X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Fextract.c;h=9cbac117f54438d25372c90cb43dd4df4c327bb5;hp=d4efc24e2ff826b4983b7bc3d1359010dddb58b9;hb=e10a0fd1f906c342b0b45bc393ec10043c6a8936;hpb=ca1be480fd209d8c24e19350e440c01832310a2d diff --git a/src/extract.c b/src/extract.c index d4efc24e..9cbac117 100644 --- a/src/extract.c +++ b/src/extract.c @@ -26,10 +26,9 @@ /* * This file provides the API functions wimlib_extract_image(), - * wimlib_extract_files(), and wimlib_extract_image_from_pipe(). Internally, - * all end up calling extract_tree() zero or more times to extract a tree of - * files from the currently selected WIM image to the specified target directory - * or NTFS volume. + * wimlib_extract_image_from_pipe(), wimlib_extract_files(), + * wimlib_extract_paths(), and wimlib_extract_pathlist(). Internally, all end + * up calling do_wimlib_extract_paths() and extract_trees(). * * Although wimlib supports multiple extraction modes/backends (NTFS-3g, UNIX, * Win32), this file does not itself have code to extract files or directories @@ -49,15 +48,16 @@ #include "wimlib/error.h" #include "wimlib/lookup_table.h" #include "wimlib/metadata.h" +#include "wimlib/pathlist.h" #include "wimlib/paths.h" #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 #include "wimlib/xml.h" +#include "wimlib/wildcard.h" #include "wimlib/wim.h" #include @@ -68,38 +68,38 @@ #define WIMLIB_EXTRACT_FLAG_MULTI_IMAGE 0x80000000 #define WIMLIB_EXTRACT_FLAG_FROM_PIPE 0x40000000 -#define WIMLIB_EXTRACT_MASK_PUBLIC 0x3fffffff +#define WIMLIB_EXTRACT_FLAG_FILEMODE 0x20000000 +#define WIMLIB_EXTRACT_FLAG_IMAGEMODE 0x10000000 + +/* Keep in sync with wimlib.h */ +#define WIMLIB_EXTRACT_MASK_PUBLIC \ + (WIMLIB_EXTRACT_FLAG_NTFS | \ + WIMLIB_EXTRACT_FLAG_HARDLINK | \ + WIMLIB_EXTRACT_FLAG_SYMLINK | \ + WIMLIB_EXTRACT_FLAG_VERBOSE | \ + WIMLIB_EXTRACT_FLAG_SEQUENTIAL | \ + WIMLIB_EXTRACT_FLAG_UNIX_DATA | \ + WIMLIB_EXTRACT_FLAG_NO_ACLS | \ + WIMLIB_EXTRACT_FLAG_STRICT_ACLS | \ + WIMLIB_EXTRACT_FLAG_RPFIX | \ + WIMLIB_EXTRACT_FLAG_NORPFIX | \ + WIMLIB_EXTRACT_FLAG_TO_STDOUT | \ + WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES | \ + WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS | \ + WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS | \ + WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES | \ + WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS | \ + WIMLIB_EXTRACT_FLAG_RESUME | \ + WIMLIB_EXTRACT_FLAG_FILE_ORDER | \ + WIMLIB_EXTRACT_FLAG_GLOB_PATHS | \ + WIMLIB_EXTRACT_FLAG_STRICT_GLOB | \ + WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES | \ + WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE) -/* Given a WIM dentry in the tree to be extracted, resolve all streams in the - * corresponding inode and set 'out_refcnt' in each to 0. */ -static int -dentry_resolve_and_zero_lte_refcnt(struct wim_dentry *dentry, void *_ctx) +static bool +dentry_in_list(const struct wim_dentry *dentry) { - struct apply_ctx *ctx = _ctx; - struct wim_inode *inode = dentry->d_inode; - struct wim_lookup_table_entry *lte; - int ret; - bool force = false; - - if (dentry->extraction_skipped) - return 0; - - /* Special case: when extracting from a pipe, the WIM lookup table is - * initially empty, so "resolving" an inode's streams is initially not - * possible. However, we still need to keep track of which streams, - * identified by SHA1 message digests, need to be extracted, so we - * "resolve" the inode's streams anyway by allocating new entries. */ - if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) - force = true; - ret = inode_resolve_ltes(inode, ctx->wim->lookup_table, force); - if (ret) - return ret; - for (unsigned i = 0; i <= inode->i_num_ads; i++) { - lte = inode_stream_lte_resolved(inode, i); - if (lte) - lte->out_refcnt = 0; - } - return 0; + return dentry->extraction_list.next != NULL; } static inline bool @@ -115,116 +115,6 @@ can_extract_named_data_streams(const struct apply_ctx *ctx) return ctx->supported_features.named_data_streams && !is_linked_extraction(ctx); } - -static int -ref_stream_to_extract(struct wim_lookup_table_entry *lte, - struct wim_dentry *dentry, struct apply_ctx *ctx) -{ - if (!lte) - return 0; - - if (likely(!is_linked_extraction(ctx)) || (lte->out_refcnt == 0 && - lte->extracted_file == NULL)) - { - ctx->progress.extract.total_bytes += wim_resource_size(lte); - ctx->progress.extract.num_streams++; - } - - if (lte->out_refcnt == 0) { - list_add_tail(<e->extraction_list, &ctx->stream_list); - ctx->num_streams_remaining++; - } - - if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_SEQUENTIAL) { - struct wim_dentry **lte_dentries; - - /* Append dentry to this stream's array of dentries referencing - * it. Use inline array to avoid memory allocation until the - * number of dentries becomes too large. */ - if (lte->out_refcnt < ARRAY_LEN(lte->inline_lte_dentries)) { - lte_dentries = lte->inline_lte_dentries; - } else { - struct wim_dentry **prev_lte_dentries; - size_t alloc_lte_dentries; - - if (lte->out_refcnt == ARRAY_LEN(lte->inline_lte_dentries)) { - prev_lte_dentries = NULL; - alloc_lte_dentries = ARRAY_LEN(lte->inline_lte_dentries); - } else { - prev_lte_dentries = lte->lte_dentries; - alloc_lte_dentries = lte->alloc_lte_dentries; - } - - if (lte->out_refcnt == alloc_lte_dentries) { - alloc_lte_dentries *= 2; - lte_dentries = REALLOC(prev_lte_dentries, - alloc_lte_dentries * - sizeof(lte_dentries[0])); - if (!lte_dentries) - return WIMLIB_ERR_NOMEM; - if (prev_lte_dentries == NULL) { - memcpy(lte_dentries, - lte->inline_lte_dentries, - sizeof(lte->inline_lte_dentries)); - } - lte->lte_dentries = lte_dentries; - lte->alloc_lte_dentries = alloc_lte_dentries; - } - lte_dentries = lte->lte_dentries; - } - lte_dentries[lte->out_refcnt] = dentry; - } - lte->out_refcnt++; - return 0; -} - -/* Given a WIM dentry in the tree to be extracted, iterate through streams that - * need to be extracted. For each one, add it to the list of streams to be - * 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. */ -static int -dentry_add_streams_to_extract(struct wim_dentry *dentry, void *_ctx) -{ - struct apply_ctx *ctx = _ctx; - struct wim_inode *inode = dentry->d_inode; - int ret; - - /* Don't process dentries marked as skipped. */ - 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)) { - ret = ref_stream_to_extract(inode_unnamed_lte_resolved(inode), - dentry, ctx); - if (ret) - return ret; - } - - /* Named data streams will be extracted only if supported in the current - * extraction mode and volume, and to avoid complications, if not doing - * a linked extraction. */ - if (can_extract_named_data_streams(ctx)) { - for (u16 i = 0; i < inode->i_num_ads; i++) { - if (!ads_entry_is_named_stream(&inode->i_ads_entries[i])) - continue; - ret = ref_stream_to_extract(inode->i_ads_entries[i].lte, - dentry, ctx); - if (ret) - return ret; - } - } - inode->i_visited = 1; - return 0; -} - /* Inform library user of progress of stream extraction following the successful * extraction of a copy of the stream specified by @lte. */ static void @@ -234,7 +124,7 @@ 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) { @@ -333,13 +223,22 @@ extract_inode(const tchar *path, struct apply_ctx *ctx, struct wim_inode *inode) } else #endif /* !__WIN32__ */ if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY) { - ret = ctx->ops->create_directory(path, ctx); + ret = ctx->ops->create_directory(path, ctx, &inode->extract_cookie); if (ret) { ERROR_WITH_ERRNO("Failed to create the directory " "\"%"TS"\"", path); } + } else if ((inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) && + ctx->ops->extract_encrypted_stream_creates_file && + ctx->supported_features.encrypted_files) { + ret = ctx->ops->extract_encrypted_stream( + path, inode_unnamed_lte_resolved(inode), ctx); + if (ret) { + ERROR_WITH_ERRNO("Failed to create and extract " + "encrypted file \"%"TS"\"", path); + } } else { - ret = ctx->ops->create_file(path, ctx); + ret = ctx->ops->create_file(path, ctx, &inode->extract_cookie); if (ret) { ERROR_WITH_ERRNO("Failed to create the file " "\"%"TS"\"", path); @@ -522,10 +421,10 @@ error: * This function operates slightly differently depending on whether @lte_spec is * NULL or not. When @lte_spec is NULL, the behavior is to extract the default * file contents (unnamed stream), and, if named data streams are supported in - * the extract mode and volume, any named data streams. When @lte_spec is NULL, - * the behavior is to extract only all copies of the stream @lte_spec, and in - * addition use @lte_spec to set the reparse data or create the symbolic link if - * appropriate. + * the extract mode and volume, any named data streams. When @lte_spec is not + * NULL, the behavior is to extract only all copies of the stream @lte_spec, and + * in addition use @lte_spec to set the reparse data or create the symbolic link + * if appropriate. * * @path * Path to file to extract (as can be passed to apply_operations @@ -553,9 +452,10 @@ extract_streams(const tchar *path, struct apply_ctx *ctx, { struct wim_inode *inode = dentry->d_inode; struct wim_lookup_table_entry *lte; + file_spec_t file_spec; int ret; - if (dentry->was_hardlinked) + if (dentry->was_linked) return 0; #ifdef ENABLE_DEBUG @@ -571,6 +471,11 @@ extract_streams(const tchar *path, struct apply_ctx *ctx, } #endif + if (ctx->ops->uses_cookies) + file_spec.cookie = inode->extract_cookie; + else + file_spec.path = path; + /* Unnamed data stream. */ lte = inode_unnamed_lte_resolved(inode); if (lte && (!lte_spec || lte == lte_spec)) { @@ -579,13 +484,20 @@ extract_streams(const tchar *path, struct apply_ctx *ctx, if (!(inode->i_attributes & (FILE_ATTRIBUTE_DIRECTORY | FILE_ATTRIBUTE_REPARSE_POINT))) { - if ((inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) && - ctx->supported_features.encrypted_files) - ret = ctx->ops->extract_encrypted_stream(path, lte, ctx); - else - ret = ctx->ops->extract_unnamed_stream(path, lte, ctx); - if (ret) - goto error; + if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED && + ctx->supported_features.encrypted_files) { + if (!ctx->ops->extract_encrypted_stream_creates_file) { + ret = ctx->ops->extract_encrypted_stream( + path, lte, ctx); + if (ret) + goto error; + } + } else { + ret = ctx->ops->extract_unnamed_stream( + file_spec, lte, ctx); + if (ret) + goto error; + } update_extract_progress(ctx, lte); } else if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) @@ -617,7 +529,7 @@ extract_streams(const tchar *path, struct apply_ctx *ctx, continue; if (lte_spec) lte = lte_override; - ret = ctx->ops->extract_named_stream(path, entry->stream_name, + ret = ctx->ops->extract_named_stream(file_spec, entry->stream_name, entry->stream_name_nbytes / 2, lte, ctx); if (ret) @@ -636,16 +548,30 @@ error: * extraction mode. */ static int extract_file_attributes(const tchar *path, struct apply_ctx *ctx, - struct wim_dentry *dentry) + struct wim_dentry *dentry, unsigned pass) { int ret; - if (ctx->ops->set_file_attributes) { - if (dentry == ctx->extract_root && ctx->root_dentry_is_special) - return 0; - ret = ctx->ops->set_file_attributes(path, - dentry->d_inode->i_attributes, - ctx); + if (ctx->ops->set_file_attributes && + !(ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES) && + !(dentry == ctx->target_dentry && ctx->root_dentry_is_special)) { + u32 attributes = dentry->d_inode->i_attributes; + + /* Clear unsupported attributes. */ + attributes &= ctx->supported_attributes_mask; + + if ((attributes & FILE_ATTRIBUTE_DIRECTORY && + !ctx->supported_features.encrypted_directories) || + (!