diff options
Diffstat (limited to 'src/or/conscache.c')
| -rw-r--r-- | src/or/conscache.c | 626 |
1 files changed, 626 insertions, 0 deletions
diff --git a/src/or/conscache.c b/src/or/conscache.c new file mode 100644 index 0000000000..e25ac5f40b --- /dev/null +++ b/src/or/conscache.c @@ -0,0 +1,626 @@ +/* Copyright (c) 2017, The Tor Project, Inc. */ +/* See LICENSE for licensing information */ + +#include "or.h" + +#include "config.h" +#include "conscache.h" +#include "storagedir.h" + +#define CCE_MAGIC 0x17162253 + +#ifdef _WIN32 +/* On Windows, unlink won't work on a file if the file is actively mmap()ed. + * That forces us to be less aggressive about unlinking files, and causes other + * changes throughout our logic. + */ +#define MUST_UNMAP_TO_UNLINK +#endif /* defined(_WIN32) */ + +/** + * A consensus_cache_entry_t is a reference-counted handle to an + * item in a consensus_cache_t. It can be mmapped into RAM, or not, + * depending whether it's currently in use. + */ +struct consensus_cache_entry_t { + uint32_t magic; /**< Must be set to CCE_MAGIC */ + HANDLE_ENTRY(consensus_cache_entry, consensus_cache_entry_t); + int32_t refcnt; /**< Reference count. */ + unsigned can_remove : 1; /**< If true, we want to delete this file. */ + /** If true, we intend to unmap this file as soon as we're done with it. */ + unsigned release_aggressively : 1; + + /** Filename for this object within the storage_dir_t */ + char *fname; + /** Labels associated with this object. Immutable once the object + * is created. */ + config_line_t *labels; + /** Pointer to the cache that includes this entry (if any). */ + consensus_cache_t *in_cache; + + /** Since what time has this object been mapped into RAM, but with the cache + * being the only having a reference to it? */ + time_t unused_since; + /** mmaped contents of the underlying file. May be NULL */ + tor_mmap_t *map; + /** Length of the body within <b>map</b>. */ + size_t bodylen; + /** Pointer to the body within <b>map</b>. */ + const uint8_t *body; +}; + +/** + * A consensus_cache_t holds a directory full of labeled items. + */ +struct consensus_cache_t { + /** Underling storage_dir_t to handle persistence */ + storage_dir_t *dir; + /** List of all the entries in the directory. */ + smartlist_t *entries; + + /** The maximum number of entries that we'd like to allow in this cache. + * This is the same as the storagedir limit when MUST_UNMAP_TO_UNLINK is + * not defined. */ + unsigned max_entries; +}; + +static void consensus_cache_clear(consensus_cache_t *cache); +static void consensus_cache_rescan(consensus_cache_t *); +static void consensus_cache_entry_map(consensus_cache_t *, + consensus_cache_entry_t *); +static void consensus_cache_entry_unmap(consensus_cache_entry_t *ent); + +/** + * Helper: Open a consensus cache in subdirectory <b>subdir</b> of the + * data directory, to hold up to <b>max_entries</b> of data. + */ +consensus_cache_t * +consensus_cache_open(const char *subdir, int max_entries) +{ + int storagedir_max_entries; + consensus_cache_t *cache = tor_malloc_zero(sizeof(consensus_cache_t)); + char *directory = get_cachedir_fname(subdir); + cache->max_entries = max_entries; + +#ifdef MUST_UNMAP_TO_UNLINK + /* If we can't unlink the files that we're still using, then we need to + * tell the storagedir backend to allow far more files than this consensus + * cache actually wants, so that it can hold files which, from this cache's + * perspective, have become useless. + */ +#define VERY_LARGE_STORAGEDIR_LIMIT (1000*1000) + storagedir_max_entries = VERY_LARGE_STORAGEDIR_LIMIT; +#else /* !