X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Fblob_table.c;h=b3eb63fcad750d17ea7dbe471a543748738a1043;hp=5551c1635549af9451d0be01c466478da447c200;hb=b82856cb22783df0e19990eb68d3694753f52220;hpb=3de1ec66f778edda19865482d685bc6f4e17faf7 diff --git a/src/blob_table.c b/src/blob_table.c index 5551c163..b3eb63fc 100644 --- a/src/blob_table.c +++ b/src/blob_table.c @@ -99,18 +99,8 @@ free_blob_table(struct blob_table *table) struct blob_descriptor * new_blob_descriptor(void) { - struct blob_descriptor *blob; - - blob = CALLOC(1, sizeof(struct blob_descriptor)); - if (blob == NULL) - return NULL; - - blob->refcnt = 1; - - /* blob->blob_location = BLOB_NONEXISTENT */ BUILD_BUG_ON(BLOB_NONEXISTENT != 0); - - return blob; + return CALLOC(1, sizeof(struct blob_descriptor)); } struct blob_descriptor * @@ -148,27 +138,11 @@ clone_blob_descriptor(const struct blob_descriptor *old) break; #ifdef WITH_NTFS_3G case BLOB_IN_NTFS_VOLUME: - if (old->ntfs_loc) { - struct ntfs_location *loc; - loc = memdup(old->ntfs_loc, sizeof(struct ntfs_location)); - if (loc == NULL) - goto out_free; - loc->path = NULL; - loc->attr_name = NULL; - new->ntfs_loc = loc; - loc->path = STRDUP(old->ntfs_loc->path); - if (loc->path == NULL) - goto out_free; - if (loc->attr_name_nchars != 0) { - loc->attr_name = utf16le_dup(old->ntfs_loc->attr_name); - if (loc->attr_name == NULL) - goto out_free; - } - } + new->ntfs_loc = clone_ntfs_location(old->ntfs_loc); + if (!new->ntfs_loc) + goto out_free; break; #endif - default: - break; } return new; @@ -203,15 +177,10 @@ blob_release_location(struct blob_descriptor *blob) break; #ifdef WITH_NTFS_3G case BLOB_IN_NTFS_VOLUME: - if (blob->ntfs_loc) { - FREE(blob->ntfs_loc->path); - FREE(blob->ntfs_loc->attr_name); - FREE(blob->ntfs_loc); - } + if (blob->ntfs_loc) + free_ntfs_location(blob->ntfs_loc); break; #endif - default: - break; } } @@ -264,10 +233,21 @@ finalize_blob(struct blob_descriptor *blob) void blob_decrement_refcnt(struct blob_descriptor *blob, struct blob_table *table) { - if (unlikely(blob->refcnt == 0)) /* See comment above */ + blob_subtract_refcnt(blob, table, 1); +} + +void +blob_subtract_refcnt(struct blob_descriptor *blob, struct blob_table *table, + u32 count) +{ + if (unlikely(blob->refcnt < count)) { + blob->refcnt = 0; /* See comment above */ return; + } + + blob->refcnt -= count; - if (--blob->refcnt != 0) + if (blob->refcnt != 0) return; if (blob->unhashed) { @@ -318,7 +298,7 @@ enlarge_blob_table(struct blob_table *table) size_t old_capacity, new_capacity; struct hlist_head *old_array, *new_array; struct blob_descriptor *blob; - struct hlist_node *cur, *tmp; + struct hlist_node *tmp; size_t i; old_capacity = table->capacity; @@ -331,7 +311,7 @@ enlarge_blob_table(struct blob_table *table) table->capacity = new_capacity; for (i = 0; i < old_capacity; i++) { - hlist_for_each_entry_safe(blob, cur, tmp, &old_array[i], hash_list) { + hlist_for_each_entry_safe(blob, tmp, &old_array[i], hash_list) { hlist_del(&blob->hash_list); blob_table_insert_raw(table, blob); } @@ -366,10 +346,9 @@ lookup_blob(const struct blob_table *table, const u8 *hash) { size_t i; struct blob_descriptor *blob; - struct hlist_node *pos; i = load_size_t_unaligned(hash) % table->capacity; - hlist_for_each_entry(blob, pos, &table->array[i], hash_list) + hlist_for_each_entry(blob, &table->array[i], hash_list) if (hashes_equal(hash, blob->hash)) return blob; return NULL; @@ -382,11 +361,11 @@ for_blob_in_table(struct blob_table *table, int (*visitor)(struct blob_descriptor *, void *), void *arg) { struct blob_descriptor *blob; - struct hlist_node *pos, *tmp; + struct hlist_node *tmp; int ret; for (size_t i = 0; i < table->capacity; i++) { - hlist_for_each_entry_safe(blob, pos, tmp, &table->array[i], + hlist_for_each_entry_safe(blob, tmp, &table->array[i], hash_list) { ret = visitor(blob, arg); @@ -416,7 +395,7 @@ cmp_blobs_by_sequential_order(const void *p1, const