_entry_fd, uint64_t n_entries, + MMapFileDescriptor *cache_entry_array_fd, uint64_t n_entry_arrays, + usec_t *last_usec, + bool show_progress) { + + uint64_t i, n; + int r; + + assert(f); + assert(cache_data_fd); + assert(cache_entry_fd); + assert(cache_entry_array_fd); + assert(last_usec); + + n = le64toh(f->header->data_hash_table_size) / sizeof(HashItem); + if (n <= 0) + return 0; + + r = journal_file_map_data_hash_table(f); + if (r < 0) + return log_error_errno(r, "Failed to map data hash table: %m"); + + for (i = 0; i < n; i++) { + uint64_t last = 0, p; + + if (show_progress) + draw_progress(0xC000 + scale_progress(0x3FFF, i, n), last_usec); + + p = le64toh(f->data_hash_table[i].head_hash_offset); + while (p != 0) { + Object *o; + uint64_t next; + + if (!contains_uint64(cache_data_fd, n_data, p)) { + error(p, "Invalid data object at hash entry %"PRIu64" of %"PRIu64, i, n); + return -EBADMSG; + } + + r = journal_file_move_to_object(f, OBJECT_DATA, p, &o); + if (r < 0) + return r; + + next = le64toh(o->data.next_hash_offset); + if (next != 0 && next <= p) { + error(p, "Hash chain has a cycle in hash entry %"PRIu64" of %"PRIu64, i, n); + return -EBADMSG; + } + + if (le64toh(o->data.hash) % n != i) { + error(p, "Hash value mismatch in hash entry %"PRIu64" of %"PRIu64, i, n); + return -EBADMSG; + } + + r = verify_data(f, o, p, cache_entry_fd, n_entries, cache_entry_array_fd, n_entry_arrays); + if (r < 0) + return r; + + last = p; + p = next; + } + + if (last != le64toh(f->data_hash_table[i].tail_hash_offset)) { + error(p, + "Tail hash pointer mismatch in hash table (%"PRIu64" != %"PRIu64")", + last, + le64toh(f->data_hash_table[i].tail_hash_offset)); + return -EBADMSG; + } + } + + return 0; +} + +static int data_object_in_hash_table(JournalFile *f, uint64_t hash, uint64_t p) { + uint64_t n, h, q; + int r; + assert(f); + + n = le64toh(f->header->data_hash_table_size) / sizeof(HashItem); + if (n <= 0) + return 0; + + r = journal_file_map_data_hash_table(f); + if (r < 0) + return log_error_errno(r, "Failed to map data hash table: %m"); + + h = hash % n; + + q = le64toh(f->data_hash_table[h].head_hash_offset); + while (q != 0) { + Object *o; + + if (p == q) + return 1; + + r = journal_file_move_to_object(f, OBJECT_DATA, q, &o); + if (r < 0) + return r; + + q = le64toh(o->data.next_hash_offset); + } + + return 0; +} + +static int verify_entry( + JournalFile *f, + Object *o, uint64_t p, + MMapFileDescriptor *cache_data_fd, uint64_t n_data, + bool last) { + + uint64_t i, n; + int r; + + assert(f); + assert(o); + assert(cache_data_fd); + + n = journal_file_entry_n_items(f, o); + for (i = 0; i < n; i++) { + uint64_t q; + Object *u; + + q = journal_file_entry_item_object_offset(f, o, i); + + if (!contains_uint64(cache_data_fd, n_data, q)) { + error(p, "Invalid data object of entry"); + return -EBADMSG; + } + + r = journal_file_move_to_object(f, OBJECT_DATA, q, &u); + if (r < 0) + return r; + + r = data_object_in_hash_table(f, le64toh(u->data.hash), q); + if (r < 0) + return r; + if (r == 0) { + error(p, "Data object missing from hash table"); + return -EBADMSG; + } + + /* Pointer might have moved, reposition */ + r = journal_file_move_to_object(f, OBJECT_DATA, q, &u); + if (r < 0) + return r; + + r = journal_file_move_to_entry_by_offset_for_data(f, u, p, DIRECTION_DOWN, NULL, NULL); + if (r < 0) + return r; + + /* The last entry object has a very high chance of not being referenced as journal files + * almost always run out of space during linking of entry items when trying to add a new + * entry array so let's not error in that scenario. */ + if (r == 0 && !last) { + error(p, "Entry object not referenced by linked data object at "OFSfmt, q); + return -EBADMSG; + } + } + + return 0; +} + +static int verify_entry_array( + JournalFile *f, + MMapFileDescriptor *cache_data_fd, uint64_t n_data, + MMapFileDescriptor *cache_entry_fd, uint64_t n_entries, + MMapFileDescriptor *cache_entry_array_fd, uint64_t n_entry_arrays, + usec_t *last_usec, + bool show_progress) { + + uint64_t i = 0, a, n, last = 0; + int r; + + assert(f); + assert(cache_data_fd); + assert(cache_entry_fd); + assert(cache_entry_array_fd); + assert(last_usec); + + n = le64toh(f->header->n_entries); + a = le64toh(f->header->entry_array_offset); + while (i < n) { + uint64_t next, m, j; + Object *o; + + if (show_progress) + draw_progress(0x8000 + scale_progress(0x3FFF, i, n), last_usec); + + if (a == 0) { + error(a, "Array chain too short at %"PRIu64" of %"PRIu64, i, n); + return -EBADMSG; + } + + if (!contains_uint64(cache_entry_array_fd, n_entry_arrays, a)) { + error(a, "Invalid array %"PRIu64" of %"PRIu64, i, n); + return -EBADMSG; + } + + r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o); + if (r < 0) + return r; + + next = le64toh(o->entry_array.next_entry_array_offset); + if (next != 0 && next <= a) { + error(a, "Array chain has cycle at %"PRIu64" of %"PRIu64" (jumps back from to "OFSfmt")", i, n, next); + return -EBADMSG; + } + + m = journal_file_entry_array_n_items(f, o); + for (j = 0; i < n && j < m; i++, j++) { + uint64_t p; + + p = journal_file_entry_array_item(f, o, j); + if (p <= last) { + error(a, "Entry array not sorted at %"PRIu64" of %"PRIu64, i, n); + return -EBADMSG; + } + last = p; + + if (!contains_uint64(cache_entry_fd, n_entries, p)) { + error(a, "Invalid array entry at %"PRIu64" of %"PRIu64, i, n); + return -EBADMSG; + } + + r = journal_file_move_to_object(f, OBJECT_ENTRY, p, &o); + if (r < 0) + return r; + + r = verify_entry(f, o, p, cache_data_fd, n_data, /*last=*/ i + 1 == n); + if (r < 0) + return r; + + /* Pointer might have moved, reposition */ + r = journal_file_move_to_object(f, OBJECT_ENTRY_ARRAY, a, &o); + if (r < 0) + return r; + } + + a = next; + } + + return 0; +} + +static int verify_hash_table( + Object *o, uint64_t p, uint64_t *n_hash_tables, uint64_t header_offset, uint64_t header_size) { + + assert(o); + assert(n_hash_tables); + + if (*n_hash_tables > 1) { + error(p, + "More than one %s: %" PRIu64, + journal_object_type_to_string(o->object.type), + *n_hash_tables); + return -EBADMSG; + } + + if (header_offset != p + offsetof(Object, hash_table.items)) { + error(p, + "Header offset for %s invalid (%" PRIu64 " != %" PRIu64 ")", + journal_object_type_to_string(o->object.type), + header_offset, + p + offsetof(Object, hash_table.items)); + return -EBADMSG; + } + + if (header_size != le64toh(o->object.size) - offsetof(Object, hash_table.items)) { + error(p, + "Header size for %s invalid (%" PRIu64 " != %" PRIu64 ")", + journal_object_type_to_string(o->object.type), + header_size, + le64toh(o->object.size) - offsetof(Object, hash_table.items)); + return -EBADMSG; + } + + (*n_hash_tables)++; + + return 0; +} + +int journal_file_verify( + JournalFile *f, + const char *key, + usec_t *first_contained, usec_t *last_validated, usec_t *last_contained, + bool show_progress) { + int r; + Object *o; + uint64_t p = 0, last_epoch = 0, last_tag_realtime = 0, last_sealed_realtime = 0; + + uint64_t entry_seqnum = 0, entry_monotonic = 0, entry_realtime = 0; + usec_t min_entry_realtime = USEC_INFINITY, max_entry_realtime = 0; + sd_id128_t entry_boot_id = {}; /* Unnecessary initialization to appease gcc */ + bool entry_seqnum_set = false, entry_monotonic_set = false, entry_realtime_set = false, found_main_entry_array = false; + uint64_t n_objects = 0, n_entries = 0, n_data = 0, n_fields = 0, n_data_hash_tables = 0, n_field_hash_tables = 0, n_entry_arrays = 0, n_tags = 0; + usec_t last_usec = 0; + _cleanup_close_ int data_fd = -1, entry_fd = -1, entry_array_fd = -1; + _cleanup_fclose_ FILE *data_fp = NULL, *entry_fp = NULL, *entry_array_fp = NULL; + MMapFileDescriptor *cache_data_fd = NULL, *cache_entry_fd = NULL, *cache_entry_array_fd = NULL; + unsigned i; + bool found_last = false; + const char *tmp_dir = NULL; + MMapCache *m; + +#if HAVE_GCRYPT + uint64_t last_tag = 0; +#endif + assert(f); + + if (key) { +#if HAVE_GCRYPT + r = journal_file_parse_verification_key(f, key); + if (r < 0) { + log_error("Failed to parse seed."); + return r; + } +#else + return -EOPNOTSUPP; +#endif + } else if (JOURNAL_HEADER_SEALED(f->header)) + return -ENOKEY; + + r = var_tmp_dir(&tmp_dir); + if (r < 0) { + log_error_errno(r, "Failed to determine temporary directory: %m"); + goto fail; + } + + data_fd = open_tmpfile_unlinkable(tmp_dir, O_RDWR | O_CLOEXEC); + if (data_fd < 0) { + r = log_error_errno(data_fd, "Failed to create data file: %m"); + goto fail; + } + + entry_fd = open_tmpfile_unlinkable(tmp_dir, O_RDWR | O_CLOEXEC); + if (entry_fd < 0) { + r = log_error_errno(entry_fd, "Failed to create entry file: %m"); + goto fail; + } + + entry_array_fd = open_tmpfile_unlinkable(tmp_dir, O_RDWR | O_CLOEXEC); + if (entry_array_fd < 0) { + r = log_error_errno(entry_array_fd, + "Failed to create entry array file: %m"); + goto fail; + } + + m = mmap_cache_fd_cache(f->cache_fd); + cache_data_fd = mmap_cache_add_fd(m, data_fd, PROT_READ|PROT_WRITE); + if (!cache_data_fd) { + r = log_oom(); + goto fail; + } + + cache_entry_fd = mmap_cache_add_fd(m, entry_fd, PROT_READ|PROT_WRITE); + if (!cache_entry_fd) { + r = log_oom(); + goto fail; + } + + cache_entry_array_fd = mmap_cache_add_fd(m, entry_array_fd, PROT_READ|PROT_WRITE); + if (!cache_entry_array_fd) { + r = log_oom(); + goto fail; + } + + r = take_fdopen_unlocked(&data_fd, "w+", &data_fp); + if (r < 0) { + log_error_errno(r, "Failed to open data file stream: %m"); + goto fail; + } + + r = take_fdopen_unlocked(&entry_fd, "w+", &entry_fp); + if (r < 0) { + log_error_errno(r, "Failed to open entry file stream: %m"); + goto fail; + } + + r = take_fdopen_unlocked(&entry_array_fd, "w+", &entry_array_fp); + if (r < 0) { + log_error_errno(r, "Failed to open entry array file stream: %m"); + goto fail; + } + + if (le32toh(f->header->compatible_flags) & ~HEADER_COMPATIBLE_SUPPORTED) { + log_error("Cannot verify file with unknown extensions."); + r = -EOPNOTSUPP; + goto fail; + } + + for (i = 0; i < sizeof(f->header->reserved); i++) + if (f->header->reserved[i] != 0) { + error(offsetof(Header, reserved[i]), "Reserved field is non-zero"); + r = -EBADMSG; + goto fail; + } + + /* First iteration: we go through all objects, verify the + * superficial structure, headers, hashes. */ + + p = le64toh(f->header->header_size); + for (;;) { + /* Early exit if there are no objects in the file, at all */ + if (le64toh(f->header->tail_object_offset) == 0) + break; + + if (show_progress) + draw_progress(scale_progress(0x7FFF, p, le64toh(f->header->tail_object_offset)), &last_usec); + + r = journal_file_move_to_object(f, OBJECT_UNUSED, p, &o); + if (r < 0) { + error_errno(p, r, "Invalid object: %m"); + goto fail; + } + + if (p > le64toh(f->header->tail_object_offset)) { + error(offsetof(Header, tail_object_offset), + "Invalid tail object pointer (%"PRIu64" > %"PRIu64")", + p, + le64toh(f->header->tail_object_offset)); + r = -EBADMSG; + goto fail; + } + + n_objects++; + + r = journal_file_object_verify(f, p, o); + if (r < 0) { + error_errno(p, r, "Invalid object contents: %m"); + goto fail; + } + + if (!!(o->object.flags & OBJECT_COMPRESSED_XZ) + + !!(o->object.flags & OBJECT_COMPRESSED_LZ4) + + !!(o->object.flags & OBJECT_COMPRESSED_ZSTD) > 1) { + error(p, "Object has multiple compression flags set (flags: 0x%x)", o->object.flags); + r = -EINVAL; + goto fail; + } + + if ((o->object.flags & OBJECT_COMPRESSED_XZ) && !JOURNAL_HEADER_COMPRESSED_XZ(f->header)) { + error(p, "XZ compressed object in file without XZ compression"); + r = -EBADMSG; + goto fail; + } + + if ((o->object.flags & OBJECT_COMPRESSED_LZ4) && !JOURNAL_HEADER_COMPRESSED_LZ4(f->header)) { + error(p, "LZ4 compressed object in file without LZ4 compression"); + r = -EBADMSG; + goto fail; + } + + if ((o->object.flags & OBJECT_COMPRESSED_ZSTD) && !JOURNAL_HEADER_COMPRESSED_ZSTD(f->header)) { + error(p, "ZSTD compressed object in file without ZSTD compression"); + r = -EBADMSG; + goto fail; + } + + switch (o->object.type) { + + case OBJECT_DATA: + r = write_uint64(data_fp, p); + if (r < 0) + goto fail; + + n_data++; + break; + + case OBJECT_FIELD: + n_fields++; + break; + + case OBJECT_ENTRY: + if (JOURNAL_HEADER_SEALED(f->header) && n_tags <= 0) { + error(p, "First entry before first tag"); + r = -EBADMSG; + goto fail; + } + + r = write_uint64(entry_fp, p); + if (r < 0) + goto fail; + + if (le64toh(o->entry.realtime) < last_tag_realtime) { + error(p, + "Older entry after newer tag (%"PRIu64" < %"PRIu64")", + le64toh(o->entry.realtime), + last_tag_realtime); + r = -EBADMSG; + goto fail; + } + + if (!entry_seqnum_set && + le64toh(o->entry.seqnum) != le64toh(f->header->head_entry_seqnum)) { + error(p, + "Head entry sequence number incorrect (%"PRIu64" != %"PRIu64")", + le64toh(o->entry.seqnum), + le64toh(f->header->head_entry_seqnum)); + r = -EBADMSG; + goto fail; + } + + if (entry_seqnum_set && + entry_seqnum >= le64toh(o->entry.seqnum)) { + error(p, + "Entry sequence number out of synchronization (%"PRIu64" >= %"PRIu64")", + entry_seqnum, + le64toh(o->entry.seqnum)); + r = -EBADMSG; + goto fail; + } + + entry_seqnum = le64toh(o->entry.seqnum); + entry_seqnum_set = true; + + if (entry_monotonic_set && + sd_id128_equal(entry_boot_id, o->entry.boot_id) && + entry_monotonic > le64toh(o->entry.monotonic)) { + error(p, + "Entry timestamp out of synchronization (%"PRIu64" > %"PRIu64")", + entry_monotonic, + le64toh(o->entry.monotonic)); + r = -EBADMSG; + goto fail; + } + + entry_monotonic = le64toh(o->entry.monotonic); + entry_boot_id = o->entry.boot_id; + entry_monotonic_set = true; + + if (!entry_realtime_set && + le64toh(o->entry.realtime) != le64toh(f->header->head_entry_realtime)) { + error(p, + "Head entry realtime timestamp incorrect (%"PRIu64" != %"PRIu64")", + le64toh(o->entry.realtime), + le64toh(f->header->head_entry_realtime)); + r = -EBADMSG; + goto fail; + } + + entry_realtime = le64toh(o->entry.realtime); + entry_realtime_set = true; + + max_entry_realtime = MAX(max_entry_realtime, le64toh(o->entry.realtime)); + min_entry_realtime = MIN(min_entry_realtime, le64toh(o->entry.realtime)); + + n_entries++; + break; + + case OBJECT_DATA_HASH_TABLE: + r = verify_hash_table(o, p, &n_data_hash_tables, + le64toh(f->header->data_hash_table_offset), + le64toh(f->header->data_hash_table_size)); + if (r < 0) + goto fail; + break; + + case OBJECT_FIELD_HASH_TABLE: + r = verify_hash_table(o, p, &n_field_hash_tables, + le64toh(f->header->field_hash_table_offset), + le64toh(f->header->field_hash_table_size)); + if (r < 0) + goto fail; + + break; + + case OBJECT_ENTRY_ARRAY: + r = write_uint64(entry_array_fp, p); + if (r < 0) + goto fail; + + if (p == le64toh(f->header->entry_array_offset)) { + if (found_main_entry_array) { + error(p, "More than one main entry array"); + r = -EBADMSG; + goto fail; + } + + found_main_entry_array = true; + } + + n_entry_arrays++; + break; + + case OBJECT_TAG: + if (!JOURNAL_HEADER_SEALED(f->header)) { + error(p, "Tag object in file without sealing"); + r = -EBADMSG; + goto fail; + } + + if (le64toh(o->tag.seqnum) != n_tags + 1) { + error(p, + "Tag sequence number out of synchronization (%"PRIu64" != %"PRIu64")", + le64toh(o->tag.seqnum), + n_tags + 1); + r = -EBADMSG; + goto fail; + } + + if (le64toh(o->tag.epoch) < last_epoch) { + error(p, + "Epoch sequence out of synchronization (%"PRIu64" < %"PRIu64")", + le64toh(o->tag.epoch), + last_epoch); + r = -EBADMSG; + goto fail; + } + +#if HAVE_GCRYPT + if (JOURNAL_HEADER_SEALED(f->header)) { + uint64_t q, rt, rt_end; + + debug(p, "Checking tag %"PRIu64"...", le64toh(o->tag.seqnum)); + + rt = f->fss_start_usec + le64toh(o->tag.epoch) * f->fss_interval_usec; + rt_end = usec_add(rt, f->fss_interval_usec); + if (entry_realtime_set && entry_realtime >= rt_end) { + error(p, + "tag/entry realtime timestamp out of synchronization (%"PRIu64" >= %"PRIu64")", + entry_realtime, + rt + f->fss_interval_usec); + r = -EBADMSG; + goto fail; + } + if (max_entry_realtime >= rt_end) { + error(p, + "Entry realtime (%"PRIu64", %s) is too late with respect to tag (%"PRIu64", %s)", + max_entry_realtime, FORMAT_TIMESTAMP(max_entry_realtime), + rt_end, FORMAT_TIMESTAMP(rt_end)); + r = -EBADMSG; + goto fail; + } + if (min_entry_realtime < rt) { + error(p, + "Entry realtime (%"PRIu64", %s) is too early with respect to tag (%"PRIu64", %s)", + min_entry_realtime, FORMAT_TIMESTAMP(min_entry_realtime), + rt, FORMAT_TIMESTAMP(rt)); + r = -EBADMSG; + goto fail; + } + min_entry_realtime = USEC_INFINITY; + + /* OK, now we know the epoch. So let's now set + * it, and calculate the HMAC for everything + * since the last tag. */ + r = journal_file_fsprg_seek(f, le64toh(o->tag.epoch)); + if (r < 0) + goto fail; + + r = journal_file_hmac_start(f); + if (r < 0) + goto fail; + + if (last_tag == 0) { + r = journal_file_hmac_put_header(f); + if (r < 0) + goto fail; + + q = le64toh(f->header->header_size); + } else + q = last_tag; + + while (q <= p) { + r = journal_file_move_to_object(f, OBJECT_UNUSED, q, &o); + if (r < 0) + goto fail; + + r = journal_file_hmac_put_object(f, OBJECT_UNUSED, o, q); + if (r < 0) + goto fail; + + q = q + ALIGN64(le64toh(o->object.size)); + } + + /* Position might have changed, let's reposition things */ + r = journal_file_move_to_object(f, OBJECT_UNUSED, p, &o); + if (r < 0) + goto fail; + + if (memcmp(o->tag.tag, gcry_md_read(f->hmac, 0), TAG_LENGTH) != 0) { + error(p, "Tag failed verification"); + r = -EBADMSG; + goto fail; + } + + f->hmac_running = false; + last_tag_realtime = rt; + last_sealed_realtime = entry_realtime; + } + + last_tag = p + ALIGN64(le64toh(o->object.size)); +#endif + + last_epoch = le64toh(o->tag.epoch); + + n_tags++; + break; + } + + if (p == le64toh(f->header->tail_object_offset)) { + found_last = true; + break; + } + + p = p + ALIGN64(le64toh(o->object.size)); + }; + + if (!found_last && le64toh(f->header->tail_object_offset) != 0) { + error(le64toh(f->header->tail_object_offset), + "Tail object pointer dead (%"PRIu64" != 0)", + le64toh(f->header->tail_object_offset)); + r = -EBADMSG; + goto fail; + } + + if (n_objects != le64toh(f->header->n_objects)) { + error(offsetof(Header, n_objects), + "Object number mismatch (%"PRIu64" != %"PRIu64")", + n_objects, + le64toh(f->header->n_objects)); + r = -EBADMSG; + goto fail; + } + + if (n_entries != le64toh(f->header->n_entries)) { + error(offsetof(Header, n_entries), + "Entry number mismatch (%"PRIu64" != %"PRIu64")", + n_entries, + le64toh(f->header->n_entries)); + r = -EBADMSG; + goto fail; + } + + if (JOURNAL_HEADER_CONTAINS(f->header, n_data) && + n_data != le64toh(f->header->n_data)) { + error(offsetof(Header, n_data), + "Data number mismatch (%"PRIu64" != %"PRIu64")", + n_data, + le64toh(f->header->n_data)); + r = -EBADMSG; + goto fail; + } + + if (JOURNAL_HEADER_CONTAINS(f->header, n_fields) && + n_fields != le64toh(f->header->n_fields)) { + error(offsetof(Header, n_fields), + "Field number mismatch (%"PRIu64" != %"PRIu64")", + n_fields, + le64toh(f->header->n_fields)); + r = -EBADMSG; + goto fail; + } + + if (JOURNAL_HEADER_CONTAINS(f->header, n_tags) && + n_tags != le64toh(f->header->n_tags)) { + error(offsetof(Header, n_tags), + "Tag number mismatch (%"PRIu64" != %"PRIu64")", + n_tags, + le64toh(f->header->n_tags)); + r = -EBADMSG; + goto fail; + } + + if (JOURNAL_HEADER_CONTAINS(f->header, n_entry_arrays) && + n_entry_arrays != le64toh(f->header->n_entry_arrays)) { + error(offsetof(Header, n_entry_arrays), + "Entry array number mismatch (%"PRIu64" != %"PRIu64")", + n_entry_arrays, + le64toh(f->header->n_entry_arrays)); + r = -EBADMSG; + goto fail; + } + + if (!found_main_entry_array && le64toh(f->header->entry_array_offset) != 0) { + error(0, "Missing main entry array"); + r = -EBADMSG; + goto fail; + } + + if (entry_seqnum_set && + entry_seqnum != le64toh(f->header->tail_entry_seqnum)) { + error(offsetof(Header, tail_entry_seqnum), + "Tail entry