diff options
Diffstat (limited to 'src/util/dict_open.c')
-rw-r--r-- | src/util/dict_open.c | 600 |
1 files changed, 600 insertions, 0 deletions
diff --git a/src/util/dict_open.c b/src/util/dict_open.c new file mode 100644 index 0000000..afea391 --- /dev/null +++ b/src/util/dict_open.c @@ -0,0 +1,600 @@ +/*++ +/* NAME +/* dict_open 3 +/* SUMMARY +/* low-level dictionary interface +/* SYNOPSIS +/* #include <dict.h> +/* +/* DICT *dict_open(dict_spec, open_flags, dict_flags) +/* const char *dict_spec; +/* int open_flags; +/* int dict_flags; +/* +/* DICT *dict_open3(dict_type, dict_name, open_flags, dict_flags) +/* const char *dict_type; +/* const char *dict_name; +/* int open_flags; +/* int dict_flags; +/* +/* int dict_put(dict, key, value) +/* DICT *dict; +/* const char *key; +/* const char *value; +/* +/* const char *dict_get(dict, key) +/* DICT *dict; +/* const char *key; +/* +/* int dict_del(dict, key) +/* DICT *dict; +/* const char *key; +/* +/* int dict_seq(dict, func, key, value) +/* DICT *dict; +/* int func; +/* const char **key; +/* const char **value; +/* +/* void dict_close(dict) +/* DICT *dict; +/* +/* typedef DICT *(*DICT_OPEN_FN) (const char *, int, int); +/* +/* dict_open_register(type, open) +/* const char *type; +/* DICT_OPEN_FN open; +/* +/* typedef DICT_OPEN_FN (*DICT_OPEN_EXTEND_FN)(const char *type); +/* +/* DICT_OPEN_EXTEND_FN dict_open_extend(call_back) +/* DICT_OPEN_EXTEND_FN call_back; +/* +/* ARGV *dict_mapnames() +/* +/* typedef ARGV *(*DICT_MAPNAMES_EXTEND_FN)(ARGV *names); +/* +/* DICT_MAPNAMES_EXTEND_FN dict_mapnames_extend(call_back) +/* DICT_MAPNAMES_EXTEND_FN call_back; +/* +/* int dict_isjmp(dict) +/* DICT *dict; +/* +/* int dict_setjmp(dict) +/* DICT *dict; +/* +/* int dict_longjmp(dict, val) +/* DICT *dict; +/* int val; +/* +/* void dict_type_override(dict, type) +/* DICT *dict; +/* const char *type; +/* DESCRIPTION +/* This module implements a low-level interface to multiple +/* physical dictionary types. +/* +/* dict_open() takes a type:name pair that specifies a dictionary type +/* and dictionary name, opens the dictionary, and returns a dictionary +/* handle. The \fIopen_flags\fR arguments are as in open(2). The +/* \fIdict_flags\fR are the bit-wise OR of zero or more of the following: +/* .IP DICT_FLAG_DUP_WARN +/* Warn about duplicate keys, if the underlying database does not +/* support duplicate keys. The default is to terminate with a fatal +/* error. +/* .IP DICT_FLAG_DUP_IGNORE +/* Ignore duplicate keys if the underlying database does not +/* support duplicate keys. The default is to terminate with a fatal +/* error. +/* .IP DICT_FLAG_DUP_REPLACE +/* Replace duplicate keys if the underlying database supports such +/* an operation. The default is to terminate with a fatal error. +/* .IP DICT_FLAG_TRY0NULL +/* With maps where this is appropriate, append no null byte to +/* keys and values. +/* When neither DICT_FLAG_TRY0NULL nor DICT_FLAG_TRY1NULL are +/* specified, the software guesses what format to use for reading; +/* and in the absence of definite information, a system-dependent +/* default is chosen for writing. +/* .IP DICT_FLAG_TRY1NULL +/* With maps where this is appropriate, append one null byte to +/* keys and values. +/* When neither DICT_FLAG_TRY0NULL nor DICT_FLAG_TRY1NULL are +/* specified, the