void *p2) v = (int)blob1->blob_location - (int)blob2->blob_location; - /* Different resource locations? */ + /* Different locations? */ if (v) return v; @@ -427,7 +406,7 @@ cmp_blobs_by_sequential_order(const void *p1, const void *p2) /* Different (possibly split) WIMs? */ if (wim1 != wim2) { - v = memcmp(wim1->hdr.guid, wim2->hdr.guid, WIM_GUID_LEN); + v = cmp_guids(wim1->hdr.guid, wim2->hdr.guid); if (v) return v; } @@ -450,13 +429,17 @@ cmp_blobs_by_sequential_order(const void *p1, const void *p2) #ifdef __WIN32__ case BLOB_IN_WINNT_FILE_ON_DISK: case BLOB_WIN32_ENCRYPTED: + /* Windows: compare by starting LCN (logical cluster number) */ + v = cmp_u64(blob1->sort_key, blob2->sort_key); + if (v) + return v; #endif /* Compare files by path: just a heuristic that will place files * in the same directory next to each other. */ return tstrcmp(blob1->file_on_disk, blob2->file_on_disk); #ifdef WITH_NTFS_3G case BLOB_IN_NTFS_VOLUME: - return tstrcmp(blob1->ntfs_loc->path, blob2->ntfs_loc->path); + return cmp_ntfs_locations(blob1->ntfs_loc, blob2->ntfs_loc); #endif default: /* No additional sorting order defined for this resource @@ -628,7 +611,7 @@ do_load_solid_info(WIMStruct *wim, struct wim_resource_descriptor **rdescs, rdesc = rdescs[i]; - wim_res_hdr_to_desc(&reshdr, wim, rdesc); + wim_reshdr_to_desc(&reshdr, wim, rdesc); /* For solid resources, the uncompressed size, compression type, * and chunk size are stored in the resource itself, not in the @@ -729,11 +712,9 @@ assign_blob_to_solid_resource(const struct wim_reshdr *reshdr, /* XXX: This linear search will be slow in the degenerate case where the * number of solid resources in the run is huge. */ blob->size = reshdr->size_in_wim; - blob->flags = reshdr->flags; for (size_t i = 0; i < num_rdescs; i++) { if (offset + blob->size <= rdescs[i]->uncompressed_size) { - blob->offset_in_res = offset; - blob_set_is_located_in_wim_resource(blob, rdescs[i]); + blob_set_is_located_in_wim_resource(blob, rdescs[i], offset); return 0; } offset -= rdescs[i]->uncompressed_size; @@ -875,6 +856,7 @@ read_blob_table(WIMStruct *wim) struct blob_table *table = NULL; struct blob_descriptor *cur_blob = NULL; size_t num_duplicate_blobs = 0; + size_t num_empty_blobs = 0; size_t num_wrong_part_blobs = 0; u32 image_index = 0; struct wim_resource_descriptor **cur_solid_rdescs = NULL; @@ -972,27 +954,13 @@ read_blob_table(WIMStruct *wim) goto out; } - /* How to handle an uncompressed resource with its - * uncompressed size different from its compressed size? - * - * Based on a simple test, WIMGAPI seems to handle this - * as follows: - * - * if (size_in_wim > uncompressed_size) { - * Ignore uncompressed_size; use size_in_wim - * instead. - * } else { - * Honor uncompressed_size, but treat the part of - * the file data above size_in_wim as all zeros. - * } - * - * So we will do the same. */ - if (unlikely(!(reshdr.flags & - WIM_RESHDR_FLAG_COMPRESSED) && - (reshdr.size_in_wim > - reshdr.uncompressed_size))) + if (unlikely(!(reshdr.flags & WIM_RESHDR_FLAG_COMPRESSED) && + (reshdr.size_in_wim != reshdr.uncompressed_size))) { - reshdr.uncompressed_size = reshdr.size_in_wim; + ERROR("Uncompressed resource has " + "size_in_wim != uncompressed_size"); + ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY; + goto out; } /* Set up a resource descriptor for this blob. */ @@ -1001,30 +969,32 @@ read_blob_table(WIMStruct *wim) if (!rdesc) goto oom; - wim_res_hdr_to_desc(&reshdr, wim, rdesc); - - cur_blob->offset_in_res = 0; - cur_blob->size = reshdr.uncompressed_size; - cur_blob->flags = reshdr.flags; - - blob_set_is_located_in_wim_resource(cur_blob, rdesc); + wim_reshdr_to_desc_and_blob(&reshdr, wim, rdesc, cur_blob); } /* cur_blob is now a blob bound to a resource. */ /* Ignore entries with all zeroes in the hash field. */ - if (is_zero_hash(cur_blob->hash)) + if (unlikely(is_zero_hash(cur_blob->hash))) goto free_cur_blob_and_continue; + /* Verify