(attributes & FILE_ATTRIBUTE_DIRECTORY) && + !ctx->supported_features.encrypted_files)) + { + attributes &= ~FILE_ATTRIBUTE_ENCRYPTED; + } + + if (attributes == 0) + attributes = FILE_ATTRIBUTE_NORMAL; + + ret = ctx->ops->set_file_attributes(path, attributes, ctx, pass); if (ret) { ERROR_WITH_ERRNO("Failed to set attributes on " "\"%"TS"\"", path); @@ -670,7 +596,7 @@ extract_short_name(const tchar *path, struct apply_ctx *ctx, /* The root of the dentry tree being extracted may not be extracted to * its original name, so its short name should be ignored. */ - if (dentry == ctx->extract_root) + if (dentry == ctx->target_dentry) return 0; if (ctx->supported_features.short_names) { @@ -701,7 +627,7 @@ extract_security(const tchar *path, struct apply_ctx *ctx, if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS) return 0; - if ((ctx->extract_root == dentry) && ctx->root_dentry_is_special) + if ((ctx->target_dentry == dentry) && ctx->root_dentry_is_special) return 0; #ifndef __WIN32__ @@ -747,11 +673,14 @@ extract_security(const tchar *path, struct apply_ctx *ctx, "descriptor on \"%"TS"\"", path); return ret; } else { + #if 0 if (errno != EACCES) { WARNING_WITH_ERRNO("Failed to set " "security descriptor " "on \"%"TS"\"", path); } + #endif + ctx->no_security_descriptors++; } } } @@ -767,7 +696,7 @@ extract_timestamps(const tchar *path, struct apply_ctx *ctx, struct wim_inode *inode = dentry->d_inode; int ret; - if ((ctx->extract_root == dentry) && ctx->root_dentry_is_special) + if ((ctx->target_dentry == dentry) && ctx->root_dentry_is_special) return 0; if (ctx->ops->set_timestamps) { @@ -798,12 +727,15 @@ dentry_is_supported(struct wim_dentry *dentry, struct wim_inode *inode = dentry->d_inode; if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) { - if (supported_features->reparse_points) - return true; - if (supported_features->symlink_reparse_points && - inode_is_symlink(inode)) - return true; - return false; + return supported_features->reparse_points || + (inode_is_symlink(inode) && + supported_features->symlink_reparse_points); + } + if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) { + if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY) + return supported_features->encrypted_directories != 0; + else + return supported_features->encrypted_files != 0; } return true; } @@ -812,7 +744,7 @@ dentry_is_supported(struct wim_dentry *dentry, * format understood by the callbacks in the apply_operations being used. * * Write the resulting path into @path, which must have room for at least - * ctx->ops->max_path characters including the null-terminator. + * ctx->ops->path_max characters. * * Return %true if successful; %false if this WIM dentry doesn't actually need * to be extracted or if the calculated path exceeds ctx->ops->max_path @@ -822,7 +754,7 @@ dentry_is_supported(struct wim_dentry *dentry, * until the extraction root. */ static bool build_extraction_path(tchar path[], struct wim_dentry *dentry, - struct apply_ctx *ctx) + const struct apply_ctx *ctx) { size_t path_nchars; LIST_HEAD(ancestor_list); @@ -831,9 +763,6 @@ build_extraction_path(tchar path[], struct wim_dentry *dentry, size_t target_prefix_nchars; struct wim_dentry *d; - if (dentry->extraction_skipped) - return false; - path_nchars = ctx->ops->path_prefix_nchars; if (ctx->ops->requires_realtarget_in_paths) { @@ -848,7 +777,10 @@ build_extraction_path(tchar path[], struct wim_dentry *dentry, } path_nchars += target_prefix_nchars; - for (d = dentry; d != ctx->extract_root; d = d->parent) { + for (d = dentry; d != ctx->target_dentry; d = d->parent) { + if (!dentry_in_list(d)) + break; + path_nchars += d->extraction_name_nchars + 1; list_add(&d->tmp_list, &ancestor_list); } @@ -876,6 +808,11 @@ static unsigned get_num_path_components(const tchar *path, tchar path_separator) { unsigned num_components = 0; +#ifdef __WIN32__ + /* Ignore drive letter. */ + if (path[0] != L'\0' && path[1] == L':') + path += 2; +#endif while (*path) { while (*path == path_separator) @@ -900,7 +837,7 @@ extract_multiimage_symlink(const tchar *oldpath, const tchar *newpath, int ret; num_raw_path_components = 0; - for (d = dentry; d != ctx->extract_root; d = d->parent) + for (d = dentry; d != ctx->target_dentry; d = d->parent) num_raw_path_components++; if (ctx->ops->requires_realtarget_in_paths) @@ -918,7 +855,11 @@ extract_multiimage_symlink(const tchar *oldpath, const tchar *newpath, num_target_path_components--; } - p_old = oldpath; + p_old = oldpath + ctx->ops->path_prefix_nchars; +#ifdef __WIN32__ + if (p_old[0] != L'\0' && p_old[1] == ':') + p_old += 2; +#endif while (*p_old == ctx->ops->path_separator) p_old++; while (--num_target_path_components) { @@ -988,7 +929,7 @@ do_dentry_extract_skeleton(tchar path[], struct wim_dentry *dentry, /* Create this file or directory unless it's the extraction root, which * was already created if necessary. */ - if (dentry != ctx->extract_root) { + if (dentry != ctx->target_dentry) { ret = extract_inode(path, ctx, inode); if (ret) return ret; @@ -997,13 +938,18 @@ do_dentry_extract_skeleton(tchar path[], struct wim_dentry *dentry, /* Create empty named data streams. */ if (can_extract_named_data_streams(ctx)) { for (u16 i = 0; i < inode->i_num_ads; i++) { + file_spec_t file_spec; struct wim_ads_entry *entry = &inode->i_ads_entries[i]; if (!ads_entry_is_named_stream(entry)) continue; if (entry->lte) continue; - ret = ctx->ops->extract_named_stream(path, + if (ctx->ops->uses_cookies) + file_spec.cookie = inode->extract_cookie; + else + file_spec.path = path; + ret = ctx->ops->extract_named_stream(file_spec, entry->stream_name, entry->stream_name_nbytes / 2, entry->lte, ctx); @@ -1017,7 +963,7 @@ do_dentry_extract_skeleton(tchar path[], struct wim_dentry *dentry, } /* Set file attributes (if supported). */ - ret = extract_file_attributes(path, ctx, dentry); + ret = extract_file_attributes(path, ctx, dentry, 0); if (ret) return ret; @@ -1049,52 +995,43 @@ symlink: ret = extract_multiimage_symlink(oldpath, path, ctx, dentry); if (ret) return ret; - dentry->was_hardlinked = 1; + dentry->was_linked = 1; return 0; hardlink: ret = extract_hardlink(oldpath, path, ctx); if (ret) return ret; - dentry->was_hardlinked = 1; + dentry->was_linked = 1; 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) +dentry_extract_skeleton(struct wim_dentry *dentry, struct apply_ctx *ctx) { - struct apply_ctx *ctx = _ctx; tchar path[ctx->ops->path_max]; struct wim_dentry *orig_dentry; 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) + && dentry_in_list(other_dentry)) { DEBUG("Creating %"TS" before %"TS" " "to guarantee correct DOS name extraction", @@ -1124,13 +1061,11 @@ again: 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 +/* Create a file or directory, then immediately extract all streams. The WIM * may not be read sequentially by this function. */ static int -dentry_extract(struct wim_dentry *dentry, void *_ctx) +dentry_extract(struct wim_dentry *dentry, struct apply_ctx *ctx) { - struct apply_ctx *ctx = _ctx; tchar path[ctx->ops->path_max]; int ret; @@ -1144,207 +1079,351 @@ dentry_extract(struct wim_dentry *dentry, void *_ctx) return extract_streams(path, ctx, dentry, NULL, NULL); } -/* Extract all instances of the stream @lte that are being extracted in this - * call of extract_tree(). @can_seek specifies whether the WIM file descriptor - * is seekable or not (e.g. is a pipe). If not and the stream needs to be - * extracted multiple times, it is extracted to a temporary file first. - * - * This is intended for use with sequential extraction of a WIM image - * (WIMLIB_EXTRACT_FLAG_SEQUENTIAL specified). */ +/* Finish extracting a file, directory, or symbolic link by setting file + * security and timestamps. */ static int -extract_stream_instances(struct wim_lookup_table_entry *lte, - struct apply_ctx *ctx, bool can_seek) +dentry_extract_final(struct wim_dentry *dentry, struct apply_ctx *ctx) { - struct wim_dentry **lte_dentries; - struct wim_lookup_table_entry *lte_tmp = NULL; - struct wim_lookup_table_entry *lte_override; - tchar *stream_tmp_filename = NULL; - tchar path[ctx->ops->path_max]; - unsigned i; int ret; + tchar path[ctx->ops->path_max]; - if (lte->out_refcnt <= ARRAY_LEN(lte->inline_lte_dentries)) - lte_dentries = lte->inline_lte_dentries; - else - lte_dentries = lte->lte_dentries; + if (!build_extraction_path(path, dentry, ctx)) + return 0; - if (likely(can_seek || lte->out_refcnt < 2)) { - lte_override = lte; - } else { - /* Need to extract stream to temporary file. */ - struct filedes fd; - int raw_fd; - - stream_tmp_filename = ttempnam(NULL, T("wimlib")); - if (!stream_tmp_filename) { - ERROR_WITH_ERRNO("Failed to create temporary filename"); - ret = WIMLIB_ERR_OPEN; - goto out; - } + ret = extract_security(path, ctx, dentry); + if (ret) + return ret; - lte_tmp = memdup(lte, sizeof(struct wim_lookup_table_entry)); - if (!lte_tmp) { - ret = WIMLIB_ERR_NOMEM; - goto out_free_stream_tmp_filename; - } - lte_tmp->resource_location = RESOURCE_IN_FILE_ON_DISK; - lte_tmp->file_on_disk = stream_tmp_filename; - lte_override = lte_tmp; - - raw_fd = topen(stream_tmp_filename, - O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, 0600); - if (raw_fd < 0) { - ERROR_WITH_ERRNO("Failed to open temporary file"); - ret = WIMLIB_ERR_OPEN; - goto out_free_lte_tmp; - } - filedes_init(&fd, raw_fd); - ret = extract_wim_resource_to_fd(lte, &fd, - wim_resource_size(lte)); - if (filedes_close(&fd) && !ret) - ret = WIMLIB_ERR_WRITE; + if (ctx->ops->requires_final_set_attributes_pass) { + /* Set file attributes (if supported). */ + ret = extract_file_attributes(path, ctx, dentry, 1); if (ret) - goto out_unlink_stream_tmp_file; + return ret; } - /* Extract all instances of the stream, reading either from the stream - * in the WIM file or from the temporary file containing the stream. - * dentry->tmp_flag is used to ensure that each dentry is processed only - * once regardless of how many times this stream appears in the streams - * of the corresponding inode. */ - for (i = 0; i < lte->out_refcnt; i++) { - struct wim_dentry *dentry = lte_dentries[i]; - - if (dentry->tmp_flag) - continue; - if (!build_extraction_path(path, dentry, ctx)) - continue; - ret = extract_streams(path, ctx, dentry, - lte, lte_override); - if (ret) - goto out_clear_tmp_flags; - dentry->tmp_flag = 1; - } - ret = 0; -out_clear_tmp_flags: - for (i = 0; i < lte->out_refcnt; i++) - lte_dentries[i]->tmp_flag = 0; -out_unlink_stream_tmp_file: - if (stream_tmp_filename) - tunlink(stream_tmp_filename); -out_free_lte_tmp: - FREE(lte_tmp); -out_free_stream_tmp_filename: - FREE(stream_tmp_filename); -out: - return ret; + return extract_timestamps(path, ctx, dentry); } -/* Extracts a list of streams (ctx.stream_list), assuming that the directory - * structure and empty files were already created. This relies on the - * per-`struct wim_lookup_table_entry' list of dentries that reference each - * stream that was constructed earlier. Streams are extracted exactly in the - * order of the stream list; however, unless the WIM's file descriptor is - * detected to be non-seekable, streams may be read from the WIM file more than - * one time if multiple copies need to be extracted. */ static int -extract_stream_list(struct apply_ctx *ctx) +extract_structure(struct list_head *dentry_list, struct apply_ctx *ctx) { - struct wim_lookup_table_entry *lte; - bool can_seek; + struct wim_dentry *dentry; int ret; - can_seek = (lseek(ctx->wim->in_fd.fd, 0, SEEK_CUR) != -1); - list_for_each_entry(lte, &ctx->stream_list, extraction_list) { - ret = extract_stream_instances(lte, ctx, can_seek); + list_for_each_entry(dentry, dentry_list, extraction_list) { + ret = dentry_extract_skeleton(dentry, ctx); if (ret) return ret; } return 0; } -#define PWM_ALLOW_WIM_HDR 0x00001 -#define PWM_SILENT_EOF 0x00002 - -/* Read the header from a stream in a pipable WIM. */ static int -read_pwm_stream_header(WIMStruct *pwm, struct wim_lookup_table_entry *lte, - int flags, struct wim_header_disk *hdr_ret) +extract_dir_structure(struct list_head *dentry_list, struct apply_ctx *ctx) { - union { - struct pwm_stream_hdr stream_hdr; - struct wim_header_disk pwm_hdr; - } buf; + struct wim_dentry *dentry; int ret; - ret = full_read(&pwm->in_fd, &buf.stream_hdr, sizeof(buf.stream_hdr)); - if (ret) - goto read_error; + list_for_each_entry(dentry, dentry_list, extraction_list) { + if (dentry_is_directory(dentry)) { + ret = dentry_extract_skeleton(dentry, ctx); + if (ret) + return ret; + } + } + return 0; +} - if ((flags & PWM_ALLOW_WIM_HDR) && buf.stream_hdr.magic == PWM_MAGIC) { - BUILD_BUG_ON(sizeof(buf.pwm_hdr) < sizeof(buf.stream_hdr)); - ret = full_read(&pwm->in_fd, &buf.stream_hdr + 1, - sizeof(buf.pwm_hdr) - sizeof(buf.stream_hdr)); +static int +extract_dentries(struct list_head *dentry_list, struct apply_ctx *ctx) +{ + struct wim_dentry *dentry; + int ret; + list_for_each_entry(dentry, dentry_list, extraction_list) { + ret = dentry_extract(dentry, ctx); if (ret) - goto read_error; - lte->resource_location = RESOURCE_NONEXISTENT; - memcpy(hdr_ret, &buf.pwm_hdr, sizeof(buf.pwm_hdr)); - return 0; + return ret; } + return 0; +} - if (buf.stream_hdr.magic != PWM_STREAM_MAGIC) { - ERROR("Data read on pipe is invalid (expected stream header)."); - return WIMLIB_ERR_INVALID_PIPABLE_WIM; - } +static int +extract_final_metadata(struct list_head *dentry_list, struct apply_ctx *ctx) +{ + struct wim_dentry *dentry; + int ret; - 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; + list_for_each_entry_reverse(dentry, dentry_list, extraction_list) { + ret = dentry_extract_final(dentry, ctx); + if (ret) + return ret; } - lte->is_pipable = 1; return 0; - -read_error: - if (ret != WIMLIB_ERR_UNEXPECTED_END_OF_FILE || !