software guesses what format to use for reading; +/* and in the absence of definite information, a system-dependent +/* default is chosen for writing. +/* .IP DICT_FLAG_LOCK +/* With maps where this is appropriate, acquire an exclusive lock +/* before writing, and acquire a shared lock before reading. +/* Release the lock when the operation completes. +/* .IP DICT_FLAG_OPEN_LOCK +/* The behavior of this flag depends on whether a database +/* sets the DICT_FLAG_MULTI_WRITER flag to indicate that it +/* is multi-writer safe. +/* +/* With databases that are not multi-writer safe, dict_open() +/* acquires a persistent exclusive lock, or it terminates with +/* a fatal run-time error. +/* +/* With databases that are multi-writer safe, dict_open() +/* downgrades the DICT_FLAG_OPEN_LOCK flag (persistent lock) +/* to DICT_FLAG_LOCK (temporary lock). +/* .IP DICT_FLAG_FOLD_FIX +/* With databases whose lookup fields are fixed-case strings, +/* fold the search string to lower case before accessing the +/* database. This includes hash:, cdb:, dbm:. nis:, ldap:, +/* *sql. WARNING: case folding is supported only for ASCII or +/* valid UTF-8. +/* .IP DICT_FLAG_FOLD_MUL +/* With databases where one lookup field can match both upper +/* and lower case, fold the search key to lower case before +/* accessing the database. This includes regexp: and pcre:. +/* WARNING: case folding is supported only for ASCII or valid +/* UTF-8. +/* .IP DICT_FLAG_FOLD_ANY +/* Short-hand for (DICT_FLAG_FOLD_FIX | DICT_FLAG_FOLD_MUL). +/* .IP DICT_FLAG_SYNC_UPDATE +/* With file-based maps, flush I/O buffers to file after each update. +/* Thus feature is not supported with some file-based dictionaries. +/* .IP DICT_FLAG_NO_REGSUB +/* Disallow regular expression substitution from the lookup string +/* into the lookup result, to block data injection attacks. +/* .IP DICT_FLAG_NO_PROXY +/* Disallow access through the unprivileged \fBproxymap\fR +/* service, to block privilege escalation attacks. +/* .IP DICT_FLAG_NO_UNAUTH +/* Disallow lookup mechanisms that lack any form of authentication, +/* to block privilege escalation attacks (example: tcp_table; +/* even NIS can be secured to some extent by requiring that +/* the server binds to a privileged port). +/* .IP DICT_FLAG_PARANOID +/* A combination of all the paranoia flags: DICT_FLAG_NO_REGSUB, +/* DICT_FLAG_NO_PROXY and DICT_FLAG_NO_UNAUTH. +/* .IP DICT_FLAG_BULK_UPDATE +/* Enable preliminary code for bulk-mode database updates. +/* The caller must create an exception handler with dict_jmp_alloc() +/* and must trap exceptions from the database client with dict_setjmp(). +/* .IP DICT_FLAG_DEBUG +/* Enable additional logging. +/* .IP DICT_FLAG_UTF8_REQUEST +/* With util_utf8_enable != 0, require that lookup/update/delete +/* keys and values are valid UTF-8. Skip a lookup/update/delete +/* request with a non-UTF-8 key, skip an update request with +/* a non-UTF-8 value, and fail a lookup request with a non-UTF-8 +/* value. +/* .IP DICT_FLAG_SRC_RHS_IS_FILE +/* With dictionaries that are created from source text, each +/* value in the source of a dictionary specifies a list of +/* file names separated by comma and/or whitespace. The file +/* contents are concatenated with a newline inserted between +/* files, and the base64-encoded result is stored