that the blob has nonzero size. */ + if (unlikely(cur_blob->size == 0)) { + num_empty_blobs++; + goto free_cur_blob_and_continue; + } + /* Verify that the part number matches that of the underlying * WIM file. */ - if (part_number != wim->hdr.part_number) { + if (unlikely(part_number != wim->hdr.part_number)) { num_wrong_part_blobs++; goto free_cur_blob_and_continue; } if (reshdr.flags & WIM_RESHDR_FLAG_METADATA) { + cur_blob->is_metadata = 1; + /* Blob table entry for a metadata resource. */ /* Metadata entries with no references must be ignored. @@ -1043,6 +1013,13 @@ read_blob_table(WIMStruct *wim) goto out; } + if (reshdr.flags & WIM_RESHDR_FLAG_SOLID) { + ERROR("Image metadata in solid resources " + "is unsupported."); + ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY; + goto out; + } + if (wim->hdr.part_number != 1) { WARNING("Ignoring metadata resource found in a " "non-first part of the split WIM"); @@ -1112,6 +1089,9 @@ read_blob_table(WIMStruct *wim) if (num_duplicate_blobs > 0) WARNING("Ignoring %zu duplicate blobs", num_duplicate_blobs); + if (num_empty_blobs > 0) + WARNING("Ignoring %zu empty blobs", num_empty_blobs); + if (num_wrong_part_blobs > 0) { WARNING("Ignoring %zu blobs with wrong part number", num_wrong_part_blobs); @@ -1146,9 +1126,9 @@ write_blob_descriptor(struct blob_descriptor_disk *disk_entry, } /* Note: the list of blob descriptors must be sorted so that all entries for the - * same solid resource are consecutive. In addition, blob descriptors with - * WIM_RESHDR_FLAG_METADATA set must be in the same order as the indices of the - * underlying images. */ + * same solid resource are consecutive. In addition, blob descriptors for + * metadata resources must be in the same order as the indices of the underlying + * images. */ int write_blob_table_from_blob_list(struct list_head *blob_list, struct filedes *out_fd, @@ -1230,7 +1210,7 @@ write_blob_table_from_blob_list(struct list_head *blob_list, * compressed blob table, MS software cannot. */ ret = write_wim_resource_from_buffer(table_buf, table_size, - WIM_RESHDR_FLAG_METADATA, + true, out_fd, WIMLIB_COMPRESSION_TYPE_NONE, 0, @@ -1249,31 +1229,58 @@ new_blob_from_data_buffer(const void *buffer, size_t size, struct blob_table *blob_table) { u8 hash[SHA1_HASH_SIZE]; - struct blob_descriptor *blob, *existing_blob; + struct blob_descriptor *blob; + void *buffer_copy; sha1_buffer(buffer, size, hash); - existing_blob = lookup_blob(blob_table, hash); - if (existing_blob) { - wimlib_assert(existing_blob->size == size); - blob = existing_blob; - blob->refcnt++; + + blob = lookup_blob(blob_table, hash); + if (blob) + return blob; + + blob = new_blob_descriptor(); + if (!blob) + return NULL; + + buffer_copy = memdup(buffer, size); + if (!buffer_copy) { + free_blob_descriptor(blob); + return NULL; + } + blob_set_is_located_in_attached_buffer(blob, buffer_copy, size); + copy_hash(blob->hash, hash); + blob_table_insert(blob_table, blob); + return blob; +} + +struct blob_descriptor * +after_blob_hashed(struct blob_descriptor *blob, + struct blob_descriptor **back_ptr, + struct blob_table *blob_table) +{ + struct blob_descriptor *duplicate_blob; + + list_del(&blob->unhashed_list); + blob->unhashed = 0; + + /* Look for a duplicate blob */ + duplicate_blob = lookup_blob(blob_table, blob->hash); + if (duplicate_blob) { + /* We have a duplicate blob. Transfer the reference counts from + * this blob to the duplicate and update the reference to this + * blob (from a stream) to point to the duplicate. The caller + * is responsible for freeing @blob if needed. */ + wimlib_assert(duplicate_blob->size == blob->size); + duplicate_blob->refcnt += blob->refcnt; + blob->refcnt = 0; + *back_ptr = duplicate_blob; + return duplicate_blob; } else { - void *buffer_copy; - blob = new_blob_descriptor(); - if (blob == NULL) - return NULL; - buffer_copy = memdup(buffer, size); - if (buffer_copy == NULL) { - free_blob_descriptor(blob); - return