sequence number incorrect (%"PRIu64" != %"PRIu64")", + entry_seqnum, + le64toh(f->header->tail_entry_seqnum)); + r = -EBADMSG; + goto fail; + } + + if (entry_monotonic_set && + (sd_id128_equal(entry_boot_id, f->header->boot_id) && + entry_monotonic != le64toh(f->header->tail_entry_monotonic))) { + error(0, + "Invalid tail monotonic timestamp (%"PRIu64" != %"PRIu64")", + entry_monotonic, + le64toh(f->header->tail_entry_monotonic)); + r = -EBADMSG; + goto fail; + } + + if (entry_realtime_set && entry_realtime != le64toh(f->header->tail_entry_realtime)) { + error(0, + "Invalid tail realtime timestamp (%"PRIu64" != %"PRIu64")", + entry_realtime, + le64toh(f->header->tail_entry_realtime)); + r = -EBADMSG; + goto fail; + } + + if (fflush(data_fp) != 0) { + r = log_error_errno(errno, "Failed to flush data file stream: %m"); + goto fail; + } + + if (fflush(entry_fp) != 0) { + r = log_error_errno(errno, "Failed to flush entry file stream: %m"); + goto fail; + } + + if (fflush(entry_array_fp) != 0) { + r = log_error_errno(errno, "Failed to flush entry array file stream: %m"); + goto fail; + } + + /* Second iteration: we follow all objects referenced from the + * two entry points: the object hash table and the entry + * array. We also check that everything referenced (directly + * or indirectly) in the data hash table also exists in the + * entry array, and vice versa. Note that we do not care for + * unreferenced objects. We only care that everything that is + * referenced is consistent. */ + + r = verify_entry_array(f, + cache_data_fd, n_data, + cache_entry_fd, n_entries, + cache_entry_array_fd, n_entry_arrays, + &last_usec, + show_progress); + if (r < 0) + goto fail; + + r = verify_data_hash_table(f, + cache_data_fd, n_data, + cache_entry_fd, n_entries, + cache_entry_array_fd, n_entry_arrays, + &last_usec, + show_progress); + if (r < 0) + goto fail; + + if (show_progress) + flush_progress(); + + mmap_cache_fd_free(cache_data_fd); + mmap_cache_fd_free(cache_entry_fd); + mmap_cache_fd_free(cache_entry_array_fd); + + if (first_contained) + *first_contained = le64toh(f->header->head_entry_realtime); + if (last_validated) + *last_validated = last_sealed_realtime; + if (last_contained) + *last_contained = le64toh(f->header->tail_entry_realtime); + + return 0; + +fail: + if (show_progress) + flush_progress(); + + log_error("File corruption detected at %s:"OFSfmt" (of %llu bytes, %"PRIu64"%%).", + f->path, + p, + (unsigned long long) f->last_stat.st_size, + 100 * p / f->last_stat.st_size); + + if (cache_data_fd) + mmap_cache_fd_free(cache_data_fd); + + if (cache_entry_fd) + mmap_cache_fd_free(cache_entry_fd); + + if (cache_entry_array_fd) + mmap_cache_fd_free(cache_entry_array_fd); + + return r; +} diff --git a/src/libsystemd/sd-journal/journal-verify.h b/src/libsystemd/sd-journal/journal-verify.h new file mode 100644 index 0000000..5790330 --- /dev/null +++ b/src/libsystemd/sd-journal/journal-verify.h @@ -0,0 +1,6 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ +#pragma once + +#include "journal-file.h" + +int journal_file_verify(JournalFile *f, const char *key, usec_t *first_contained, usec_t *last_validated, usec_t *last_contained, bool show_progress); diff --git a/src/libsystemd/sd-journal/lookup3.c b/src/libsystemd/sd-journal/lookup3.c new file mode 100644 index 0000000..39967f2 --- /dev/null +++ b/src/libsystemd/sd-journal/lookup3.c @@ -0,0 +1,1006 @@ +/* SPDX-License-Identifier: LicenseRef-lookup3-public-domain */ +/* Slightly modified by Lennart Poettering, to avoid name clashes, and + * unexport a few functions. */ + +#include "lookup3.h" + +/* +------------------------------------------------------------------------------- +lookup3.c, by Bob Jenkins, May 2006, Public Domain. + +These are functions for producing 32-bit hashes for hash table lookup. +hashword(), hashlittle(), hashlittle2(), hashbig(), mix(), and final() +are externally useful functions. Routines to test the hash are included +if SELF_TEST is defined. You can use this free for any purpose. It's in +the public domain. It has no warranty. + +You probably want to use hashlittle(). hashlittle() and hashbig() +hash byte arrays. hashlittle() is faster than hashbig() on +little-endian machines. Intel and AMD are little-endian machines. +On second thought, you probably want hashlittle2(), which is identical to +hashlittle() except it returns two 32-bit hashes for the price of one. +You could implement hashbig2() if you wanted but I haven't bothered here. + +If you want to find a hash of, say, exactly 7 integers, do + a = i1; b = i2; c = i3; + mix(a,b,c); + a += i4; b += i5; c += i6; + mix(a,b,c); + a += i7; + final(a,b,c); +then use c as the hash value. If you have a variable length array of +4-byte integers to hash, use hashword(). If you have a byte array (like +a character string), use hashlittle(). If you have several byte arrays, or +a mix of things, see the comments above hashlittle(). + +Why is this so big? I read 12 bytes at a time into 3 4-byte integers, +then mix those integers. This is fast (you can do a lot more thorough +mixing with 12*3 instructions on 3 integers than you can with 3 instructions +on 1 byte), but shoehorning those bytes into integers efficiently is messy. +------------------------------------------------------------------------------- +*/ +/* #define SELF_TEST 1 */ + +#include <stdint.h> /* defines uint32_t etc */ +#include <stdio.h> /* defines printf for tests */ +#include <sys/param.h> /* attempt to define endianness */ +#include <time.h> /* defines time_t for timings in the test */ +#ifdef linux +# include <endian.h> /* attempt to define endianness */ +#endif + +#if __GNUC__ >= 7 +_Pragma("GCC diagnostic ignored \"-Wimplicit-fallthrough\"") +#endif + +/* + * My best guess at if you are big-endian or little-endian. This may + * need adjustment. + */ +#if (defined(__BYTE_ORDER) && defined(__LITTLE_ENDIAN) && \ + __BYTE_ORDER == __LITTLE_ENDIAN) || \ + (defined(i386) || defined(__i386__) || defined(__i486__) || \ + defined(__i586__) || defined(__i686__) || defined(vax) || defined(MIPSEL)) +# define HASH_LITTLE_ENDIAN 1 +# define HASH_BIG_ENDIAN 0 +#elif (defined(__BYTE_ORDER) && defined(__BIG_ENDIAN) && \ + __BYTE_ORDER == __BIG_ENDIAN) || \ + (defined(sparc) || defined(POWERPC) || defined(mc68000) || defined(sel)) +# define HASH_LITTLE_ENDIAN 0 +# define HASH_BIG_ENDIAN 1 +#else +# define HASH_LITTLE_ENDIAN 0 +# define HASH_BIG_ENDIAN 0 +#endif + +#define hashsize(n) ((uint32_t)1<<(n)) +#define hashmask(n) (hashsize(n)-1) +#define rot(x,k) (((x)<<(k)) | ((x)>>(32-(k)))) + +/* +------------------------------------------------------------------------------- +mix -- mix 3 32-bit values reversibly. + +This is reversible, so any information in (a,b,c) before mix() is +still in (a,b,c) after mix(). + +If four pairs of (a,b,c) inputs are run through mix(), or through +mix() in reverse, there are at least 32 bits of the output that +are sometimes the same for one pair and different for another pair. +This was tested for: +* pairs that differed by one bit, by two bits, in any combination + of top bits of (a,b,c), or in any combination of bottom bits of + (a,b,c). +* "differ" is defined as +, -, ^, or ~^. For + and -, I transformed + the output delta to a Gray code (a^(a>>1)) so a string of 1's (as + is commonly produced by subtraction) look like a single 1-bit + difference. +* the base values were pseudorandom, all zero but one bit set, or + all zero plus a counter that starts at zero. + +Some k values for my "a-=c; a^=rot(c,k); c+=b;" arrangement that +satisfy this are + 4 6 8 16 19 4 + 9 15 3 18 27 15 + 14 9 3 7 17 3 +Well, "9 15 3 18 27 15" didn't quite get 32 bits diffing +for "differ" defined as + with a one-bit base and a two-bit delta. I +used http://burtleburtle.net/bob/hash/avalanche.html to choose +the operations, constants, and arrangements of the variables. + +This does not achieve avalanche. There are input bits of (a,b,c) +that fail to affect some output bits of (a,b,c), especially of a. The +most thoroughly mixed value is c, but it doesn't really even achieve +avalanche in c. + +This allows some parallelism. Read-after-writes are good at doubling +the number of bits affected, so the goal of mixing pulls in the opposite +direction as the goal of parallelism. I did what I could. Rotates +seem to cost as much as shifts on every machine I could lay my hands +on, and rotates are much kinder to the top and bottom bits, so I used +rotates. +------------------------------------------------------------------------------- +*/ +#define mix(a,b,c) \ +{ \ + a -= c; a ^= rot(c, 4); c += b; \ + b -= a; b ^= rot(a, 6); a += c; \ + c -= b; c ^= rot(b, 8); b += a; \ + a -= c; a ^= rot(c,16); c += b; \ + b -= a; b ^= rot(a,19); a += c; \ + c -= b; c ^= rot(b, 4); b += a; \ +} + +/* +------------------------------------------------------------------------------- +final -- final mixing of 3 32-bit values (a,b,c) into c + +Pairs of (a,b,c) values differing in only a few bits will usually +produce values of c that look totally different. This was tested for +* pairs that differed by one bit, by two bits, in any combination + of top bits of (a,b,c), or in any combination of bottom bits of + (a,b,c). +* "differ" is defined as +, -, ^, or ~^. For + and -, I transformed + the output delta to a Gray code (a^(a>>1)) so a string of 1's (as + is commonly produced by subtraction) look like a single 1-bit + difference. +* the base values were pseudorandom, all zero but one bit set, or + all zero plus a counter that starts at zero. + +These constants passed: + 14 11 25 16 4 14 24 + 12 14 25 16 4 14 24 +and these came close: + 4 8 15 26 3 22 24 + 10 8 15 26 3 22 24 + 11 8 15 26 3 22 24 +------------------------------------------------------------------------------- +*/ +#define final(a,b,c) \ +{ \ + c ^= b; c -= rot(b,14); \ + a ^= c; a -= rot(c,11); \ + b ^= a; b -= rot(a,25); \ + c ^= b; c -= rot(b,16); \ + a ^= c; a -= rot(c,4); \ + b ^= a; b -= rot(a,14); \ + c ^= b; c -= rot(b,24); \ +} + +/* +-------------------------------------------------------------------- + This works on all machines. To be useful, it requires + -- that the key be an array of uint32_t's, and + -- that the length be the number of uint32_t's in the key + + The function hashword() is identical to hashlittle() on little-endian + machines, and identical to hashbig() on big-endian machines, + except that the length has to be measured in uint32_ts rather than in + bytes. hashlittle() is more complicated than hashword() only because + hashlittle() has to dance around fitting the key bytes into registers. +-------------------------------------------------------------------- +*/ +uint32_t jenkins_hashword( +const uint32_t *k, /* the key, an array of uint32_t values */ +size_t length, /* the length of the key, in uint32_ts */ +uint32_t initval) /* the previous hash, or an arbitrary value */ +{ + uint32_t a,b,c; + + /* Set up the internal state */ + a = b = c = 0xdeadbeef + (((uint32_t)length)<<2) + initval; + + /*------------------------------------------------- handle most of the key */ + while (length > 3) + { + a += k[0]; + b += k[1]; + c += k[2]; + mix(a,b,c); + length -= 3; + k += 3; + } + + /*------------------------------------------- handle the last 3 uint32_t's */ + switch(length) /* all the case statements fall through */ + { + case 3 : c+=k[2]; + case 2 : b+=k[1]; + case 1 : a+=k[0]; + final(a,b,c); + case 0: /* case 0: nothing left to add */ + break; + } + /*------------------------------------------------------ report the result */ + return c; +} + +/* +-------------------------------------------------------------------- +hashword2() -- same as hashword(), but take two seeds and return two +32-bit values. pc and pb must both be nonnull, and *pc and *pb must +both be initialized with seeds. If you pass in (*pb)==0, the output +(*pc) will be the same as the return value from hashword(). +-------------------------------------------------------------------- +*/ +void jenkins_hashword2 ( +const uint32_t *k, /* the key, an array of uint32_t values */ +size_t length, /* the length of the key, in uint32_ts */ +uint32_t *pc, /* IN: seed OUT: primary hash value */ +uint32_t *pb) /* IN: more seed OUT: secondary hash value */ +{ + uint32_t a,b,c; + + /* Set up the internal state */ + a = b = c = 0xdeadbeef + ((uint32_t)(length<<2)) + *pc; + c += *pb; + + /*------------------------------------------------- handle most of the key */ + while (length > 3) + { + a += k[0]; + b += k[1]; + c += k[2]; + mix(a,b,c); + length -= 3; + k += 3; + } + + /*------------------------------------------- handle the last 3 uint32_t's */ + switch(length) /* all the case statements fall through */ + { + case 3 : c+=k[2]; + case 2 : b+=k[1]; + case 1 : a+=k[0]; + final(a,b,c); + case 0: /* case 0: nothing left to add */ + break; + } + /*------------------------------------------------------ report the result */ + *pc=c; *pb=b; +} + +/* +------------------------------------------------------------------------------- +hashlittle() -- hash a variable-length key into a 32-bit value + k : the key (the unaligned variable-length array of bytes) + length : the length of the key, counting by bytes + initval : can be any 4-byte value +Returns a 32-bit value. Every bit of the key affects every bit of +the return value. Two keys differing by one or two bits will have +totally different hash values. + +The best hash table sizes are powers of 2. There is no need to do +mod a prime (mod is sooo slow!). If you need less than 32 bits, +use a bitmask. For example, if you need only 10 bits, do + h = (h & hashmask(10)); +In which case, the hash table should have hashsize(10) elements. + +If you are hashing n strings (uint8_t **)k, do it like this: + for (i=0, h=0; i<n; ++i) h = hashlittle( k[i], len[i], h); + +By Bob Jenkins, 2006. bob_jenkins@burtleburtle.net. You may use this +code any way you wish, private, educational, or commercial. It's free. + +Use for hash table lookup, or anything where one collision in 2^^32 is +acceptable. Do NOT use for cryptographic purposes. +------------------------------------------------------------------------------- +*/ + +uint32_t jenkins_hashlittle( const void *key, size_t length, uint32_t initval) +{ + uint32_t a,b,c; /* internal state */ + union { const void *ptr; size_t i; } u; /* needed for Mac Powerbook G4 */ + + /* Set up the internal state */ + a = b = c = 0xdeadbeef + ((uint32_t)length) + initval; + + u.ptr = key; + if (HASH_LITTLE_ENDIAN && ((u.i & 0x3) == 0)) { + const uint32_t *k = (const uint32_t *)key; /* read 32-bit chunks */ + + /*------ all but last block: aligned reads and affect 32 bits of (a,b,c) */ + while (length > 12) + { + a += k[0]; + b += k[1]; + c += k[2]; + mix(a,b,c); + length -= 12; + k += 3; + } + + /*----------------------------- handle the last (probably partial) block */ + /* + * "k[2]&0xffffff" actually reads beyond the end of the string, but + * then masks off the part it's not allowed to read. Because the + * string is aligned, the masked-off tail is in the same word as the + * rest of the string. Every machine with memory protection I've seen + * does it on word boundaries, so is OK with this. But valgrind will + * still catch it and complain. The masking trick does make the hash + * noticeably faster for short strings (like English words). + */ +#if !VALGRIND && !HAS_FEATURE_ADDRESS_SANITIZER && !HAS_FEATURE_MEMORY_SANITIZER + + switch(length) + { + case 12: c+=k[2]; b+=k[1]; a+=k[0]; break; + case 11: c+=k[2]&0xffffff; b+=k[1]; a+=k[0]; break; + case 10: c+=k[2]&0xffff; b+=k[1]; a+=k[0]; break; + case 9 : c+=k[2]&0xff; b+=k[1]; a+=k[0]; break; + case 8 : b+=k[1]; a+=k[0]; break; + case 7 : b+=k[1]&0xffffff; a+=k[0]; break; + case 6 : b+=k[1]&0xffff; a+=k[0]; break; + case 5 : b+=k[1]&0xff; a+=k[0]; break; + case 4 : a+=k[0]; break; + case 3 : a+=k[0]&0xffffff; break; + case 2 : a+=k[0]&0xffff; break; + case 1 : a+=k[0]&0xff; break; + case 0 : return c; /* zero length strings require no mixing */ + } + +#else /* make valgrind happy */ + { + const uint8_t *k8 = (const uint8_t *) k; + + switch(length) + { + case 12: c+=k[2]; b+=k[1]; a+=k[0]; break; + case 11: c+=((uint32_t)k8[10])<<16; /* fall through */ + case 10: c+=((uint32_t)k8[9])<<8; /* fall through */ + case 9 : c+=k8[8]; /* fall through */ + case 8 : b+=k[1]; a+=k[0]; break; + case 7 : b+=((uint32_t)k8[6])<<16; /* fall through */ + case 6 : b+=((uint32_t)k8[5])<<8; /* fall through */ + case 5 : b+=k8[4]; /* fall through */ + case 4 : a+=k[0]; break; + case 3 : a+=((uint32_t)k8[2])<<16; /* fall through */ + case 2 : a+=((uint32_t)k8[1])<<8; /* fall through */ + case 1 : a+=k8[0]; break; + case 0 : return c; + } + } + +#endif /* !valgrind */ + + } else if (HASH_LITTLE_ENDIAN && ((u.i & 0x1) == 0)) { + const uint16_t *k = (const uint16_t *)key; /* read 16-bit chunks */ + const uint8_t *k8; + + /*--------------- all but last block: aligned reads and different mixing */ + while (length > 12) + { + a += k[0] + (((uint32_t)k[1])<<16); + b += k[2] + (((uint32_t)k[3])<<16); + c += k[4] + (((uint32_t)k[5])<<16); + mix(a,b,c); + length -= 12; + k += 6; + } + + /*----------------------------- handle the last (probably partial) block */ + k8 = (const uint8_t *)k; + switch(length) + { + case 12: c+=k[4]+(((uint32_t)k[5])<<16); + b+=k[2]+(((uint32_t)k[3])<<16); + a+=k[0]+(((uint32_t)k[1])<<16); + break; + case 11: c+=((uint32_t)k8[10])<<16; /* fall through */ + case 10: c+=k[4]; + b+=k[2]+(((uint32_t)k[3])<<16); + a+=k[0]+(((uint32_t)k[1])<<16); + break; + case 9 : c+=k8[8]; /* fall through */ + case 8 : b+=k[2]+(((uint32_t)k[3])<<16); + a+=k[0]+(((uint32_t)k[1])<<16); + break; + case 7 : b+=((uint32_t)k8[6])<<16; /* fall through */ + case 6 : b+=k[2]; + a+=k[0]+(((uint32_t)k[1])<<16); + break; + case 5 : b+=k8[4]; /* fall through */ + case 4 : a+=k[0]+(((uint32_t)k[1])<<16); + break; + case 3 : a+=((uint32_t)k8[2])<<16; /* fall through */ + case 2 : a+=k[0]; + break; + case 1 : a+=k8[0]; + break; + case 0 : return c; /* zero length requires no mixing */ + } + + } else { /* need to read the key one byte at a time */ + const uint8_t *k = (const uint8_t *)key; + + /*--------------- all but the last block: affect some 32 bits of (a,b,c) */ + while (length > 12) + { + a += k[0]; + a += ((uint32_t)k[1])<<8; + a += ((uint32_t)k[2])<<16; + a += ((uint32_t)k[3])<<24; + b += k[4]; + b += ((uint32_t)k[5])<<8; + b += ((uint32_t)k[6])<<16; + b += ((uint32_t)k[7])<<24; + c += k[8]; + c += ((uint32_t)k[9])<<8; + c += ((uint32_t)k[10])<<16; + c += ((uint32_t)k[11])<<24; + mix(a,b,c); + length -= 12; + k += 12; + } + + /*-------------------------------- last block: affect all 32 bits of (c) */ + switch(length) /* all the case statements fall through */ + { + case 12: c+=((uint32_t)k[11])<<24; + case 11: c+=((uint32_t)k[10])<<16; + case 10: c+=((uint32_t)k[9])<<8; + case 9 : c+=k[8]; + case 8 : b+=((uint32_t)k[7])<<24; + case 7 : b+=((uint32_t)k[6])<<16; + case 6 : b+=((uint32_t)k[5])<<8; + case 5 : b+=k[4]; + case 4 : a+=((uint32_t)k[3])<<24; + case 3 : a+=((uint32_t)k[2])<<16; + case 2 : a+=((uint32_t)k[1])<<8; + case 1 : a+=k[0]; + break; + case 0 : return c; + } + } + + final(a,b,c); + return c; +} + +/* + * hashlittle2: return 2 32-bit hash values + * + * This is identical to hashlittle(), except it returns two 32-bit hash + * values instead of just one. This is good enough for hash table + * lookup with 2^^64 buckets, or if you want a second hash if you're not + * happy with the first, or if you want a probably-unique 64-bit ID for + * the key. *pc is better mixed than *pb, so use *pc first. If you want + * a 64-bit value do something like "*pc + (((uint64_t)*pb)<<32)". + */ +void jenkins_hashlittle2( + const void *key, /* the key to hash */ + size_t length, /* length of the key */ + uint32_t *pc, /* IN: primary