RmHasherTask *rm_hasher_task_new(RmHasher *hasher, RmDigest *digest,
gpointer task_user_data) {
g_mutex_lock(&hasher->lock);
{ hasher->active_tasks++; }
g_mutex_unlock(&hasher->lock);
RmHasherTask *self = g_slice_new0(RmHasherTask);
self->hasher = hasher;
if(digest) {
self->digest = digest;
} else {
self->digest = rm_digest_new(hasher->digest_type, 0, 0, 0,
hasher->digest_type == RM_DIGEST_PARANOID);
}
/* get a recycled hashpipe if available */
self->hashpipe = g_async_queue_try_pop(hasher->hashpipe_pool);
if(!self->hashpipe) {
if(g_atomic_int_get(&hasher->unalloc_hashpipes) > 0) {
/* create a new hashpipe */
g_atomic_int_dec_and_test(&hasher->unalloc_hashpipes);
self->hashpipe =
rm_util_thread_pool_new((GFunc)rm_hasher_hashpipe_worker, hasher, 1);
} else {
/* already at thread limit - wait for a hashpipe to come available */
self->hashpipe = g_async_queue_pop(hasher->hashpipe_pool);
}
}
rm_assert_gentle(self->hashpipe);
self->task_user_data = task_user_data;
return self;
}
RmDigest *rm_digest_copy(RmDigest *digest) {
g_assert(digest);
RmDigest *self = NULL;
switch(digest->type) {
case RM_DIGEST_MD5:
case RM_DIGEST_SHA512:
case RM_DIGEST_SHA256:
case RM_DIGEST_SHA1:
self = g_slice_new0(RmDigest);
self->bytes = digest->bytes;
self->type = digest->type;
self->glib_checksum = g_checksum_copy(digest->glib_checksum);
break;
case RM_DIGEST_PARANOID:
case RM_DIGEST_SPOOKY:
case RM_DIGEST_SPOOKY32:
case RM_DIGEST_SPOOKY64:
case RM_DIGEST_MURMUR:
case RM_DIGEST_CITY:
case RM_DIGEST_CITY256:
case RM_DIGEST_MURMUR256:
case RM_DIGEST_CITY512:
case RM_DIGEST_MURMUR512:
case RM_DIGEST_BASTARD:
case RM_DIGEST_CUMULATIVE:
self = rm_digest_new(
digest->type, digest->initial_seed1, digest->initial_seed2, digest->bytes
);
if(self->type == RM_DIGEST_PARANOID) {
self->paranoid_offset = digest->paranoid_offset;
rm_digest_free(self->shadow_hash);
self->shadow_hash = rm_digest_copy(digest->shadow_hash);
}
if(self->checksum && digest->checksum) {
memcpy((char *)self->checksum, (char *)digest->checksum, self->bytes);
}
break;
default:
g_assert_not_reached();
}
return self;
}
RmDigest *rm_digest_copy(RmDigest *digest) {
rm_assert_gentle(digest);
RmDigest *self = NULL;
switch(digest->type) {
case RM_DIGEST_MD5:
case RM_DIGEST_SHA512:
case RM_DIGEST_SHA256:
case RM_DIGEST_SHA1:
self = g_slice_new0(RmDigest);
self->bytes = digest->bytes;
self->type = digest->type;
self->glib_checksum = g_checksum_copy(digest->glib_checksum);
break;
case RM_DIGEST_SPOOKY:
case RM_DIGEST_SPOOKY32:
case RM_DIGEST_SPOOKY64:
case RM_DIGEST_MURMUR:
case RM_DIGEST_CITY:
case RM_DIGEST_CITY256:
case RM_DIGEST_MURMUR256:
case RM_DIGEST_CITY512:
case RM_DIGEST_MURMUR512:
case RM_DIGEST_XXHASH:
case RM_DIGEST_FARMHASH:
case RM_DIGEST_BASTARD:
case RM_DIGEST_CUMULATIVE:
case RM_DIGEST_EXT:
self = rm_digest_new(digest->type, 0, 0, digest->bytes, FALSE);
if(self->checksum && digest->checksum) {
memcpy((char *)self->checksum, (char *)digest->checksum, self->bytes);
}
break;
case RM_DIGEST_PARANOID:
default:
rm_assert_gentle_not_reached();
}
return self;
}
static RmDirectory *rm_directory_new(char *dirname) {
RmDirectory *self = g_new0(RmDirectory, 1);
self->file_count = 0;
self->dupe_count = 0;
self->prefd_files = 0;
self->was_merged = false;
self->was_inserted = false;
self->mergeups = 0;
self->dirname = dirname;
self->finished = false;
self->depth = 0;
for(char *s = dirname; *s; s++) {
self->depth += (*s == G_DIR_SEPARATOR);
}
RmStat dir_stat;
if(rm_sys_stat(self->dirname, &dir_stat) == -1) {
rm_log_perror("stat(2) failed during sort");
} else {
self->metadata.dir_mtime = dir_stat.st_mtime;
self->metadata.dir_inode = dir_stat.st_ino;
self->metadata.dir_dev = dir_stat.st_dev;
}
/* Special cumulative hashsum, that is not dependent on the
* order in which the file hashes were added.
