X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Fextract.c;h=1907e699315e16e7c94029af63c4872b2e887af8;hp=b40d708d6c6e8b5fc4f64ef8e61e31514e4b2edf;hb=4377b51bd25d329e2f2c68772385f62e9c867b5c;hpb=57117039d9fe296fb160217aea226856c44c5d10 diff --git a/src/extract.c b/src/extract.c index b40d708d..1907e699 100644 --- a/src/extract.c +++ b/src/extract.c @@ -53,7 +53,6 @@ #include "wimlib/reparse.h" #include "wimlib/resource.h" #include "wimlib/security.h" -#include "wimlib/swm.h" #ifdef __WIN32__ # include "wimlib/win32.h" /* for realpath() equivalent */ #endif @@ -123,13 +122,19 @@ ref_stream_to_extract(struct wim_lookup_table_entry *lte, if (!lte) return 0; - if (likely(!is_linked_extraction(ctx)) || (lte->out_refcnt == 0 && - lte->extracted_file == NULL)) + /* Tally the size only for each extraction of the stream (not hard + * links). */ + if (!(dentry->d_inode->i_visited && + ctx->supported_features.hard_links) && + (!is_linked_extraction(ctx) || (lte->out_refcnt == 0 && + lte->extracted_file == NULL))) { - ctx->progress.extract.total_bytes += wim_resource_size(lte); + ctx->progress.extract.total_bytes += lte->size; ctx->progress.extract.num_streams++; } + /* Add stream to the extraction_list only one time, even if it's going + * to be extracted to multiple locations. */ if (lte->out_refcnt == 0) { list_add_tail(<e->extraction_list, &ctx->stream_list); ctx->num_streams_remaining++; @@ -183,7 +188,9 @@ ref_stream_to_extract(struct wim_lookup_table_entry *lte, * extracted (ctx->stream_list) if not already done so, and also update the * progress information (ctx->progress) with the stream. Furthermore, if doing * a sequential extraction, build a mapping from each the stream to the dentries - * referencing it. */ + * referencing it. + * + * This uses the i_visited member of the inodes (assumed to be 0 initially). */ static int dentry_add_streams_to_extract(struct wim_dentry *dentry, void *_ctx) { @@ -195,10 +202,6 @@ dentry_add_streams_to_extract(struct wim_dentry *dentry, void *_ctx) if (dentry->extraction_skipped) return 0; - /* Don't process additional hard links. */ - if (inode->i_visited && ctx->supported_features.hard_links) - return 0; - /* The unnamed data stream will always be extracted, except in an * unlikely case. */ if (!inode_is_encrypted_directory(inode)) { @@ -234,7 +237,7 @@ update_extract_progress(struct apply_ctx *ctx, wimlib_progress_func_t progress_func = ctx->progress_func; union wimlib_progress_info *progress = &ctx->progress; - progress->extract.completed_bytes += wim_resource_size(lte); + progress->extract.completed_bytes += lte->size; if (progress_func && progress->extract.completed_bytes >= ctx->next_progress) { @@ -1115,6 +1118,11 @@ hardlink: return 0; } +/* This is a wrapper around do_dentry_extract_skeleton() that handles building + * the path, doing short name reordering. This is also idempotent; dentries + * already processed have skeleton_extracted set and no action is taken. See + * apply_operations.requires_short_name_reordering for more details about short + * name reordering. */ static int dentry_extract_skeleton(struct wim_dentry *dentry, void *_ctx) { @@ -1124,32 +1132,19 @@ dentry_extract_skeleton(struct wim_dentry *dentry, void *_ctx) struct wim_dentry *other_dentry; int ret; - /* Here we may re-order the extraction of multiple names (hard links) - * for the same file in the same directory in order to ensure the short - * (DOS) name is set correctly. A short name is always associated with - * exactly one long name, and at least on NTFS, only one long name for a - * file can have a short name associated with it. (More