initval, OUT: primary hash */ + uint32_t *pb) /* IN: secondary initval, OUT: secondary hash */ +{ + uint32_t a,b,c; /* internal state */ + union { const void *ptr; size_t i; } u; /* needed for Mac Powerbook G4 */ + + /* Set up the internal state */ + a = b = c = 0xdeadbeef + ((uint32_t)length) + *pc; + c += *pb; + + u.ptr = key; + if (HASH_LITTLE_ENDIAN && ((u.i & 0x3) == 0)) { + const uint32_t *k = (const uint32_t *)key; /* read 32-bit chunks */ + + /*------ all but last block: aligned reads and affect 32 bits of (a,b,c) */ + while (length > 12) + { + a += k[0]; + b += k[1]; + c += k[2]; + mix(a,b,c); + length -= 12; + k += 3; + } + + /*----------------------------- handle the last (probably partial) block */ + /* + * "k[2]&0xffffff" actually reads beyond the end of the string, but + * then masks off the part it's not allowed to read. Because the + * string is aligned, the masked-off tail is in the same word as the + * rest of the string. Every machine with memory protection I've seen + * does it on word boundaries, so is OK with this. But valgrind will + * still catch it and complain. The masking trick does make the hash + * noticeably faster for short strings (like English words). + */ +#if !VALGRIND && !HAS_FEATURE_ADDRESS_SANITIZER && !HAS_FEATURE_MEMORY_SANITIZER + + switch(length) + { + case 12: c+=k[2]; b+=k[1]; a+=k[0]; break; + case 11: c+=k[2]&0xffffff; b+=k[1]; a+=k[0]; break; + case 10: c+=k[2]&0xffff; b+=k[1]; a+=k[0]; break; + case 9 : c+=k[2]&0xff; b+=k[1]; a+=k[0]; break; + case 8 : b+=k[1]; a+=k[0]; break; + case 7 : b+=k[1]&0xffffff; a+=k[0]; break; + case 6 : b+=k[1]&0xffff; a+=k[0]; break; + case 5 : b+=k[1]&0xff; a+=k[0]; break; + case 4 : a+=k[0]; break; + case 3 : a+=k[0]&0xffffff; break; + case 2 : a+=k[0]&0xffff; break; + case 1 : a+=k[0]&0xff; break; + case 0 : *pc=c; *pb=b; return; /* zero length strings require no mixing */ + } + +#else /* make valgrind happy */ + + { + const uint8_t *k8 = (const uint8_t *)k; + switch(length) + { + case 12: c+=k[2]; b+=k[1]; a+=k[0]; break; + case 11: c+=((uint32_t)k8[10])<<16; /* fall through */ + case 10: c+=((uint32_t)k8[9])<<8; /* fall through */ + case 9 : c+=k8[8]; /* fall through */ + case 8 : b+=k[1]; a+=k[0]; break; + case 7 : b+=((uint32_t)k8[6])<<16; /* fall through */ + case 6 : b+=((uint32_t)k8[5])<<8; /* fall through */ + case 5 : b+=k8[4]; /* fall through */ + case 4 : a+=k[0]; break; + case 3 : a+=((uint32_t)k8[2])<<16; /* fall through */ + case 2 : a+=((uint32_t)k8[1])<<8; /* fall through */ + case 1 : a+=k8[0]; break; + case 0 : *pc=c; *pb=b; return; /* zero length strings require no mixing */ + } + } + +#endif /* !valgrind */ + + } else if (HASH_LITTLE_ENDIAN && ((u.i & 0x1) == 0)) { + const uint16_t *k = (const uint16_t *)key; /* read 16-bit chunks */ + const uint8_t *k8; + + /*--------------- all but last block: aligned reads and different mixing */ + while (length > 12) + { + a += k[0] + (((uint32_t)k[1])<<16); + b += k[2] + (((uint32_t)k[3])<<16); + c += k[4] + (((uint32_t)k[5])<<16); + mix(a,b,c); + length -= 12; + k += 6; + } + + /*----------------------------- handle the last (probably partial) block */ + k8 = (const uint8_t *)k; + switch(length) + { + case 12: c+=k[4]+(((uint32_t)k[5])<<16); + b+=k[2]+(((uint32_t)k[3])<<16); + a+=k[0]+(((uint32_t)k[1])<<16); + break; + case 11: c+=((uint32_t)k8[10])<<16; /* fall through */ + case 10: c+=k[4]; + b+=k[2]+(((uint32_t)k[3])<<16); + a+=k[0]+(((uint32_t)k[1])<<16); + break; + case 9 : c+=k8[8]; /* fall through */ + case 8 : b+=k[2]+(((uint32_t)k[3])<<16); + a+=k[0]+(((uint32_t)k[1])<<16); + break; + case 7 : b+=((uint32_t)k8[6])<<16; /* fall through */ + case 6 : b+=k[2]; + a+=k[0]+(((uint32_t)k[1])<<16); + break; + case 5 : b+=k8[4]; /* fall through */ + case 4 : a+=k[0]+(((uint32_t)k[1])<<16); + break; + case 3 : a+=((uint32_t)k8[2])<<16; /* fall through */ + case 2 : a+=k[0]; + break; + case 1 : a+=k8[0]; + break; + case 0 : *pc=c; *pb=b; return; /* zero length strings require no mixing */ + } + + } else { /* need to read the key one byte at a time */ + const uint8_t *k = (const uint8_t *)key; + + /*--------------- all but the last block: affect some 32 bits of (a,b,c) */ + while (length > 12) + { + a += k[0]; + a += ((uint32_t)k[1])<<8; + a += ((uint32_t)k[2])<<16; + a += ((uint32_t)k[3])<<24; + b += k[4]; + b += ((uint32_t)k[5])<<8; + b += ((uint32_t)k[6])<<16; + b += ((uint32_t)k[7])<<24; + c += k[8]; + c += ((uint32_t)k[9])<<8; + c += ((uint32_t)k[10])<<16; + c += ((uint32_t)k[11])<<24; + mix(a,b,c); + length -= 12; + k += 12; + } + + /*-------------------------------- last block: affect all 32 bits of (c) */ + switch(length) /* all the case statements fall through */ + { + case 12: c+=((uint32_t)k[11])<<24; + case 11: c+=((uint32_t)k[10])<<16; + case 10: c+=((uint32_t)k[9])<<8; + case 9 : c+=k[8]; + case 8 : b+=((uint32_t)k[7])<<24; + case 7 : b+=((uint32_t)k[6])<<16; + case 6 : b+=((uint32_t)k[5])<<8; + case 5 : b+=k[4]; + case 4 : a+=((uint32_t)k[3])<<24; + case 3 : a+=((uint32_t)k[2])<<16; + case 2 : a+=((uint32_t)k[1])<<8; + case 1 : a+=k[0]; + break; + case 0 : *pc=c; *pb=b; return; /* zero length strings require no mixing */ + } + } + + final(a,b,c); + *pc=c; *pb=b; +} + +/* + * hashbig(): + * This is the same as hashword() on big-endian machines. It is different + * from hashlittle() on all machines. hashbig() takes advantage of + * big-endian byte ordering. + */ +uint32_t jenkins_hashbig( const void *key, size_t length, uint32_t initval) +{ + uint32_t a,b,c; + union { const void *ptr; size_t i; } u; /* to cast key to (size_t) happily */ + + /* Set up the internal state */ + a = b = c = 0xdeadbeef + ((uint32_t)length) + initval; + + u.ptr = key; + if (HASH_BIG_ENDIAN && ((u.i & 0x3) == 0)) { + const uint32_t *k = (const uint32_t *)key; /* read 32-bit chunks */ + + /*------ all but last block: aligned reads and affect 32 bits of (a,b,c) */ + while (length > 12) + { + a += k[0]; + b += k[1]; + c += k[2]; + mix(a,b,c); + length -= 12; + k += 3; + } + + /*----------------------------- handle the last (probably partial) block */ + /* + * "k[2]<<8" actually reads beyond the end of the string, but + * then shifts out the part it's not allowed to read. Because the + * string is aligned, the illegal read is in the same word as the + * rest of the string. Every machine with memory protection I've seen + * does it on word boundaries, so is OK with this. But valgrind will + * still catch it and complain. The masking trick does make the hash + * noticeably faster for short strings (like English words). + */ +#if !VALGRIND && !HAS_FEATURE_ADDRESS_SANITIZER && !HAS_FEATURE_MEMORY_SANITIZER + + switch(length) + { + case 12: c+=k[2]; b+=k[1]; a+=k[0]; break; + case 11: c+=k[2]&0xffffff00; b+=k[1]; a+=k[0]; break; + case 10: c+=k[2]&0xffff0000; b+=k[1]; a+=k[0]; break; + case 9 : c+=k[2]&0xff000000; b+=k[1]; a+=k[0]; break; + case 8 : b+=k[1]; a+=k[0]; break; + case 7 : b+=k[1]&0xffffff00; a+=k[0]; break; + case 6 : b+=k[1]&0xffff0000; a+=k[0]; break; + case 5 : b+=k[1]&0xff000000; a+=k[0]; break; + case 4 : a+=k[0]; break; + case 3 : a+=k[0]&0xffffff00; break; + case 2 : a+=k[0]&0xffff0000; break; + case 1 : a+=k[0]&0xff000000; break; + case 0 : return c; /* zero length strings require no mixing */ + } + +#else /* make valgrind happy */ + + { + const uint8_t *k8 = (const uint8_t *)k; + switch(length) /* all the case statements fall through */ + { + case 12: c+=k[2]; b+=k[1]; a+=k[0]; break; + case 11: c+=((uint32_t)k8[10])<<8; /* fall through */ + case 10: c+=((uint32_t)k8[9])<<16; /* fall through */ + case 9 : c+=((uint32_t)k8[8])<<24; /* fall through */ + case 8 : b+=k[1]; a+=k[0]; break; + case 7 : b+=((uint32_t)k8[6])<<8; /* fall through */ + case 6 : b+=((uint32_t)k8[5])<<16; /* fall through */ + case 5 : b+=((uint32_t)k8[4])<<24; /* fall through */ + case 4 : a+=k[0]; break; + case 3 : a+=((uint32_t)k8[2])<<8; /* fall through */ + case 2 : a+=((uint32_t)k8[1])<<16; /* fall through */ + case 1 : a+=((uint32_t)k8[0])<<24; break; + case 0 : return c; + } + } + +#endif /* !VALGRIND */ + + } else { /* need to read the key one byte at a time */ + const uint8_t *k = (const uint8_t *)key; + + /*--------------- all but the last block: affect some 32 bits of (a,b,c) */ + while (length > 12) + { + a += ((uint32_t)k[0])<<24; + a += ((uint32_t)k[1])<<16; + a += ((uint32_t)k[2])<<8; + a += ((uint32_t)k[3]); + b += ((uint32_t)k[4])<<24; + b += ((uint32_t)k[5])<<16; + b += ((uint32_t)k[6])<<8; + b += ((uint32_t)k[7]); + c += ((uint32_t)k[8])<<24; + c += ((uint32_t)k[9])<<16; + c += ((uint32_t)k[10])<<8; + c += ((uint32_t)k[11]); + mix(a,b,c); + length -= 12; + k += 12; + } + + /*-------------------------------- last block: affect all 32 bits of (c) */ + switch(length) /* all the case statements fall through */ + { + case 12: c+=k[11]; + case 11: c+=((uint32_t)k[10])<<8; + case 10: c+=((uint32_t)k[9])<<16; + case 9 : c+=((uint32_t)k[8])<<24; + case 8 : b+=k[7]; + case 7 : b+=((uint32_t)k[6])<<8; + case 6 : b+=((uint32_t)k[5])<<16; + case 5 : b+=((uint32_t)k[4])<<24; + case 4 : a+=k[3]; + case 3 : a+=((uint32_t)k[2])<<8; + case 2 : a+=((uint32_t)k[1])<<16; + case 1 : a+=((uint32_t)k[0])<<24; + break; + case 0 : return c; + } + } + + final(a,b,c); + return c; +} + +#ifdef SELF_TEST + +/* used for timings */ +void driver1() +{ + uint8_t buf[256]; + uint32_t i; + uint32_t h=0; + time_t a,z; + + time(&a); + for (i=0; i<256; ++i) buf[i] = 'x'; + for (i=0; i<1; ++i) + { + h = hashlittle(&buf[0],1,h); + } + time(&z); + if (z-a > 0) printf("time %d %.8x\n", z-a, h); +} + +/* check that every input bit changes every output bit half the time */ +#define HASHSTATE 1 +#define HASHLEN 1 +#define MAXPAIR 60 +#define MAXLEN 70 +void driver2() +{ + uint8_t qa[MAXLEN+1], qb[MAXLEN+2], *a = &qa[0], *b = &qb[1]; + uint32_t c[HASHSTATE], d[HASHSTATE], i=0, j=0, k, l, m=0, z; + uint32_t e[HASHSTATE],f[HASHSTATE],g[HASHSTATE],h[HASHSTATE]; + uint32_t x[HASHSTATE],y[HASHSTATE]; + uint32_t hlen; + + printf("No more than %d trials should ever be needed \n",MAXPAIR/2); + for (hlen=0; hlen < MAXLEN; ++hlen) + { + z=0; + for (i=0; i<hlen; ++i) /*----------------------- for each input byte, */ + { + for (j=0; j<8; ++j) /*------------------------ for each input bit, */ + { + for (m=1; m<8; ++m) /*------------- for several possible initvals, */ + { + for (l=0; l<HASHSTATE; ++l) + e[l]=f[l]=g[l]=h[l]=x[l]=y[l]=~((uint32_t)0); + + /*---- check that every output bit is affected by that input bit */ + for (k=0; k<MAXPAIR; k+=2) + { + uint32_t finished=1; + /* keys have one bit different */ + for (l=0; l<hlen+1; ++l) {a[l] = b[l] = (uint8_t)0;} + /* have a and b be two keys differing in only one bit */ + a[i] ^= (k<<j); + a[i] ^= (k>>(8-j)); + c[0] = hashlittle(a, hlen, m); + b[i] ^= ((k+1)<<j); + b[i] ^= ((k+1)>>(8-j)); + d[0] = hashlittle(b, hlen, m); + /* check every bit is 1, 0, set, and not set at least once */ + for (l=0; l<HASHSTATE; ++l) + { + e[l] &= (c[l]^d[l]); + f[l] &= ~(c[l]^d[l]); + g[l] &= c[l]; + h[l] &= ~c[l]; + x[l] &= d[l]; + y[l] &= ~d[l]; + if (e[l]|f[l]|g[l]|h[l]|x[l]|y[l]) finished=0; + } + if (finished) break; + } + if (k>z) z=k; + if (k==MAXPAIR) + { + printf("Some bit didn't change: "); + printf("%.8x %.8x %.8x %.8x %.8x %.8x ", + e[0],f[0],g[0],h[0],x[0],y[0]); + printf("i %d j %d m %d len %d\n", i, j, m, hlen); + } + if (z==MAXPAIR) goto done; + } + } + } + done: + if (z < MAXPAIR) + { + printf("Mix success %2d bytes %2d initvals ",i,m); + printf("required %d trials\n", z/2); + } + } + printf("\n"); +} + +/* Check for reading beyond the end of the buffer and alignment problems */ +void driver3() +{ + uint8_t buf[MAXLEN+20], *b; + uint32_t len; + uint8_t q[] = "This is the time for all good men to come to the aid of their country..."; + uint32_t h; + uint8_t qq[] = "xThis is the time for all good men to come to the aid of their country..."; + uint32_t i; + uint8_t qqq[] = "xxThis is the time for all good men to come to the aid of their country..."; + uint32_t j; + uint8_t qqqq[] = "xxxThis is the time for all good men to come to the aid of their country..."; + uint32_t ref,x,y; + uint8_t *p; + + printf("Endianness. These lines should all be the same (for values filled in):\n"); + printf("%.8x %.8x %.8x\n", + hashword((const uint32_t *)q, (sizeof(q)-1)/4, 13), + hashword((const uint32_t *)q, (sizeof(q)-5)/4, 13), + hashword((const uint32_t *)q, (sizeof(q)-9)/4, 13)); + p = q; + printf("%.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x\n", + hashlittle(p, sizeof(q)-1, 13), hashlittle(p, sizeof(q)-2, 13), + hashlittle(p, sizeof(q)-3, 13), hashlittle(p, sizeof(q)-4, 13), + hashlittle(p, sizeof(q)-5, 13), hashlittle(p, sizeof(q)-6, 13), + hashlittle(p, sizeof(q)-7, 13), hashlittle(p, sizeof(q)-8, 13), + hashlittle(p, sizeof(q)-9, 13), hashlittle(p, sizeof(q)-10, 13), + hashlittle(p, sizeof(q)-11, 13), hashlittle(p, sizeof(q)-12, 13)); + p = &qq[1]; + printf("%.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x\n", + hashlittle(p, sizeof(q)-1, 13), hashlittle(p, sizeof(q)-2, 13), + hashlittle(p, sizeof(q)-3, 13), hashlittle(p, sizeof(q)-4, 13), + hashlittle(p, sizeof(q)-5, 13), hashlittle(p, sizeof(q)-6, 13), + hashlittle(p, sizeof(q)-7, 13), hashlittle(p, sizeof(q)-8, 13), + hashlittle(p, sizeof(q)-9, 13), hashlittle(p, sizeof(q)-10, 13), + hashlittle(p, sizeof(q)-11, 13), hashlittle(p, sizeof(q)-12, 13)); + p = &qqq[2]; + printf("%.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x\n", + hashlittle(p, sizeof(q)-1, 13), hashlittle(p, sizeof(q)-2, 13), + hashlittle(p, sizeof(q)-3, 13), hashlittle(p, sizeof(q)-4, 13), + hashlittle(p, sizeof(q)-5, 13), hashlittle(p, sizeof(q)-6, 13), + hashlittle(p, sizeof(q)-7, 13), hashlittle(p, sizeof(q)-8, 13), + hashlittle(p, sizeof(q)-9, 13), hashlittle(p, sizeof(q)-10, 13), + hashlittle(p, sizeof(q)-11, 13), hashlittle(p, sizeof(q)-12, 13)); + p = &qqqq[3]; + printf("%.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x %.8x\n", + hashlittle(p, sizeof(q)-1, 13), hashlittle(p, sizeof(q)-2, 13), + hashlittle(p, sizeof(q)-3, 13), hashlittle(p, sizeof(q)-4, 13), + hashlittle(p, sizeof(q)-5, 13), hashlittle(p, sizeof(q)-6, 13), + hashlittle(p, sizeof(q)-7, 13), hashlittle(p, sizeof(q)-8, 13), + hashlittle(p, sizeof(q)-9, 13), hashlittle(p, sizeof(q)-10, 13), + hashlittle(p, sizeof(q)-11, 13), hashlittle(p, sizeof(q)-12, 13)); + printf("\n"); + + /* check that hashlittle2 and hashlittle produce the same results */ + i=47; j=0; + hashlittle2(q, sizeof(q), &i, &j); + if (hashlittle(q, sizeof(q), 47) != i) + printf("hashlittle2 and hashlittle mismatch\n"); + + /* check that hashword2 and hashword produce the same results */ + len = 0xdeadbeef; + i=47, j=0; + hashword2(&len, 1, &i, &j); + if (hashword(&len, 1, 47) != i) + printf("hashword2 and hashword mismatch %x %x\n", + i, hashword(&len, 1, 47)); + + /* check hashlittle doesn't read before or after the ends of the string */ + for (h=0, b=buf+1; h<8; ++h, ++b) + { + for (i=0; i<MAXLEN; ++i) + { + len = i; + for (j=0; j<i; ++j) *(b+j)=0; + + /* these should all be equal */ + ref = hashlittle(b, len, (uint32_t)1); + *(b+i)=(uint8_t)~0; + *(b-1)=(uint8_t)~0; + x = hashlittle(b, len, (uint32_t)1); + y = hashlittle(b, len, (uint32_t)1); + if ((ref != x) || (ref != y)) + { + printf("alignment error: %.8x %.8x %.8x %d %d\n",ref,x,y, + h, i); + } + } + } +} + +/* check for problems with nulls */ + void driver4() +{ + uint8_t buf[1]; + uint32_t h,i,state[HASHSTATE]; + + buf[0] = ~0; + for (i=0; i<HASHSTATE; ++i) state[i] = 1; + printf("These should all be different\n"); + for (i=0, h=0; i<8; ++i) + { + h = hashlittle(buf, 0, h); + printf("%2ld 0-byte strings, hash is %.8x\n", i, h); + } +} + +void driver5() +{ + uint32_t b,c; + b=0, c=0, hashlittle2("", 0, &c, &b); + printf("hash is %.8lx %.8lx\n", c, b); /* deadbeef deadbeef */ + b=0xdeadbeef, c=0, hashlittle2("", 0, &c, &b); + printf("hash is %.8lx %.8lx\n", c, b); /* bd5b7dde deadbeef */ + b=0xdeadbeef, c=0xdeadbeef, hashlittle2("", 0, &c, &b); + printf("hash is %.8lx %.8lx\n", c, b); /* 9c093ccd bd5b7dde */ + b=0, c=0, hashlittle2("Four score and seven years ago", 30, &c, &b); + printf("hash is %.8lx %.8lx\n", c, b); /* 17770551 ce7226e6 */ + b=1, c=0, hashlittle2("Four score and seven years ago", 30, &c, &b); + printf("hash is %.8lx %.8lx\n", c, b); /* e3607cae bd371de4 */ + b=0, c=1, hashlittle2("Four score and seven years ago", 30, &c, &b); + printf("hash is %.8lx %.8lx\n", c, b); /* cd628161 6cbea4b3 */ + c = hashlittle("Four score and seven years ago", 30, 0); + printf("hash is %.8lx\n", c); /* 17770551 */ + c = hashlittle("Four score and seven years ago", 30, 1); + printf("hash is %.8lx\n", c); /* cd628161 */ +} + +int main() +{ + driver1(); /* test that the key is hashed: used for timings */ + driver2(); /* test that whole key is hashed thoroughly */ + driver3(); /* test that nothing but the key is hashed */ + driver4(); /* test hashing multiple buffers (all buffers are null) */ + driver5(); /* test the hash against known vectors */ + return 1; +} + +#endif /* SELF_TEST */ diff --git a/src/libsystemd/sd-journal/lookup3.h b/src/libsystemd/sd-journal/lookup3.h new file mode 100644 index 0000000..04e493e --- /dev/null +++ b/src/libsystemd/sd-journal/lookup3.h @@ -0,0 +1,23 @@ +/* SPDX-License-Identifier: LicenseRef-lookup3-public-domain */ +#pragma once + +#include <inttypes.h> +#include <sys/types.h> + +#include "macro.h" + +uint32_t jenkins_hashword(const uint32_t *k, size_t length, uint32_t initval) _pure_; +void jenkins_hashword2(const uint32_t *k, size_t length, uint32_t *pc, uint32_t *pb); + +uint32_t jenkins_hashlittle(const void *key, size_t length, uint32_t initval) _pure_; +void jenkins_hashlittle2(const void *key, size_t length, uint32_t *pc, uint32_t *pb); + +uint32_t jenkins_hashbig(const void *key, size_t length, uint32_t initval) _pure_; + +static inline uint64_t jenkins_hash64(const void *data, size_t length) { + uint32_t a = 0, b = 0; + + jenkins_hashlittle2(data, length, &a, &b); + + return ((uint64_t) a << 32ULL) | (uint64_t) b; +} diff --git a/src/libsystemd/sd-journal/mmap-cache.c b/src/libsystemd/sd-journal/mmap-cache.c new file mode 100644 index 0000000..82407f9 --- /dev/null +++ b/src/libsystemd/sd-journal/mmap-cache.c @@ -0,0 +1,596 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ + +#include <errno.h> +#include <stdlib.h> +#include <sys/mman.h> + +#include "alloc-util.h" +#include "errno-util.h" +#include "fd-util.h" +#include "hashmap.h" +#include "list.h" +#include "log.h" +#include "macro.h" +#include "memory-util.h" +#include "mmap-cache.h" +#include "sigbus.h" + +typedef struct Window Window; +typedef struct Context Context; + +struct Window { + MMapCache *cache; + + bool invalidated:1; + bool keep_always:1; + bool in_unused:1; + + void *ptr; + uint64_t offset; + size_t size; + + MMapFileDescriptor *fd; + + LIST_FIELDS(Window, by_fd); + LIST_FIELDS(Window, unused); + + LIST_HEAD(Context, contexts); +}; + +struct Context { + Window *window; + + LIST_FIELDS(Context, by_window); +}; + +struct MMapFileDescriptor { + MMapCache *cache; + int fd; + int prot; + bool sigbus; + LIST_HEAD(Window, windows); +}; + +struct MMapCache { + unsigned n_ref; + unsigned n_windows; + + unsigned n_context_cache_hit, n_window_list_hit, n_missed; + + Hashmap *fds; + + LIST_HEAD(Window, unused); + Window *last_unused; + + Context contexts[MMAP_CACHE_MAX_CONTEXTS]; +}; + +#define WINDOWS_MIN 64 + +#if ENABLE_DEBUG_MMAP_CACHE +/* Tiny windows increase mmap activity and the chance of exposing unsafe use. */ +# define WINDOW_SIZE (page_size()) +#else +# define WINDOW_SIZE (8ULL*1024ULL*1024ULL) +#endif + +MMapCache* mmap_cache_new(void) { + MMapCache *m; + + m = new0(MMapCache, 1); + if (!m) + return NULL; + + m->n_ref = 1; + return m; +} + +static void window_unlink(Window *w) { + + assert(w); + + if (w->ptr) + munmap(w->ptr, w->size); + + if (w->fd) + LIST_REMOVE(by_fd, w->fd->windows, w); + + if (w->in_unused) { + if (w->cache->last_unused == w) + w->cache->last_unused = w->unused_prev; + + LIST_REMOVE(unused, w->cache->unused, w); + } + + LIST_FOREACH(by_window, c, w->contexts) { + assert(c->window == w); + c->window = NULL; + } +} + +static void window_invalidate(Window *w) { + assert(w); + assert(w->fd); + + if (w->invalidated) + return; + + /* Replace the window with anonymous pages. This is useful + * when we hit a SIGBUS and want to make sure the file cannot + * trigger any further SIGBUS, possibly overrunning the sigbus + * queue. */ + + assert_se(mmap(w->ptr, w->size, w->fd->prot, MAP_PRIVATE|MAP_ANONYMOUS|MAP_FIXED, -1, 0) == w->ptr); + w->invalidated = true; +} + +static void window_free(Window *w) { + assert(w); + + window_unlink(w); + w->cache->n_windows--; + free(w); +} + +_pure_ static bool window_matches(Window *w, uint64_t offset, size_t size) { + assert(w); + assert(size > 0); + + return + offset >= w->offset && + offset + size <= w->offset + w->size; +} + +_pure_ static bool window_matches_fd(Window *w, MMapFileDescriptor *f, uint64_t offset, size_t size) { + assert(w); + assert(f); + + return + w->fd == f && + window_matches(w, offset, size); +} + +static Window *window_add(MMapCache *m, MMapFileDescriptor *f, bool keep_always, uint64_t offset, size_t size, void *ptr) { + Window *w; + + assert(m); + assert(f); + + if (!m->last_unused || m->n_windows <= WINDOWS_MIN) { + + /* Allocate a new window */ + w = new(Window, 1); + if (!w) + return NULL; + m->n_windows++; + } else { + + /* Reuse an existing one */ + w = m->last_unused; + window_unlink(w); + } + + *w = (Window) { + .cache = m, + .fd = f, + .keep_always = keep_always, + .offset = offset, + .size = size, + .ptr = ptr, + }; + + LIST_PREPEND(by_fd, f->windows, w); + + return w; +} + +static void context_detach_window(MMapCache *m, Context *c) { + Window *w; + + assert(m); + assert(c); + + if (!c->window) + return; + + w = TAKE_PTR(c->window); + LIST_REMOVE(by_window, w->contexts, c); + + if (!w->contexts && !w->keep_always) { + /* Not used anymore? */ +#if ENABLE_DEBUG_MMAP_CACHE + /* Unmap unused windows immediately to expose use-after-unmap + * by SIGSEGV. */ + window_free(w); +#else + LIST_PREPEND(unused, m->unused, w); + if (!m->last_unused) + m->last_unused = w; + + w->in_unused = true; +#endif + } +} + +static void context_attach_window(MMapCache *m, Context *c, Window *w) { + assert(m); + assert(c); + assert(w); + + if (c->window == w) + return; + + context_detach_window(m, c); + + if (w->in_unused) { + /* Used again? */ + if (m->last_unused == w) + m->last_unused = w->unused_prev; + LIST_REMOVE(unused, m->unused, w); + + w->in_unused = false; + } + + c->window = w; + LIST_PREPEND(by_window, w->contexts, c); +} + +static MMapCache *mmap_cache_free(MMapCache *m) { + assert(m); + + for (int i = 0; i < MMAP_CACHE_MAX_CONTEXTS; i++) + context_detach_window(m, &m->contexts[i]); + + hashmap_free(m->fds); + + while (m->unused) + window_free(m->unused); + + return mfree(m); +} + +DEFINE_TRIVIAL_REF_UNREF_FUNC(MMapCache, mmap_cache, mmap_cache_free); + +static int make_room(MMapCache *m) { + assert(m); + + if (!m->last_unused) + return 0; + + window_free(m->last_unused); + return 1; +} + +static int try_context( + MMapFileDescriptor *f, + Context *c, + bool keep_always, + uint64_t offset, + size_t size, + void **ret) { + + assert(f); + assert(f->cache); + assert(f->cache->n_ref > 0); + assert(c); + assert(size > 0); + assert(ret); + + if (!c->window) + return 0; + + if (!window_matches_fd(c->window, f, offset, size)) { + + /* Drop the reference to the window, since it's unnecessary now */ + context_detach_window(f->cache, c); + return 0; + } + + if (c->window->fd->sigbus) + return -EIO; + + c->window->keep_always = c->window->keep_always || keep_always; + + *ret = (uint8_t*) c->window->ptr + (offset - c->window->offset); + f->cache->n_context_cache_hit++; + + return 1; +} + +static int find_mmap( + MMapFileDescriptor *f, + Context *c, + bool keep_always, + uint64_t offset, + size_t size, + void **ret) { + + Window *found = NULL; + + assert(f); + assert(f->cache); + assert(f->cache->n_ref > 0); + assert(c); + assert(size > 0); + + if (f->sigbus) + return -EIO; + + LIST_FOREACH(by_fd, w, f->windows) + if (window_matches(w, offset, size)) { + found = w; + break; + } + + if (!found) + return 0; + + context_attach_window(f->cache, c, found); + found->keep_always = found->keep_always || keep_always; + + *ret = (uint8_t*) found->ptr + (offset - found->offset); + f->cache->n_window_list_hit++; + + return 1; +} + +static int mmap_try_harder(MMapFileDescriptor *f, void *addr, int flags, uint64_t offset, size_t size, void **res) { + void *ptr; + + assert(f); + assert(res); + + for (;;) { + int r; + + ptr = mmap(addr, size, f->prot, flags, f->fd, offset); + if (ptr != MAP_FAILED) + break; + if (errno != ENOMEM) + return negative_errno(); + + r = make_room(f->cache); + if (r < 0) + return r; + if (r == 0) + return -ENOMEM; + } + + *res = ptr; + return 0; +} + +static int add_mmap( + MMapFileDescriptor *f, + Context *c, + bool keep_always, + uint64_t offset, + size_t size, + struct stat *st, + void **ret) { + + uint64_t woffset, wsize; + Window *w; + void *d; + int r; + + assert(f); + assert(f->cache); + assert(f->cache->n_ref > 0); + assert(c); + assert(size > 0); + assert(ret); + + woffset = offset & ~((uint64_t) page_size() - 1ULL); + wsize = size + (offset - woffset); + wsize = PAGE_ALIGN(wsize); + + if (wsize < WINDOW_SIZE) { + uint64_t delta; + + delta = PAGE_ALIGN((WINDOW_SIZE - wsize) / 2); + + if (delta > offset) + woffset = 0; + else + woffset -= delta; + + wsize = WINDOW_SIZE; + } + + if (st) { + /* Memory maps that are larger then the files + underneath have undefined behavior. Hence, clamp + things to the file size if we know it */ + + if (woffset >= (uint64_t) st->st_size) + return -EADDRNOTAVAIL; + + if (woffset + wsize > (uint64_t) st->st_size) + wsize = PAGE_ALIGN(st->st_size - woffset); + } + + r = mmap_try_harder(f, NULL, MAP_SHARED, woffset, wsize, &d); + if (r < 0) + return r; + + w = window_add(f->cache, f, keep_always, woffset, wsize, d); + if (!w) + goto outofmem; + + context_attach_window(f->cache, c, w); + + *ret = (uint8_t*) w->ptr + (offset - w->offset); + + return 1; + +outofmem: + (void) munmap(d, wsize); + return -ENOMEM; +} + +int mmap_cache_fd_get( + MMapFileDescriptor *f, + unsigned context, + bool keep_always, + uint64_t offset, + size_t size, + struct stat *st, + void **ret) { + + Context *c; + int r; + + assert(f); + assert(f->cache); + assert(f->cache->n_ref > 0); + assert(size > 0); + assert(ret); + assert(context < MMAP_CACHE_MAX_CONTEXTS); + + c = &f->cache->contexts[context]; + + /* Check whether the current context is the right one already */ + r = try_context(f, c, keep_always, offset, size, ret); + if (r != 0) + return r; + + /* Search for a matching mmap */ + r = find_mmap(f, c, keep_always, offset, size, ret); + if (r != 0) + return r; + + f->cache->n_missed++; + + /* Create a new mmap */ + return add_mmap(f, c, keep_always, offset, size, st, ret); +} + +void mmap_cache_stats_log_debug(MMapCache *m) { + assert(m); + + log_debug("mmap cache statistics: %u context cache hit, %u window list hit, %u miss", m->n_context_cache_hit, m->n_window_list_hit, m->n_missed); +} + +static void mmap_cache_process_sigbus(MMapCache *m) { + bool found = false; + MMapFileDescriptor *f; + int r; + + assert(m); + + /* Iterate through all triggered pages and mark their files as + * invalidated */ + for (;;) { + bool ours; + void *addr; + + r = sigbus_pop(&addr); + if (_likely_(r == 0)) + break; + if (r < 0) { + log_error_errno(r, "SIGBUS handling failed: %m"); + abort(); + } + + ours = false; + HASHMAP_FOREACH(f, m->fds) { + LIST_FOREACH(by_fd, w, f->windows) { + if ((uint8_t*) addr >= (uint8_t*) w->ptr && + (uint8_t*) addr < (uint8_t*) w->ptr + w->size) { + found = ours = f->sigbus = true; + break; + } + } + + if (ours) + break; + } + + /* Didn't find a matching window, give up */ + if (!ours) { + log_error("Unknown SIGBUS page, aborting."); + abort(); + } + } + + /* The list of triggered pages is now empty. Now, let's remap + * all windows of the triggered file to anonymous maps, so + * that no page of the file in question is triggered again, so + * that we can be sure not to hit the queue size limit. */ + if (_likely_(!found)) + return; + + HASHMAP_FOREACH(f, m->fds) { + if (!f->sigbus) + continue; + + LIST_FOREACH(by_fd, w, f->windows) + window_invalidate(w); + } +} + +bool mmap_cache_fd_got_sigbus(MMapFileDescriptor *f) { + assert(f); + + mmap_cache_process_sigbus(f->cache); + + return f->sigbus; +} + +MMapFileDescriptor* mmap_cache_add_fd(MMapCache *m, int fd, int prot) { + MMapFileDescriptor *f; + int r; + + assert(m); + assert(fd >= 0); + + f = hashmap_get(m->fds, FD_TO_PTR(fd)); + if (f) + return f; + + r = hashmap_ensure_allocated(&m->fds, NULL); + if (r < 0) + return NULL; + + f = new0(MMapFileDescriptor, 1); + if (!f) + return NULL; + + r = hashmap_put(m->fds, FD_TO_PTR(fd), f); + if (r < 0) + return mfree(f); + + f->cache = mmap_cache_ref(m); + f->fd = fd; + f->prot = prot; + + return f; +} + +void mmap_cache_fd_free(MMapFileDescriptor *f) { + assert(f); + assert(f->cache); + + /* Make sure that any queued SIGBUS are first dispatched, so + * that we don't end up with a SIGBUS entry we cannot relate + * to any existing memory map */ + + mmap_cache_process_sigbus(f->cache); + + while (f->windows) + window_free(f->windows); + + if (f->cache) { + assert_se(hashmap_remove(f->cache->fds, FD_TO_PTR(f->fd))); + f->cache = mmap_cache_unref(f->cache); + } + + free(f); +} + +MMapCache* mmap_cache_fd_cache(MMapFileDescriptor *f) { + assert(f); + + return f->cache; +} diff --git a/src/libsystemd/sd-journal/mmap-cache.h b/src/libsystemd/sd-journal/mmap-cache.h new file mode 100644 index 0000000..4769414 --- /dev/null +++ b/src/libsystemd/sd-journal/mmap-cache.h @@ -0,0 +1,32 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ +#pragma once + +#include <stdbool.h> +#include <sys/stat.h> + +/* One context per object type, plus one of the header, plus one "additional" one */ +#define MMAP_CACHE_MAX_CONTEXTS 9 + +typedef struct MMapCache MMapCache; +typedef struct MMapFileDescriptor MMapFileDescriptor; + +MMapCache* mmap_cache_new(void); +MMapCache* mmap_cache_ref(MMapCache *m); +MMapCache* mmap_cache_unref(MMapCache *m); +DEFINE_TRIVIAL_CLEANUP_FUNC(MMapCache*, mmap_cache_unref); + +int mmap_cache_fd_get( + MMapFileDescriptor *f, + unsigned context, + bool keep_always, + uint64_t offset, + size_t size, + struct stat *st, + void **ret); +MMapFileDescriptor* mmap_cache_add_fd(MMapCache *m, int fd, int prot); +MMapCache* mmap_cache_fd_cache(MMapFileDescriptor *f); +void mmap_cache_fd_free(MMapFileDescriptor *f); + +void mmap_cache_stats_log_debug(MMapCache *m); + +bool mmap_cache_fd_got_sigbus(MMapFileDescriptor *f); diff --git a/src/libsystemd/sd-journal/sd-journal.c b/src/libsystemd/sd-journal/sd-journal.c new file mode 100644 index 0000000..c970876 --- /dev/null +++ b/src/libsystemd/sd-journal/sd-journal.c @@ -0,0 +1,3206 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ + +#include <errno.h> +#include <fcntl.h> +#include <inttypes.h> +#include <linux/magic.h> +#include <poll.h> +#include <stddef.h> +#include <sys/inotify.h> +#include <sys/vfs.h> +#include <unistd.h> + +#include "sd-journal.h" + +#include "alloc-util.h" +#include "catalog.h" +#include "compress.h" +#include "dirent-util.h" +#include "env-file.h" +#include "escape.h" +#include "fd-util.h" +#include "fileio.h" +#include "format-util.h" +#include "fs-util.h" +#include "hashmap.h" +#include "hostname-util.h" +#include "id128-util.h" +#include "inotify-util.h" +#include "io-util.h" +#include "journal-def.h" +#include "journal-file.h" +#include "journal-internal.h" +#include "list.h" +#include "lookup3.h" +#include "nulstr-util.h" +#include "path-util.h" +#include "process-util.h" +#include "replace-var.h" +#include "stat-util.h" +#include "stdio-util.h" +#include "string-util.h" +#include "strv.h" +#include "syslog-util.h" + +#define JOURNAL_FILES_RECHECK_USEC (2 * USEC_PER_SEC) + +/* The maximum size of variable values we'll expand in catalog entries. We bind this to PATH_MAX for now, as + * we want to be able to show all officially valid paths at least */ +#define REPLACE_VAR_MAX PATH_MAX + +#define DEFAULT_DATA_THRESHOLD (64*1024) + +static void remove_file_real(sd_journal *j, JournalFile *f); + +static bool journal_pid_changed(sd_journal *j) { + assert(j); + + /* We don't support people creating a journal object and + * keeping it around over a fork(). Let's complain. */ + + return j->original_pid != getpid_cached(); +} + +static int journal_put_error(sd_journal *j, int r, const char *path) { + _cleanup_free_ char *copy = NULL; + int k; + + /* Memorize an error we encountered, and store which + * file/directory it was generated from. Note that we store + * only *one* path per error code, as the error code is the + * key into the hashmap, and the path is the value. This means + * we keep track only of all error kinds, but not of all error + * locations. This has the benefit that the hashmap cannot + * grow beyond bounds. + * + * We return an error here only if we didn't manage to + * memorize the real error. */ + + if (r >= 0) + return r; + + if (path) { + copy = strdup(path); + if (!copy) + return -ENOMEM; + } + + k = hashmap_ensure_put(&j->errors, NULL, INT_TO_PTR(r), copy); + if (k < 0) { + if (k == -EEXIST) + return 0; + + return k; + } + + TAKE_PTR(copy); + return 0; +} + +static void detach_location(sd_journal *j) { + JournalFile *f; + + assert(j); + + j->current_file = NULL; + j->current_field = 0; + + ORDERED_HASHMAP_FOREACH(f, j->files) + journal_file_reset_location(f); +} + +static void init_location(Location *l, LocationType type, JournalFile *f, Object *o) { + assert(l); + assert(IN_SET(type, LOCATION_DISCRETE, LOCATION_SEEK)); + assert(f); + + *l = (Location) { + .type = type, + .seqnum = le64toh(o->entry.seqnum), + .seqnum_id = f->header->seqnum_id, + .realtime = le64toh(o->entry.realtime), + .monotonic = le64toh(o->entry.monotonic), + .boot_id = o->entry.boot_id, + .xor_hash = le64toh(o->entry.xor_hash), + .seqnum_set = true, + .realtime_set = true, + .monotonic_set = true, + .xor_hash_set = true, + }; +} + +static void set_location(sd_journal *j, JournalFile *f, Object *o) { + assert(j); + assert(f); + assert(o); + + init_location(&j->current_location, LOCATION_DISCRETE, f, o); + + j->current_file = f; + j->current_field = 0; + + /* Let f know its candidate entry was picked. */ + assert(f->location_type == LOCATION_SEEK); + f->location_type = LOCATION_DISCRETE; +} + +static int match_is_valid(const void *data, size_t size) { + const char *b = ASSERT_PTR(data); + + if (size < 2) + return false; + + if (((char*) data)[0] == '_' && ((char*) data)[1] == '_') + return false; + + for (const char *p = b; p < b + size; p++) { + + if (*p == '=') + return p > b; + + if (*p == '_') + continue; + + if (*p >= 'A' && *p <= 'Z') + continue; + + if (ascii_isdigit(*p)) + continue; + + return false; + } + + return false; +} + +static bool same_field(const void *_a, size_t s, const void *_b, size_t t) { + const uint8_t *a = _a, *b = _b; + + for (size_t j = 0; j < s && j < t; j++) { + + if (a[j] != b[j]) + return false; + + if (a[j] == '=') + return true; + } + + assert_not_reached(); +} + +static Match *match_new(Match *p, MatchType t) { + Match *m; + + m = new(Match, 1); + if (!m) + return NULL; + + *m = (Match) { + .type = t, + .parent = p, + }; + + if (p) + LIST_PREPEND(matches, p->matches, m); + + return m; +} + +static Match *match_free(Match *m) { + assert(m); + + while (m->matches) + match_free(m->matches); + + if (m->parent) + LIST_REMOVE(matches, m->parent->matches, m); + + free(m->data); + return mfree(m); +} + +static Match *match_free_if_empty(Match *m) { + if (!m || m->matches) + return m; + + return match_free(m); +} + +_public_ int sd_journal_add_match(sd_journal *j, const void *data, size_t size) { + Match *add_here = NULL, *m = NULL; + uint64_t hash; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(data, -EINVAL); + + if (size == 0) + size = strlen(data); + + if (!match_is_valid(data, size)) + return -EINVAL; + + /* level 0: AND term + * level 1: OR terms + * level 2: AND terms + * level 3: OR terms + * level 4: concrete matches */ + + if (!j->level0) { + j->level0 = match_new(NULL, MATCH_AND_TERM); + if (!j->level0) + return -ENOMEM; + } + + if (!j->level1) { + j->level1 = match_new(j->level0, MATCH_OR_TERM); + if (!j->level1) + return -ENOMEM; + } + + if (!j->level2) { + j->level2 = match_new(j->level1, MATCH_AND_TERM); + if (!j->level2) + return -ENOMEM; + } + + assert(j->level0->type == MATCH_AND_TERM); + assert(j->level1->type == MATCH_OR_TERM); + assert(j->level2->type == MATCH_AND_TERM); + + /* Old-style Jenkins (unkeyed) hashing only here. We do not cover new-style siphash (keyed) hashing + * here, since it's different for each file, and thus can't be pre-calculated in the Match object. */ + hash = jenkins_hash64(data, size); + + LIST_FOREACH(matches, l3, j->level2->matches) { + assert(l3->type == MATCH_OR_TERM); + + LIST_FOREACH(matches, l4, l3->matches) { + assert(l4->type == MATCH_DISCRETE); + + /* Exactly the same match already? Then ignore + * this addition */ + if (l4->hash == hash && + l4->size == size && + memcmp(l4->data, data, size) == 0) + return 0; + + /* Same field? Then let's add this to this OR term */ + if (same_field(data, size, l4->data, l4->size)) { + add_here = l3; + break; + } + } + + if (add_here) + break; + } + + if (!add_here) { + add_here = match_new(j->level2, MATCH_OR_TERM); + if (!add_here) + goto fail; + } + + m = match_new(add_here, MATCH_DISCRETE); + if (!m) + goto fail; + + m->hash = hash; + m->size = size; + m->data = memdup(data, size); + if (!m->data) + goto fail; + + detach_location(j); + + return 0; + +fail: + match_free(m); + match_free_if_empty(add_here); + j->level2 = match_free_if_empty(j->level2); + j->level1 = match_free_if_empty(j->level1); + j->level0 = match_free_if_empty(j->level0); + + return -ENOMEM; +} + +_public_ int sd_journal_add_conjunction(sd_journal *j) { + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + if (!j->level0) + return 0; + + if (!j->level1) + return 0; + + if (!j->level1->matches) + return 0; + + j->level1 = NULL; + j->level2 = NULL; + + return 0; +} + +_public_ int sd_journal_add_disjunction(sd_journal *j) { + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + if (!j->level0) + return 0; + + if (!j->level1) + return 0; + + if (!j->level2) + return 0; + + if (!j->level2->matches) + return 0; + + j->level2 = NULL; + return 0; +} + +static char *match_make_string(Match *m) { + char *p = NULL, *r; + bool enclose = false; + + if (!m) + return strdup("none"); + + if (m->type == MATCH_DISCRETE) + return cescape_length(m->data, m->size); + + LIST_FOREACH(matches, i, m->matches) { + char *t, *k; + + t = match_make_string(i); + if (!t) + return mfree(p); + + if (p) { + k = strjoin(p, m->type == MATCH_OR_TERM ? " OR " : " AND ", t); + free(p); + free(t); + + if (!k) + return NULL; + + p = k; + + enclose = true; + } else + p = t; + } + + if (enclose) { + r = strjoin("(", p, ")"); + free(p); + return r; + } + + return p; +} + +char *journal_make_match_string(sd_journal *j) { + assert(j); + + return match_make_string(j->level0); +} + +_public_ void sd_journal_flush_matches(sd_journal *j) { + if (!j) + return; + + if (j->level0) + match_free(j->level0); + + j->level0 = j->level1 = j->level2 = NULL; + + detach_location(j); +} + +_pure_ static int compare_with_location(const JournalFile *f, const Location *l, const JournalFile *current_file) { + int r; + + assert(f); + assert(l); + assert(f->location_type == LOCATION_SEEK); + assert(IN_SET(l->type, LOCATION_DISCRETE, LOCATION_SEEK)); + + if (l->monotonic_set && + sd_id128_equal(f->current_boot_id, l->boot_id) && + l->realtime_set && + f->current_realtime == l->realtime && + l->xor_hash_set && + f->current_xor_hash == l->xor_hash && + l->seqnum_set && + sd_id128_equal(f->header->seqnum_id, l->seqnum_id) && + f->current_seqnum == l->seqnum && + f != current_file) + return 0; + + if (l->seqnum_set && + sd_id128_equal(f->header->seqnum_id, l->seqnum_id)) { + + r = CMP(f->current_seqnum, l->seqnum); + if (r != 0) + return r; + } + + if (l->monotonic_set && + sd_id128_equal(f->current_boot_id, l->boot_id)) { + + r = CMP(f->current_monotonic, l->monotonic); + if (r != 0) + return r; + } + + if (l->realtime_set) { + + r = CMP(f->current_realtime, l->realtime); + if (r != 0) + return r; + } + + if (l->xor_hash_set) { + + r = CMP(f->current_xor_hash, l->xor_hash); + if (r != 0) + return r; + } + + return 0; +} + +static int next_for_match( + sd_journal *j, + Match *m, + JournalFile *f, + uint64_t after_offset, + direction_t direction, + Object **ret, + uint64_t *offset) { + + int r; + uint64_t np = 0; + + assert(j); + assert(m); + assert(f); + + if (m->type == MATCH_DISCRETE) { + Object *d; + uint64_t hash; + + /* If the keyed hash logic is used, we need to calculate the hash fresh per file. Otherwise + * we can use what we pre-calculated. */ + if (JOURNAL_HEADER_KEYED_HASH(f->header)) + hash = journal_file_hash_data(f, m->data, m->size); + else + hash = m->hash; + + r = journal_file_find_data_object_with_hash(f, m->data, m->size, hash, &d, NULL); + if (r <= 0) + return r; + + return journal_file_move_to_entry_by_offset_for_data(f, d, after_offset, direction, ret, offset); + + } else if (m->type == MATCH_OR_TERM) { + + /* Find the earliest match beyond after_offset */ + + LIST_FOREACH(matches, i, m->matches) { + uint64_t cp; + + r = next_for_match(j, i, f, after_offset, direction, NULL, &cp); + if (r < 0) + return r; + else if (r > 0) { + if (np == 0 || (direction == DIRECTION_DOWN ? cp < np : cp > np)) + np = cp; + } + } + + if (np == 0) + return 0; + + } else if (m->type == MATCH_AND_TERM) { + Match *last_moved; + + /* Always jump to the next matching entry and repeat + * this until we find an offset that matches for all + * matches. */ + + if (!m->matches) + return 0; + + r = next_for_match(j, m->matches, f, after_offset, direction, NULL, &np); + if (r <= 0) + return r; + + assert(direction == DIRECTION_DOWN ? np >= after_offset : np <= after_offset); + last_moved = m->matches; + + LIST_LOOP_BUT_ONE(matches, i, m->matches, last_moved) { + uint64_t cp; + + r = next_for_match(j, i, f, np, direction, NULL, &cp); + if (r <= 0) + return r; + + assert(direction == DIRECTION_DOWN ? cp >= np : cp <= np); + if (direction == DIRECTION_DOWN ? cp > np : cp < np) { + np = cp; + last_moved = i; + } + } + } + + assert(np > 0); + + if (ret) { + r = journal_file_move_to_object(f, OBJECT_ENTRY, np, ret); + if (r < 0) + return r; + } + + if (offset) + *offset = np; + + return 1; +} + +static int find_location_for_match( + sd_journal *j, + Match *m, + JournalFile *f, + direction_t direction, + Object **ret, + uint64_t *offset) { + + int r; + + assert(j); + assert(m); + assert(f); + + if (m->type == MATCH_DISCRETE) { + Object *d; + uint64_t dp, hash; + + if (JOURNAL_HEADER_KEYED_HASH(f->header)) + hash = journal_file_hash_data(f, m->data, m->size); + else + hash = m->hash; + + r = journal_file_find_data_object_with_hash(f, m->data, m->size, hash, &d, &dp); + if (r <= 0) + return r; + + /* FIXME: missing: find by monotonic */ + + if (j->current_location.type == LOCATION_HEAD) + return journal_file_next_entry_for_data(f, d, DIRECTION_DOWN, ret, offset); + if (j->current_location.type == LOCATION_TAIL) + return journal_file_next_entry_for_data(f, d, DIRECTION_UP, ret, offset); + if (j->current_location.seqnum_set && sd_id128_equal(j->current_location.seqnum_id, f->header->seqnum_id)) + return journal_file_move_to_entry_by_seqnum_for_data(f, d, j->current_location.seqnum, direction, ret, offset); + if (j->current_location.monotonic_set) { + r = journal_file_move_to_entry_by_monotonic_for_data(f, d, j->current_location.boot_id, j->current_location.monotonic, direction, ret, offset); + if (r != 0) + return r; + + /* The data object might have been invalidated. */ + r = journal_file_move_to_object(f, OBJECT_DATA, dp, &d); + if (r < 0) + return r; + } + if (j->current_location.realtime_set) + return journal_file_move_to_entry_by_realtime_for_data(f, d, j->current_location.realtime, direction, ret, offset); + + return journal_file_next_entry_for_data(f, d, direction, ret, offset); + + } else if (m->type == MATCH_OR_TERM) { + uint64_t np = 0; + + /* Find the earliest match */ + + LIST_FOREACH(matches, i, m->matches) { + uint64_t cp; + + r = find_location_for_match(j, i, f, direction, NULL, &cp); + if (r < 0) + return r; + else if (r > 0) { + if (np == 0 || (direction == DIRECTION_DOWN ? np > cp : np < cp)) + np = cp; + } + } + + if (np == 0) + return 0; + + if (ret) { + r = journal_file_move_to_object(f, OBJECT_ENTRY, np, ret); + if (r < 0) + return r; + } + + if (offset) + *offset = np; + + return 1; + + } else { + uint64_t np = 0; + + assert(m->type == MATCH_AND_TERM); + + /* First jump to the last match, and then find the + * next one where all matches match */ + + if (!m->matches) + return 0; + + LIST_FOREACH(matches, i, m->matches) { + uint64_t cp; + + r = find_location_for_match(j, i, f, direction, NULL, &cp); + if (r <= 0) + return r; + + if (np == 0 || (direction == DIRECTION_DOWN ? cp > np : cp < np)) + np = cp; + } + + return next_for_match(j, m, f, np, direction, ret, offset); + } +} + +static int find_location_with_matches( + sd_journal *j, + JournalFile *f, + direction_t direction, + Object **ret, + uint64_t *offset) { + + int r; + + assert(j); + assert(f); + assert(ret); + assert(offset); + + if (!j->level0) { + /* No matches is simple */ + + if (j->current_location.type == LOCATION_HEAD) + return journal_file_next_entry(f, 0, DIRECTION_DOWN, ret, offset); + if (j->current_location.type == LOCATION_TAIL) + return journal_file_next_entry(f, 0, DIRECTION_UP, ret, offset); + if (j->current_location.seqnum_set && sd_id128_equal(j->current_location.seqnum_id, f->header->seqnum_id)) + return journal_file_move_to_entry_by_seqnum(f, j->current_location.seqnum, direction, ret, offset); + if (j->current_location.monotonic_set) { + r = journal_file_move_to_entry_by_monotonic(f, j->current_location.boot_id, j->current_location.monotonic, direction, ret, offset); + if (r != 0) + return r; + } + if (j->current_location.realtime_set) + return journal_file_move_to_entry_by_realtime(f, j->current_location.realtime, direction, ret, offset); + + return journal_file_next_entry(f, 0, direction, ret, offset); + } else + return find_location_for_match(j, j->level0, f, direction, ret, offset); +} + +static int next_with_matches( + sd_journal *j, + JournalFile *f, + direction_t direction, + Object **ret, + uint64_t *offset) { + + assert(j); + assert(f); + assert(ret); + assert(offset); + + /* No matches is easy. We simple advance the file + * pointer by one. */ + if (!j->level0) + return journal_file_next_entry(f, f->current_offset, direction, ret, offset); + + /* If we have a match then we look for the next matching entry + * with an offset at least one step larger */ + return next_for_match(j, j->level0, f, + direction == DIRECTION_DOWN ? f->current_offset + 1 + : f->current_offset - 1, + direction, ret, offset); +} + +static int next_beyond_location(sd_journal *j, JournalFile *f, direction_t direction) { + Object *c; + uint64_t cp, n_entries; + int r; + + assert(j); + assert(f); + + n_entries = le64toh(f->header->n_entries); + + /* If we hit EOF before, we don't need to look into this file again + * unless direction changed or new entries appeared. */ + if (f->last_direction == direction && f->location_type == LOCATION_TAIL && + n_entries == f->last_n_entries) + return 0; + + f->last_n_entries = n_entries; + + if (f->last_direction == direction && f->current_offset > 0) { + /* LOCATION_SEEK here means we did the work in a previous + * iteration and the current location already points to a + * candidate entry. */ + if (f->location_type != LOCATION_SEEK) { + r = next_with_matches(j, f, direction, &c, &cp); + if (r <= 0) + return r; + + journal_file_save_location(f, c, cp); + } + } else { + f->last_direction = direction; + + r = find_location_with_matches(j, f, direction, &c, &cp); + if (r <= 0) + return r; + + journal_file_save_location(f, c, cp); + } + + /* OK, we found the spot, now let's advance until an entry + * that is actually different from what we were previously + * looking at. This is necessary to handle entries which exist + * in two (or more) journal files, and which shall all be + * suppressed but one. */ + + for (;;) { + bool found; + + if (j->current_location.type == LOCATION_DISCRETE) { + int k; + + k = compare_with_location(f, &j->current_location, j->current_file); + + found = direction == DIRECTION_DOWN ? k > 0 : k < 0; + } else + found = true; + + if (found) + return 1; + + r = next_with_matches(j, f, direction, &c, &cp); + if (r <= 0) + return r; + + journal_file_save_location(f, c, cp); + } +} + +static int real_journal_next(sd_journal *j, direction_t direction) { + JournalFile *new_file = NULL; + unsigned n_files; + const void **files; + Object *o; + int r; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + r = iterated_cache_get(j->files_cache, NULL, &files, &n_files); + if (r < 0) + return r; + + for (unsigned i = 0; i < n_files; i++) { + JournalFile *f = (JournalFile *)files[i]; + bool found; + + r = next_beyond_location(j, f, direction); + if (r < 0) { + log_debug_errno(r, "Can't iterate through %s, ignoring: %m", f->path); + remove_file_real(j, f); + continue; + } else if (r == 0) { + f->location_type = LOCATION_TAIL; + continue; + } + + if (!new_file) + found = true; + else { + int k; + + k = journal_file_compare_locations(f, new_file); + + found = direction == DIRECTION_DOWN ? k < 0 : k > 0; + } + + if (found) + new_file = f; + } + + if (!new_file) + return 0; + + r = journal_file_move_to_object(new_file, OBJECT_ENTRY, new_file->current_offset, &o); + if (r < 0) + return r; + + set_location(j, new_file, o); + + return 1; +} + +_public_ int sd_journal_next(sd_journal *j) { + return real_journal_next(j, DIRECTION_DOWN); +} + +_public_ int sd_journal_previous(sd_journal *j) { + return real_journal_next(j, DIRECTION_UP); +} + +static int real_journal_next_skip(sd_journal *j, direction_t direction, uint64_t skip) { + int c = 0, r; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(skip <= INT_MAX, -ERANGE); + + if (skip == 0) { + /* If this is not a discrete skip, then at least + * resolve the current location */ + if (j->current_location.type != LOCATION_DISCRETE) { + r = real_journal_next(j, direction); + if (r < 0) + return r; + } + + return 0; + } + + do { + r = real_journal_next(j, direction); + if (r < 0) + return r; + + if (r == 0) + return c; + + skip--; + c++; + } while (skip > 0); + + return c; +} + +_public_ int sd_journal_next_skip(sd_journal *j, uint64_t skip) { + return real_journal_next_skip(j, DIRECTION_DOWN, skip); +} + +_public_ int sd_journal_previous_skip(sd_journal *j, uint64_t skip) { + return real_journal_next_skip(j, DIRECTION_UP, skip); +} + +_public_ int sd_journal_get_cursor(sd_journal *j, char **cursor) { + Object *o; + int r; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(cursor, -EINVAL); + + if (!j->current_file || j->current_file->current_offset <= 0) + return -EADDRNOTAVAIL; + + r = journal_file_move_to_object(j->current_file, OBJECT_ENTRY, j->current_file->current_offset, &o); + if (r < 0) + return r; + + if (asprintf(cursor, + "s=%s;i=%"PRIx64";b=%s;m=%"PRIx64";t=%"PRIx64";x=%"PRIx64, + SD_ID128_TO_STRING(j->current_file->header->seqnum_id), le64toh(o->entry.seqnum), + SD_ID128_TO_STRING(o->entry.boot_id), le64toh(o->entry.monotonic), + le64toh(o->entry.realtime), + le64toh(o->entry.xor_hash)) < 0) + return -ENOMEM; + + return 0; +} + +_public_ int sd_journal_seek_cursor(sd_journal *j, const char *cursor) { + unsigned long long seqnum, monotonic, realtime, xor_hash; + bool seqnum_id_set = false, + seqnum_set = false, + boot_id_set = false, + monotonic_set = false, + realtime_set = false, + xor_hash_set = false; + sd_id128_t seqnum_id, boot_id; + int r; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(!isempty(cursor), -EINVAL); + + for (const char *p = cursor;;) { + _cleanup_free_ char *word = NULL; + + r = extract_first_word(&p, &word, ";", EXTRACT_DONT_COALESCE_SEPARATORS); + if (r < 0) + return r; + if (r == 0) + break; + + if (word[0] == '\0' || word[1] != '=') + return -EINVAL; + + switch (word[0]) { + case 's': + seqnum_id_set = true; + r = sd_id128_from_string(word + 2, &seqnum_id); + if (r < 0) + return r; + break; + + case 'i': + seqnum_set = true; + if (sscanf(word + 2, "%llx", &seqnum) != 1) + return -EINVAL; + break; + + case 'b': + boot_id_set = true; + r = sd_id128_from_string(word + 2, &boot_id); + if (r < 0) + return r; + break; + + case 'm': + monotonic_set = true; + if (sscanf(word + 2, "%llx", &monotonic) != 1) + return -EINVAL; + break; + + case 't': + realtime_set = true; + if (sscanf(word + 2, "%llx", &realtime) != 1) + return -EINVAL; + break; + + case 'x': + xor_hash_set = true; + if (sscanf(word + 2, "%llx", &xor_hash) != 1) + return -EINVAL; + break; + } + } + + if ((!seqnum_set || !seqnum_id_set) && + (!monotonic_set || !boot_id_set) && + !realtime_set) + return -EINVAL; + + detach_location(j); + j->current_location = (Location) { + .type = LOCATION_SEEK, + }; + + if (realtime_set) { + j->current_location.realtime = (uint64_t) realtime; + j->current_location.realtime_set = true; + } + + if (seqnum_set && seqnum_id_set) { + j->current_location.seqnum = (uint64_t) seqnum; + j->current_location.seqnum_id = seqnum_id; + j->current_location.seqnum_set = true; + } + + if (monotonic_set && boot_id_set) { + j->current_location.monotonic = (uint64_t) monotonic; + j->current_location.boot_id = boot_id; + j->current_location.monotonic_set = true; + } + + if (xor_hash_set) { + j->current_location.xor_hash = (uint64_t) xor_hash; + j->current_location.xor_hash_set = true; + } + + return 0; +} + +_public_ int sd_journal_test_cursor(sd_journal *j, const char *cursor) { + int r; + Object *o; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(!isempty(cursor), -EINVAL); + + if (!j->current_file || j->current_file->current_offset <= 0) + return -EADDRNOTAVAIL; + + r = journal_file_move_to_object(j->current_file, OBJECT_ENTRY, j->current_file->current_offset, &o); + if (r < 0) + return r; + + for (;;) { + _cleanup_free_ char *item = NULL; + unsigned long long ll; + sd_id128_t id; + int k = 0; + + r = extract_first_word(&cursor, &item, ";", EXTRACT_DONT_COALESCE_SEPARATORS); + if (r < 0) + return r; + + if (r == 0) + break; + + if (strlen(item) < 2 || item[1] != '=') + return -EINVAL; + + switch (item[0]) { + + case 's': + k = sd_id128_from_string(item+2, &id); + if (k < 0) + return k; + if (!sd_id128_equal(id, j->current_file->header->seqnum_id)) + return 0; + break; + + case 'i': + if (sscanf(item+2, "%llx", &ll) != 1) + return -EINVAL; + if (ll != le64toh(o->entry.seqnum)) + return 0; + break; + + case 'b': + k = sd_id128_from_string(item+2, &id); + if (k < 0) + return k; + if (!sd_id128_equal(id, o->entry.boot_id)) + return 0; + break; + + case 'm': + if (sscanf(item+2, "%llx", &ll) != 1) + return -EINVAL; + if (ll != le64toh(o->entry.monotonic)) + return 0; + break; + + case 't': + if (sscanf(item+2, "%llx", &ll) != 1) + return -EINVAL; + if (ll != le64toh(o->entry.realtime)) + return 0; + break; + + case 'x': + if (sscanf(item+2, "%llx", &ll) != 1) + return -EINVAL; + if (ll != le64toh(o->entry.xor_hash)) + return 0; + break; + } + } + + return 1; +} + +_public_ int sd_journal_seek_monotonic_usec(sd_journal *j, sd_id128_t boot_id, uint64_t usec) { + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + detach_location(j); + + j->current_location = (Location) { + .type = LOCATION_SEEK, + .boot_id = boot_id, + .monotonic = usec, + .monotonic_set = true, + }; + + return 0; +} + +_public_ int sd_journal_seek_realtime_usec(sd_journal *j, uint64_t usec) { + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + detach_location(j); + + j->current_location = (Location) { + .type = LOCATION_SEEK, + .realtime = usec, + .realtime_set = true, + }; + + return 0; +} + +_public_ int sd_journal_seek_head(sd_journal *j) { + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + detach_location(j); + + j->current_location = (Location) { + .type = LOCATION_HEAD, + }; + + return 0; +} + +_public_ int sd_journal_seek_tail(sd_journal *j) { + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + detach_location(j); + + j->current_location = (Location) { + .type = LOCATION_TAIL, + }; + + return 0; +} + +static void check_network(sd_journal *j, int fd) { + assert(j); + + if (j->on_network) + return; + + j->on_network = fd_is_network_fs(fd); +} + +static bool file_has_type_prefix(const char *prefix, const char *filename) { + const char *full, *tilded, *atted; + + full = strjoina(prefix, ".journal"); + tilded = strjoina(full, "~"); + atted = strjoina(prefix, "@"); + + return STR_IN_SET(filename, full, tilded) || + startswith(filename, atted); +} + +static bool file_type_wanted(int flags, const char *filename) { + assert(filename); + + if (!endswith(filename, ".journal") && !endswith(filename, ".journal~")) + return false; + + /* no flags set → every type is OK */ + if (!