(flags & PWM_SILENT_EOF)) - ERROR_WITH_ERRNO("Error reading pipable WIM from pipe"); - return ret; } -/* Skip over an unneeded stream in a pipable WIM being read from a pipe. */ +/* Creates a temporary file opened for writing. The open file descriptor is + * returned in @fd_ret and its name is returned in @name_ret (dynamically + * allocated). */ static int -skip_pwm_stream(struct wim_lookup_table_entry *lte) +create_temporary_file(struct filedes *fd_ret, tchar **name_ret) { - return read_partial_wim_resource(lte, wim_resource_size(lte), - NULL, NULL, - WIMLIB_READ_RESOURCE_FLAG_SEEK_ONLY, - 0); + tchar *name; + int raw_fd; + +retry: + name = ttempnam(NULL, T("wimlib")); + if (name == NULL) { + ERROR_WITH_ERRNO("Failed to create temporary filename"); + return WIMLIB_ERR_NOMEM; + } + + raw_fd = topen(name, O_WRONLY | O_CREAT | O_EXCL | O_BINARY, 0600); + + if (raw_fd < 0) { + if (errno == EEXIST) { + FREE(name); + goto retry; + } + ERROR_WITH_ERRNO("Failed to open temporary file \"%"TS"\"", name); + FREE(name); + return WIMLIB_ERR_OPEN; + } + + filedes_init(fd_ret, raw_fd); + *name_ret = name; + return 0; } +/* Extract all instances of the stream @lte that are being extracted in this + * call of extract_tree(), but actually read the stream data from @lte_override. + */ static int -extract_streams_from_pipe(struct apply_ctx *ctx) +extract_stream_instances(struct wim_lookup_table_entry *lte, + struct wim_lookup_table_entry *lte_override, + struct apply_ctx *ctx) { - struct wim_lookup_table_entry *found_lte; - struct wim_lookup_table_entry *needed_lte; - struct wim_lookup_table *lookup_table; - struct wim_header_disk pwm_hdr; - int ret; - int pwm_flags; + struct wim_dentry **lte_dentries; + tchar path[ctx->ops->path_max]; + size_t i; + int ret; + + if (lte->out_refcnt <= ARRAY_LEN(lte->inline_lte_dentries)) + lte_dentries = lte->inline_lte_dentries; + else + lte_dentries = lte->lte_dentries; + + for (i = 0; i < lte->out_refcnt; i++) { + struct wim_dentry *dentry = lte_dentries[i]; + + if (dentry->tmp_flag) + continue; + if (!build_extraction_path(path, dentry, ctx)) + continue; + ret = extract_streams(path, ctx, dentry, lte, lte_override); + if (ret) + goto out_clear_tmp_flags; + dentry->tmp_flag = 1; + } + ret = 0; +out_clear_tmp_flags: + for (i = 0; i < lte->out_refcnt; i++) + lte_dentries[i]->tmp_flag = 0; + return ret; +} + +/* Determine whether the specified stream needs to be extracted to a temporary + * file or not. + * + * @lte->out_refcnt specifies the number of instances of this stream that must + * be extracted. + * + * @is_partial_res is %true if this stream is just one of multiple in a single + * WIM resource being extracted. */ +static bool +need_tmpfile_to_extract(struct wim_lookup_table_entry *lte, + bool is_partial_res) +{ + /* Temporary file is always required when reading a partial resource, + * since in that case we retrieve all the contained streams in one pass. + * */ + if (is_partial_res) + return true; + + /* Otherwise we don't need a temporary file if only a single instance of + * the stream is needed. */ + if (lte->out_refcnt == 1) + return false; + + wimlib_assert(lte->out_refcnt >= 2); + + /* We also don't need a temporary file if random access to the stream is + * allowed. */ + if (lte->resource_location != RESOURCE_IN_WIM || + filedes_is_seekable(<e->rspec->wim->in_fd)) + return false; + + return true; +} + +static int +begin_extract_stream_to_tmpfile(struct wim_lookup_table_entry *lte, + bool is_partial_res, + void *_ctx) +{ + struct apply_ctx *ctx = _ctx; + int ret; + + if (!need_tmpfile_to_extract(lte, is_partial_res)) { + DEBUG("Temporary file not needed " + "for stream (size=%"PRIu64")", lte->size); + ret = extract_stream_instances(lte, lte, ctx); + if (ret) + return ret; + + return BEGIN_STREAM_STATUS_SKIP_STREAM; + } + + DEBUG("Temporary file needed for stream (size=%"PRIu64")", lte->size); + return create_temporary_file(&ctx->tmpfile_fd, &ctx->tmpfile_name); +} + +static int +end_extract_stream_to_tmpfile(struct wim_lookup_table_entry *lte, + int status, void *_ctx) +{ + struct apply_ctx *ctx = _ctx; + struct wim_lookup_table_entry lte_override; + int ret; + int errno_save = errno; + + ret = filedes_close(&ctx->tmpfile_fd); + + if (status) { + ret = status; + errno = errno_save; + goto out_delete_tmpfile; + } + + if (ret) { + ERROR_WITH_ERRNO("Error writing temporary file %"TS, ctx->tmpfile_name); + ret = WIMLIB_ERR_WRITE; + goto out_delete_tmpfile; + } + + /* Now that a full stream has been extracted to a temporary file, + * extract all instances of it to the actual target. */ + + memcpy(<e_override, lte, sizeof(struct wim_lookup_table_entry)); + lte_override.resource_location = RESOURCE_IN_FILE_ON_DISK; + lte_override.file_on_disk = ctx->tmpfile_name; + + ret = extract_stream_instances(lte, <e_override, ctx); + +out_delete_tmpfile: + errno_save = errno; + tunlink(ctx->tmpfile_name); + FREE(ctx->tmpfile_name); + errno = errno_save; + return ret; +} + +/* Extracts a list of streams (ctx.stream_list), assuming that the directory + * structure and empty files were already created. This relies on the + * per-`struct wim_lookup_table_entry' list of dentries that reference each + * stream that was constructed earlier. */ +static int +extract_stream_list(struct apply_ctx *ctx) +{ + struct read_stream_list_callbacks cbs = { + .begin_stream = begin_extract_stream_to_tmpfile, + .begin_stream_ctx = ctx, + .consume_chunk = extract_chunk_to_fd, + .consume_chunk_ctx = &ctx->tmpfile_fd, + .end_stream = end_extract_stream_to_tmpfile, + .end_stream_ctx = ctx, + }; + return read_stream_list(&ctx->stream_list, + offsetof(struct wim_lookup_table_entry, extraction_list), + &cbs, VERIFY_STREAM_HASHES); +} + +#define PWM_ALLOW_WIM_HDR 0x00001 +#define PWM_SILENT_EOF 0x00002 + +/* 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)); + if (ret) + goto read_error; + + if ((flags & PWM_ALLOW_WIM_HDR) && buf.stream_hdr.magic == PWM_MAGIC) { + BUILD_BUG_ON(sizeof(buf.pwm_hdr) < sizeof(buf.stream_hdr)); + ret = full_read(&pwm->in_fd, &buf.stream_hdr + 1, + sizeof(buf.pwm_hdr) - sizeof(buf.stream_hdr)); + + if (ret) + goto read_error; + lte->resource_location = RESOURCE_NONEXISTENT; + memcpy(hdr_ret, &buf.pwm_hdr, sizeof(buf.pwm_hdr)); + return 0; + } + + if (le64_to_cpu(buf.stream_hdr.magic) != PWM_STREAM_MAGIC) { + ERROR("Data read on pipe is invalid (expected stream header)."); + return WIMLIB_ERR_INVALID_PIPABLE_WIM; + } + + copy_hash(lte->hash, buf.stream_hdr.hash); + + 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: + if (ret != WIMLIB_ERR_UNEXPECTED_END_OF_FILE || !(flags & PWM_SILENT_EOF)) + ERROR_WITH_ERRNO("Error reading pipable WIM from pipe"); + return ret; +} + +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; + int ret; + int pwm_flags; 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)) @@ -1356,8 +1435,10 @@ 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)) @@ -1368,23 +1449,66 @@ 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_stream(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; - - ret = extract_stream_instances(needed_lte, ctx, false); + tchar *tmpfile_name = NULL; + struct wim_lookup_table_entry *lte_override; + struct wim_lookup_table_entry tmpfile_lte; + + needed_lte->offset_in_res = found_lte->offset_in_res; + needed_lte->flags = found_lte->flags; + needed_lte->size = found_lte->size; + + lte_unbind_wim_resource_spec(found_lte); + lte_bind_wim_resource_spec(needed_lte, rspec); + + if (needed_lte->out_refcnt > 1) { + + struct filedes tmpfile_fd; + + /* Extract stream to temporary file. */ + ret = create_temporary_file(&tmpfile_fd, &tmpfile_name); + if (ret) { + lte_unbind_wim_resource_spec(needed_lte); + goto out_free_found_lte; + } + + ret = extract_full_stream_to_fd(needed_lte, + &tmpfile_fd); + if (ret) { + filedes_close(&tmpfile_fd); + goto delete_tmpfile; + } + + if (filedes_close(&tmpfile_fd)) { + ERROR_WITH_ERRNO("Error writing to temporary " + "file \"%"TS"\"", tmpfile_name); + ret = WIMLIB_ERR_WRITE; + goto delete_tmpfile; + } + memcpy(&tmpfile_lte, needed_lte, + sizeof(struct wim_lookup_table_entry)); + tmpfile_lte.resource_location = RESOURCE_IN_FILE_ON_DISK; + tmpfile_lte.file_on_disk = tmpfile_name; + lte_override = &tmpfile_lte; + } else { + lte_override = needed_lte; + } + + ret = extract_stream_instances(needed_lte, lte_override, ctx); + delete_tmpfile: + lte_unbind_wim_resource_spec(needed_lte); + if (tmpfile_name) { + tunlink(tmpfile_name); + FREE(tmpfile_name); + } if (ret) goto out_free_found_lte; ctx->num_streams_remaining--; } else if (found_lte->resource_location != RESOURCE_NONEXISTENT) { - ret = skip_pwm_stream(found_lte); + ret = skip_wim_stream(found_lte); if (ret) goto out_free_found_lte; } else { @@ -1411,6 +1535,8 @@ 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; @@ -1420,54 +1546,185 @@ resume_done: return 0; } -/* Finish extracting a file, directory, or symbolic link by setting file - * security and timestamps. */ +/* Extract a WIM dentry to standard output. + * + * This obviously doesn't make sense in all cases. We return an error if the + * dentry does not correspond to a regular file. Otherwise we extract the + * unnamed data stream only. */ static int -dentry_extract_final(struct wim_dentry *dentry, void *_ctx) +extract_dentry_to_stdout(struct wim_dentry *dentry, + const struct wim_lookup_table *lookup_table) { - struct apply_ctx *ctx = _ctx; - int ret; - tchar path[ctx->ops->path_max]; + struct wim_inode *inode = dentry->d_inode; + struct wim_lookup_table_entry *lte; + struct filedes _stdout; - if (!build_extraction_path(path, dentry, ctx)) + if (inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT | + FILE_ATTRIBUTE_DIRECTORY)) + { + ERROR("\"%"TS"\" is not a regular file and therefore cannot be " + "extracted to standard output", dentry_full_path(dentry)); + return WIMLIB_ERR_NOT_A_REGULAR_FILE; + } + + lte = inode_unnamed_lte(inode, lookup_table); + if (!lte) { + const u8 *hash = inode_unnamed_stream_hash(inode); + if (!is_zero_hash(hash)) + return stream_not_found_error(inode, hash); return 0; + } - ret = extract_security(path, ctx, dentry); - if (ret) - return ret; + filedes_init(&_stdout, STDOUT_FILENO); + return extract_full_stream_to_fd(lte, &_stdout); +} - return extract_timestamps(path, ctx, dentry); +static int +extract_dentries_to_stdout(struct wim_dentry **dentries, size_t num_dentries, + const struct wim_lookup_table *lookup_table) +{ + for (size_t i = 0; i < num_dentries; i++) { + int ret = extract_dentry_to_stdout(dentries[i], lookup_table); + if (ret) + return ret; + } + return 0; } +/**********************************************************************/ + /* - * Extract a WIM dentry to standard output. + * Removes duplicate dentries from the array. * - * This obviously doesn't make sense in all cases. We return an error if the - * dentry does not correspond to a regular file. Otherwise we extract the - * unnamed data stream only. + * Returns the new number of dentries, packed at the front of the array. + */ +static size_t +remove_duplicate_trees(struct wim_dentry **trees, size_t num_trees) +{ + size_t i, j = 0; + for (i = 0; i < num_trees; i++) { + if (!trees[i]->tmp_flag) { + /* Found distinct dentry. */ + trees[i]->tmp_flag = 1; + trees[j++] = trees[i]; + } + } + for (i = 