* It is not used as full hash, but as sorting speedup.
*/
self->digest = rm_digest_new(RM_DIGEST_CUMULATIVE, 0, 0, 0, false);
g_queue_init(&self->known_files);
g_queue_init(&self->children);
self->hash_set =
g_hash_table_new((GHashFunc)rm_digest_hash, (GEqualFunc)rm_digest_equal);
return self;
}
RmDigest *rm_digest_new(RmDigestType type, RmOff seed1, RmOff seed2, RmOff ext_size,
bool use_shadow_hash) {
RmDigest *digest = g_slice_new0(RmDigest);
digest->checksum = NULL;
digest->type = type;
digest->bytes = 0;
switch(type) {
case RM_DIGEST_SPOOKY32:
/* cannot go lower than 64, since we read 8 byte in some places.
* simulate by leaving the part at the end empty
*/
digest->bytes = 64 / 8;
break;
case RM_DIGEST_XXHASH:
case RM_DIGEST_FARMHASH:
case RM_DIGEST_SPOOKY64:
digest->bytes = 64 / 8;
break;
case RM_DIGEST_MD5:
digest->glib_checksum = g_checksum_new(G_CHECKSUM_MD5);
ADD_SEED(digest, seed1);
digest->bytes = 128 / 8;
return digest;
#if HAVE_SHA512
case RM_DIGEST_SHA512:
digest->glib_checksum = g_checksum_new(G_CHECKSUM_SHA512);
ADD_SEED(digest, seed1);
digest->bytes = 512 / 8;
return digest;
#endif
case RM_DIGEST_SHA256:
digest->glib_checksum = g_checksum_new(G_CHECKSUM_SHA256);
ADD_SEED(digest, seed1);
digest->bytes = 256 / 8;
return digest;
case RM_DIGEST_SHA1:
digest->glib_checksum = g_checksum_new(G_CHECKSUM_SHA1);
ADD_SEED(digest, seed1);
digest->bytes = 160 / 8;
return digest;
case RM_DIGEST_MURMUR512:
case RM_DIGEST_CITY512:
digest->bytes = 512 / 8;
break;
case RM_DIGEST_EXT:
/* gets allocated on rm_digest_update() */
digest->bytes = ext_size;
break;
case RM_DIGEST_MURMUR256:
case RM_DIGEST_CITY256:
case RM_DIGEST_BASTARD:
digest->bytes = 256 / 8;
break;
case RM_DIGEST_SPOOKY:
case RM_DIGEST_MURMUR:
case RM_DIGEST_CITY:
case RM_DIGEST_CUMULATIVE:
digest->bytes = 128 / 8;
break;
case RM_DIGEST_PARANOID:
digest->bytes = 0;
digest->paranoid = g_slice_new0(RmParanoid);
digest->paranoid->incoming_twin_candidates = g_async_queue_new();
if(use_shadow_hash) {
digest->paranoid->shadow_hash =
rm_digest_new(RM_DIGEST_XXHASH, seed1, seed2, 0, false);
}
break;
default:
rm_assert_gentle_not_reached();
}
/* starting values to let us generate up to 4 different hashes in parallel with
* different starting seeds:
* */
static const RmOff seeds[4] = {0x0000000000000000, 0xf0f0f0f0f0f0f0f0,
0x3333333333333333, 0xaaaaaaaaaaaaaaaa};
if(digest->bytes > 0 && type != RM_DIGEST_PARANOID) {
const int n_seeds = sizeof(seeds) / sizeof(seeds[0]);
/* checksum type - allocate memory and initialise */
digest->checksum = g_slice_alloc0(digest->bytes);
for(gsize block = 0; block < (digest->bytes / 16); block++) {
digest->checksum[block].first = seeds[block % n_seeds] ^ seed1;
digest->checksum[block].second = seeds[block % n_seeds] ^ seed2;
}
}
if(digest->type == RM_DIGEST_BASTARD) {
/* bastard type *always* has *pure* murmur hash for first checksum
* and seeded city for second checksum */
digest->checksum[0].first = digest->checksum[0].second = 0;
}
return digest;
}