specifically, - * there can be unlimited names in the POSIX namespace, but only one - * name can be in the Win32+DOS namespace, or one name in the Win32 - * namespace with a corresponding name in the DOS namespace.) To ensure - * the short name of a file is associated with the correct long name in - * a directory, we extract the long name with a corresponding short name - * before any additional names. This can affect NTFS-3g extraction - * (which uses ntfs_set_ntfs_dos_name(), which doesn't allow specifying - * the long name to associate with a short name) and may affect Win32 - * extraction as well (which uses SetFileShortName()). */ - if (dentry->skeleton_extracted) return 0; + orig_dentry = NULL; if (ctx->supported_features.short_names + && ctx->ops->requires_short_name_reordering && !dentry_has_short_name(dentry) && !dentry->d_inode->i_dos_name_extracted) { inode_for_each_dentry(other_dentry, dentry->d_inode) { if (dentry_has_short_name(other_dentry) && !other_dentry->skeleton_extracted - && other_dentry->parent == dentry->parent) + && other_dentry->in_extraction_tree) { DEBUG("Creating %"TS" before %"TS" " "to guarantee correct DOS name extraction", @@ -1179,6 +1174,14 @@ again: return 0; } +static int +dentry_extract_dir_skeleton(struct wim_dentry *dentry, void *_ctx) +{ + if (dentry->d_inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY) + return dentry_extract_skeleton(dentry, _ctx); + return 0; +} + /* Create a file or directory, then immediately extract all streams. This * assumes that WIMLIB_EXTRACT_FLAG_SEQUENTIAL is not specified, since the WIM * may not be read sequentially by this function. */ @@ -1254,8 +1257,7 @@ extract_stream_instances(struct wim_lookup_table_entry *lte, goto out_free_lte_tmp; } filedes_init(&fd, raw_fd); - ret = extract_wim_resource_to_fd(lte, &fd, - wim_resource_size(lte)); + ret = extract_stream_to_fd(lte, &fd, lte->size); if (filedes_close(&fd) && !ret) ret = WIMLIB_ERR_WRITE; if (ret) @@ -1324,12 +1326,14 @@ extract_stream_list(struct apply_ctx *ctx) /* Read the header from a stream in a pipable WIM. */ static int read_pwm_stream_header(WIMStruct *pwm, struct wim_lookup_table_entry *lte, + struct wim_resource_spec *rspec, int flags, struct wim_header_disk *hdr_ret) { union { struct pwm_stream_hdr stream_hdr; struct wim_header_disk pwm_hdr; } buf; + struct wim_reshdr reshdr; int ret; ret = full_read(&pwm->in_fd, &buf.stream_hdr, sizeof(buf.stream_hdr)); @@ -1353,20 +1357,17 @@ read_pwm_stream_header(WIMStruct *pwm, struct wim_lookup_table_entry *lte, return WIMLIB_ERR_INVALID_PIPABLE_WIM; } - lte->resource_entry.original_size = le64_to_cpu(buf.stream_hdr.uncompressed_size); copy_hash(lte->hash, buf.stream_hdr.hash); - lte->resource_entry.flags = le32_to_cpu(buf.stream_hdr.flags); - lte->resource_entry.offset = pwm->in_fd.offset; - lte->resource_location = RESOURCE_IN_WIM; - lte->wim = pwm; - if (lte->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED) { - lte->compression_type = pwm->compression_type; - lte->resource_entry.size = 0; - } else { - lte->compression_type = WIMLIB_COMPRESSION_TYPE_NONE; - lte->resource_entry.size = lte->resource_entry.original_size; - } - lte->is_pipable = 1; + + reshdr.size_in_wim = 0; + reshdr.flags = le32_to_cpu(buf.stream_hdr.flags); + reshdr.offset_in_wim = pwm->in_fd.offset; + reshdr.uncompressed_size = le64_to_cpu(buf.stream_hdr.uncompressed_size); + wim_res_hdr_to_spec(&reshdr, pwm, rspec); + lte_bind_wim_resource_spec(lte, rspec); + lte->flags = rspec->flags; + lte->size = rspec->uncompressed_size; + lte->offset_in_res = 0; return 0; read_error: @@ -1375,13 +1376,22 @@ read_error: return ret; } +static int +skip_pwm_chunk_cb(const void *chunk, size_t chunk_size, void *_ctx) +{ + return 0; +} + /* Skip over an unneeded stream in a pipable WIM being read from a pipe. */ static int skip_pwm_stream(struct wim_lookup_table_entry *lte) { - return read_partial_wim_resource(lte, wim_resource_size(lte), - NULL, NULL, - WIMLIB_READ_RESOURCE_FLAG_SEEK_ONLY, + return read_partial_wim_resource(lte, + lte->size, + skip_pwm_chunk_cb, + lte_cchunk_size(lte), + NULL, + WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS, 0); } @@ -1389,6 +1399,7 @@ static int extract_streams_from_pipe(struct apply_ctx *ctx) { struct wim_lookup_table_entry *found_lte; + struct wim_resource_spec *rspec; struct wim_lookup_table_entry *needed_lte; struct wim_lookup_table *lookup_table; struct wim_header_disk pwm_hdr; @@ -1397,9 +1408,13 @@ extract_streams_from_pipe(struct apply_ctx *ctx) ret = WIMLIB_ERR_NOMEM; found_lte = new_lookup_table_entry(); - if (!found_lte) + if (found_lte == NULL) goto out; + rspec = MALLOC(sizeof(struct wim_resource_spec)); + if (rspec == NULL) + goto out_free_found_lte; + lookup_table = ctx->wim->lookup_table; pwm_flags = PWM_ALLOW_WIM_HDR; if ((ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RESUME)) @@ -1411,8 +1426,10 @@ extract_streams_from_pipe(struct apply_ctx *ctx) ctx->progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN, &ctx->progress); while (ctx->num_streams_remaining) { - ret = read_pwm_stream_header(ctx->wim, found_lte, pwm_flags, - &pwm_hdr); + if (found_lte->resource_location != RESOURCE_NONEXISTENT) + lte_unbind_wim_resource_spec(found_lte); + ret = read_pwm_stream_header(ctx->wim, found_lte, rspec, + pwm_flags, &pwm_hdr); if (ret) { if (ret == WIMLIB_ERR_UNEXPECTED_END_OF_FILE && (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RESUME)) @@ -1423,18 +1440,18 @@ extract_streams_from_pipe(struct apply_ctx *ctx) } if ((found_lte->resource_location != RESOURCE_NONEXISTENT) - && !(found_lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) - && (needed_lte = __lookup_resource(lookup_table, found_lte->hash)) + && !(found_lte->flags & WIM_RESHDR_FLAG_METADATA) + && (needed_lte = lookup_resource(lookup_table, found_lte->hash)) && (needed_lte->out_refcnt)) { - copy_resource_entry(&needed_lte->resource_entry, - &found_lte->resource_entry); - needed_lte->resource_location = found_lte->resource_location; - needed_lte->wim = found_lte->wim; - needed_lte->compression_type = found_lte->compression_type; - needed_lte->is_pipable = found_lte->is_pipable; - + lte_unbind_wim_resource_spec(found_lte); + lte_bind_wim_resource_spec(needed_lte, rspec); + needed_lte->offset_in_res = found_lte->offset_in_res; + needed_lte->flags = found_lte->flags; + needed_lte->size = found_lte->size; ret = extract_stream_instances(needed_lte, ctx, false); + lte_unbind_wim_resource_spec(needed_lte); + if (ret) goto out_free_found_lte; ctx->num_streams_remaining--; @@ -1466,6 +1483,8 @@ extract_streams_from_pipe(struct apply_ctx *ctx) } ret = 0; out_free_found_lte: + if (found_lte->resource_location != RESOURCE_IN_WIM) + FREE(rspec); free_lookup_table_entry(found_lte); out: return ret; @@ -1525,8 +1544,7 @@ extract_dentry_to_stdout(struct wim_dentry *dentry) if (lte) { struct filedes _stdout; filedes_init(&_stdout, STDOUT_FILENO); - ret = extract_wim_resource_to_fd(lte, &_stdout, - wim_resource_size(lte)); + ret = extract_stream_to_fd(lte, &_stdout, lte->size); } } return ret; @@ -1621,6 +1639,8 @@ dentry_calculate_extraction_path(struct wim_dentry *dentry, void *_args) struct apply_ctx *ctx = _args; int ret; + dentry->in_extraction_tree = 1; + if (dentry == ctx->extract_root || dentry->extraction_skipped) return 0; @@ -1740,6 +1760,7 @@ dentry_reset_needs_extraction(struct wim_dentry *dentry, void *_ignore) { struct wim_inode *inode = dentry->d_inode; + dentry->in_extraction_tree = 0; dentry->extraction_skipped = 0; dentry->was_hardlinked = 0; dentry->skeleton_extracted = 0; @@ -1799,20 +1820,13 @@ dentry_tally_features(struct wim_dentry *dentry, void *_features) return 0; } -static int -dentry_clear_inode_visited(struct wim_dentry *dentry, void *_ignore) -{ - dentry->d_inode->i_visited = 0; - return 0; -} - /* Tally the features necessary to extract a dentry tree. */ static void dentry_tree_get_features(struct wim_dentry *root, struct wim_features *features) { memset(features, 0, sizeof(struct wim_features)); for_dentry_in_tree(root, dentry_tally_features, features); - for_dentry_in_tree(root, dentry_clear_inode_visited, NULL); + dentry_tree_clear_inode_visited(root); } static u32 @@ -2066,7 +2080,7 @@ do_extract_warnings(struct apply_ctx *ctx) ctx->no_security_descriptors == 0) return; - WARNING("Extraction of \"%"TS"\" complete, but with one or more warnings:", + WARNING("Extraction to \"%"TS"\" complete, but with one or more warnings:", ctx->target); if (ctx->partial_security_descriptors != 0) { WARNING("- Could only partially set the security descriptor\n" @@ -2318,6 +2332,13 @@ extract_tree(WIMStruct *wim, const tchar *wim_source_path, const tchar *target, ctx.realtarget_nchars = tstrlen(ctx.realtarget); } + if (ctx.ops->requires_short_name_reordering) { + ret = for_dentry_in_tree(root, dentry_extract_dir_skeleton, + &ctx); + if (ret) + goto out_free_realtarget; + } + /* Finally, the important part: extract the tree of files. */ if (extract_flags & (WIMLIB_EXTRACT_FLAG_SEQUENTIAL | WIMLIB_EXTRACT_FLAG_FROM_PIPE)) { @@ -2558,8 +2579,6 @@ wimlib_extract_files(WIMStruct *wim, const struct wimlib_extract_command *cmds, size_t num_cmds, int default_extract_flags, - WIMStruct **additional_swms, - unsigned num_additional_swms, wimlib_progress_func_t progress_func) { int ret; @@ -2568,21 +2587,12 @@ wimlib_extract_files(WIMStruct *wim, default_extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC; - ret = verify_swm_set(wim, additional_swms, num_additional_swms); - if (ret) - goto out; - if (num_cmds == 0) - goto out; - - if (num_additional_swms) - merge_lookup_tables(wim, additional_swms, num_additional_swms); + return 0; cmds_copy = CALLOC(num_cmds, sizeof(cmds[0])); - if (!cmds_copy) { - ret = WIMLIB_ERR_NOMEM; - goto out_restore_lookup_table; - } + if (!cmds_copy) + return WIMLIB_ERR_NOMEM; for (size_t i = 0; i < num_cmds; i++) { cmds_copy[i].extract_flags = (default_extract_flags | @@ -2619,10 +2629,6 @@ out_free_cmds_copy: FREE(cmds_copy[i].fs_dest_path); } FREE(cmds_copy); -out_restore_lookup_table: - if (num_additional_swms) - unmerge_lookup_table(wim); -out: return ret; } @@ -2649,7 +2655,7 @@ extract_single_image(WIMStruct *wim, int image, { int ret; tchar *target_copy = canonicalize_fs_path(target); - if (!target_copy) + if (target_copy == NULL) return WIMLIB_ERR_NOMEM; struct wimlib_extract_command cmd = { .wim_source_path = T(""), @@ -2743,24 +2749,10 @@ do_wimlib_extract_image(WIMStruct *wim, int image, const tchar *target, int extract_flags, - WIMStruct **additional_swms, - unsigned num_additional_swms, wimlib_progress_func_t progress_func) { int ret; - if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) { - wimlib_assert(wim->hdr.part_number == 1); - wimlib_assert(num_additional_swms == 0); - } else { - ret = verify_swm_set(wim, additional_swms, num_additional_swms); - if (ret) - return ret; - - if (num_additional_swms) - merge_lookup_tables(wim, additional_swms, num_additional_swms); - } - if (image == WIMLIB_ALL_IMAGES) { ret = extract_all_images(wim, target, extract_flags, progress_func); @@ -2776,8 +2768,6 @@ do_wimlib_extract_image(WIMStruct *wim, lte_free_extracted_file, NULL); } - if (num_additional_swms) - unmerge_lookup_table(wim); return ret; } @@ -2805,8 +2795,7 @@ wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name, * wimlib_open_wim(), getting a WIMStruct in this way will result in * an empty lookup table, no XML data read, and no filename set. */ ret = open_wim_as_WIMStruct(&pipe_fd, - WIMLIB_OPEN_FLAG_FROM_PIPE | - WIMLIB_OPEN_FLAG_SPLIT_OK, + WIMLIB_OPEN_FLAG_FROM_PIPE, &pwm, progress_func); if (ret) return ret; @@ -2838,11 +2827,12 @@ wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name, * WIMs.) */ { struct wim_lookup_table_entry xml_lte; - ret = read_pwm_stream_header(pwm, &xml_lte, 0, NULL); + struct wim_resource_spec xml_rspec; + ret = read_pwm_stream_header(pwm, &xml_lte, &xml_rspec, 0, NULL); if (ret) goto out_wimlib_free; - if (!(xml_lte.resource_entry.flags & WIM_RESHDR_FLAG_METADATA)) + if (!(xml_lte.flags & WIM_RESHDR_FLAG_METADATA)) { ERROR("Expected XML data, but found non-metadata " "stream."); @@ -2850,12 +2840,12 @@ wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name, goto out_wimlib_free; } - copy_resource_entry(&pwm->hdr.xml_res_entry, - &xml_lte.resource_entry); + wim_res_spec_to_hdr(&xml_rspec, &pwm->hdr.xml_data_reshdr); ret = read_wim_xml_data(pwm); if (ret) goto out_wimlib_free; + if (wim_info_get_num_images(pwm->wim_info) != pwm->hdr.image_count) { ERROR("Image count in XML data is not the same as in WIM header."); ret = WIMLIB_ERR_XML; @@ -2891,22 +2881,29 @@ wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name, for (i = 1; i <= pwm->hdr.image_count; i++) { struct wim_lookup_table_entry *metadata_lte; struct wim_image_metadata *imd; + struct wim_resource_spec *metadata_rspec; metadata_lte = new_lookup_table_entry(); - if (!metadata_lte) { + if (metadata_lte == NULL) { ret = WIMLIB_ERR_NOMEM; goto out_wimlib_free; } + metadata_rspec = MALLOC(sizeof(struct wim_resource_spec)); + if (metadata_rspec == NULL) { + ret = WIMLIB_ERR_NOMEM; + free_lookup_table_entry(metadata_lte); + goto out_wimlib_free; + } - ret = read_pwm_stream_header(pwm, metadata_lte, 0, NULL); + ret = read_pwm_stream_header(pwm, metadata_lte, metadata_rspec, 0, NULL); imd = pwm->image_metadata[i - 1]; imd->metadata_lte = metadata_lte; - if (ret) + if (ret) { + FREE(metadata_rspec); goto out_wimlib_free; + } - if (!(metadata_lte->resource_entry.flags & - WIM_RESHDR_FLAG_METADATA)) - { + if (!(metadata_lte->flags & WIM_RESHDR_FLAG_METADATA)) { ERROR("Expected metadata resource, but found " "non-metadata stream."); ret = WIMLIB_ERR_INVALID_PIPABLE_WIM; @@ -2931,7 +2928,7 @@ wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name, /* Extract the image. */ extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE; ret = do_wimlib_extract_image(pwm, image, target, - extract_flags, NULL, 0, progress_func); + extract_flags, progress_func); /* Clean up and return. */ out_wimlib_free: wimlib_free(pwm); @@ -2944,12 +2941,9 @@ wimlib_extract_image(WIMStruct *wim, int image, const tchar *target, int extract_flags, - WIMStruct **additional_swms, - unsigned num_additional_swms, wimlib_progress_func_t progress_func) { extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC; return do_wimlib_extract_image(wim, image, target, extract_flags, - additional_swms, num_additional_swms, progress_func); }