0; i < j; i++) + trees[i]->tmp_flag = 0; + return j; +} + +/* + * Remove dentries that are descendants of other dentries in the array. + * + * Returns the new number of dentries, packed at the front of the array. */ +static size_t +remove_contained_trees(struct wim_dentry **trees, size_t num_trees) +{ + size_t i, j = 0; + for (i = 0; i < num_trees; i++) + trees[i]->tmp_flag = 1; + for (i = 0; i < num_trees; i++) { + struct wim_dentry *d = trees[i]; + while (!dentry_is_root(d)) { + d = d->parent; + if (d->tmp_flag) + goto tree_contained; + } + trees[j++] = trees[i]; + continue; + + tree_contained: + trees[i]->tmp_flag = 0; + } + + for (i = 0; i < j; i++) + trees[i]->tmp_flag = 0; + return j; +} + static int -extract_dentry_to_stdout(struct wim_dentry *dentry) +dentry_append_to_list(struct wim_dentry *dentry, void *_dentry_list) { - int ret = 0; - if (dentry->d_inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT | - FILE_ATTRIBUTE_DIRECTORY)) - { - ERROR("\"%"TS"\" is not a regular file and therefore cannot be " - "extracted to standard output", dentry_full_path(dentry)); - ret = WIMLIB_ERR_NOT_A_REGULAR_FILE; - } else { - struct wim_lookup_table_entry *lte; - - lte = inode_unnamed_lte_resolved(dentry->d_inode); - if (lte) { - struct filedes _stdout; - filedes_init(&_stdout, STDOUT_FILENO); - ret = extract_wim_resource_to_fd(lte, &_stdout, - wim_resource_size(lte)); + struct list_head *dentry_list = _dentry_list; + list_add_tail(&dentry->extraction_list, dentry_list); + return 0; +} + +static void +dentry_reset_extraction_list_node(struct wim_dentry *dentry) +{ + dentry->extraction_list = (struct list_head){NULL, NULL}; +} + +static int +dentry_delete_from_list(struct wim_dentry *dentry, void *_ignore) +{ + list_del(&dentry->extraction_list); + dentry_reset_extraction_list_node(dentry); + return 0; +} + +/* + * Build the preliminary list of dentries to be extracted. + * + * The list maintains the invariant that if d1 and d2 are in the list and d1 is + * an ancestor of d2, then d1 appears before d2 in the list. + */ +static void +build_dentry_list(struct list_head *dentry_list, struct wim_dentry **trees, + size_t num_trees, bool add_ancestors) +{ + INIT_LIST_HEAD(dentry_list); + + /* Add the trees recursively. */ + for (size_t i = 0; i < num_trees; i++) + for_dentry_in_tree(trees[i], dentry_append_to_list, dentry_list); + + /* If requested, add ancestors of the trees. */ + if (add_ancestors) { + for (size_t i = 0; i < num_trees; i++) { + struct wim_dentry *dentry = trees[i]; + struct wim_dentry *ancestor; + struct list_head *place_after; + + if (dentry_is_root(dentry)) + continue; + + place_after = dentry_list; + ancestor = dentry; + do { + ancestor = ancestor->parent; + if (dentry_in_list(ancestor)) { + place_after = &ancestor->extraction_list; + break; + } + } while (!dentry_is_root(ancestor)); + + ancestor = dentry; + do { + ancestor = ancestor->parent; + if (dentry_in_list(ancestor)) + break; + list_add(&ancestor->extraction_list, place_after); + } while (!dentry_is_root(ancestor)); } } - return ret; +} + +static const struct apply_operations * +select_apply_operations(int extract_flags) +{ +#ifdef WITH_NTFS_3G + if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) + return &ntfs_3g_apply_ops; +#endif +#ifdef __WIN32__ + return &win32_apply_ops; +#else + return &unix_apply_ops; +#endif } #ifdef __WIN32__ @@ -1517,94 +1774,48 @@ file_name_valid(utf16lechar *name, size_t num_chars, bool fix) return true; } -static bool -dentry_is_dot_or_dotdot(const struct wim_dentry *dentry) -{ - const utf16lechar *file_name = dentry->file_name; - return file_name != NULL && - file_name[0] == cpu_to_le16('.') && - (file_name[1] == cpu_to_le16('\0') || - (file_name[1] == cpu_to_le16('.') && - file_name[2] == cpu_to_le16('\0'))); -} - -static int -dentry_mark_skipped(struct wim_dentry *dentry, void *_ignore) -{ - dentry->extraction_skipped = 1; - return 0; -} - -/* - * dentry_calculate_extraction_path- - * - * Calculate the actual filename component at which a WIM dentry will be - * extracted, handling invalid filenames "properly". - * - * dentry->extraction_name usually will be set the same as dentry->file_name (on - * UNIX, converted into the platform's multibyte encoding). However, if the - * file name contains characters that are not valid on the current platform or - * has some other format that is not valid, leave dentry->extraction_name as - * NULL and set dentry->extraction_skipped to indicate that this dentry should - * not be extracted, unless the appropriate flag - * WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES is set in the extract flags, in - * which case a substitute filename will be created and set instead. - * - * Conflicts with case-insensitive names on Windows are handled similarly; see - * below. - */ static int -dentry_calculate_extraction_path(struct wim_dentry *dentry, void *_args) +dentry_calculate_extraction_name(struct wim_dentry *dentry, + struct apply_ctx *ctx) { - struct apply_ctx *ctx = _args; int ret; - if (dentry == ctx->extract_root || dentry->extraction_skipped) + if (dentry == ctx->target_dentry) return 0; if (!dentry_is_supported(dentry, &ctx->supported_features)) goto skip_dentry; - if (dentry_is_dot_or_dotdot(dentry)) { - /* WIM files shouldn't contain . or .. entries. But if they are - * there, don't attempt to extract them. */ - WARNING("Skipping extraction of unexpected . or .. file " - "\"%"TS"\"", dentry_full_path(dentry)); - goto skip_dentry; - } - -#ifdef __WIN32__ if (!ctx->ops->supports_case_sensitive_filenames) { struct wim_dentry *other; list_for_each_entry(other, &dentry->case_insensitive_conflict_list, case_insensitive_conflict_list) { - if (ctx->extract_flags & - WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) { - WARNING("\"%"TS"\" has the same " - "case-insensitive name as " - "\"%"TS"\"; extracting " - "dummy name instead", - dentry_full_path(dentry), - dentry_full_path(other)); - goto out_replace; - } else { - WARNING("Not extracting \"%"TS"\": " - "has same case-insensitive " - "name as \"%"TS"\"", - dentry_full_path(dentry), - dentry_full_path(other)); - goto skip_dentry; + if (dentry_in_list(other)) { + if (ctx->extract_flags & + WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) { + WARNING("\"%"TS"\" has the same " + "case-insensitive name as " + "\"%"TS"\"; extracting " + "dummy name instead", + dentry_full_path(dentry), + dentry_full_path(other)); + goto out_replace; + } else { + WARNING("Not extracting \"%"TS"\": " + "has same case-insensitive " + "name as \"%"TS"\"", + dentry_full_path(dentry), + dentry_full_path(other)); + goto skip_dentry; + } } } } -#else /* __WIN32__ */ - wimlib_assert(ctx->ops->supports_case_sensitive_filenames); -#endif /* !__WIN32__ */ if (file_name_valid(dentry->file_name, dentry->file_name_nbytes / 2, false)) { -#ifdef __WIN32__ +#if TCHAR_IS_UTF16LE dentry->extraction_name = dentry->file_name; dentry->extraction_name_nchars = dentry->file_name_nbytes / 2; return 0; @@ -1639,7 +1850,7 @@ out_replace: tchar *tchar_name; size_t tchar_nchars; - #ifdef __WIN32__ + #if TCHAR_IS_UTF16LE tchar_name = utf16_name_copy; tchar_nchars = dentry->file_name_nbytes / 2; #else @@ -1656,7 +1867,7 @@ out_replace: fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars, T(" (invalid filename #%lu)"), ++ctx->invalid_sequence); - #ifndef __WIN32__ + #if !TCHAR_IS_UTF16LE FREE(tchar_name); #endif dentry->extraction_name = memdup(fixed_name, @@ -1668,34 +1879,244 @@ out_replace: return 0; skip_dentry: - for_dentry_in_tree(dentry, dentry_mark_skipped, NULL); + for_dentry_in_tree(dentry, dentry_delete_from_list, NULL); return 0; } -/* Clean up dentry and inode structure after extraction. */ -static int -dentry_reset_needs_extraction(struct wim_dentry *dentry, void *_ignore) -{ +/* + * Calculate the actual filename component at which each WIM dentry will be + * extracted, with special handling for dentries that are unsupported by the + * extraction backend or have invalid names. + * + * Note: this has a dependency on start_extract() being called because + * ctx.supported_features must be filled in in order to determine whether each + * dentry is supported. + * + * Possible error codes: WIMLIB_ERR_NOMEM, WIMLIB_ERR_INVALID_UTF16_STRING + */ +static int +dentry_list_calculate_extraction_names(struct list_head *dentry_list, + struct apply_ctx *ctx) +{ + struct list_head *prev, *cur; + + /* Can't use list_for_each_entry() because a call to + * dentry_calculate_extraction_name() may delete the current dentry and + * its children from the list. */ + + prev = dentry_list; + for (;;) { + struct wim_dentry *dentry; + int ret; + + cur = prev->next; + if (cur == dentry_list) + break; + + dentry = list_entry(cur, struct wim_dentry, extraction_list); + + ret = dentry_calculate_extraction_name(dentry, ctx); + if (ret) + return ret; + + if (prev->next == cur) + prev = cur; + else + ; /* Current dentry and its children (which follow in + the list) were deleted. prev stays the same. */ + } + return 0; +} + +static int +dentry_resolve_streams(struct wim_dentry *dentry, int extract_flags, + struct wim_lookup_table *lookup_table) +{ struct wim_inode *inode = dentry->d_inode; + struct wim_lookup_table_entry *lte; + int ret; + bool force = false; - dentry->extraction_skipped = 0; - dentry->was_hardlinked = 0; - dentry->skeleton_extracted = 0; - inode->i_visited = 0; - FREE(inode->i_extracted_file); - inode->i_extracted_file = NULL; - inode->i_dos_name_extracted = 0; - if ((void*)dentry->extraction_name != (void*)dentry->file_name) - FREE(dentry->extraction_name); - dentry->extraction_name = NULL; + /* Special case: when extracting from a pipe, the WIM lookup table is + * initially empty, so "resolving" an inode's streams is initially not + * possible. However, we still need to keep track of which streams, + * identified by SHA1 message digests, need to be extracted, so we + * "resolve" the inode's streams anyway by allocating new entries. */ + if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) + force = true; + ret = inode_resolve_streams(inode, lookup_table, force); + if (ret) + return ret; + for (u32 i = 0; i <= inode->i_num_ads; i++) { + lte = inode_stream_lte_resolved(inode, i); + if (lte) + lte->out_refcnt = 0; + } return 0; } -/* Tally features necessary to extract a dentry and the corresponding inode. */ +/* + * For each dentry to be extracted, resolve all streams in the corresponding + * inode and set 'out_refcnt' in each to 0. + * + * Possible error codes: WIMLIB_ERR_RESOURCE_NOT_FOUND, WIMLIB_ERR_NOMEM. + */ +static int +dentry_list_resolve_streams(struct list_head *dentry_list, + struct apply_ctx *ctx) +{ + struct wim_dentry *dentry; + int ret; + + list_for_each_entry(dentry, dentry_list, extraction_list) { + ret = dentry_resolve_streams(dentry, + ctx->extract_flags, + ctx->wim->lookup_table); + if (ret) + return ret; + } + return 0; +} + +static int +ref_stream(struct wim_lookup_table_entry *lte, + struct wim_dentry *dentry, struct apply_ctx *ctx) +{ + if (!lte) + return 0; + + /* 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 += lte->size; + ctx->progress.extract.num_streams++; + } + + /* Add stream to the dentry_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++; + } + + if (!(ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FILE_ORDER)) { + struct wim_dentry **lte_dentries; + + /* Append dentry to this stream's array of dentries referencing + * it. Use inline array to avoid memory allocation until the + * number of dentries becomes too large. */ + if (lte->out_refcnt < ARRAY_LEN(lte->inline_lte_dentries)) { + lte_dentries = lte->inline_lte_dentries; + } else { + struct wim_dentry **prev_lte_dentries; + size_t alloc_lte_dentries; + + if (lte->out_refcnt == ARRAY_LEN(lte->inline_lte_dentries)) { + prev_lte_dentries = NULL; + alloc_lte_dentries = ARRAY_LEN(lte->inline_lte_dentries); + } else { + prev_lte_dentries = lte->lte_dentries; + alloc_lte_dentries = lte->alloc_lte_dentries; + } + + if (lte->out_refcnt == alloc_lte_dentries) { + alloc_lte_dentries *= 2; + lte_dentries = REALLOC(prev_lte_dentries, + alloc_lte_dentries * + sizeof(lte_dentries[0])); + if (lte_dentries == NULL) + return WIMLIB_ERR_NOMEM; + if (prev_lte_dentries == NULL) { + memcpy(lte_dentries, + lte->inline_lte_dentries, + sizeof(lte->inline_lte_dentries)); + } + lte->lte_dentries = lte_dentries; + lte->alloc_lte_dentries = alloc_lte_dentries; + } + lte_dentries = lte->lte_dentries; + } + lte_dentries[lte->out_refcnt] = dentry; + } + lte->out_refcnt++; + return 0; +} + +static int +dentry_ref_streams(struct wim_dentry *dentry, struct apply_ctx *ctx) +{ + struct wim_inode *inode = dentry->d_inode; + int ret; + + /* The unnamed data stream will always be extracted, except in an + * unlikely case. */ + if (!inode_is_encrypted_directory(inode)) { + ret = ref_stream(inode_unnamed_lte_resolved(inode), + dentry, ctx); + if (ret) + return ret; + } + + /* Named data streams will be extracted only if supported in the current + * extraction mode and volume, and to avoid complications, if not doing + * a linked extraction. */ + if (can_extract_named_data_streams(ctx)) { + for (u16 i = 0; i < inode->i_num_ads; i++) { + if (!ads_entry_is_named_stream(&inode->i_ads_entries[i])) + continue; + ret = ref_stream(inode->i_ads_entries[i].lte, + dentry, ctx); + if (ret) + return ret; + } + } + inode->i_visited = 1; + return 0; +} + +/* + * For each dentry to be extracted, iterate through the data streams of the + * corresponding inode. For each such stream that is not to be ignored due to + * the supported features or extraction flags, add it to the list of streams to + * be extracted (ctx->stream_list) if not already done so. + * + * Also, if doing a sequential extraction, build a mapping from each stream to + * the dentries referencing it. + * + * This also initializes the extract progress info with byte and stream + * information. + * + * Note: This has a dependency on start_extract being called because + * ctx.supported_features must be filled in in order to determine whether named + * data streams are supported. + * + * Note: this uses the i_visited member of the inodes (assumed to be 0 + * initially), but does not reset it. + * + * Possible error codes: WIMLIB_ERR_NOMEM. + */ static int -dentry_tally_features(struct wim_dentry *dentry, void *_features) +dentry_list_ref_streams(struct list_head *dentry_list, struct apply_ctx *ctx) +{ + struct wim_dentry *dentry; + int ret; + + list_for_each_entry(dentry, dentry_list, extraction_list) { + ret = dentry_ref_streams(dentry, ctx); + if (ret) + return ret; + } + return 0; +} + +/* Tally features necessary to extract a dentry and the corresponding inode. */ +static void +dentry_tally_features(struct wim_dentry *dentry, struct wim_features *features) { - struct wim_features *features = _features; struct wim_inode *inode = dentry->d_inode; if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE) @@ -1706,8 +2127,12 @@ dentry_tally_features(struct wim_dentry *dentry, void *_features) features->system_files++; if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED) features->compressed_files++; - if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) - features->encrypted_files++; + if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) { + if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY) + features->encrypted_directories++; + else + features->encrypted_files++; + } if (inode->i_attributes & FILE_ATTRIBUTE_NOT_CONTENT_INDEXED) features->not_context_indexed_files++; if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE) @@ -1730,268 +2155,267 @@ dentry_tally_features(struct wim_dentry *dentry, void *_features) if (inode_has_unix_data(inode)) features->unix_data++; inode->i_visited = 1; - 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. */ +/* Tally the features necessary to extract the specified dentries. */ static void -dentry_tree_get_features(struct wim_dentry *root, struct wim_features *features) +dentry_list_get_features(struct list_head *dentry_list, + struct wim_features *features) { + struct wim_dentry *dentry; + 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); + + list_for_each_entry(dentry, dentry_list, extraction_list) + dentry_tally_features(dentry, features); + + list_for_each_entry(dentry, dentry_list, extraction_list) + dentry->d_inode->i_visited = 0; } -static int -do_feature_check(const struct wim_features *required_features, - const struct wim_features *supported_features, - int extract_flags, - const struct apply_operations *ops, - const tchar *wim_source_path) +static u32 +compute_supported_attributes_mask(const struct wim_features *supported_features) { - const tchar *loc; - const tchar *mode = "this extraction mode"; + u32 mask = (u32)~0UL; - if (wim_source_path[0] == '\0') - loc = "the WIM image"; - else - loc = wim_source_path; + if (!supported_features->archive_files) + mask &= ~FILE_ATTRIBUTE_ARCHIVE; - /* We're an archive program, so theoretically we can do what we want - * with FILE_ATTRIBUTE_ARCHIVE (which is a dumb flag anyway). Don't - * bother the user about it. */ -#if 0 - if (required_features->archive_files && !supported_features->archive_files) - { - WARNING( - "%lu files in %"TS" are marked as archived, but this attribute\n" -" is not supported in %"TS".", - required_features->archive_files, loc, mode); - } -#endif + if (!supported_features->hidden_files) + mask &= ~FILE_ATTRIBUTE_HIDDEN; - if (required_features->hidden_files && !supported_features->hidden_files) - { - WARNING( - "%lu files in %"TS" are marked as hidden, but this\n" -" attribute is not supported in %"TS".", - required_features->hidden_files, loc, mode); - } + if (!supported_features->system_files) + mask &= ~FILE_ATTRIBUTE_SYSTEM; - if (required_features->system_files && !supported_features->system_files) - { - WARNING( - "%lu files in %"TS" are marked as system files,\n" -" but this attribute is not supported in %"TS".", - required_features->system_files, loc, mode); - } + if (!supported_features->not_context_indexed_files) + mask &= ~FILE_ATTRIBUTE_NOT_CONTENT_INDEXED; - if (required_features->compressed_files && !supported_features->compressed_files) - { - WARNING( - "%lu files in %"TS" are marked as being transparently\n" -" compressed, but transparent compression is not supported in\n" -" %"TS". These files will be extracted as uncompressed.", - required_features->compressed_files, loc, mode); - } + if (!supported_features->compressed_files) + mask &= ~FILE_ATTRIBUTE_COMPRESSED; - if (required_features->encrypted_files && !supported_features->encrypted_files) - { - WARNING( - "%lu files in %"TS" are marked as being encrypted,\n" -" but encryption is not supported in %"TS". These files\n" -" will be extracted as raw encrypted data instead.", - required_features->encrypted_files, loc, mode); - } + if (!supported_features->sparse_files) + mask &= ~FILE_ATTRIBUTE_SPARSE_FILE; - if (required_features->not_context_indexed_files && - !supported_features->not_context_indexed_files) - { - WARNING( - "%lu files in %"TS" are marked as not content indexed,\n" -" but this attribute is not supported in %"TS".", - required_features->not_context_indexed_files, loc, mode); - } + if (!supported_features->reparse_points) + mask &= ~FILE_ATTRIBUTE_REPARSE_POINT; - if (required_features->sparse_files && !supported_features->sparse_files) - { - WARNING( - "%lu files in %"TS" are marked as sparse, but creating\n" -" sparse files is not supported in %"TS". These files\n" -" will be extracted as non-sparse.", - required_features->sparse_files, loc, mode); - } + return mask; +} +static int +do_feature_check(const struct wim_features *required_features, + const struct wim_features *supported_features, + int extract_flags, const struct apply_operations *ops) +{ + /* File attributes. */ + if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES)) { + /* Note: Don't bother the user about FILE_ATTRIBUTE_ARCHIVE. + * We're an archive program, so theoretically we can do what we + * want with it. */ + + if (required_features->hidden_files && + !supported_features->hidden_files) + WARNING("Ignoring FILE_ATTRIBUTE_HIDDEN of %lu files", + required_features->hidden_files); + + if (required_features->system_files && + !supported_features->system_files) + WARNING("Ignoring FILE_ATTRIBUTE_SYSTEM of %lu files", + required_features->system_files); + + if (required_features->compressed_files && + !supported_features->compressed_files) + WARNING("Ignoring FILE_ATTRIBUTE_COMPRESSED of %lu files", + required_features->compressed_files); + + if (required_features->not_context_indexed_files && + !supported_features->not_context_indexed_files) + WARNING("Ignoring FILE_ATTRIBUTE_NOT_CONTENT_INDEXED of %lu files", + required_features->not_context_indexed_files); + + if (required_features->sparse_files && + !supported_features->sparse_files) + WARNING("Ignoring FILE_ATTRIBUTE_SPARSE_FILE of %lu files", + required_features->sparse_files); + + if (required_features->encrypted_directories && + !supported_features->encrypted_directories) + WARNING("Ignoring FILE_ATTRIBUTE_ENCRYPTED of %lu directories", + required_features->encrypted_directories); + } + + /* Encrypted files. */ + if (required_features->encrypted_files && + !supported_features->encrypted_files) + WARNING("Ignoring %lu encrypted files", + required_features->encrypted_files); + + /* Named data streams. */ if (required_features->named_data_streams && - !supported_features->named_data_streams) + (!supported_features->named_data_streams || + (extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK | + WIMLIB_EXTRACT_FLAG_HARDLINK)))) + WARNING("Ignoring named data streams of %lu files", + required_features->named_data_streams); + + /* Hard links. */ + if ((extract_flags & WIMLIB_EXTRACT_FLAG_HARDLINK) && + !supported_features->hard_links) { - WARNING( - "%lu files in %"TS" contain one or more alternate (named)\n" -" data streams, which are not supported in %"TS".\n" -" Alternate data streams will NOT be extracted.", - required_features->named_data_streams, loc, mode); + ERROR("Extraction backend does not support hard links!"); + return WIMLIB_ERR_UNSUPPORTED; } + if (required_features->hard_links && !supported_features->hard_links) + WARNING("Extracting %lu hard links as independent files", + required_features->hard_links); - if (unlikely(extract_flags & (WIMLIB_EXTRACT_FLAG_HARDLINK | - WIMLIB_EXTRACT_FLAG_SYMLINK)) && - required_features->named_data_streams && - supported_features->named_data_streams) + /* Symbolic links and reparse points. */ + if ((extract_flags & WIMLIB_EXTRACT_FLAG_SYMLINK) && + !supported_features->symlink_reparse_points) { - WARNING( - "%lu files in %"TS" contain one or more alternate (named)\n" -" data streams, which are not supported in linked extraction mode.\n" -" Alternate data streams will NOT be extracted.", - required_features->named_data_streams, loc); + ERROR("Extraction backend does not support symbolic links!"); + return WIMLIB_ERR_UNSUPPORTED; } - - if (required_features->hard_links && !supported_features->hard_links) + if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS) && + required_features->symlink_reparse_points && + !supported_features->symlink_reparse_points && + !supported_features->reparse_points) { - WARNING( - "%lu files in %"TS" are hard links, but hard links are\n" -" not supported in %"TS". Hard links will be extracted as\n" -" duplicate copies of the linked files.", - required_features->hard_links, loc, mode); + ERROR("Extraction backend does not support symbolic links!"); + return WIMLIB_ERR_UNSUPPORTED; } - - if (required_features->reparse_points && !supported_features->reparse_points) + if (required_features->reparse_points && + !supported_features->reparse_points) { if (supported_features->symlink_reparse_points) { if (required_features->other_reparse_points) { - WARNING( - "%lu files in %"TS" are reparse points that are neither\n" -" symbolic links nor junction points and are not supported in\n" -" %"TS". These reparse points will not be extracted.", - required_features->other_reparse_points, loc, - mode); + WARNING("Ignoring %lu non-symlink/junction " + "reparse point files", + required_features->other_reparse_points); } } else { - WARNING( - "%lu files in %"TS" are reparse points, which are\n" -" not supported in %"TS" and will not be extracted.", - required_features->reparse_points, loc, mode); + WARNING("Ignoring %lu reparse point files", + required_features->reparse_points); } } - if (required_features->security_descriptors && + /* Security descriptors. */ + if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS | + WIMLIB_EXTRACT_FLAG_UNIX_DATA)) + == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) && + required_features->security_descriptors && !supported_features->security_descriptors) { - WARNING( - "%lu files in %"TS" have Windows NT security descriptors,\n" -" but extracting security descriptors is not supported in\n" -" %"TS". No security descriptors will be extracted.", - required_features->security_descriptors, loc, mode); - } - - if (required_features->short_names && !supported_features->short_names) - { - WARNING( - "%lu files in %"TS" have short (DOS) names, but\n" -" extracting short names is not supported in %"TS".