NULL; - } - blob->blob_location = BLOB_IN_ATTACHED_BUFFER; - blob->attached_buffer = buffer_copy; - blob->size = size; - copy_hash(blob->hash, hash); + /* No duplicate blob, so we need to insert this blob into the + * blob table and treat it as a hashed blob. */ blob_table_insert(blob_table, blob); + return blob; } - return blob; } /* @@ -1296,42 +1303,16 @@ int hash_unhashed_blob(struct blob_descriptor *blob, struct blob_table *blob_table, struct blob_descriptor **blob_ret) { - int ret; - struct blob_descriptor *duplicate_blob; struct blob_descriptor **back_ptr; + int ret; - wimlib_assert(blob->unhashed); - - /* back_ptr must be saved because @back_inode and @back_stream_id are in - * union with the SHA-1 message digest and will no longer be valid once - * the SHA-1 has been calculated. */ back_ptr = retrieve_pointer_to_unhashed_blob(blob); ret = sha1_blob(blob); if (ret) return ret; - list_del(&blob->unhashed_list); - blob->unhashed = 0; - - /* Look for a duplicate blob */ - duplicate_blob = lookup_blob(blob_table, blob->hash); - if (duplicate_blob) { - /* We have a duplicate blob. Transfer the reference counts from - * this blob to the duplicate and update the reference to this - * blob (from an stream) to point to the duplicate. The caller - * is responsible for freeing @blob if needed. */ - wimlib_assert(duplicate_blob->size == blob->size); - duplicate_blob->refcnt += blob->refcnt; - blob->refcnt = 0; - *back_ptr = duplicate_blob; - blob = duplicate_blob; - } else { - /* No duplicate blob, so we need to insert this blob into the - * blob table and treat it as a hashed blob. */ - blob_table_insert(blob_table, blob); - } - *blob_ret = blob; + *blob_ret = after_blob_hashed(blob, back_ptr, blob_table); return 0; } @@ -1343,8 +1324,10 @@ blob_to_wimlib_resource_entry(const struct blob_descriptor *blob, wentry->uncompressed_size = blob->size; if (blob->blob_location == BLOB_IN_WIM) { + unsigned res_flags = blob->rdesc->flags; + wentry->part_number = blob->rdesc->wim->hdr.part_number; - if (blob->flags & WIM_RESHDR_FLAG_SOLID) { + if (res_flags & WIM_RESHDR_FLAG_SOLID) { wentry->offset = blob->offset_in_res; } else { wentry->compressed_size = blob->rdesc->size_in_wim; @@ -1353,14 +1336,16 @@ blob_to_wimlib_resource_entry(const struct blob_descriptor *blob, wentry->raw_resource_offset_in_wim = blob->rdesc->offset_in_wim; wentry->raw_resource_compressed_size = blob->rdesc->size_in_wim; wentry->raw_resource_uncompressed_size = blob->rdesc->uncompressed_size; + + wentry->is_compressed = (res_flags & WIM_RESHDR_FLAG_COMPRESSED) != 0; + wentry->is_free = (res_flags & WIM_RESHDR_FLAG_FREE) != 0; + wentry->is_spanned = (res_flags & WIM_RESHDR_FLAG_SPANNED) != 0; + wentry->packed = (res_flags & WIM_RESHDR_FLAG_SOLID) != 0; } - copy_hash(wentry->sha1_hash, blob->hash); + if (!blob->unhashed) + copy_hash(wentry->sha1_hash, blob->hash); wentry->reference_count = blob->refcnt; - wentry->is_compressed = (blob->flags & WIM_RESHDR_FLAG_COMPRESSED) != 0; - wentry->is_metadata = (blob->flags & WIM_RESHDR_FLAG_METADATA) != 0; - wentry->is_free = (blob->flags & WIM_RESHDR_FLAG_FREE) != 0; - wentry->is_spanned = (blob->flags & WIM_RESHDR_FLAG_SPANNED) != 0; - wentry->packed = (blob->flags & WIM_RESHDR_FLAG_SOLID) != 0; + wentry->is_metadata = blob->is_metadata; } struct iterate_blob_context { @@ -1394,10 +1379,17 @@ wimlib_iterate_lookup_table(WIMStruct *wim, int flags, if (wim_has_metadata(wim)) { int ret; for (int i = 0; i < wim->hdr.image_count; i++) { - ret = do_iterate_blob(wim->image_metadata[i]->metadata_blob, - &ctx); + struct blob_descriptor *blob; + struct wim_image_metadata *imd = wim->image_metadata[i]; + + ret = do_iterate_blob(imd->metadata_blob, &ctx); if (ret) return ret; + image_for_each_unhashed_blob(blob, imd) { + ret = do_iterate_blob(blob, &ctx); + if (ret) + return ret; + } } } return for_blob_in_table(wim->blob_table, do_iterate_blob, &ctx);