under the +/* key. +/* .sp +/* NOTE 1: it is up to the application to decode lookup results +/* with dict_file_lookup() or equivalent (this requires that +/* the dictionary is opened with DICT_FLAG_SRC_RHS_IS_FILE). +/* Decoding is not built into the normal dictionary lookup +/* method, because that would complicate dictionary nesting, +/* pipelining, and proxying. +/* .sp +/* NOTE 2: it is up to the application to convert file names +/* into base64-encoded file content before calling the dictionary +/* update method (see dict_file(3) for support). Automatic +/* file content encoding is available only when a dictionary +/* is created from source text. +/* .PP +/* Specify DICT_FLAG_NONE for no special processing. +/* +/* The dictionary types are as follows: +/* .IP environ +/* The process environment array. The \fIdict_name\fR argument is ignored. +/* .IP dbm +/* DBM file. +/* .IP hash +/* Berkeley DB file in hash format. +/* .IP btree +/* Berkeley DB file in btree format. +/* .IP nis +/* NIS map. Only read access is supported. +/* .IP nisplus +/* NIS+ map. Only read access is supported. +/* .IP netinfo +/* NetInfo table. Only read access is supported. +/* .IP ldap +/* LDAP ("light-weight" directory access protocol) database access. +/* .IP pcre +/* PERL-compatible regular expressions. +/* .IP regexp +/* POSIX-compatible regular expressions. +/* .IP texthash +/* Flat text in postmap(1) input format. +/* .PP +/* dict_open3() takes separate arguments for dictionary type and +/* name, but otherwise performs the same functions as dict_open(). +/* +/* The dict_get(), dict_put(), dict_del(), and dict_seq() +/* macros evaluate their first argument multiple times. +/* These names should have been in uppercase. +/* +/* dict_get() retrieves the value stored in the named dictionary +/* under the given key. A null pointer means the value was not found. +/* As with dict_lookup(), the result is owned by the lookup table +/* implementation. Make a copy if the result is to be modified, +/* or if the result is to survive multiple table lookups. +/* +/* dict_put() stores the specified key and value into the named +/* dictionary. A zero (DICT_STAT_SUCCESS) result means the +/* update was made. +/* +/* dict_del() removes a dictionary entry, and returns +/* DICT_STAT_SUCCESS in case of success. +/* +/* dict_seq() iterates over all members in the named dictionary. +/* func is define DICT_SEQ_FUN_FIRST (select first member) or +/* DICT_SEQ_FUN_NEXT (select next member). A zero (DICT_STAT_SUCCESS) +/* result means that an entry was found. +/* +/* dict_close() closes the specified dictionary and cleans up the +/* associated data structures. +/* +/* dict_open_register() adds support for a new dictionary type. +/* +/* dict_open_extend() registers a call-back function that looks +/* up the dictionary open() function for a type that is not +/* registered, or null in case of error. The result value is +/* the last previously-registered call-back or null. +/* +/* dict_mapnames() returns a sorted list with the names of all available +/* dictionary types. +/* +/* dict_mapnames_extend() registers a call-back function that +/* enumerates additional dictionary type names. The result +/* will be sorted by dict_mapnames(). The result value +/* is the last previously-registered call-back