(defined(MUST_UNMAP_TO_UNLINK)) */ + /* Otherwise, we can just tell the storagedir to use the same limits + * as this cache. */ + storagedir_max_entries = max_entries; +#endif /* defined(MUST_UNMAP_TO_UNLINK) */ + + cache->dir = storage_dir_new(directory, storagedir_max_entries); + tor_free(directory); + if (!cache->dir) { + tor_free(cache); + return NULL; + } + + consensus_cache_rescan(cache); + return cache; +} + +/** Return true if it's okay to put more entries in this cache than + * its official file limit. + * + * (We need this method on Windows, where we can't unlink files that are still + * in use, and therefore might need to temporarily exceed the file limit until + * the no-longer-wanted files are deletable.) + */ +int +consensus_cache_may_overallocate(consensus_cache_t *cache) +{ + (void) cache; +#ifdef MUST_UNMAP_TO_UNLINK + return 1; +#else + return 0; +#endif +} + +/** + * Tell the sandbox (if any) configured by <b>cfg</b> to allow the + * operations that <b>cache</b> will need. + */ +int +consensus_cache_register_with_sandbox(consensus_cache_t *cache, + struct sandbox_cfg_elem **cfg) +{ +#ifdef MUST_UNMAP_TO_UNLINK + /* Our Linux sandbox doesn't support huge file lists like the one that would + * be generated by using VERY_LARGE_STORAGEDIR_LIMIT above in + * consensus_cache_open(). Since the Linux sandbox is the only one we have + * right now, we just assert that we never reach this point when we've had + * to use VERY_LARGE_STORAGEDIR_LIMIT. + * + * If at some point in the future we have a different sandbox mechanism that + * can handle huge file lists, we can remove this assertion or make it + * conditional. + */ + tor_assert_nonfatal_unreached(); +#endif /* defined(MUST_UNMAP_TO_UNLINK) */ + return storage_dir_register_with_sandbox(cache->dir, cfg); +} + +/** + * Helper: clear all entries from <b>cache</b> (but do not delete + * any that aren't marked for removal + */ +static void +consensus_cache_clear(consensus_cache_t *cache) +{ + consensus_cache_delete_pending(cache, 0); + + SMARTLIST_FOREACH_BEGIN(cache->entries, consensus_cache_entry_t *, ent) { + ent->in_cache = NULL; + consensus_cache_entry_decref(ent); + } SMARTLIST_FOREACH_END(ent); + smartlist_free(cache->entries); + cache->entries = NULL; +} + +/** + * Drop all storage held by <b>cache</b>. + */ +void +consensus_cache_free_(consensus_cache_t *cache) +{ + if (! cache) + return; + + if (cache->entries) { + consensus_cache_clear(cache); + } + storage_dir_free(cache->dir); + tor_free(cache); +} + +/** + * Write <b>datalen</b> bytes of data at <b>data</b> into the <b>cache</b>, + * labeling that data with <b>labels</b>. On failure, return NULL. On + * success, return a newly created consensus_cache_entry_t. + * + * The returned value will be owned by the cache, and you will have a + * reference to it. Call consensus_cache_entry_decref() when you are + * done with it. + * + * The provided <b>labels</b> MUST have distinct keys: if they don't, + * this API does not specify which values (if any) for the duplicate keys + * will be considered. + */ +consensus_cache_entry_t * +consensus_cache_add(consensus_cache_t *cache, + const config_line_t *labels, + const uint8_t *data, + size_t datalen) +{ + char *fname = NULL; + int r = storage_dir_save_labeled_to_file(cache->dir, + labels, data, datalen, &fname); + if (r < 0 || fname == NULL) { + return NULL; + } + consensus_cache_entry_t *ent = + tor_malloc_zero(sizeof(consensus_cache_entry_t)); + ent->magic = CCE_MAGIC; + ent->fname = fname; + ent->labels = config_lines_dup(labels); + ent->in_cache = cache; + ent->unused_since = TIME_MAX; + smartlist_add(cache->entries, ent); + /* Start the reference count at 2: the caller owns one copy, and the + * cache owns another. + */ + ent->refcnt = 2; + + return ent; +} + +/** + * Given a <b>cache</b>, return some entry for which <b>key</b>=<b>value</b>. + * Return NULL if no such entry exists. + * + * Does not adjust reference counts. + */ +consensus_cache_entry_t * +consensus_cache_find_first(consensus_cache_t *cache, + const char *key, + const char *value) +{ + smartlist_t *tmp = smartlist_new(); + consensus_cache_find_all(tmp, cache, key, value); + consensus_cache_entry_t *ent = NULL; + if (smartlist_len(tmp)) + ent = smartlist_get(tmp, 0); + smartlist_free(tmp); + return ent; +} + +/** + * Given a <b>cache</b>, add every entry to <b>out<b> for which + * <b>key</b>=<b>value</b>. If <b>key</b> is NULL, add every entry. + * + * Do not add any entry that has been marked for removal. + * + * Does not adjust reference counts. + */ +void +consensus_cache_find_all(smartlist_t *out, + consensus_cache_t *cache, + const char *key, + const char *value) +{ + SMARTLIST_FOREACH_BEGIN(cache->entries, consensus_cache_entry_t *, ent) { + if (ent->can_remove == 1) { + /* We want to delete this; pretend it isn't there. */ + continue; + } + if (! key) { + smartlist_add(out, ent); + continue; + } + const char *found_val = consensus_cache_entry_get_value(ent, key); + if (found_val && !strcmp(value, found_val)) { + smartlist_add(out, ent); + } + } SMARTLIST_FOREACH_END(ent); +} + +/** + * Given a list of consensus_cache_entry_t, remove all those entries + * that do not have <b>key</b>=<b>value</b> in their labels. + * + * Does not adjust reference counts. + */ +void +consensus_cache_filter_list(smartlist_t *lst, + const char *key, + const char *value) +{ + if (BUG(lst == NULL)) + return; // LCOV_EXCL_LINE + if (key == NULL) + return; + SMARTLIST_FOREACH_BEGIN(lst, consensus_cache_entry_t *, ent) { + const char *found_val = consensus_cache_entry_get_value(ent, key); + if (! found_val || strcmp(value, found_val)) { + SMARTLIST_DEL_CURRENT(lst, ent); + } + } SMARTLIST_FOREACH_END(ent); +} + +/** + * If <b>ent</b> has a label with the given <b>key</b>, return its + * value. Otherwise return NULL. + * + * The return value is only guaranteed to be valid for as long as you + * hold a reference to <b>ent</b>. + */ +const char * +consensus_cache_entry_get_value(const consensus_cache_entry_t *ent, + const char *key) +{ + const config_line_t *match = config_line_find(ent->labels, key); + if (match) + return match->value; + else + return NULL; +} + +/** + * Return a pointer to the labels in <b>ent</b>. + * + * This pointer is only guaranteed to be valid for as long as you + * hold a reference to <b>ent</b>. + */ +const config_line_t * +consensus_cache_entry_get_labels(const consensus_cache_entry_t *ent) +{ + return ent->labels; +} + +/** + * Increase the reference count of <b>ent</b>. + */ +void +consensus_cache_entry_incref(consensus_cache_entry_t *ent) +{ + if (BUG(ent->magic != CCE_MAGIC)) + return; // LCOV_EXCL_LINE + ++ent->refcnt; + ent->unused_since = TIME_MAX; +} + +/** + * Release a reference held to <b>ent</b>. + * + * If it was the last reference, ent will be freed. Therefore, you must not + * use <b>ent</b> after calling this function. + */ +void +consensus_cache_entry_decref(consensus_cache_entry_t *ent) +{ + if (! ent) + return; + if (BUG(ent->refcnt <= 