\n" -" Short names will not be extracted.\n", - required_features->short_names, loc, mode); + ERROR("Extraction backend does not support security descriptors!"); + return WIMLIB_ERR_UNSUPPORTED; } + if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS) && + required_features->security_descriptors && + !supported_features->security_descriptors) + WARNING("Ignoring Windows NT security descriptors of %lu files", + required_features->security_descriptors); + /* UNIX data. */ if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) && required_features->unix_data && !supported_features->unix_data) { - ERROR("Extracting UNIX data is not supported in %"TS, mode); + ERROR("Extraction backend does not support UNIX data!"); return WIMLIB_ERR_UNSUPPORTED; } - if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) && - required_features->short_names && !supported_features->short_names) + + /* DOS Names. */ + if (required_features->short_names && + !supported_features->short_names) { - ERROR("Extracting short names is not supported in %"TS"", mode); - return WIMLIB_ERR_UNSUPPORTED; + if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) { + ERROR("Extraction backend does not support DOS names!"); + return WIMLIB_ERR_UNSUPPORTED; + } + WARNING("Ignoring DOS names of %lu files", + required_features->short_names); } + + /* Timestamps. */ if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) && !ops->set_timestamps) { - ERROR("Extracting timestamps is not supported in %"TS"", mode); - return WIMLIB_ERR_UNSUPPORTED; - } - if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS | - WIMLIB_EXTRACT_FLAG_UNIX_DATA)) - == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) && - required_features->security_descriptors && - !supported_features->security_descriptors) - { - ERROR("Extracting security descriptors is not supported in %"TS, mode); + ERROR("Extraction backend does not support timestamps!"); return WIMLIB_ERR_UNSUPPORTED; } - if ((extract_flags & WIMLIB_EXTRACT_FLAG_HARDLINK) && - !supported_features->hard_links) - { - ERROR("Hard link extraction mode requested, but " - "%"TS" does not support hard links!", mode); - return WIMLIB_ERR_UNSUPPORTED; + return 0; +} + +static void +do_extract_warnings(struct apply_ctx *ctx) +{ + if (ctx->partial_security_descriptors == 0 && + ctx->no_security_descriptors == 0) + return; + + 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" + " on %lu files or directories.", + ctx->partial_security_descriptors); + } + if (ctx->no_security_descriptors != 0) { + WARNING("- Could not set security descriptor at all\n" + " on %lu files or directories.", + ctx->no_security_descriptors); } +#ifdef __WIN32__ + WARNING("To fully restore all security descriptors, run the program\n" + " with Administrator rights."); +#endif +} - if ((extract_flags & WIMLIB_EXTRACT_FLAG_SYMLINK) && - !supported_features->symlink_reparse_points) - { - ERROR("Symbolic link extraction mode requested, but " - "%"TS" does not support symbolic " - "links!", mode); - return WIMLIB_ERR_UNSUPPORTED; +static void +destroy_dentry_list(struct list_head *dentry_list) +{ + struct wim_dentry *dentry, *tmp; + struct wim_inode *inode; + + list_for_each_entry_safe(dentry, tmp, dentry_list, extraction_list) { + inode = dentry->d_inode; + dentry_reset_extraction_list_node(dentry); + dentry->was_linked = 0; + dentry->skeleton_extracted = 0; + inode->i_visited = 0; + FREE(inode->i_extracted_file); + inode->i_extracted_file = NULL; + inode->i_dos_name_extracted = 0; + if ((void*)dentry->extraction_name != (void*)dentry->file_name) + FREE(dentry->extraction_name); + dentry->extraction_name = NULL; } - return 0; } -/* - * extract_tree - Extract a file or directory tree from the currently selected - * WIM image. - * - * @wim: WIMStruct for the WIM file, with the desired image selected - * (as wim->current_image). - * - * @wim_source_path: - * "Canonical" (i.e. no leading or trailing slashes, path - * separators WIM_PATH_SEPARATOR) path inside the WIM image to - * extract. An empty string means the full image. - * - * @target: - * Filesystem path to extract the file or directory tree to. - * (Or, with WIMLIB_EXTRACT_FLAG_NTFS: the name of a NTFS volume.) - * - * @extract_flags: - * WIMLIB_EXTRACT_FLAG_*. Also, the private flag - * WIMLIB_EXTRACT_FLAG_MULTI_IMAGE will be set if this is being - * called through wimlib_extract_image() with WIMLIB_ALL_IMAGES as - * the image. - * - * @progress_func: - * If non-NULL, progress function for the extraction. The messages - * that may be sent in this function are: - * - * WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN or - * WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN; - * WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN; - * WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END; - * WIMLIB_PROGRESS_MSG_EXTRACT_DENTRY; - * WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS; - * WIMLIB_PROGRESS_MSG_APPLY_TIMESTAMPS; - * WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END or - * WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END. - * - * Returns 0 on success; a positive WIMLIB_ERR_* code on failure. - */ +static void +destroy_stream_list(struct list_head *stream_list) +{ + struct wim_lookup_table_entry *lte; + + list_for_each_entry(lte, stream_list, extraction_list) + if (lte->out_refcnt > ARRAY_LEN(lte->inline_lte_dentries)) + FREE(lte->lte_dentries); +} + static int -extract_tree(WIMStruct *wim, const tchar *wim_source_path, const tchar *target, - int extract_flags, wimlib_progress_func_t progress_func) +extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees, + const tchar *target, int extract_flags, + wimlib_progress_func_t progress_func) { - struct wim_dentry *root; - struct wim_features required_features; - struct apply_ctx ctx; int ret; - struct wim_lookup_table_entry *lte; + struct apply_ctx ctx; + struct list_head dentry_list; + struct wim_features required_features; + + /* Handle stdout extraction as a separate case. */ + if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) + return extract_dentries_to_stdout(trees, num_trees, + wim->lookup_table); /* Start initializing the apply_ctx. */ memset(&ctx, 0, sizeof(struct apply_ctx)); @@ -2007,80 +2431,85 @@ extract_tree(WIMStruct *wim, const tchar *wim_source_path, const tchar *target, WIMLIB_EXTRACT_MASK_PUBLIC); ctx.progress.extract.image_name = wimlib_get_image_name(wim, wim->current_image); - ctx.progress.extract.extract_root_wim_source_path = wim_source_path; ctx.progress.extract.target = target; } - INIT_LIST_HEAD(&ctx.stream_list); - /* Translate the path to extract into the corresponding - * `struct wim_dentry', which will be the root of the - * "dentry tree" to extract. */ - root = get_dentry(wim, wim_source_path); - if (!root) { - ERROR("Path \"%"TS"\" does not exist in WIM image %d", - wim_source_path, wim->current_image); - ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST; - goto out; - } + if (extract_flags & WIMLIB_EXTRACT_FLAG_FILEMODE) { + /* Called from wimlib_extract_files(). There should be only 1 + * tree, and directory structure should not be preserved. */ + wimlib_assert(num_trees == 1); + wimlib_assert(extract_flags & + WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE); + ret = calculate_dentry_full_path(trees[0]); + if (ret) + return ret; + ctx.progress.extract.extract_root_wim_source_path = trees[0]->_full_path; + ctx.target_dentry = trees[0]; + } else { + ctx.progress.extract.extract_root_wim_source_path = T(""); + ctx.target_dentry = wim_root_dentry(wim); + } + /* Note: ctx.target_dentry represents the dentry that gets extracted to + * @target. There may be none, in which case it gets set to the image + * root and never matches any of the dentries actually being extracted. + */ + + num_trees = remove_duplicate_trees(trees, num_trees); - ctx.extract_root = root; + /* All trees are now distinct. */ - /* Select the appropriate apply_operations based on the - * platform and extract_flags. */ -#ifdef __WIN32__ - ctx.ops = &win32_apply_ops; -#else - ctx.ops = &unix_apply_ops; -#endif + num_trees = remove_contained_trees(trees, num_trees); -#ifdef WITH_NTFS_3G - if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) - ctx.ops = &ntfs_3g_apply_ops; -#endif + /* All trees are now distinct and non-overlapping. */ + + /* Build list of dentries to be extracted. */ + build_dentry_list(&dentry_list, trees, num_trees, + !(extract_flags & WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE)); + + /* Select the appropriate apply_operations based on the platform and + * extract_flags. */ + ctx.ops = select_apply_operations(extract_flags); + + /* Figure out whether the root dentry is being extracted to the root of + * a volume and therefore needs to be treated "specially", for example + * not being explicitly created and not having attributes set. */ + if (ctx.ops->target_is_root && ctx.ops->root_directory_is_special) + ctx.root_dentry_is_special = ctx.ops->target_is_root(target); /* Call the start_extract() callback. This gives the apply_operations * implementation a chance to do any setup needed to access the volume. - * Furthermore, it's expected to set the supported features of this - * extraction mode (ctx.supported_features), which are determined at - * runtime as they may vary depending on the actual volume. These - * features are then compared with the actual features extracting this - * dentry tree requires. Some mismatches will merely produce warnings - * and the unsupported data will be ignored; others will produce errors. - */ + * Furthermore, start_extract() is expected to set the supported + * features of this extraction mode (ctx.supported_features), which are + * determined at runtime as they may vary depending on the actual + * volume. */ ret = ctx.ops->start_extract(target, &ctx); if (ret) - goto out; + goto out_destroy_dentry_list; - dentry_tree_get_features(root, &required_features); + /* Get and check the features required to extract the dentries. */ + dentry_list_get_features(&dentry_list, &required_features); ret = do_feature_check(&required_features, &ctx.supported_features, - extract_flags, ctx.ops, wim_source_path); + extract_flags, ctx.ops); if (ret) goto out_finish_or_abort_extract; - /* Figure out whether the root dentry is being extracted to the root of - * a volume and therefore needs to be treated "specially", for example - * not being explicitly created and not having attributes set. */ - if (ctx.ops->target_is_root && ctx.ops->root_directory_is_special) - ctx.root_dentry_is_special = ctx.ops->target_is_root(target); + ctx.supported_attributes_mask = + compute_supported_attributes_mask(&ctx.supported_features); - /* Calculate the actual filename component of each extracted dentry. In - * the process, set the dentry->extraction_skipped flag on dentries that - * are being skipped for some reason (e.g. invalid filename). */ - ret = for_dentry_in_tree(root, dentry_calculate_extraction_path, &ctx); + /* Calculate extraction name for each dentry and remove subtrees that + * can't be extracted due to naming problems. */ + ret = dentry_list_calculate_extraction_names(&dentry_list, &ctx); if (ret) - goto out_dentry_reset_needs_extraction; + goto out_finish_or_abort_extract; - /* Build the list of the streams that need to be extracted and - * initialize ctx.progress.extract with