or null. +/* +/* dict_setjmp() saves processing context and makes that context +/* available for use with dict_longjmp(). Normally, dict_setjmp() +/* returns zero. A non-zero result means that dict_setjmp() +/* returned through a dict_longjmp() call; the result is the +/* \fIval\fR argument given to dict_longjmp(). dict_isjmp() +/* returns non-zero when dict_setjmp() and dict_longjmp() +/* are enabled for a given dictionary. +/* +/* NB: non-local jumps such as dict_longjmp() are not safe for +/* jumping out of any routine that manipulates DICT data. +/* longjmp() like calls are best avoided in signal handlers. +/* +/* dict_type_override() changes the symbolic dictionary type. +/* This is used by dictionaries whose internals are based on +/* some other dictionary type. +/* DIAGNOSTICS +/* Fatal error: open error, unsupported dictionary type, attempt to +/* update non-writable dictionary. +/* +/* The lookup routine returns non-null when the request is +/* satisfied. The update, delete and sequence routines return +/* zero (DICT_STAT_SUCCESS) when the request is satisfied. +/* The dict->errno value is non-zero only when the last operation +/* was not satisfied due to a dictionary access error. This +/* can have the following values: +/* .IP DICT_ERR_NONE(zero) +/* There was no dictionary access error. For example, the +/* request was satisfied, the requested information did not +/* exist in the dictionary, or the information already existed +/* when it should not exist (collision). +/* .IP DICT_ERR_RETRY(<0) +/* The dictionary was temporarily unavailable. This can happen +/* with network-based services. +/* .IP DICT_ERR_CONFIG(<0) +/* The dictionary was unavailable due to a configuration error. +/* .PP +/* Generally, a program is expected to test the function result +/* value for "success" first. If the operation was not successful, +/* a program is expected to test for a non-zero dict->error +/* status to distinguish between a data notfound/collision +/* condition or a dictionary access error. +/* LICENSE +/* .ad +/* .fi +/* The Secure Mailer license must be distributed with this software. +/* AUTHOR(S) +/* Wietse Venema +/* IBM T.J. Watson Research +/* P.O. Box 704 +/* Yorktown Heights, NY 10598, USA +/* +/* Wietse Venema +/* Google, Inc. +/* 111 8th Avenue +/* New York, NY 10011, USA +/*--*/ + +/* System library. */ + +#include <sys_defs.h> +#include <string.h> +#include <stdlib.h> + +/* Utility library. */ + +#include <argv.h> +#include <mymalloc.h> +#include <msg.h> +#include <dict.h> +#include <dict_cdb.h> +#include <dict_env.h> +#include <dict_unix.h> +#include <dict_tcp.h> +#include <dict_sdbm.h> +#include <dict_dbm.h> +#include <dict_db.h> +#include <dict_lmdb.h> +#include <dict_nis.h> +#include <dict_nisplus.h> +#include <dict_ni.h> +#include <dict_pcre.h> +#include <dict_regexp.h> +#include <dict_static.h> +#include <dict_cidr.h> +#include <dict_ht.h> +#include <dict_thash.h> +#include <dict_sockmap.h> +#include <dict_fail.h> +#include <dict_pipe.h> +#include <dict_random.h> +#include <dict_union.h> +#include <dict_inline.h> +#include <stringops.h> +#include <split_at.h> +#include <htable.h> +#include <myflock.h> + + /* + * lookup table for available map types. + */ +typedef struct { + char *type; + DICT_OPEN_FN open; +} DICT_OPEN_INFO; + +static