0)) + return; // LCOV_EXCL_LINE + if (BUG(ent->magic != CCE_MAGIC)) + return; // LCOV_EXCL_LINE + + --ent->refcnt; + + if (ent->refcnt == 1 && ent->in_cache) { + /* Only the cache has a reference: we don't need to keep the file + * mapped */ + if (ent->map) { + if (ent->release_aggressively) { + consensus_cache_entry_unmap(ent); + } else { + ent->unused_since = approx_time(); + } + } + return; + } + + if (ent->refcnt > 0) + return; + + /* Refcount is zero; we can free it. */ + if (ent->map) { + consensus_cache_entry_unmap(ent); + } + tor_free(ent->fname); + config_free_lines(ent->labels); + consensus_cache_entry_handles_clear(ent); + memwipe(ent, 0, sizeof(consensus_cache_entry_t)); + tor_free(ent); +} + +/** + * Mark <b>ent</b> for deletion from the cache. Deletion will not occur + * until the cache is the only place that holds a reference to <b>ent</b>. + */ +void +consensus_cache_entry_mark_for_removal(consensus_cache_entry_t *ent) +{ + ent->can_remove = 1; +} + +/** + * Mark <b>ent</b> as the kind of entry that we don't need to keep mmap'd for + * any longer than we're actually using it. + */ +void +consensus_cache_entry_mark_for_aggressive_release(consensus_cache_entry_t *ent) +{ + ent->release_aggressively = 1; +} + +/** + * Try to read the body of <b>ent</b> into memory if it isn't already + * loaded. On success, set *<b>body_out</b> to the body, *<b>sz_out</b> + * to its size, and return 0. On failure return -1. + * + * The resulting body pointer will only be valid for as long as you + * hold a reference to <b>ent</b>. + */ +int +consensus_cache_entry_get_body(const consensus_cache_entry_t *ent, + const uint8_t **body_out, + size_t *sz_out) +{ + if (BUG(ent->magic != CCE_MAGIC)) + return -1; // LCOV_EXCL_LINE + + if (! ent->map) { + if (! ent->in_cache) + return -1; + + consensus_cache_entry_map((consensus_cache_t *)ent->in_cache, + (consensus_cache_entry_t *)ent); + if (! ent->map) { + return -1; + } + } + + *body_out = ent->body; + *sz_out = ent->bodylen; + return 0; +} + +/** + * Unmap every mmap'd element of <b>cache</b> that has been unused + * since <b>cutoff</b>. + */ +void +consensus_cache_unmap_lazy(consensus_cache_t *cache, time_t cutoff) +{ + SMARTLIST_FOREACH_BEGIN(cache->entries, consensus_cache_entry_t *, ent) { + tor_assert_nonfatal(ent->in_cache == cache); + if (ent->refcnt > 1 || BUG(ent->in_cache == NULL)) { + /* Somebody is using this entry right now */ + continue; + } + if (ent->unused_since > cutoff) { + /* Has been unused only for a little while */ + continue; + } + if (ent->map == NULL) { + /* Not actually mapped. */ + continue; + } + consensus_cache_entry_unmap(ent); + } SMARTLIST_FOREACH_END(ent); +} + +/** + * Return the number of currently unused filenames available in this cache. + */ +int +consensus_cache_get_n_filenames_available(consensus_cache_t *cache) +{ + tor_assert(cache); + int max = cache->max_entries; + int used = smartlist_len(storage_dir_list(cache->dir)); +#ifdef MUST_UNMAP_TO_UNLINK + if (used > max) + return 0; +#else + tor_assert_nonfatal(max >= used); +#endif /* defined(MUST_UNMAP_TO_UNLINK) */ + return max - used; +} + +/** + * Delete every element of <b>cache</b> has been marked with + * consensus_cache_entry_mark_for_removal. If <b>force</b> is false, + * retain those entries which are in use by something other than the cache. + */ +void +consensus_cache_delete_pending(consensus_cache_t *cache, int force) +{ + SMARTLIST_FOREACH_BEGIN(cache->entries, consensus_cache_entry_t *, ent) { + tor_assert_nonfatal(ent->in_cache == cache); + int force_ent = force; +#ifdef MUST_UNMAP_TO_UNLINK + /* We cannot delete anything with an active mmap on win32, so no + * force-deletion. */ + if (ent->map) { + force_ent = 0; + } +#endif /* defined(MUST_UNMAP_TO_UNLINK) */ + if (! force_ent) { + if (ent->refcnt > 1 || BUG(ent->in_cache == NULL)) { + /* Somebody is using this entry right now */ + continue; + } + } + if (ent->can_remove == 0) { + /* Don't want to delete this. */ + continue; + } + if (BUG(ent->refcnt <= 0)) { + continue; // LCOV_EXCL_LINE + } + + SMARTLIST_DEL_CURRENT(cache->entries, ent); + ent->in_cache = NULL; + char *fname = tor_strdup(ent->fname); /* save a copy */ + consensus_cache_entry_decref(ent); + storage_dir_remove_file(cache->dir, fname); + tor_free(fname); + } SMARTLIST_FOREACH_END(ent); +} + +/** + * Internal helper: rescan <b>cache</b> and rebuild its list of entries. + */ +static void +consensus_cache_rescan(consensus_cache_t *cache) +{ + if (cache->entries) { + consensus_cache_clear(cache); + } + + cache->entries = smartlist_new(); + const smartlist_t *fnames = storage_dir_list(cache->dir); + SMARTLIST_FOREACH_BEGIN(fnames, const char *, fname) { + tor_mmap_t *map = NULL; + config_line_t *labels = NULL; + const uint8_t *body; + size_t bodylen; + map = storage_dir_map_labeled(cache->dir, fname, + &labels, &body, &bodylen); + if (! map) { + /* The ERANGE error might come from tor_mmap_file() -- it means the file + * was empty. EINVAL might come from ..map_labeled() -- it means the + * file was misformatted. In both cases, we should just delete it. + */ + if (errno == ERANGE || errno == EINVAL) { + log_warn(LD_FS, "Found %s file %s in consensus cache; removing it.", + errno == ERANGE ? "empty" : "misformatted", + escaped(fname)); + storage_dir_remove_file(cache->dir, fname); + } else { + /* Can't load this; continue */ + log_warn(LD_FS, "Unable to map file %s from consensus cache: %s", + escaped(fname), strerror(errno)); + } + continue; + } + consensus_cache_entry_t *ent = + tor_malloc_zero(sizeof(consensus_cache_entry_t)); + ent->magic = CCE_MAGIC; + ent->fname = tor_strdup(fname); + ent->labels = labels; + ent->refcnt = 1; + ent->in_cache = cache; + ent->unused_since = TIME_MAX; + smartlist_add(cache->entries, ent); + tor_munmap_file(map); /* don't actually need to keep this around */ + } SMARTLIST_FOREACH_END(fname); +} + +/** + * Make sure that <b>ent</b> is mapped into RAM. + */ +static void +consensus_cache_entry_map(consensus_cache_t *cache, + consensus_cache_entry_t *ent) +{ + if (ent->map) + return; + + ent->map = storage_dir_map_labeled(cache->dir, ent->fname, + NULL, &ent->body, &ent->bodylen); + ent->unused_since = TIME_MAX; +} + +/** + * Unmap <b>ent</b> from RAM. + * + * Do not call this if something other than the cache is holding a reference + * to <b>ent</b> + */ +static void +consensus_cache_entry_unmap(consensus_cache_entry_t *ent) +{ + ent->unused_since = TIME_MAX; + if (!ent->map) + return; + + tor_munmap_file(ent->map); + ent->map = NULL; + ent->body = NULL; + ent->bodylen = 0; + ent->unused_since = TIME_MAX; +} + +HANDLE_IMPL(consensus_cache_entry, consensus_cache_entry_t, ) + +#ifdef TOR_UNIT_TESTS +/** + * Testing only: Return true iff <b>ent</b> is mapped into memory. + * + * (In normal operation, this information is not exposed.) + */ +int +consensus_cache_entry_is_mapped(consensus_cache_entry_t *ent) +{ + if (ent->map) { + tor_assert(ent->body); + return 1; + } else { + tor_assert(!ent->body); + return 0; + } +} +#endif /* defined(TOR_UNIT_TESTS) */ + |