stream information. */ - ret = for_dentry_in_tree(ctx.extract_root, - dentry_resolve_and_zero_lte_refcnt, &ctx); + /* Build list of streams to extract. */ + ret = dentry_list_resolve_streams(&dentry_list, &ctx); if (ret) - goto out_dentry_reset_needs_extraction; - - ret = for_dentry_in_tree(ctx.extract_root, - dentry_add_streams_to_extract, &ctx); + goto out_finish_or_abort_extract; + INIT_LIST_HEAD(&ctx.stream_list); + ret = dentry_list_ref_streams(&dentry_list, &ctx); if (ret) - goto out_teardown_stream_list; + goto out_destroy_stream_list; if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) { /* When extracting from a pipe, the number of bytes of data to @@ -2103,30 +2532,6 @@ extract_tree(WIMStruct *wim, const tchar *wim_source_path, const tchar *target, } } - /* Handle the special case of extracting a file to standard - * output. In that case, "root" should be a single file, not a - * directory tree. (If not, extract_dentry_to_stdout() will - * return an error.) */ - if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) { - ret = extract_dentry_to_stdout(root); - goto out_teardown_stream_list; - } - - /* If a sequential extraction was specified, sort the streams to be - * extracted by their position in the WIM file so that the WIM file can - * be read sequentially. */ - if ((extract_flags & (WIMLIB_EXTRACT_FLAG_SEQUENTIAL | - WIMLIB_EXTRACT_FLAG_FROM_PIPE)) - == WIMLIB_EXTRACT_FLAG_SEQUENTIAL) - { - ret = sort_stream_list_by_sequential_order( - &ctx.stream_list, - offsetof(struct wim_lookup_table_entry, - extraction_list)); - if (ret) - goto out_teardown_stream_list; - } - if (ctx.ops->realpath_works_on_nonexisting_files && ((extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) || ctx.ops->requires_realtarget_in_paths)) @@ -2134,23 +2539,24 @@ extract_tree(WIMStruct *wim, const tchar *wim_source_path, const tchar *target, ctx.realtarget = realpath(target, NULL); if (!ctx.realtarget) { ret = WIMLIB_ERR_NOMEM; - goto out_teardown_stream_list; + goto out_destroy_stream_list; } ctx.realtarget_nchars = tstrlen(ctx.realtarget); } if (progress_func) { - progress_func(*wim_source_path ? WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN : - WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN, - &ctx.progress); + int msg; + if (extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) + msg = WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN; + else + msg = WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN; + progress_func(msg, &ctx.progress); } - if (!ctx.root_dentry_is_special) - { + if (!ctx.root_dentry_is_special) { tchar path[ctx.ops->path_max]; - if (build_extraction_path(path, root, &ctx)) - { - ret = extract_inode(path, &ctx, root->d_inode); + if (build_extraction_path(path, ctx.target_dentry, &ctx)) { + ret = extract_inode(path, &ctx, ctx.target_dentry->d_inode); if (ret) goto out_free_realtarget; } @@ -2176,17 +2582,25 @@ 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) { + if (progress_func) + progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN, + &ctx.progress); + ret = extract_dir_structure(&dentry_list, &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)) { + if (!(extract_flags & WIMLIB_EXTRACT_FLAG_FILE_ORDER)) { /* Sequential extraction requested, so two passes are needed - * (one for directory structure, one for streams.) */ - if (progress_func) + * (one for file structure, one for streams.) */ + if (progress_func && !ctx.ops->requires_short_name_reordering) progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN, &ctx.progress); if (!(extract_flags & WIMLIB_EXTRACT_FLAG_RESUME)) { - ret = for_dentry_in_tree(root, dentry_extract_skeleton, &ctx); + ret = extract_structure(&dentry_list, &ctx); if (ret) goto out_free_realtarget; } @@ -2203,10 +2617,10 @@ extract_tree(WIMStruct *wim, const tchar *wim_source_path, const tchar *target, /* Sequential extraction was not requested, so we can make do * with one pass where we both create the files and extract * streams. */ - if (progress_func) + if (progress_func && !ctx.ops->requires_short_name_reordering) progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN, &ctx.progress); - ret = for_dentry_in_tree(root, dentry_extract, &ctx); + ret = extract_dentries(&dentry_list, &ctx); if (ret) goto out_free_realtarget; if (progress_func) @@ -2238,28 +2652,25 @@ extract_tree(WIMStruct *wim, const tchar *wim_source_path, const tchar *target, if (progress_func) progress_func(WIMLIB_PROGRESS_MSG_APPLY_TIMESTAMPS, &ctx.progress); - ret = for_dentry_in_tree_depth(root, dentry_extract_final, &ctx); + ret = extract_final_metadata(&dentry_list, &ctx); if (ret) goto out_free_realtarget; if (progress_func) { - progress_func(*wim_source_path ? WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END : - WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END, - &ctx.progress); + int msg; + if (extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) + msg = WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END; + else + msg = WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END; + progress_func(msg, &ctx.progress); } - + do_extract_warnings(&ctx); ret = 0; out_free_realtarget: FREE(ctx.realtarget); -out_teardown_stream_list: - /* Free memory allocated as part of the mapping from each - * wim_lookup_table_entry to the dentries that reference it. */ - if (ctx.extract_flags & WIMLIB_EXTRACT_FLAG_SEQUENTIAL) - list_for_each_entry(lte, &ctx.stream_list, extraction_list) - if (lte->out_refcnt > ARRAY_LEN(lte->inline_lte_dentries)) - FREE(lte->lte_dentries); -out_dentry_reset_needs_extraction: - for_dentry_in_tree(root, dentry_reset_needs_extraction, NULL); +out_destroy_stream_list: + if (!(ctx.extract_flags & WIMLIB_EXTRACT_FLAG_FILE_ORDER)) + destroy_stream_list(&ctx.stream_list); out_finish_or_abort_extract: if (ret) { if (ctx.ops->abort_extract) @@ -2268,22 +2679,38 @@ out_finish_or_abort_extract: if (ctx.ops->finish_extract) ret = ctx.ops->finish_extract(&ctx); } -out: +out_destroy_dentry_list: + destroy_dentry_list(&dentry_list); return ret; } -/* Validates a single wimlib_extract_command, mostly checking to make sure the - * extract flags make sense. */ static int -check_extract_command(struct wimlib_extract_command *cmd, int wim_header_flags) +mkdir_if_needed(const tchar *target) { - int extract_flags; - - /* Empty destination path? */ - if (cmd->fs_dest_path[0] == T('\0')) - return WIMLIB_ERR_INVALID_PARAM; + struct stat stbuf; + if (tstat(target, &stbuf)) { + if (errno == ENOENT) { + if (tmkdir(target, 0755)) { + ERROR_WITH_ERRNO("Failed to create directory " + "\"%"TS"\"", target); + return WIMLIB_ERR_MKDIR; + } + } else { + ERROR_WITH_ERRNO("Failed to stat \"%"TS"\"", target); + return WIMLIB_ERR_STAT; + } + } else if (!S_ISDIR(stbuf.st_mode)) { + ERROR("\"%"TS"\" is not a directory", target); + return WIMLIB_ERR_NOTDIR; + } + return 0; +} - extract_flags = cmd->extract_flags; +/* Make sure the extraction flags make sense, and update them if needed. */ +static int +check_extract_flags(const WIMStruct *wim, int *extract_flags_p) +{ + int extract_flags = *extract_flags_p; /* Check for invalid flag combinations */ if ((extract_flags & @@ -2309,20 +2736,22 @@ check_extract_command(struct wimlib_extract_command *cmd, int wim_header_flags) WIMLIB_EXTRACT_FLAG_FROM_PIPE)) == WIMLIB_EXTRACT_FLAG_RESUME) return WIMLIB_ERR_INVALID_PARAM; - if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) { #ifndef WITH_NTFS_3G + if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) { ERROR("wimlib was compiled without support for NTFS-3g, so\n" - " we cannot apply a WIM image directly to a NTFS volume."); + " it cannot apply a WIM image directly to a NTFS volume."); return WIMLIB_ERR_UNSUPPORTED; -#endif } +#endif if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX | - WIMLIB_EXTRACT_FLAG_NORPFIX)) == 0) + WIMLIB_EXTRACT_FLAG_NORPFIX | + WIMLIB_EXTRACT_FLAG_IMAGEMODE)) == + WIMLIB_EXTRACT_FLAG_IMAGEMODE) { - /* Do reparse point fixups by default if the WIM header says - * they are enabled and we are extracting a full image. */ - if (wim_header_flags & WIM_HDR_FLAG_RP_FIX) + /* For full-image extraction, do reparse point fixups by default + * if the WIM header says they are enabled. */ + if (wim->hdr.flags & WIM_HDR_FLAG_RP_FIX) extract_flags |= WIMLIB_EXTRACT_FLAG_RPFIX; } @@ -2331,9 +2760,8 @@ check_extract_command(struct wimlib_extract_command *cmd, int wim_header_flags) * corresponding file or symbolic link data. This needs to be handled * better. */ if ((extract_flags & (WIMLIB_EXTRACT_FLAG_UNIX_DATA | - WIMLIB_EXTRACT_FLAG_SEQUENTIAL)) - == (WIMLIB_EXTRACT_FLAG_UNIX_DATA | - WIMLIB_EXTRACT_FLAG_SEQUENTIAL)) + WIMLIB_EXTRACT_FLAG_FILE_ORDER)) + == WIMLIB_EXTRACT_FLAG_UNIX_DATA) { if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) { WARNING("Setting UNIX file/owner group may " @@ -2341,180 +2769,172 @@ check_extract_command(struct wimlib_extract_command *cmd, int wim_header_flags) " symbolic links " "when applying from a pipe."); } else { - extract_flags &= ~WIMLIB_EXTRACT_FLAG_SEQUENTIAL; + extract_flags |= WIMLIB_EXTRACT_FLAG_FILE_ORDER; WARNING("Disabling sequential extraction for " "UNIX data mode"); } } - cmd->extract_flags = extract_flags; + *extract_flags_p = extract_flags; return 0; } +static u32 +get_wildcard_flags(int extract_flags) +{ + u32 wildcard_flags = 0; + + if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_GLOB) + wildcard_flags |= WILDCARD_FLAG_ERROR_IF_NO_MATCH; + else + wildcard_flags |= WILDCARD_FLAG_WARN_IF_NO_MATCH; + + if (default_ignore_case) + wildcard_flags |= WILDCARD_FLAG_CASE_INSENSITIVE; + + return wildcard_flags; +} + +struct append_dentry_ctx { + struct wim_dentry **dentries; + size_t num_dentries; + size_t num_alloc_dentries; +}; -/* Internal function to execute extraction commands for a WIM image. The paths - * in the extract commands are expected to be already "canonicalized". */ static int -do_wimlib_extract_files(WIMStruct *wim, - int image, - struct wimlib_extract_command *cmds, - size_t num_cmds, - wimlib_progress_func_t progress_func) +append_dentry_cb(struct wim_dentry *dentry, void *_ctx) +{ + struct append_dentry_ctx *ctx = _ctx; + + if (ctx->num_dentries == ctx->num_alloc_dentries) { + struct wim_dentry **new_dentries; + size_t new_length; + + new_length = max(ctx->num_alloc_dentries + 8, + ctx->num_alloc_dentries * 3 / 2); + new_dentries = REALLOC(ctx->dentries, + new_length * sizeof(ctx->dentries[0])); + if (new_dentries == NULL) + return WIMLIB_ERR_NOMEM; + ctx->dentries = new_dentries; + ctx->num_alloc_dentries = new_length; + } + ctx->dentries[ctx->num_dentries++] = dentry; + return 0; +} + +static int +do_wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target, + const tchar * const *paths, size_t num_paths, + int extract_flags, wimlib_progress_func_t progress_func) { int ret; - bool found_link_cmd = false; - bool found_nolink_cmd = false; + struct wim_dentry **trees; + size_t num_trees; + + if (wim == NULL || target == NULL || target[0] == T('\0') || + (num_paths != 0 && paths == NULL)) + return WIMLIB_ERR_INVALID_PARAM; + + ret = check_extract_flags(wim, &extract_flags); + if (ret) + return ret; - /* Select the image from which we are extracting files */ ret = select_wim_image(wim, image); if (ret) return ret; - /* Make sure there are no streams in the WIM that have not been - * checksummed yet. */ ret = wim_checksum_unhashed_streams(wim); if (ret) return ret; - /* Check for problems with the extraction commands */ - for (size_t i = 0; i < num_cmds; i++) { - ret = check_extract_command(&cmds[i], wim->hdr.flags); - if (ret) - return ret; - if (cmds[i].extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK | - WIMLIB_EXTRACT_FLAG_HARDLINK)) { - found_link_cmd = true; - } else { - found_nolink_cmd = true; - } - if (found_link_cmd && found_nolink_cmd) { - ERROR("Symlink or hardlink extraction mode must " - "be set on all extraction commands"); - return WIMLIB_ERR_INVALID_PARAM; - } - } - - /* Execute the extraction commands */ - for (size_t i = 0; i < num_cmds; i++) { - ret = extract_tree(wim, - cmds[i].wim_source_path, - cmds[i].fs_dest_path, - cmds[i].extract_flags, - progress_func); + if ((extract_flags & (WIMLIB_EXTRACT_FLAG_NTFS | + WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE | + WIMLIB_EXTRACT_FLAG_FILEMODE)) == + (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE)) + { + ret = mkdir_if_needed(target); if (ret) return ret; } - return 0; -} - -/* API function documented in wimlib.h */ -WIMLIBAPI int -wimlib_extract_files(WIMStruct *wim, - int image, - 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; - struct wimlib_extract_command *cmds_copy; - int all_flags = 0; - default_extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC; + if (extract_flags & WIMLIB_EXTRACT_FLAG_GLOB_PATHS) { - ret = verify_swm_set(wim, additional_swms, num_additional_swms); - if (ret) - goto out; - - if (num_cmds == 0) - goto out; + struct append_dentry_ctx append_dentry_ctx = { + .dentries = NULL, + .num_dentries = 0, + .num_alloc_dentries = 0, + }; - if (num_additional_swms) - merge_lookup_tables(wim, additional_swms, num_additional_swms); + u32 wildcard_flags = get_wildcard_flags(extract_flags); - cmds_copy = CALLOC(num_cmds, sizeof(cmds[0])); - if (!cmds_copy) { - ret = WIMLIB_ERR_NOMEM; - goto out_restore_lookup_table; - } + for (size_t i = 0; i < num_paths; i++) { + tchar *path = canonicalize_wim_path(paths[i]); + if (path == NULL) { + ret = WIMLIB_ERR_NOMEM; + trees = append_dentry_ctx.dentries; + goto out_free_trees; + } + ret = expand_wildcard(wim, path, + append_dentry_cb, + &append_dentry_ctx, + wildcard_flags); + FREE(path); + if (ret) { + trees = append_dentry_ctx.dentries; + goto out_free_trees; + } + } + trees = append_dentry_ctx.dentries; + num_trees = append_dentry_ctx.num_dentries; + } else { + trees = MALLOC(num_paths * sizeof(trees[0])); + if (trees == NULL) + return WIMLIB_ERR_NOMEM; - for (size_t i = 0; i < num_cmds; i++) { - cmds_copy[i].extract_flags = (default_extract_flags | - cmds[i].extract_flags) - & WIMLIB_EXTRACT_MASK_PUBLIC; - all_flags |= cmds_copy[i].extract_flags; + for (size_t i = 0; i < num_paths; i++) { - cmds_copy[i].wim_source_path = canonicalize_wim_path(cmds[i].wim_source_path); - if (!cmds_copy[i].wim_source_path) { - ret = WIMLIB_ERR_NOMEM; - goto out_free_cmds_copy; - } + tchar *path = canonicalize_wim_path(paths[i]); + if (path == NULL) { + ret = WIMLIB_ERR_NOMEM; + goto out_free_trees; + } - cmds_copy[i].fs_dest_path = canonicalize_fs_path(cmds[i].fs_dest_path); - if (!cmds_copy[i].fs_dest_path) { - ret = WIMLIB_ERR_NOMEM; - goto out_free_cmds_copy; + trees[i] = get_dentry(wim, path, + WIMLIB_CASE_PLATFORM_DEFAULT); + FREE(path); + if (trees[i] == NULL) { + ERROR("Path \"%"TS"\" does not exist " + "in WIM image %d", + paths[i], wim->current_image); + ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST; + goto out_free_trees; + } } - + num_trees = num_paths; } - ret = do_wimlib_extract_files(wim, image, - cmds_copy, num_cmds, - progress_func); - if (all_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK | - WIMLIB_EXTRACT_FLAG_HARDLINK)) - { - for_lookup_table_entry(wim->lookup_table, - lte_free_extracted_file, NULL); - } -out_free_cmds_copy: - for (size_t i = 0; i < num_cmds; i++) { - FREE(cmds_copy[i].wim_source_path); - FREE(cmds_copy[i].fs_dest_path); + if (num_trees == 0) { + ret = 0; + goto out_free_trees; } - FREE(cmds_copy); -out_restore_lookup_table: - if (num_additional_swms) - unmerge_lookup_table(wim); -out: + + ret = extract_trees(wim, trees, num_trees, + target, extract_flags, progress_func); +out_free_trees: + FREE(trees); return ret; } -/* - * Extracts an image from a WIM file. - * - * @wim: WIMStruct for the WIM file. - * - * @image: Number of the single image to extract. - * - * @target: Directory or NTFS volume to extract the image to. - * - * @extract_flags: Bitwise or of WIMLIB_EXTRACT_FLAG_*. - * - * @progress_func: If non-NULL, a progress function to be called - * periodically. - * - * Returns 0 on success; nonzero on failure. - */ static int extract_single_image(WIMStruct *wim, int image, const tchar *target, int extract_flags, wimlib_progress_func_t progress_func) { - int ret; - tchar *target_copy = canonicalize_fs_path(target); - if (!target_copy) - return WIMLIB_ERR_NOMEM; - struct wimlib_extract_command cmd = { - .wim_source_path = T(""), - .fs_dest_path = target_copy, - .extract_flags = extract_flags, - }; - ret = do_wimlib_extract_files(wim, image, &cmd, 1, progress_func); - FREE(target_copy); - return ret; + const tchar *path = T(""); + extract_flags |= WIMLIB_EXTRACT_FLAG_IMAGEMODE; + return do_wimlib_extract_paths(wim, image, target, &path, 1, + extract_flags, progress_func); } static const tchar * const filename_forbidden_chars = @@ -2551,7 +2971,6 @@ extract_all_images(WIMStruct *wim, int ret; int image; const tchar *image_name; - struct stat stbuf; extract_flags |= WIMLIB_EXTRACT_FLAG_MULTI_IMAGE; @@ -2560,21 +2979,9 @@ extract_all_images(WIMStruct *wim, return WIMLIB_ERR_INVALID_PARAM; } - if (tstat(target, &stbuf)) { - if (errno == ENOENT) { - if (tmkdir(target, 0755)) { - ERROR_WITH_ERRNO("Failed to create directory \"%"TS"\"", target); - return WIMLIB_ERR_MKDIR; - } - } else { - ERROR_WITH_ERRNO("Failed to stat \"%"TS"\"", target); - return WIMLIB_ERR_STAT; - } - } else if (!S_ISDIR(stbuf.st_mode)) { - ERROR("\"%"TS"\" is not a directory", target); - return WIMLIB_ERR_NOTDIR; - } - + ret = mkdir_if_needed(target); + if (ret) + return ret; tmemcpy(buf, target, output_path_len); buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR; for (image = 1; image <= wim->hdr.image_count; image++) { @@ -2594,50 +3001,147 @@ extract_all_images(WIMStruct *wim, return 0; } +static void +clear_lte_extracted_file(WIMStruct *wim, int extract_flags) +{ + if (unlikely(extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK | + WIMLIB_EXTRACT_FLAG_HARDLINK))) + for_lookup_table_entry(wim->lookup_table, + lte_free_extracted_file, NULL); +} + static int 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 (extract_flags & (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE | + WIMLIB_EXTRACT_FLAG_TO_STDOUT | + WIMLIB_EXTRACT_FLAG_GLOB_PATHS)) + return WIMLIB_ERR_INVALID_PARAM; - if (image == WIMLIB_ALL_IMAGES) { + if (image == WIMLIB_ALL_IMAGES) ret = extract_all_images(wim, target, extract_flags, progress_func); - } else { + else ret = extract_single_image(wim, image, target, extract_flags, progress_func); + + clear_lte_extracted_file(wim, extract_flags); + return ret; +} + + +/**************************************************************************** + * Extraction API * + ****************************************************************************/ + +/* Note: new code should use wimlib_extract_paths() instead of + * wimlib_extract_files() if possible. */ +WIMLIBAPI int +wimlib_extract_files(WIMStruct *wim, int image, + const struct wimlib_extract_command *cmds, size_t num_cmds, + int default_extract_flags, + wimlib_progress_func_t progress_func) +{ + int all_flags = 0; + int link_flags; + int ret; + + if (num_cmds == 0) + return 0; + + default_extract_flags |= WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE; + + for (size_t i = 0; i < num_cmds; i++) { + int cmd_flags = (cmds[i].extract_flags | + default_extract_flags); + + if (cmd_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC) + return WIMLIB_ERR_INVALID_PARAM; + + int cmd_link_flags = (cmd_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK | + WIMLIB_EXTRACT_FLAG_HARDLINK)); + if (i == 0) { + link_flags = cmd_link_flags; + } else { + if (cmd_link_flags != link_flags) { + ERROR("The same symlink or hardlink extraction mode " + "must be set on all extraction commands!"); + return WIMLIB_ERR_INVALID_PARAM; + } + } + all_flags |= cmd_flags; + } + if (all_flags & WIMLIB_EXTRACT_FLAG_GLOB_PATHS) { + ERROR("Glob paths not supported for wimlib_extract_files(). " + "Use wimlib_extract_paths() instead."); + return WIMLIB_ERR_INVALID_PARAM; } - if (extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK | - WIMLIB_EXTRACT_FLAG_HARDLINK)) - { - for_lookup_table_entry(wim->lookup_table, - lte_free_extracted_file, - NULL); + for (size_t i = 0; i < num_cmds; i++) { + int extract_flags = (cmds[i].extract_flags | + default_extract_flags); + const tchar *target = cmds[i].fs_dest_path; + const tchar *wim_source_path = cmds[i].wim_source_path; + + ret = do_wimlib_extract_paths(wim, image, target, + &wim_source_path, 1, + extract_flags | WIMLIB_EXTRACT_FLAG_FILEMODE, + progress_func); + if (ret) + break; + } + + clear_lte_extracted_file(wim, all_flags); + return ret; +} + +WIMLIBAPI int +wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target, + const tchar * const *paths, size_t num_paths, + int extract_flags, wimlib_progress_func_t progress_func) +{ + int ret; + + if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC) + return WIMLIB_ERR_INVALID_PARAM; + + ret = do_wimlib_extract_paths(wim, image, target, paths, num_paths, + extract_flags, progress_func); + clear_lte_extracted_file(wim, extract_flags); + return ret; +} + +WIMLIBAPI int +wimlib_extract_pathlist(WIMStruct *wim, int image, const tchar *target, + const tchar *path_list_file, int extract_flags, + wimlib_progress_func_t progress_func) +{ + int ret; + tchar **paths; + size_t num_paths; + void *mem; + + ret = read_path_list_file(path_list_file, &paths, &num_paths, &mem); + if (ret) { + ERROR("Failed to read path list file \"%"TS"\"", + path_list_file); + return ret; } - if (num_additional_swms) - unmerge_lookup_table(wim); + + ret = wimlib_extract_paths(wim, image, target, + (const tchar * const *)paths, num_paths, + extract_flags, progress_func); + FREE(paths); + FREE(mem); return ret; } -/* API function documented in wimlib.h */ WIMLIBAPI int wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name, const tchar *target, int extract_flags, @@ -2649,20 +3153,18 @@ wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name, int image; unsigned i; - extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC; - - if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) + if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC) return WIMLIB_ERR_INVALID_PARAM; - extract_flags |= WIMLIB_EXTRACT_FLAG_SEQUENTIAL; + if (extract_flags & WIMLIB_EXTRACT_FLAG_FILE_ORDER) + return WIMLIB_ERR_INVALID_PARAM; /* Read the WIM header from the pipe and get a WIMStruct to represent * the pipable WIM. Caveats: Unlike getting a WIMStruct with * 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; @@ -2694,11 +3196,12 @@ 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."); @@ -2706,15 +3209,15 @@ 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; + ret = WIMLIB_ERR_IMAGE_COUNT; goto out_wimlib_free; } } @@ -2729,7 +3232,7 @@ wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name, ret = WIMLIB_ERR_INVALID_IMAGE; goto out_wimlib_free; } else if (image == WIMLIB_ALL_IMAGES) { - ERROR("Applying all images from a pipe is not supported."); + ERROR("Applying all images from a pipe is not supported!"); ret = WIMLIB_ERR_INVALID_IMAGE; goto out_wimlib_free; } @@ -2747,22 +3250,29 @@ 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; @@ -2770,7 +3280,7 @@ wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name, } if (i == image) { - /* Metadata resource is for the images being extracted. + /* Metadata resource is for the image being extracted. * Parse it and save the metadata in memory. */ ret = read_metadata_resource(pwm, imd); if (ret) @@ -2779,7 +3289,7 @@ wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name, } else { /* Metadata resource is not for the image being * extracted. Skip over it. */ - ret = skip_pwm_stream(metadata_lte); + ret = skip_wim_stream(metadata_lte); if (ret) goto out_wimlib_free; } @@ -2787,25 +3297,19 @@ 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); return ret; } -/* API function documented in wimlib.h */ WIMLIBAPI int -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) +wimlib_extract_image(WIMStruct *wim, int image, const tchar *target, + int extract_flags, wimlib_progress_func_t progress_func) { - extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC; + if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC) + return WIMLIB_ERR_INVALID_PARAM; return do_wimlib_extract_image(wim, image, target, extract_flags, - additional_swms, num_additional_swms, progress_func); }