const DICT_OPEN_INFO dict_open_info[] = { + DICT_TYPE_ENVIRON, dict_env_open, + DICT_TYPE_HT, dict_ht_open, + DICT_TYPE_UNIX, dict_unix_open, + DICT_TYPE_TCP, dict_tcp_open, +#ifdef HAS_DBM + DICT_TYPE_DBM, dict_dbm_open, +#endif +#ifdef HAS_DB + DICT_TYPE_HASH, dict_hash_open, + DICT_TYPE_BTREE, dict_btree_open, +#endif +#ifdef HAS_NIS + DICT_TYPE_NIS, dict_nis_open, +#endif +#ifdef HAS_NISPLUS + DICT_TYPE_NISPLUS, dict_nisplus_open, +#endif +#ifdef HAS_NETINFO + DICT_TYPE_NETINFO, dict_ni_open, +#endif +#ifdef HAS_POSIX_REGEXP + DICT_TYPE_REGEXP, dict_regexp_open, +#endif + DICT_TYPE_STATIC, dict_static_open, + DICT_TYPE_CIDR, dict_cidr_open, + DICT_TYPE_THASH, dict_thash_open, + DICT_TYPE_SOCKMAP, dict_sockmap_open, + DICT_TYPE_FAIL, dict_fail_open, + DICT_TYPE_PIPE, dict_pipe_open, + DICT_TYPE_RANDOM, dict_random_open, + DICT_TYPE_UNION, dict_union_open, + DICT_TYPE_INLINE, dict_inline_open, +#ifndef USE_DYNAMIC_MAPS +#ifdef HAS_PCRE + DICT_TYPE_PCRE, dict_pcre_open, +#endif +#ifdef HAS_CDB + DICT_TYPE_CDB, dict_cdb_open, +#endif +#ifdef HAS_SDBM + DICT_TYPE_SDBM, dict_sdbm_open, +#endif +#ifdef HAS_LMDB + DICT_TYPE_LMDB, dict_lmdb_open, +#endif +#endif /* !USE_DYNAMIC_MAPS */ + 0, +}; + +static HTABLE *dict_open_hash; + + /* + * Extension hooks. + */ +static DICT_OPEN_EXTEND_FN dict_open_extend_hook; +static DICT_MAPNAMES_EXTEND_FN dict_mapnames_extend_hook; + + /* + * Workaround. + */ +DEFINE_DICT_LMDB_MAP_SIZE; +DEFINE_DICT_DB_CACHE_SIZE; + +/* dict_open_init - one-off initialization */ + +static void dict_open_init(void) +{ + const char *myname = "dict_open_init"; + const DICT_OPEN_INFO *dp; + + if (dict_open_hash != 0) + msg_panic("%s: multiple initialization", myname); + dict_open_hash = htable_create(10); + + for (dp = dict_open_info; dp->type; dp++) + htable_enter(dict_open_hash, dp->type, (void *) dp); +} + +/* dict_open - open dictionary */ + +DICT *dict_open(const char *dict_spec, int open_flags, int dict_flags) +{ + char *saved_dict_spec = mystrdup(dict_spec); + char *dict_name; + DICT *dict; + + if ((dict_name = split_at(saved_dict_spec, ':')) == 0) + msg_fatal("open dictionary: expecting \"type:name\" form instead of \"%s\"", + dict_spec); + + dict = dict_open3(saved_dict_spec, dict_name, open_flags, dict_flags); + myfree(saved_dict_spec); + return (dict); +} + + +/* dict_open3 - open dictionary */ + +DICT *dict_open3(const char *dict_type, const char *dict_name, + int open_flags, int dict_flags) +{ + const char *myname = "dict_open"; + DICT_OPEN_INFO *dp; + DICT_OPEN_FN open_fn; + DICT *dict; + + if (*dict_type == 0 || *dict_name == 0) + msg_fatal("open dictionary: expecting \"type:name\" form instead of \"%s:%s\"", + dict_type, dict_name); + if (dict_open_hash == 0) + dict_open_init(); + if ((dp = (DICT_OPEN_INFO *) htable_find(dict_open_hash, dict_type)) == 0) { + if (dict_open_extend_hook != 0 + && (open_fn = dict_open_extend_hook(dict_type)) != 0) { + dict_open_register(dict_type, open_fn); + dp = (DICT_OPEN_INFO *) htable_find(dict_open_hash, dict_type); + } + if (dp == 0) + return (dict_surrogate(dict_type, dict_name, open_flags, dict_flags, + "unsupported dictionary type: %s", dict_type)); + } + if ((dict = dp->open(dict_name, open_flags, dict_flags)) == 0) + return (dict_surrogate(dict_type, dict_name, open_flags, dict_flags, + "cannot