(flags & (SD_JOURNAL_SYSTEM | SD_JOURNAL_CURRENT_USER))) + return true; + + if (flags & SD_JOURNAL_SYSTEM && file_has_type_prefix("system", filename)) + return true; + + if (flags & SD_JOURNAL_CURRENT_USER) { + char prefix[5 + DECIMAL_STR_MAX(uid_t) + 1]; + + xsprintf(prefix, "user-"UID_FMT, getuid()); + + if (file_has_type_prefix(prefix, filename)) + return true; + } + + return false; +} + +static bool path_has_prefix(sd_journal *j, const char *path, const char *prefix) { + assert(j); + assert(path); + assert(prefix); + + if (j->toplevel_fd >= 0) + return false; + + return path_startswith(path, prefix); +} + +static void track_file_disposition(sd_journal *j, JournalFile *f) { + assert(j); + assert(f); + + if (!j->has_runtime_files && path_has_prefix(j, f->path, "/run")) + j->has_runtime_files = true; + else if (!j->has_persistent_files && path_has_prefix(j, f->path, "/var")) + j->has_persistent_files = true; +} + +static const char *skip_slash(const char *p) { + + if (!p) + return NULL; + + while (*p == '/') + p++; + + return p; +} + +static int add_any_file( + sd_journal *j, + int fd, + const char *path) { + + _cleanup_close_ int our_fd = -1; + JournalFile *f; + struct stat st; + int r; + + assert(j); + assert(fd >= 0 || path); + + if (fd < 0) { + assert(path); /* For gcc. */ + if (j->toplevel_fd >= 0) + /* If there's a top-level fd defined make the path relative, explicitly, since otherwise + * openat() ignores the first argument. */ + + fd = our_fd = openat(j->toplevel_fd, skip_slash(path), O_RDONLY|O_CLOEXEC|O_NONBLOCK); + else + fd = our_fd = open(path, O_RDONLY|O_CLOEXEC|O_NONBLOCK); + if (fd < 0) { + r = log_debug_errno(errno, "Failed to open journal file %s: %m", path); + goto error; + } + + r = fd_nonblock(fd, false); + if (r < 0) { + r = log_debug_errno(errno, "Failed to turn off O_NONBLOCK for %s: %m", path); + goto error; + } + } + + if (fstat(fd, &st) < 0) { + r = log_debug_errno(errno, "Failed to fstat %s: %m", path ?: "fd"); + goto error; + } + + r = stat_verify_regular(&st); + if (r < 0) { + log_debug_errno(r, "Refusing to open %s: %m", path ?: "fd"); + goto error; + } + + if (path) { + f = ordered_hashmap_get(j->files, path); + if (f) { + if (stat_inode_same(&f->last_stat, &st)) { + /* We already track this file, under the same path and with the same + * device/inode numbers, it's hence really the same. Mark this file as seen + * in this generation. This is used to GC old files in process_q_overflow() + * to detect journal files that are still there and discern them from those + * which are gone. */ + + f->last_seen_generation = j->generation; + return 0; + } + + /* So we tracked a file under this name, but it has a different inode/device. In that + * case, it got replaced (probably due to rotation?), let's drop it hence from our + * list. */ + remove_file_real(j, f); + f = NULL; + } + } + + if (ordered_hashmap_size(j->files) >= JOURNAL_FILES_MAX) { + r = log_debug_errno(SYNTHETIC_ERRNO(ETOOMANYREFS), + "Too many open journal files, not adding %s.", path ?: "fd"); + goto error; + } + + r = journal_file_open(fd, path, O_RDONLY, 0, 0, 0, NULL, j->mmap, NULL, &f); + if (r < 0) { + log_debug_errno(r, "Failed to open journal file %s: %m", path ?: "from fd"); + goto error; + } + + /* journal_file_dump(f); */ + + /* journal_file_open() generates an replacement fname if necessary, so we can use f->path. */ + r = ordered_hashmap_put(j->files, f->path, f); + if (r < 0) { + f->close_fd = false; /* Make sure journal_file_close() doesn't close the caller's fd + * (or our own). The caller or we will do that ourselves. */ + (void) journal_file_close(f); + goto error; + } + + TAKE_FD(our_fd); /* the fd is now owned by the JournalFile object */ + + f->last_seen_generation = j->generation; + + track_file_disposition(j, f); + check_network(j, f->fd); + + j->current_invalidate_counter++; + + log_debug("File %s added.", f->path); + + return 0; + +error: + (void) journal_put_error(j, r, path); /* path==NULL is OK. */ + return r; +} + +static int add_file_by_name( + sd_journal *j, + const char *prefix, + const char *filename) { + + _cleanup_free_ char *path = NULL; + + assert(j); + assert(prefix); + assert(filename); + + if (j->no_new_files) + return 0; + + if (!file_type_wanted(j->flags, filename)) + return 0; + + path = path_join(prefix, filename); + if (!path) + return -ENOMEM; + + return add_any_file(j, -1, path); +} + +static int remove_file_by_name( + sd_journal *j, + const char *prefix, + const char *filename) { + + _cleanup_free_ char *path = NULL; + JournalFile *f; + + assert(j); + assert(prefix); + assert(filename); + + path = path_join(prefix, filename); + if (!path) + return -ENOMEM; + + f = ordered_hashmap_get(j->files, path); + if (!f) + return 0; + + remove_file_real(j, f); + return 1; +} + +static void remove_file_real(sd_journal *j, JournalFile *f) { + assert(j); + assert(f); + + (void) ordered_hashmap_remove(j->files, f->path); + + log_debug("File %s removed.", f->path); + + if (j->current_file == f) { + j->current_file = NULL; + j->current_field = 0; + } + + if (j->unique_file == f) { + /* Jump to the next unique_file or NULL if that one was last */ + j->unique_file = ordered_hashmap_next(j->files, j->unique_file->path); + j->unique_offset = 0; + if (!j->unique_file) + j->unique_file_lost = true; + } + + if (j->fields_file == f) { + j->fields_file = ordered_hashmap_next(j->files, j->fields_file->path); + j->fields_offset = 0; + if (!j->fields_file) + j->fields_file_lost = true; + } + + (void) journal_file_close(f); + + j->current_invalidate_counter++; +} + +static int dirname_is_machine_id(const char *fn) { + sd_id128_t id, machine; + const char *e; + int r; + + /* Returns true if the specified directory name matches the local machine ID */ + + r = sd_id128_get_machine(&machine); + if (r < 0) + return r; + + e = strchr(fn, '.'); + if (e) { + const char *k; + + /* Looks like it has a namespace suffix. Verify that. */ + if (!log_namespace_name_valid(e + 1)) + return false; + + k = strndupa_safe(fn, e - fn); + r = sd_id128_from_string(k, &id); + } else + r = sd_id128_from_string(fn, &id); + if (r < 0) + return r; + + return sd_id128_equal(id, machine); +} + +static int dirname_has_namespace(const char *fn, const char *namespace) { + const char *e; + + /* Returns true if the specified directory name matches the specified namespace */ + + e = strchr(fn, '.'); + if (e) { + const char *k; + + if (!namespace) + return false; + + if (!streq(e + 1, namespace)) + return false; + + k = strndupa_safe(fn, e - fn); + return id128_is_valid(k); + } + + if (namespace) + return false; + + return id128_is_valid(fn); +} + +static bool dirent_is_journal_file(const struct dirent *de) { + assert(de); + + /* Returns true if the specified directory entry looks like a journal file we might be interested in */ + + if (!IN_SET(de->d_type, DT_REG, DT_LNK, DT_UNKNOWN)) + return false; + + return endswith(de->d_name, ".journal") || + endswith(de->d_name, ".journal~"); +} + +static bool dirent_is_journal_subdir(const struct dirent *de) { + const char *e, *n; + assert(de); + + /* returns true if the specified directory entry looks like a directory that might contain journal + * files we might be interested in, i.e. is either a 128bit ID or a 128bit ID suffixed by a + * namespace. */ + + if (!IN_SET(de->d_type, DT_DIR, DT_LNK, DT_UNKNOWN)) + return false; + + e = strchr(de->d_name, '.'); + if (!e) + return id128_is_valid(de->d_name); /* No namespace */ + + n = strndupa_safe(de->d_name, e - de->d_name); + if (!id128_is_valid(n)) + return false; + + return log_namespace_name_valid(e + 1); +} + +static int directory_open(sd_journal *j, const char *path, DIR **ret) { + DIR *d; + + assert(j); + assert(path); + assert(ret); + + if (j->toplevel_fd < 0) + d = opendir(path); + else + /* Open the specified directory relative to the toplevel fd. Enforce that the path specified is + * relative, by dropping the initial slash */ + d = xopendirat(j->toplevel_fd, skip_slash(path), 0); + if (!d) + return -errno; + + *ret = d; + return 0; +} + +static int add_directory(sd_journal *j, const char *prefix, const char *dirname); + +static void directory_enumerate(sd_journal *j, Directory *m, DIR *d) { + assert(j); + assert(m); + assert(d); + + FOREACH_DIRENT_ALL(de, d, goto fail) { + if (dirent_is_journal_file(de)) + (void) add_file_by_name(j, m->path, de->d_name); + + if (m->is_root && dirent_is_journal_subdir(de)) + (void) add_directory(j, m->path, de->d_name); + } + + return; +fail: + log_debug_errno(errno, "Failed to enumerate directory %s, ignoring: %m", m->path); +} + +static void directory_watch(sd_journal *j, Directory *m, int fd, uint32_t mask) { + int r; + + assert(j); + assert(m); + assert(fd >= 0); + + /* Watch this directory if that's enabled and if it not being watched yet. */ + + if (m->wd > 0) /* Already have a watch? */ + return; + if (j->inotify_fd < 0) /* Not watching at all? */ + return; + + m->wd = inotify_add_watch_fd(j->inotify_fd, fd, mask); + if (m->wd < 0) { + log_debug_errno(errno, "Failed to watch journal directory '%s', ignoring: %m", m->path); + return; + } + + r = hashmap_put(j->directories_by_wd, INT_TO_PTR(m->wd), m); + if (r == -EEXIST) + log_debug_errno(r, "Directory '%s' already being watched under a different path, ignoring: %m", m->path); + if (r < 0) { + log_debug_errno(r, "Failed to add watch for journal directory '%s' to hashmap, ignoring: %m", m->path); + (void) inotify_rm_watch(j->inotify_fd, m->wd); + m->wd = -1; + } +} + +static int add_directory( + sd_journal *j, + const char *prefix, + const char *dirname) { + + _cleanup_free_ char *path = NULL; + _cleanup_closedir_ DIR *d = NULL; + Directory *m; + int r, k; + + assert(j); + assert(prefix); + + /* Adds a journal file directory to watch. If the directory is already tracked this updates the inotify watch + * and reenumerates directory contents */ + + path = path_join(prefix, dirname); + if (!path) { + r = -ENOMEM; + goto fail; + } + + log_debug("Considering directory '%s'.", path); + + /* We consider everything local that is in a directory for the local machine ID, or that is stored in /run */ + if ((j->flags & SD_JOURNAL_LOCAL_ONLY) && + !((dirname && dirname_is_machine_id(dirname) > 0) || path_has_prefix(j, path, "/run"))) + return 0; + + if (dirname && + (!(FLAGS_SET(j->flags, SD_JOURNAL_ALL_NAMESPACES) || + dirname_has_namespace(dirname, j->namespace) > 0 || + (FLAGS_SET(j->flags, SD_JOURNAL_INCLUDE_DEFAULT_NAMESPACE) && dirname_has_namespace(dirname, NULL) > 0)))) + return 0; + + r = directory_open(j, path, &d); + if (r < 0) { + log_debug_errno(r, "Failed to open directory '%s': %m", path); + goto fail; + } + + m = hashmap_get(j->directories_by_path, path); + if (!m) { + m = new(Directory, 1); + if (!m) { + r = -ENOMEM; + goto fail; + } + + *m = (Directory) { + .is_root = false, + .path = path, + }; + + if (hashmap_put(j->directories_by_path, m->path, m) < 0) { + free(m); + r = -ENOMEM; + goto fail; + } + + path = NULL; /* avoid freeing in cleanup */ + j->current_invalidate_counter++; + + log_debug("Directory %s added.", m->path); + + } else if (m->is_root) + return 0; /* Don't 'downgrade' from root directory */ + + m->last_seen_generation = j->generation; + + directory_watch(j, m, dirfd(d), + IN_CREATE|IN_MOVED_TO|IN_MODIFY|IN_ATTRIB|IN_DELETE| + IN_DELETE_SELF|IN_MOVE_SELF|IN_UNMOUNT|IN_MOVED_FROM| + IN_ONLYDIR); + + if (!j->no_new_files) + directory_enumerate(j, m, d); + + check_network(j, dirfd(d)); + + return 0; + +fail: + k = journal_put_error(j, r, path ?: prefix); + if (k < 0) + return k; + + return r; +} + +static int add_root_directory(sd_journal *j, const char *p, bool missing_ok) { + + _cleanup_closedir_ DIR *d = NULL; + Directory *m; + int r, k; + + assert(j); + + /* Adds a root directory to our set of directories to use. If the root directory is already in the set, we + * update the inotify logic, and renumerate the directory entries. This call may hence be called to initially + * populate the set, as well as to update it later. */ + + if (p) { + /* If there's a path specified, use it. */ + + log_debug("Considering root directory '%s'.", p); + + if ((j->flags & SD_JOURNAL_RUNTIME_ONLY) && + !path_has_prefix(j, p, "/run")) + return -EINVAL; + + if (j->prefix) + p = strjoina(j->prefix, p); + + r = directory_open(j, p, &d); + if (r == -ENOENT && missing_ok) + return 0; + if (r < 0) { + log_debug_errno(r, "Failed to open root directory %s: %m", p); + goto fail; + } + } else { + _cleanup_close_ int dfd = -1; + + /* If there's no path specified, then we use the top-level fd itself. We duplicate the fd here, since + * opendir() will take possession of the fd, and close it, which we don't want. */ + + p = "."; /* store this as "." in the directories hashmap */ + + dfd = fcntl(j->toplevel_fd, F_DUPFD_CLOEXEC, 3); + if (dfd < 0) { + r = -errno; + goto fail; + } + + d = take_fdopendir(&dfd); + if (!d) { + r = -errno; + goto fail; + } + + rewinddir(d); + } + + m = hashmap_get(j->directories_by_path, p); + if (!m) { + m = new0(Directory, 1); + if (!m) { + r = -ENOMEM; + goto fail; + } + + m->is_root = true; + + m->path = strdup(p); + if (!m->path) { + free(m); + r = -ENOMEM; + goto fail; + } + + if (hashmap_put(j->directories_by_path, m->path, m) < 0) { + free(m->path); + free(m); + r = -ENOMEM; + goto fail; + } + + j->current_invalidate_counter++; + + log_debug("Root directory %s added.", m->path); + + } else if (!m->is_root) + return 0; + + directory_watch(j, m, dirfd(d), + IN_CREATE|IN_MOVED_TO|IN_MODIFY|IN_ATTRIB|IN_DELETE| + IN_ONLYDIR); + + if (!j->no_new_files) + directory_enumerate(j, m, d); + + check_network(j, dirfd(d)); + + return 0; + +fail: + k = journal_put_error(j, r, p); + if (k < 0) + return k; + + return r; +} + +static void remove_directory(sd_journal *j, Directory *d) { + assert(j); + + if (d->wd > 0) { + hashmap_remove(j->directories_by_wd, INT_TO_PTR(d->wd)); + + if (j->inotify_fd >= 0) + (void) inotify_rm_watch(j->inotify_fd, d->wd); + } + + hashmap_remove(j->directories_by_path, d->path); + + if (d->is_root) + log_debug("Root directory %s removed.", d->path); + else + log_debug("Directory %s removed.", d->path); + + free(d->path); + free(d); +} + +static int add_search_paths(sd_journal *j) { + + static const char search_paths[] = + "/run/log/journal\0" + "/var/log/journal\0"; + const char *p; + + assert(j); + + /* We ignore most errors here, since the idea is to only open + * what's actually accessible, and ignore the rest. */ + + NULSTR_FOREACH(p, search_paths) + (void) add_root_directory(j, p, true); + + if (!(j->flags & SD_JOURNAL_LOCAL_ONLY)) + (void) add_root_directory(j, "/var/log/journal/remote", true); + + return 0; +} + +static int add_current_paths(sd_journal *j) { + JournalFile *f; + + assert(j); + assert(j->no_new_files); + + /* Simply adds all directories for files we have open as directories. We don't expect errors here, so we + * treat them as fatal. */ + + ORDERED_HASHMAP_FOREACH(f, j->files) { + _cleanup_free_ char *dir = NULL; + int r; + + r = path_extract_directory(f->path, &dir); + if (r < 0) + return r; + + r = add_directory(j, dir, NULL); + if (r < 0) + return r; + } + + return 0; +} + +static int allocate_inotify(sd_journal *j) { + assert(j); + + if (j->inotify_fd < 0) { + j->inotify_fd = inotify_init1(IN_NONBLOCK|IN_CLOEXEC); + if (j->inotify_fd < 0) + return -errno; + } + + return hashmap_ensure_allocated(&j->directories_by_wd, NULL); +} + +static sd_journal *journal_new(int flags, const char *path, const char *namespace) { + _cleanup_(sd_journal_closep) sd_journal *j = NULL; + + j = new0(sd_journal, 1); + if (!j) + return NULL; + + j->original_pid = getpid_cached(); + j->toplevel_fd = -1; + j->inotify_fd = -1; + j->flags = flags; + j->data_threshold = DEFAULT_DATA_THRESHOLD; + + if (path) { + char *t; + + t = strdup(path); + if (!t) + return NULL; + + if (flags & SD_JOURNAL_OS_ROOT) + j->prefix = t; + else + j->path = t; + } + + if (namespace) { + j->namespace = strdup(namespace); + if (!j->namespace) + return NULL; + } + + j->files = ordered_hashmap_new(&path_hash_ops); + if (!j->files) + return NULL; + + j->files_cache = ordered_hashmap_iterated_cache_new(j->files); + j->directories_by_path = hashmap_new(&path_hash_ops); + j->mmap = mmap_cache_new(); + if (!j->files_cache || !j->directories_by_path || !j->mmap) + return NULL; + + return TAKE_PTR(j); +} + +#define OPEN_ALLOWED_FLAGS \ + (SD_JOURNAL_LOCAL_ONLY | \ + SD_JOURNAL_RUNTIME_ONLY | \ + SD_JOURNAL_SYSTEM | \ + SD_JOURNAL_CURRENT_USER | \ + SD_JOURNAL_ALL_NAMESPACES | \ + SD_JOURNAL_INCLUDE_DEFAULT_NAMESPACE) + +_public_ int sd_journal_open_namespace(sd_journal **ret, const char *namespace, int flags) { + _cleanup_(sd_journal_closep) sd_journal *j = NULL; + int r; + + assert_return(ret, -EINVAL); + assert_return((flags & ~OPEN_ALLOWED_FLAGS) == 0, -EINVAL); + + j = journal_new(flags, NULL, namespace); + if (!j) + return -ENOMEM; + + r = add_search_paths(j); + if (r < 0) + return r; + + *ret = TAKE_PTR(j); + return 0; +} + +_public_ int sd_journal_open(sd_journal **ret, int flags) { + return sd_journal_open_namespace(ret, NULL, flags); +} + +#define OPEN_CONTAINER_ALLOWED_FLAGS \ + (SD_JOURNAL_LOCAL_ONLY | SD_JOURNAL_SYSTEM) + +_public_ int sd_journal_open_container(sd_journal **ret, const char *machine, int flags) { + _cleanup_free_ char *root = NULL, *class = NULL; + _cleanup_(sd_journal_closep) sd_journal *j = NULL; + char *p; + int r; + + /* This is deprecated, people should use machined's OpenMachineRootDirectory() call instead in + * combination with sd_journal_open_directory_fd(). */ + + assert_return(machine, -EINVAL); + assert_return(ret, -EINVAL); + assert_return((flags & ~OPEN_CONTAINER_ALLOWED_FLAGS) == 0, -EINVAL); + assert_return(hostname_is_valid(machine, 0), -EINVAL); + + p = strjoina("/run/systemd/machines/", machine); + r = parse_env_file(NULL, p, + "ROOT", &root, + "CLASS", &class); + if (r == -ENOENT) + return -EHOSTDOWN; + if (r < 0) + return r; + if (!root) + return -ENODATA; + + if (!streq_ptr(class, "container")) + return -EIO; + + j = journal_new(flags, root, NULL); + if (!j) + return -ENOMEM; + + r = add_search_paths(j); + if (r < 0) + return r; + + *ret = TAKE_PTR(j); + return 0; +} + +#define OPEN_DIRECTORY_ALLOWED_FLAGS \ + (SD_JOURNAL_OS_ROOT | \ + SD_JOURNAL_SYSTEM | SD_JOURNAL_CURRENT_USER ) + +_public_ int sd_journal_open_directory(sd_journal **ret, const char *path, int flags) { + _cleanup_(sd_journal_closep) sd_journal *j = NULL; + int r; + + assert_return(ret, -EINVAL); + assert_return(path, -EINVAL); + assert_return((flags & ~OPEN_DIRECTORY_ALLOWED_FLAGS) == 0, -EINVAL); + + j = journal_new(flags, path, NULL); + if (!j) + return -ENOMEM; + + if (flags & SD_JOURNAL_OS_ROOT) + r = add_search_paths(j); + else + r = add_root_directory(j, path, false); + if (r < 0) + return r; + + *ret = TAKE_PTR(j); + return 0; +} + +_public_ int sd_journal_open_files(sd_journal **ret, const char **paths, int flags) { + _cleanup_(sd_journal_closep) sd_journal *j = NULL; + int r; + + assert_return(ret, -EINVAL); + assert_return(flags == 0, -EINVAL); + + j = journal_new(flags, NULL, NULL); + if (!j) + return -ENOMEM; + + STRV_FOREACH(path, paths) { + r = add_any_file(j, -1, *path); + if (r < 0) + return r; + } + + j->no_new_files = true; + + *ret = TAKE_PTR(j); + return 0; +} + +#define OPEN_DIRECTORY_FD_ALLOWED_FLAGS \ + (SD_JOURNAL_OS_ROOT | \ + SD_JOURNAL_SYSTEM | SD_JOURNAL_CURRENT_USER ) + +_public_ int sd_journal_open_directory_fd(sd_journal **ret, int fd, int flags) { + _cleanup_(sd_journal_closep) sd_journal *j = NULL; + struct stat st; + int r; + + assert_return(ret, -EINVAL); + assert_return(fd >= 0, -EBADF); + assert_return((flags & ~OPEN_DIRECTORY_FD_ALLOWED_FLAGS) == 0, -EINVAL); + + if (fstat(fd, &st) < 0) + return -errno; + + if (!S_ISDIR(st.st_mode)) + return -EBADFD; + + j = journal_new(flags, NULL, NULL); + if (!j) + return -ENOMEM; + + j->toplevel_fd = fd; + + if (flags & SD_JOURNAL_OS_ROOT) + r = add_search_paths(j); + else + r = add_root_directory(j, NULL, false); + if (r < 0) + return r; + + *ret = TAKE_PTR(j); + return 0; +} + +_public_ int sd_journal_open_files_fd(sd_journal **ret, int fds[], unsigned n_fds, int flags) { + JournalFile *f; + _cleanup_(sd_journal_closep) sd_journal *j = NULL; + int r; + + assert_return(ret, -EINVAL); + assert_return(n_fds > 0, -EBADF); + assert_return(flags == 0, -EINVAL); + + j = journal_new(flags, NULL, NULL); + if (!j) + return -ENOMEM; + + for (unsigned i = 0; i < n_fds; i++) { + struct stat st; + + if (fds[i] < 0) { + r = -EBADF; + goto fail; + } + + if (fstat(fds[i], &st) < 0) { + r = -errno; + goto fail; + } + + r = stat_verify_regular(&st); + if (r < 0) + goto fail; + + r = add_any_file(j, fds[i], NULL); + if (r < 0) + goto fail; + } + + j->no_new_files = true; + j->no_inotify = true; + + *ret = TAKE_PTR(j); + return 0; + +fail: + /* If we fail, make sure we don't take possession of the files we managed to make use of successfully, and they + * remain open */ + ORDERED_HASHMAP_FOREACH(f, j->files) + f->close_fd = false; + + return r; +} + +_public_ void sd_journal_close(sd_journal *j) { + Directory *d; + + if (!j) + return; + + sd_journal_flush_matches(j); + + ordered_hashmap_free_with_destructor(j->files, journal_file_close); + iterated_cache_free(j->files_cache); + + while ((d = hashmap_first(j->directories_by_path))) + remove_directory(j, d); + + while ((d = hashmap_first(j->directories_by_wd))) + remove_directory(j, d); + + hashmap_free(j->directories_by_path); + hashmap_free(j->directories_by_wd); + + safe_close(j->inotify_fd); + + if (j->mmap) { + mmap_cache_stats_log_debug(j->mmap); + mmap_cache_unref(j->mmap); + } + + hashmap_free_free(j->errors); + + free(j->path); + free(j->prefix); + free(j->namespace); + free(j->unique_field); + free(j->fields_buffer); + free(j); +} + +_public_ int sd_journal_get_realtime_usec(sd_journal *j, uint64_t *ret) { + Object *o; + JournalFile *f; + int r; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(ret, -EINVAL); + + f = j->current_file; + if (!f) + return -EADDRNOTAVAIL; + + if (f->current_offset <= 0) + return -EADDRNOTAVAIL; + + r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o); + if (r < 0) + return r; + + *ret = le64toh(o->entry.realtime); + return 0; +} + +_public_ int sd_journal_get_monotonic_usec(sd_journal *j, uint64_t *ret, sd_id128_t *ret_boot_id) { + Object *o; + JournalFile *f; + int r; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + f = j->current_file; + if (!f) + return -EADDRNOTAVAIL; + + if (f->current_offset <= 0) + return -EADDRNOTAVAIL; + + r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o); + if (r < 0) + return r; + + if (ret_boot_id) + *ret_boot_id = o->entry.boot_id; + else { + sd_id128_t id; + + r = sd_id128_get_boot(&id); + if (r < 0) + return r; + + if (!sd_id128_equal(id, o->entry.boot_id)) + return -ESTALE; + } + + if (ret) + *ret = le64toh(o->entry.monotonic); + + return 0; +} + +static bool field_is_valid(const char *field) { + assert(field); + + if (isempty(field)) + return false; + + if (startswith(field, "__")) + return false; + + for (const char *p = field; *p; p++) { + + if (*p == '_') + continue; + + if (*p >= 'A' && *p <= 'Z') + continue; + + if (ascii_isdigit(*p)) + continue; + + return false; + } + + return true; +} + +_public_ int sd_journal_get_data(sd_journal *j, const char *field, const void **data, size_t *size) { + JournalFile *f; + size_t field_length; + Object *o; + int r; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(field, -EINVAL); + assert_return(data, -EINVAL); + assert_return(size, -EINVAL); + assert_return(field_is_valid(field), -EINVAL); + + f = j->current_file; + if (!f) + return -EADDRNOTAVAIL; + + if (f->current_offset <= 0) + return -EADDRNOTAVAIL; + + r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o); + if (r < 0) + return r; + + field_length = strlen(field); + + uint64_t n = journal_file_entry_n_items(f, o); + for (uint64_t i = 0; i < n; i++) { + uint64_t p; + void *d; + size_t l; + + p = journal_file_entry_item_object_offset(f, o, i); + r = journal_file_data_payload(f, NULL, p, field, field_length, j->data_threshold, &d, &l); + if (r == 0) + goto next; + if (IN_SET(r, -EADDRNOTAVAIL, -EBADMSG)) { + log_debug_errno(r, "Entry item %"PRIu64" data object is bad, skipping over it: %m", i); + goto next; + } + if (r < 0) + return r; + + *data = d; + *size = l; + + return 0; + + next: + /* journal_file_data_payload() may clear or overwrite cached object. */ + r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o); + if (r < 0) + return r; + } + + return -ENOENT; +} + +_public_ int sd_journal_enumerate_data(sd_journal *j, const void **data, size_t *size) { + JournalFile *f; + Object *o; + int r; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(data, -EINVAL); + assert_return(size, -EINVAL); + + f = j->current_file; + if (!f) + return -EADDRNOTAVAIL; + + if (f->current_offset <= 0) + return -EADDRNOTAVAIL; + + r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o); + if (r < 0) + return r; + + for (uint64_t n = journal_file_entry_n_items(f, o); j->current_field < n; j->current_field++) { + uint64_t p; + void *d; + size_t l; + + p = journal_file_entry_item_object_offset(f, o, j->current_field); + r = journal_file_data_payload(f, NULL, p, NULL, 0, j->data_threshold, &d, &l); + if (IN_SET(r, -EADDRNOTAVAIL, -EBADMSG)) { + log_debug_errno(r, "Entry item %"PRIu64" data object is bad, skipping over it: %m", j->current_field); + goto next; + } + if (r < 0) + return r; + assert(r > 0); + + *data = d; + *size = l; + + j->current_field++; + + return 1; + + next: + /* journal_file_data_payload() may clear or overwrite cached object. */ + r = journal_file_move_to_object(f, OBJECT_ENTRY, f->current_offset, &o); + if (r < 0) + return r; + } + + return 0; +} + +_public_ int sd_journal_enumerate_available_data(sd_journal *j, const void **data, size_t *size) { + for (;;) { + int r; + + r = sd_journal_enumerate_data(j, data, size); + if (r >= 0) + return r; + if (!JOURNAL_ERRNO_IS_UNAVAILABLE_FIELD(r)) + return r; + j->current_field++; /* Try with the next field */ + } +} + +_public_ void sd_journal_restart_data(sd_journal *j) { + if (!j) + return; + + j->current_field = 0; +} + +static int reiterate_all_paths(sd_journal *j) { + assert(j); + + if (j->no_new_files) + return add_current_paths(j); + + if (j->flags & SD_JOURNAL_OS_ROOT) + return add_search_paths(j); + + if (j->toplevel_fd >= 0) + return add_root_directory(j, NULL, false); + + if (j->path) + return add_root_directory(j, j->path, true); + + return add_search_paths(j); +} + +_public_ int sd_journal_get_fd(sd_journal *j) { + int r; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + if (j->no_inotify) + return -EMEDIUMTYPE; + + if (j->inotify_fd >= 0) + return j->inotify_fd; + + r = allocate_inotify(j); + if (r < 0) + return r; + + log_debug("Reiterating files to get inotify watches established."); + + /* Iterate through all dirs again, to add them to the inotify */ + r = reiterate_all_paths(j); + if (r < 0) + return r; + + return j->inotify_fd; +} + +_public_ int sd_journal_get_events(sd_journal *j) { + int fd; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + fd = sd_journal_get_fd(j); + if (fd < 0) + return fd; + + return POLLIN; +} + +_public_ int sd_journal_get_timeout(sd_journal *j, uint64_t *timeout_usec) { + int fd; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(timeout_usec, -EINVAL); + + fd = sd_journal_get_fd(j); + if (fd < 0) + return fd; + + if (!j->on_network) { + *timeout_usec = UINT64_MAX; + return 0; + } + + /* If we are on the network we need to regularly check for + * changes manually */ + + *timeout_usec = j->last_process_usec + JOURNAL_FILES_RECHECK_USEC; + return 1; +} + +static void process_q_overflow(sd_journal *j) { + JournalFile *f; + Directory *m; + + assert(j); + + /* When the inotify queue overruns we need to enumerate and re-validate all journal files to bring our list + * back in sync with what's on disk. For this we pick a new generation counter value. It'll be assigned to all + * journal files we encounter. All journal files and all directories that don't carry it after reenumeration + * are subject for unloading. */ + + log_debug("Inotify queue overrun, reiterating everything."); + + j->generation++; + (void) reiterate_all_paths(j); + + ORDERED_HASHMAP_FOREACH(f, j->files) { + + if (f->last_seen_generation == j->generation) + continue; + + log_debug("File '%s' hasn't been seen in this enumeration, removing.", f->path); + remove_file_real(j, f); + } + + HASHMAP_FOREACH(m, j->directories_by_path) { + + if (m->last_seen_generation == j->generation) + continue; + + if (m->is_root) /* Never GC root directories */ + continue; + + log_debug("Directory '%s' hasn't been seen in this enumeration, removing.", f->path); + remove_directory(j, m); + } + + log_debug("Reiteration complete."); +} + +static void process_inotify_event(sd_journal *j, const struct inotify_event *e) { + Directory *d; + + assert(j); + assert(e); + + if (e->mask & IN_Q_OVERFLOW) { + process_q_overflow(j); + return; + } + + /* Is this a subdirectory we watch? */ + d = hashmap_get(j->directories_by_wd, INT_TO_PTR(e->wd)); + if (d) { + if (!(e->mask & IN_ISDIR) && e->len > 0 && + (endswith(e->name, ".journal") || + endswith(e->name, ".journal~"))) { + + /* Event for a journal file */ + + if (e->mask & (IN_CREATE|IN_MOVED_TO|IN_MODIFY|IN_ATTRIB)) + (void) add_file_by_name(j, d->path, e->name); + else if (e->mask & (IN_DELETE|IN_MOVED_FROM|IN_UNMOUNT)) + (void) remove_file_by_name(j, d->path, e->name); + + } else if (!d->is_root && e->len == 0) { + + /* Event for a subdirectory */ + + if (e->mask & (IN_DELETE_SELF|IN_MOVE_SELF|IN_UNMOUNT)) + remove_directory(j, d); + + } else if (d->is_root && (e->mask & IN_ISDIR) && e->len > 0 && id128_is_valid(e->name)) { + + /* Event for root directory */ + + if (e->mask & (IN_CREATE|IN_MOVED_TO|IN_MODIFY|IN_ATTRIB)) + (void) add_directory(j, d->path, e->name); + } + + return; + } + + if (e->mask & IN_IGNORED) + return; + + log_debug("Unexpected inotify event."); +} + +static int determine_change(sd_journal *j) { + bool b; + + assert(j); + + b = j->current_invalidate_counter != j->last_invalidate_counter; + j->last_invalidate_counter = j->current_invalidate_counter; + + return b ? SD_JOURNAL_INVALIDATE : SD_JOURNAL_APPEND; +} + +_public_ int sd_journal_process(sd_journal *j) { + bool got_something = false; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + if (j->inotify_fd < 0) /* We have no inotify fd yet? Then there's noting to process. */ + return 0; + + j->last_process_usec = now(CLOCK_MONOTONIC); + j->last_invalidate_counter = j->current_invalidate_counter; + + for (;;) { + union inotify_event_buffer buffer; + ssize_t l; + + l = read(j->inotify_fd, &buffer, sizeof(buffer)); + if (l < 0) { + if (ERRNO_IS_TRANSIENT(errno)) + return got_something ? determine_change(j) : SD_JOURNAL_NOP; + + return -errno; + } + + got_something = true; + + FOREACH_INOTIFY_EVENT(e, buffer, l) + process_inotify_event(j, e); + } +} + +_public_ int sd_journal_wait(sd_journal *j, uint64_t timeout_usec) { + int r; + uint64_t t; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + if (j->inotify_fd < 0) { + JournalFile *f; + + /* This is the first invocation, hence create the + * inotify watch */ + r = sd_journal_get_fd(j); + if (r < 0) + return r; + + /* Server might have done some vacuuming while we weren't watching. + Get rid of the deleted files now so they don't stay around indefinitely. */ + ORDERED_HASHMAP_FOREACH(f, j->files) { + r = journal_file_fstat(f); + if (r == -EIDRM) + remove_file_real(j, f); + else if (r < 0) { + log_debug_errno(r,"Failed to fstat() journal file '%s' : %m", f->path); + continue; + } + } + + /* The journal might have changed since the context + * object was created and we weren't watching before, + * hence don't wait for anything, and return + * immediately. */ + return determine_change(j); + } + + r = sd_journal_get_timeout(j, &t); + if (r < 0) + return r; + + if (t != UINT64_MAX) { + t = usec_sub_unsigned(t, now(CLOCK_MONOTONIC)); + + if (timeout_usec == UINT64_MAX || timeout_usec > t) + timeout_usec = t; + } + + do { + r = fd_wait_for_event(j->inotify_fd, POLLIN, timeout_usec); + } while (r == -EINTR); + + if (r < 0) + return r; + + return sd_journal_process(j); +} + +_public_ int sd_journal_get_cutoff_realtime_usec(sd_journal *j, uint64_t *from, uint64_t *to) { + JournalFile *f; + bool first = true; + uint64_t fmin = 0, tmax = 0; + int r; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(from || to, -EINVAL); + assert_return(from != to, -EINVAL); + + ORDERED_HASHMAP_FOREACH(f, j->files) { + usec_t fr, t; + + r = journal_file_get_cutoff_realtime_usec(f, &fr, &t); + if (r == -ENOENT) + continue; + if (r < 0) + return r; + if (r == 0) + continue; + + if (first) { + fmin = fr; + tmax = t; + first = false; + } else { + fmin = MIN(fr, fmin); + tmax = MAX(t, tmax); + } + } + + if (from) + *from = fmin; + if (to) + *to = tmax; + + return first ? 0 : 1; +} + +_public_ int sd_journal_get_cutoff_monotonic_usec( + sd_journal *j, + sd_id128_t boot_id, + uint64_t *ret_from, + uint64_t *ret_to) { + + uint64_t from = UINT64_MAX, to = UINT64_MAX; + bool found = false; + JournalFile *f; + int r; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(ret_from != ret_to, -EINVAL); + + ORDERED_HASHMAP_FOREACH(f, j->files) { + usec_t ff, tt; + + r = journal_file_get_cutoff_monotonic_usec(f, boot_id, &ff, &tt); + if (r == -ENOENT) + continue; + if (r < 0) + return r; + if (r == 0) + continue; + + if (found) { + from = MIN(ff, from); + to = MAX(tt, to); + } else { + from = ff; + to = tt; + found = true; + } + } + + if (ret_from) + *ret_from = from; + if (ret_to) + *ret_to = to; + + return found; +} + +void journal_print_header(sd_journal *j) { + JournalFile *f; + bool newline = false; + + assert(j); + + ORDERED_HASHMAP_FOREACH(f, j->files) { + if (newline) + putchar('\n'); + else + newline = true; + + journal_file_print_header(f); + } +} + +_public_ int sd_journal_get_usage(sd_journal *j, uint64_t *ret) { + JournalFile *f; + uint64_t sum = 0; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(ret, -EINVAL); + + ORDERED_HASHMAP_FOREACH(f, j->files) { + struct stat st; + uint64_t b; + + if (fstat(f->fd, &st) < 0) + return -errno; + + b = (uint64_t) st.st_blocks; + if (b > UINT64_MAX / 512) + return -EOVERFLOW; + b *= 512; + + if (sum > UINT64_MAX - b) + return -EOVERFLOW; + sum += b; + } + + *ret = sum; + return 0; +} + +_public_ int sd_journal_query_unique(sd_journal *j, const char *field) { + int r; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(!isempty(field), -EINVAL); + + if (!field_is_valid(field)) + return -EINVAL; + + r = free_and_strdup(&j->unique_field, field); + if (r < 0) + return r; + + j->unique_file = NULL; + j->unique_offset = 0; + j->unique_file_lost = false; + + return 0; +} + +_public_ int sd_journal_enumerate_unique( + sd_journal *j, + const void **ret_data, + size_t *ret_size) { + + size_t k; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(j->unique_field, -EINVAL); + + k = strlen(j->unique_field); + + if (!j->unique_file) { + if (j->unique_file_lost) + return 0; + + j->unique_file = ordered_hashmap_first(j->files); + if (!j->unique_file) + return 0; + + j->unique_offset = 0; + } + + for (;;) { + JournalFile *of; + Object *o; + void *odata; + size_t ol; + bool found; + int r; + + /* Proceed to next data object in the field's linked list */ + if (j->unique_offset == 0) { + r = journal_file_find_field_object(j->unique_file, j->unique_field, k, &o, NULL); + if (r < 0) + return r; + + j->unique_offset = r > 0 ? le64toh(o->field.head_data_offset) : 0; + } else { + r = journal_file_move_to_object(j->unique_file, OBJECT_DATA, j->unique_offset, &o); + if (r < 0) + return r; + + j->unique_offset = le64toh(o->data.next_field_offset); + } + + /* We reached the end of the list? Then start again, with the next file */ + if (j->unique_offset == 0) { + j->unique_file = ordered_hashmap_next(j->files, j->unique_file->path); + if (!j->unique_file) + return 0; + + continue; + } + + /* We do not use OBJECT_DATA context here, but OBJECT_UNUSED + * instead, so that we can look at this data object at the same + * time as one on another file */ + r = journal_file_move_to_object(j->unique_file, OBJECT_UNUSED, j->unique_offset, &o); + if (r < 0) + return r; + + /* Let's do the type check by hand, since we used 0 context above. */ + if (o->object.type != OBJECT_DATA) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "%s:offset " OFSfmt ": object has type %d, expected %d", + j->unique_file->path, + j->unique_offset, + o->object.type, OBJECT_DATA); + + r = journal_file_data_payload(j->unique_file, o, j->unique_offset, NULL, 0, + j->data_threshold, &odata, &ol); + if (r < 0) + return r; + + /* Check if we have at least the field name and "=". */ + if (ol <= k) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "%s:offset " OFSfmt ": object has size %zu, expected at least %zu", + j->unique_file->path, + j->unique_offset, ol, k + 1); + + if (memcmp(odata, j->unique_field, k) != 0 || ((const char*) odata)[k] != '=') + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "%s:offset " OFSfmt ": object does not start with \"%s=\"", + j->unique_file->path, + j->unique_offset, + j->unique_field); + + /* OK, now let's see if we already returned this data object by checking if it exists in the + * earlier traversed files. */ + found = false; + ORDERED_HASHMAP_FOREACH(of, j->files) { + if (of == j->unique_file) + break; + + /* Skip this file it didn't have any fields indexed */ + if (JOURNAL_HEADER_CONTAINS(of->header, n_fields) && le64toh(of->header->n_fields) <= 0) + continue; + + /* We can reuse the hash from our current file only on old-style journal files + * without keyed hashes. On new-style files we have to calculate the hash anew, to + * take the per-file hash seed into consideration. */ + if (!JOURNAL_HEADER_KEYED_HASH(j->unique_file->header) && !JOURNAL_HEADER_KEYED_HASH(of->header)) + r = journal_file_find_data_object_with_hash(of, odata, ol, le64toh(o->data.hash), NULL, NULL); + else + r = journal_file_find_data_object(of, odata, ol, NULL, NULL); + if (r < 0) + return r; + if (r > 0) { + found = true; + break; + } + } + + if (found) + continue; + + *ret_data = odata; + *ret_size = ol; + + return 1; + } +} + +_public_ int sd_journal_enumerate_available_unique(sd_journal *j, const void **data, size_t *size) { + for (;;) { + int r; + + r = sd_journal_enumerate_unique(j, data, size); + if (r >= 0) + return r; + if (!JOURNAL_ERRNO_IS_UNAVAILABLE_FIELD(r)) + return r; + /* Try with the next field. sd_journal_enumerate_unique() modifies state, so on the next try + * we will access the next field. */ + } +} + +_public_ void sd_journal_restart_unique(sd_journal *j) { + if (!j) + return; + + j->unique_file = NULL; + j->unique_offset = 0; + j->unique_file_lost = false; +} + +_public_ int sd_journal_enumerate_fields(sd_journal *j, const char **field) { + int r; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(field, -EINVAL); + + if (!j->fields_file) { + if (j->fields_file_lost) + return 0; + + j->fields_file = ordered_hashmap_first(j->files); + if (!j->fields_file) + return 0; + + j->fields_hash_table_index = 0; + j->fields_offset = 0; + } + + for (;;) { + JournalFile *f, *of; + uint64_t m; + Object *o; + size_t sz; + bool found; + + f = j->fields_file; + + if (j->fields_offset == 0) { + bool eof = false; + + /* We are not yet positioned at any field. Let's pick the first one */ + r = journal_file_map_field_hash_table(f); + if (r < 0) + return r; + + m = le64toh(f->header->field_hash_table_size) / sizeof(HashItem); + for (;;) { + if (j->fields_hash_table_index >= m) { + /* Reached the end of the hash table, go to the next file. */ + eof = true; + break; + } + + j->fields_offset = le64toh(f->field_hash_table[j->fields_hash_table_index].head_hash_offset); + + if (j->fields_offset != 0) + break; + + /* Empty hash table bucket, go to next one */ + j->fields_hash_table_index++; + } + + if (eof) { + /* Proceed with next file */ + j->fields_file = ordered_hashmap_next(j->files, f->path); + if (!j->fields_file) { + *field = NULL; + return 0; + } + + j->fields_offset = 0; + j->fields_hash_table_index = 0; + continue; + } + + } else { + /* We are already positioned at a field. If so, let's figure out the next field from it */ + + r = journal_file_move_to_object(f, OBJECT_FIELD, j->fields_offset, &o); + if (r < 0) + return r; + + j->fields_offset = le64toh(o->field.next_hash_offset); + if (j->fields_offset == 0) { + /* Reached the end of the hash table chain */ + j->fields_hash_table_index++; + continue; + } + } + + /* We use OBJECT_UNUSED here, so that the iterator below doesn't remove our mmap window */ + r = journal_file_move_to_object(f, OBJECT_UNUSED, j->fields_offset, &o); + if (r < 0) + return r; + + /* Because we used OBJECT_UNUSED above, we need to do our type check manually */ + if (o->object.type != OBJECT_FIELD) + return log_debug_errno(SYNTHETIC_ERRNO(EBADMSG), + "%s:offset " OFSfmt ": object has type %i, expected %i", + f->path, j->fields_offset, + o->object.type, OBJECT_FIELD); + + sz = le64toh(o->object.size) - offsetof(Object, field.payload); + + /* Let's see if we already returned this field name before. */ + found = false; + ORDERED_HASHMAP_FOREACH(of, j->files) { + if (of == f) + break; + + /* Skip this file it didn't have any fields indexed */ + if (JOURNAL_HEADER_CONTAINS(of->header, n_fields) && le64toh(of->header->n_fields) <= 0) + continue; + + if (!JOURNAL_HEADER_KEYED_HASH(f->header) && !JOURNAL_HEADER_KEYED_HASH(of->header)) + r = journal_file_find_field_object_with_hash(of, o->field.payload, sz, + le64toh(o->field.hash), NULL, NULL); + else + r = journal_file_find_field_object(of, o->field.payload, sz, NULL, NULL); + if (r < 0) + return r; + if (r > 0) { + found = true; + break; + } + } + + if (found) + continue; + + /* Check if this is really a valid string containing no NUL byte */ + if (memchr(o->field.payload, 0, sz)) + return -EBADMSG; + + if (j->data_threshold > 0 && sz > j->data_threshold) + sz = j->data_threshold; + + if (!GREEDY_REALLOC(j->fields_buffer, sz + 1)) + return -ENOMEM; + + memcpy(j->fields_buffer, o->field.payload, sz); + j->fields_buffer[sz] = 0; + + if (!field_is_valid(j->fields_buffer)) + return -EBADMSG; + + *field = j->fields_buffer; + return 1; + } +} + +_public_ void sd_journal_restart_fields(sd_journal *j) { + if (!j) + return; + + j->fields_file = NULL; + j->fields_hash_table_index = 0; + j->fields_offset = 0; + j->fields_file_lost = false; +} + +_public_ int sd_journal_reliable_fd(sd_journal *j) { + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + return !j->on_network; +} + +static char *lookup_field(const char *field, void *userdata) { + sd_journal *j = ASSERT_PTR(userdata); + const void *data; + size_t size, d; + int r; + + assert(field); + + r = sd_journal_get_data(j, field, &data, &size); + if (r < 0 || + size > REPLACE_VAR_MAX) + return strdup(field); + + d = strlen(field) + 1; + + return strndup((const char*) data + d, size - d); +} + +_public_ int sd_journal_get_catalog(sd_journal *j, char **ret) { + const void *data; + size_t size; + sd_id128_t id; + _cleanup_free_ char *text = NULL, *cid = NULL; + char *t; + int r; + + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(ret, -EINVAL); + + r = sd_journal_get_data(j, "MESSAGE_ID", &data, &size); + if (r < 0) + return r; + + cid = strndup((const char*) data + 11, size - 11); + if (!cid) + return -ENOMEM; + + r = sd_id128_from_string(cid, &id); + if (r < 0) + return r; + + r = catalog_get(CATALOG_DATABASE, id, &text); + if (r < 0) + return r; + + t = replace_var(text, lookup_field, j); + if (!t) + return -ENOMEM; + + *ret = t; + return 0; +} + +_public_ int sd_journal_get_catalog_for_message_id(sd_id128_t id, char **ret) { + assert_return(ret, -EINVAL); + + return catalog_get(CATALOG_DATABASE, id, ret); +} + +_public_ int sd_journal_set_data_threshold(sd_journal *j, size_t sz) { + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + + j->data_threshold = sz; + return 0; +} + +_public_ int sd_journal_get_data_threshold(sd_journal *j, size_t *sz) { + assert_return(j, -EINVAL); + assert_return(!journal_pid_changed(j), -ECHILD); + assert_return(sz, -EINVAL); + + *sz = j->data_threshold; + return 0; +} + +_public_ int sd_journal_has_runtime_files(sd_journal *j) { + assert_return(j, -EINVAL); + + return j->has_runtime_files; +} + +_public_ int sd_journal_has_persistent_files(sd_journal *j) { + assert_return(j, -EINVAL); + + return j->has_persistent_files; +} diff --git a/src/libsystemd/sd-journal/test-audit-type.c b/src/libsystemd/sd-journal/test-audit-type.c new file mode 100644 index 0000000..1d5003b --- /dev/null +++ b/src/libsystemd/sd-journal/test-audit-type.c @@ -0,0 +1,24 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ + +#include <stdio.h> +#include <linux/audit.h> + +#include "audit-type.h" +#include "tests.h" + +static void print_audit_label(int i) { + const char *name; + + name = audit_type_name_alloca(i); + /* This is a separate function only because of alloca */ + printf("%i → %s → %s\n", i, audit_type_to_string(i), name); +} + +TEST(audit_type) { + int i; + + for (i = 0; i <= AUDIT_KERNEL; i++) + print_audit_label(i); +} + +DEFINE_TEST_MAIN(LOG_INFO); diff --git a/src/libsystemd/sd-journal/test-catalog.c b/src/libsystemd/sd-journal/test-catalog.c new file mode 100644 index 0000000..9a15915 --- /dev/null +++ b/src/libsystemd/sd-journal/test-catalog.c @@ -0,0 +1,236 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ + +#include <errno.h> +#include <fcntl.h> +#include <locale.h> +#include <unistd.h> + +#include "sd-messages.h" + +#include "alloc-util.h" +#include "catalog.h" +#include "fd-util.h" +#include "fs-util.h" +#include "log.h" +#include "macro.h" +#include "path-util.h" +#include "string-util.h" +#include "strv.h" +#include "tests.h" +#include "tmpfile-util.h" +#include "util.h" + +static char** catalog_dirs = NULL; +static const char *no_catalog_dirs[] = { + "/bin/hopefully/with/no/catalog", + NULL +}; + +static OrderedHashmap* test_import(const char* contents, ssize_t size, int code) { + _cleanup_(unlink_tempfilep) char name[] = "/tmp/test-catalog.XXXXXX"; + _cleanup_close_ int fd = -EBADF; + OrderedHashmap *h; + + if (size < 0) + size = strlen(contents); + + assert_se(h = ordered_hashmap_new(&catalog_hash_ops)); + + fd = mkostemp_safe(name); + assert_se(fd >= 0); + assert_se(write(fd, contents, size) == size); + + assert_se(catalog_import_file(h, name) == code); + + return h; +} + +static void test_catalog_import_invalid(void) { + _cleanup_ordered_hashmap_free_free_free_ OrderedHashmap *h = NULL; + + h = test_import("xxx", -1, -EINVAL); + assert_se(ordered_hashmap_isempty(h)); +} + +static void test_catalog_import_badid(void) { + _unused_ _cleanup_ordered_hashmap_free_free_free_ OrderedHashmap *h = NULL; + const char *input = +"-- 0027229ca0644181a76c4e92458afaff dededededededededededededededede\n" \ +"Subject: message\n" \ +"\n" \ +"payload\n"; + h = test_import(input, -1, -EINVAL); +} + +static void test_catalog_import_one(void) { + _cleanup_ordered_hashmap_free_free_free_ OrderedHashmap *h = NULL; + char *payload; + + const char *input = +"-- 0027229ca0644181a76c4e92458afaff dededededededededededededededed\n" \ +"Subject: message\n" \ +"\n" \ +"payload\n"; + const char *expect = +"Subject: message\n" \ +"\n" \ +"payload\n"; + + h = test_import(input, -1, 0); + assert_se(ordered_hashmap_size(h) == 1); + + ORDERED_HASHMAP_FOREACH(payload, h) { + printf("expect: %s\n", expect); + printf("actual: %s\n", payload); + assert_se(streq(expect, payload)); + } +} + +static void test_catalog_import_merge(void) { + _cleanup_ordered_hashmap_free_free_free_ OrderedHashmap *h = NULL; + char *payload; + + const char *input = +"-- 0027229ca0644181a76c4e92458afaff dededededededededededededededed\n" \ +"Subject: message\n" \ +"Defined-By: me\n" \ +"\n" \ +"payload\n" \ +"\n" \ +"-- 0027229ca0644181a76c4e92458afaff dededededededededededededededed\n" \ +"Subject: override subject\n" \ +"X-Header: hello\n" \ +"\n" \ +"override payload\n"; + + const char *combined = +"Subject: override subject\n" \ +"X-Header: hello\n" \ +"Subject: message\n" \ +"Defined-By: me\n" \ +"\n" \ +"override payload\n"; + + h = test_import(input, -1, 0); + assert_se(ordered_hashmap_size(h) == 1); + + ORDERED_HASHMAP_FOREACH(payload, h) + assert_se(streq(combined, payload)); +} + +static void test_catalog_import_merge_no_body(void) { + _cleanup_ordered_hashmap_free_free_free_ OrderedHashmap *h = NULL; + char *payload; + + const char *input = +"-- 0027229ca0644181a76c4e92458afaff dededededededededededededededed\n" \ +"Subject: message\n" \ +"Defined-By: me\n" \ +"\n" \ +"payload\n" \ +"\n" \ +"-- 0027229ca0644181a76c4e92458afaff dededededededededededededededed\n" \ +"Subject: override subject\n" \ +"X-Header: hello\n" \ +"\n"; + + const char *combined = +"Subject: override subject\n" \ +"X-Header: hello\n" \ +"Subject: message\n" \ +"Defined-By: me\n" \ +"\n" \ +"payload\n"; + + h = test_import(input, -1, 0); + assert_se(ordered_hashmap_size(h) == 1); + + ORDERED_HASHMAP_FOREACH(payload, h) + assert_se(streq(combined, payload)); +} + +static void test_catalog_update(const char *database) { + int r; + + /* Test what happens if there are no files. */ + r = catalog_update(database, NULL, NULL); + assert_se(r == 0); + + /* Test what happens if there are no files in the directory. */ + r = catalog_update(database, NULL, no_catalog_dirs); + assert_se(r == 0); + + /* Make sure that we at least have some files loaded or the + * catalog_list below will fail. */ + r = catalog_update(database, NULL, (const char * const *) catalog_dirs); + assert_se(r == 0); +} + +static void test_catalog_file_lang(void) { + _cleanup_free_ char *lang = NULL, *lang2 = NULL, *lang3 = NULL, *lang4 = NULL; + + assert_se(catalog_file_lang("systemd.de_DE.catalog", &lang) == 1); + assert_se(streq(lang, "de_DE")); + + assert_se(catalog_file_lang("systemd..catalog", &lang2) == 0); + assert_se(lang2 == NULL); + + assert_se(catalog_file_lang("systemd.fr.catalog", &lang2) == 1); + assert_se(streq(lang2, "fr")); + + assert_se(catalog_file_lang("systemd.fr.catalog.gz", &lang3) == 0); + assert_se(lang3 == NULL); + + assert_se(catalog_file_lang("systemd.01234567890123456789012345678901.catalog", &lang3) == 0); + assert_se(lang3 == NULL); + + assert_se(catalog_file_lang("systemd.0123456789012345678901234567890.catalog", &lang3) == 1); + assert_se(streq(lang3, "0123456789012345678901234567890")); + + assert_se(catalog_file_lang("/x/y/systemd.catalog", &lang4) == 0); + assert_se(lang4 == NULL); + + assert_se(catalog_file_lang("/x/y/systemd.ru_RU.catalog", &lang4) == 1); + assert_se(streq(lang4, "ru_RU")); +} + +int main(int argc, char *argv[]) { + _cleanup_(unlink_tempfilep) char database[] = "/tmp/test-catalog.XXXXXX"; + _cleanup_close_ int fd = -1; + _cleanup_free_ char *text = NULL; + int r; + + setlocale(LC_ALL, "de_DE.UTF-8"); + + test_setup_logging(LOG_DEBUG); + + /* If test-catalog is located at the build directory, then use catalogs in that. + * If it is not, e.g. installed by systemd-tests package, then use installed catalogs. */ + catalog_dirs = STRV_MAKE(get_catalog_dir()); + + assert_se(access(catalog_dirs[0], F_OK) >= 0); + log_notice("Using catalog directory '%s'", catalog_dirs[0]); + + test_catalog_file_lang(); + + test_catalog_import_invalid(); + test_catalog_import_badid(); + test_catalog_import_one(); + test_catalog_import_merge(); + test_catalog_import_merge_no_body(); + + assert_se((fd = mkostemp_safe(database)) >= 0); + + test_catalog_update(database); + + r = catalog_list(stdout, database, true); + assert_se(r >= 0); + + r = catalog_list(stdout, database, false); + assert_se(r >= 0); + + assert_se(catalog_get(database, SD_MESSAGE_COREDUMP, &text) >= 0); + printf(">>>%s<<<\n", text); + + return 0; +} diff --git a/src/libsystemd/sd-journal/test-journal-enum.c b/src/libsystemd/sd-journal/test-journal-enum.c new file mode 100644 index 0000000..03fe8e2 --- /dev/null +++ b/src/libsystemd/sd-journal/test-journal-enum.c @@ -0,0 +1,37 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ + +#include <stdio.h> + +#include "sd-journal.h" + +#include "journal-internal.h" +#include "log.h" +#include "macro.h" +#include "tests.h" + +int main(int argc, char *argv[]) { + unsigned n = 0; + _cleanup_(sd_journal_closep) sd_journal *j = NULL; + + test_setup_logging(LOG_DEBUG); + + assert_se(sd_journal_open(&j, SD_JOURNAL_LOCAL_ONLY) >= 0); + + assert_se(sd_journal_add_match(j, "_TRANSPORT=syslog", 0) >= 0); + assert_se(sd_journal_add_match(j, "_UID=0", 0) >= 0); + + SD_JOURNAL_FOREACH_BACKWARDS(j) { + const void *d; + size_t l; + + assert_se(sd_journal_get_data(j, "MESSAGE", &d, &l) >= 0); + + printf("%.*s\n", (int) l, (char*) d); + + n++; + if (n >= 10) + break; + } + + return 0; +} diff --git a/src/libsystemd/sd-journal/test-journal-init.c b/src/libsystemd/sd-journal/test-journal-init.c new file mode 100644 index 0000000..80aff75 --- /dev/null +++ b/src/libsystemd/sd-journal/test-journal-init.c @@ -0,0 +1,50 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ + +#include "sd-journal.h" + +#include "chattr-util.h" +#include "log.h" +#include "parse-util.h" +#include "rm-rf.h" +#include "tests.h" +#include "util.h" + +int main(int argc, char *argv[]) { + sd_journal *j; + int r, i, I = 100; + char t[] = "/var/tmp/journal-stream-XXXXXX"; + + test_setup_logging(LOG_DEBUG); + + if (argc >= 2) { + r = safe_atoi(argv[1], &I); + if (r < 0) + log_info("Could not parse loop count argument. Using default."); + } + + log_info("Running %d loops", I); + + assert_se(mkdtemp(t)); + (void) chattr_path(t, FS_NOCOW_FL, FS_NOCOW_FL, NULL); + + for (i = 0; i < I; i++) { + r = sd_journal_open(&j, SD_JOURNAL_LOCAL_ONLY); + assert_se(r == 0); + + sd_journal_close(j); + + r = sd_journal_open_directory(&j, t, 0); + assert_se(r == 0); + + sd_journal_close(j); + + j = NULL; + r = sd_journal_open_directory(&j, t, SD_JOURNAL_LOCAL_ONLY); + assert_se(r == -EINVAL); + assert_se(j == NULL); + } + + assert_se(rm_rf(t, REMOVE_ROOT|REMOVE_PHYSICAL) >= 0); + + return 0; +} diff --git a/src/libsystemd/sd-journal/test-journal-match.c b/src/libsystemd/sd-journal/test-journal-match.c new file mode 100644 index 0000000..ded6756 --- /dev/null +++ b/src/libsystemd/sd-journal/test-journal-match.c @@ -0,0 +1,62 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ + +#include <stdio.h> + +#include "sd-journal.h" + +#include "alloc-util.h" +#include "journal-internal.h" +#include "log.h" +#include "string-util.h" +#include "tests.h" +#include "util.h" + +int main(int argc, char *argv[]) { + _cleanup_(sd_journal_closep) sd_journal *j = NULL; + _cleanup_free_ char *t; + + test_setup_logging(LOG_DEBUG); + + assert_se(sd_journal_open(&j, 0) >= 0); + + assert_se(sd_journal_add_match(j, "foobar", 0) < 0); + assert_se(sd_journal_add_match(j, "foobar=waldo", 0) < 0); + assert_se(sd_journal_add_match(j, "", 0) < 0); + assert_se(sd_journal_add_match(j, "=", 0) < 0); + assert_se(sd_journal_add_match(j, "=xxxxx", 0) < 0); + assert_se(sd_journal_add_match(j, (uint8_t[4]){'A', '=', '\1', '\2'}, 4) >= 0); + assert_se(sd_journal_add_match(j, (uint8_t[5]){'B', '=', 'C', '\0', 'D'}, 5) >= 0); + assert_se(sd_journal_add_match(j, "HALLO=WALDO", 0) >= 0); + assert_se(sd_journal_add_match(j, "QUUX=mmmm", 0) >= 0); + assert_se(sd_journal_add_match(j, "QUUX=xxxxx", 0) >= 0); + assert_se(sd_journal_add_match(j, "HALLO=", 0) >= 0); + assert_se(sd_journal_add_match(j, "QUUX=xxxxx", 0) >= 0); + assert_se(sd_journal_add_match(j, "QUUX=yyyyy", 0) >= 0); + assert_se(sd_journal_add_match(j, "PIFF=paff", 0) >= 0); + + assert_se(sd_journal_add_disjunction(j) >= 0); + + assert_se(sd_journal_add_match(j, "ONE=one", 0) >= 0); + assert_se(sd_journal_add_match(j, "ONE=two", 0) >= 0); + assert_se(sd_journal_add_match(j, "TWO=two", 0) >= 0); + + assert_se(sd_journal_add_conjunction(j) >= 0); + + assert_se(sd_journal_add_match(j, "L4_1=yes", 0) >= 0); + assert_se(sd_journal_add_match(j, "L4_1=ok", 0) >= 0); + assert_se(sd_journal_add_match(j, "L4_2=yes", 0) >= 0); + assert_se(sd_journal_add_match(j, "L4_2=ok", 0) >= 0); + + assert_se(sd_journal_add_disjunction(j) >= 0); + + assert_se(sd_journal_add_match(j, "L3=yes", 0) >= 0); + assert_se(sd_journal_add_match(j, "L3=ok", 0) >= 0); + + assert_se(t = journal_make_match_string(j)); + + printf("resulting match expression is: %s\n", t); + + assert_se(streq(t, "(((L3=ok OR L3=yes) OR ((L4_2=ok OR L4_2=yes) AND (L4_1=ok OR L4_1=yes))) AND ((TWO=two AND (ONE=two OR ONE=one)) OR (PIFF=paff AND (QUUX=yyyyy OR QUUX=xxxxx OR QUUX=mmmm) AND (HALLO= OR HALLO=WALDO) AND B=C\\000D AND A=\\001\\002)))")); + + return 0; +} diff --git a/src/libsystemd/sd-journal/test-journal-send.c b/src/libsystemd/sd-journal/test-journal-send.c new file mode 100644 index 0000000..ca1fe7c --- /dev/null +++ b/src/libsystemd/sd-journal/test-journal-send.c @@ -0,0 +1,111 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ + +#include <errno.h> +#include <stdlib.h> +#include <unistd.h> + +#include "sd-journal.h" + +#include "fileio.h" +#include "journal-send.h" +#include "macro.h" +#include "memory-util.h" +#include "tests.h" + +TEST(journal_print) { + assert_se(sd_journal_print(LOG_INFO, "XXX") == 0); + assert_se(sd_journal_print(LOG_INFO, "%s", "YYY") == 0); + assert_se(sd_journal_print(LOG_INFO, "X%4094sY", "ZZZ") == 0); + assert_se(sd_journal_print(LOG_INFO, "X%*sY", (int) LONG_LINE_MAX - 8 - 3, "ZZZ") == 0); + assert_se(sd_journal_print(LOG_INFO, "X%*sY", (int) LONG_LINE_MAX - 8 - 2, "ZZZ") == -ENOBUFS); +} + +TEST(journal_send) { + _cleanup_free_ char *huge = NULL; + +#define HUGE_SIZE (4096*1024) + assert_se(huge = malloc(HUGE_SIZE)); + + /* utf-8 and non-utf-8, message-less and message-ful iovecs */ + struct iovec graph1[] = { + {(char*) "GRAPH=graph", STRLEN("GRAPH=graph")} + }; + struct iovec graph2[] = { + {(char*) "GRAPH=graph\n", STRLEN("GRAPH=graph\n")} + }; + struct iovec message1[] = { + {(char*) "MESSAGE=graph", STRLEN("MESSAGE=graph")} + }; + struct iovec message2[] = { + {(char*) "MESSAGE=graph\n", STRLEN("MESSAGE=graph\n")} + }; + + assert_se(sd_journal_print(LOG_INFO, "piepapo") == 0); + + assert_se(sd_journal_send("MESSAGE=foobar", + "VALUE=%i", 7, + NULL) == 0); + + errno = ENOENT; + assert_se(sd_journal_perror("Foobar") == 0); + + assert_se(sd_journal_perror("") == 0); + + memcpy(huge, "HUGE=", STRLEN("HUGE=")); + memset(&huge[STRLEN("HUGE=")], 'x', HUGE_SIZE - STRLEN("HUGE=") - 1); + huge[HUGE_SIZE - 1] = '\0'; + + assert_se(sd_journal_send("MESSAGE=Huge field attached", + huge, + NULL) == 0); + + assert_se(sd_journal_send("MESSAGE=uiui", + "VALUE=A", + "VALUE=B", + "VALUE=C", + "SINGLETON=1", + "OTHERVALUE=X", + "OTHERVALUE=Y", + "WITH_BINARY=this is a binary value \a", + NULL) == 0); + + syslog(LOG_NOTICE, "Hello World!"); + + assert_se(sd_journal_print(LOG_NOTICE, "Hello World") == 0); + + assert_se(sd_journal_send("MESSAGE=Hello World!", + "MESSAGE_ID=52fb62f99e2c49d89cfbf9d6de5e3555", + "PRIORITY=5", + "HOME=%s", getenv("HOME"), + "TERM=%s", getenv("TERM"), + "PAGE_SIZE=%li", sysconf(_SC_PAGESIZE), + "N_CPUS=%li", sysconf(_SC_NPROCESSORS_ONLN), + NULL) == 0); + + assert_se(sd_journal_sendv(graph1, 1) == 0); + assert_se(sd_journal_sendv(graph2, 1) == 0); + assert_se(sd_journal_sendv(message1, 1) == 0); + assert_se(sd_journal_sendv(message2, 1) == 0); + + /* test without location fields */ +#undef sd_journal_sendv + assert_se(sd_journal_sendv(graph1, 1) == 0); + assert_se(sd_journal_sendv(graph2, 1) == 0); + assert_se(sd_journal_sendv(message1, 1) == 0); + assert_se(sd_journal_sendv(message2, 1) == 0); + + /* The above syslog() opens a fd which is stored in libc, and the valgrind reports the fd is + * leaked when we do not call closelog(). */ + closelog(); +} + +static int outro(void) { + /* Sleep a bit to make it easy for journald to collect metadata. */ + sleep(1); + + close_journal_fd(); + + return EXIT_SUCCESS; +} + +DEFINE_TEST_MAIN_FULL(LOG_INFO, NULL, outro); diff --git a/src/libsystemd/sd-journal/test-mmap-cache.c b/src/libsystemd/sd-journal/test-mmap-cache.c new file mode 100644 index 0000000..6ec3693 --- /dev/null +++ b/src/libsystemd/sd-journal/test-mmap-cache.c @@ -0,0 +1,66 @@ +/* SPDX-License-Identifier: LGPL-2.1-or-later */ + +#include <fcntl.h> +#include <stdlib.h> +#include <sys/mman.h> +#include <unistd.h> + +#include "fd-util.h" +#include "macro.h" +#include "mmap-cache.h" +#include "tmpfile-util.h" +#include "util.h" + +int main(int argc, char *argv[]) { + MMapFileDescriptor *fx; + int x, y, z, r; + char px[] = "/tmp/testmmapXXXXXXX", py[] = "/tmp/testmmapYXXXXXX", pz[] = "/tmp/testmmapZXXXXXX"; + MMapCache *m; + void *p, *q; + + assert_se(m = mmap_cache_new()); + + x = mkostemp_safe(px); + assert_se(x >= 0); + unlink(px); + + assert_se(fx = mmap_cache_add_fd(m, x, PROT_READ)); + + y = mkostemp_safe(py); + assert_se(y >= 0); + unlink(py); + + z = mkostemp_safe(pz); + assert_se(z >= 0); + unlink(pz); + + r = mmap_cache_fd_get(fx, 0, false, 1, 2, NULL, &p); + assert_se(r >= 0); + + r = mmap_cache_fd_get(fx, 0, false, 2, 2, NULL, &q); + assert_se(r >= 0); + + assert_se((uint8_t*) p + 1 == (uint8_t*) q); + + r = mmap_cache_fd_get(fx, 1, false, 3, 2, NULL, &q); + assert_se(r >= 0); + + assert_se((uint8_t*) p + 2 == (uint8_t*) q); + + r = mmap_cache_fd_get(fx, 0, false, 16ULL*1024ULL*1024ULL, 2, NULL, &p); + assert_se(r >= 0); + + r = mmap_cache_fd_get(fx, 1, false, 16ULL*1024ULL*1024ULL+1, 2, NULL, &q); + assert_se(r >= 0); + + assert_se((uint8_t*) p + 1 == (uint8_t*) q); + + mmap_cache_fd_free(fx); + mmap_cache_unref(m); + + safe_close(x); + safe_close(y); + safe_close(z); + + return 0; +} |