open %s:%s: %m", dict_type, dict_name)); + if (msg_verbose) + msg_info("%s: %s:%s", myname, dict_type, dict_name); + /* XXX The choice between wait-for-lock or no-wait is hard-coded. */ + if (dict->flags & DICT_FLAG_OPEN_LOCK) { + if (dict->flags & DICT_FLAG_LOCK) + msg_panic("%s: attempt to open %s:%s with both \"open\" lock and \"access\" lock", + myname, dict_type, dict_name); + /* Multi-writer safe map: downgrade persistent lock to temporary. */ + if (dict->flags & DICT_FLAG_MULTI_WRITER) { + dict->flags &= ~DICT_FLAG_OPEN_LOCK; + dict->flags |= DICT_FLAG_LOCK; + } + /* Multi-writer unsafe map: acquire exclusive lock or bust. */ + else if (dict->lock(dict, MYFLOCK_OP_EXCLUSIVE | MYFLOCK_OP_NOWAIT) < 0) + msg_fatal("%s:%s: unable to get exclusive lock: %m", + dict_type, dict_name); + } + /* Last step: insert proxy for UTF-8 syntax checks and casefolding. */ + if ((dict->flags & DICT_FLAG_UTF8_ACTIVE) == 0 + && DICT_NEED_UTF8_ACTIVATION(util_utf8_enable, dict_flags)) + dict = dict_utf8_activate(dict); + return (dict); +} + +/* dict_open_register - register dictionary type */ + +void dict_open_register(const char *type, DICT_OPEN_FN open) +{ + const char *myname = "dict_open_register"; + DICT_OPEN_INFO *dp; + HTABLE_INFO *ht; + + if (dict_open_hash == 0) + dict_open_init(); + if (htable_find(dict_open_hash, type)) + msg_panic("%s: dictionary type exists: %s", myname, type); + dp = (DICT_OPEN_INFO *) mymalloc(sizeof(*dp)); + dp->open = open; + ht = htable_enter(dict_open_hash, type, (void *) dp); + dp->type = ht->key; +} + +/* dict_open_extend - register alternate dictionary search routine */ + +DICT_OPEN_EXTEND_FN dict_open_extend(DICT_OPEN_EXTEND_FN new_cb) +{ + DICT_OPEN_EXTEND_FN old_cb; + + old_cb = dict_open_extend_hook; + dict_open_extend_hook = new_cb; + return (old_cb); +} + +/* dict_sort_alpha_cpp - qsort() callback */ + +static int dict_sort_alpha_cpp(const void *a, const void *b) +{ + return (strcmp(((char **) a)[0], ((char **) b)[0])); +} + +/* dict_mapnames - return an ARGV of available map_names */ + +ARGV *dict_mapnames() +{ + HTABLE_INFO **ht_info; + HTABLE_INFO **ht; + DICT_OPEN_INFO *dp; + ARGV *mapnames; + + if (dict_open_hash == 0) + dict_open_init(); + mapnames = argv_alloc(dict_open_hash->used + 1); + for (ht_info = ht = htable_list(dict_open_hash); *ht; ht++) { + dp = (DICT_OPEN_INFO *) ht[0]->value; + argv_add(mapnames, dp->type, ARGV_END); + } + if (dict_mapnames_extend_hook != 0) + (void) dict_mapnames_extend_hook(mapnames); + qsort((void *) mapnames->argv, mapnames->argc, sizeof(mapnames->argv[0]), + dict_sort_alpha_cpp); + myfree((void *) ht_info); + argv_terminate(mapnames); + return mapnames; +} + +/* dict_mapnames_extend - register alternate dictionary type list routine */ + +DICT_MAPNAMES_EXTEND_FN dict_mapnames_extend(DICT_MAPNAMES_EXTEND_FN new_cb) +{ + DICT_MAPNAMES_EXTEND_FN old_cb; + + old_cb = dict_mapnames_extend_hook; + dict_mapnames_extend_hook = new_cb; + return (old_cb); +} + +/* dict_type_override - disguise a dictionary type */ + +void dict_type_override(DICT *dict, const char *type) +{ + myfree(dict->type); + dict->type = mystrdup(type); +} + +#ifdef TEST + + /* + * Proof-of-concept test program. + */ +int main(int argc, char **argv) +{ + dict_test(argc, argv); + return (0); +} + +#endif |