X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Fextract.c;h=3adbca342ddfa65e50f61e9e11ca30be5c16fcc4;hp=b7a15c04306941f483bc6bc450b13976236bcb85;hb=b5b9681794d1f5f13350e3567f6f6e74f5c779cf;hpb=ce81b1490c131f0be20c21661bc8f954d45c417c diff --git a/src/extract.c b/src/extract.c index b7a15c04..3adbca34 100644 --- a/src/extract.c +++ b/src/extract.c @@ -1,17 +1,12 @@ /* * extract.c * - * Support for extracting WIM files. - * - * This code does NOT contain any filesystem-specific features. In particular, - * security information (i.e. file permissions) and alternate data streams are - * ignored, except possibly to read an alternate data stream that contains - * symbolic link data. + * Support for extracting WIM images, or files or directories contained in a WIM + * image. */ /* - * Copyright (C) 2010 Carl Thijssen - * Copyright (C) 2012 Eric Biggers + * Copyright (C) 2012, 2013, 2014 Eric Biggers * * This file is part of wimlib, a library for working with WIM files. * @@ -29,756 +24,1806 @@ * along with wimlib; if not, see http://www.gnu.org/licenses/. */ +/* + * This file provides the API functions wimlib_extract_image(), + * wimlib_extract_image_from_pipe(), 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 + * to any specific target; instead, it handles generic functionality and relies + * on lower-level callback functions declared in `struct apply_operations' to do + * the actual extraction. + */ + +#ifdef HAVE_CONFIG_H +# include "config.h" +#endif -#include "config.h" +#include "wimlib/apply.h" +#include "wimlib/dentry.h" +#include "wimlib/encoding.h" +#include "wimlib/endianness.h" +#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" +#ifdef __WIN32__ +# include "wimlib/win32.h" /* for realpath() equivalent */ +#endif +#include "wimlib/xml.h" +#include "wimlib/wildcard.h" +#include "wimlib/wim.h" -#include #include #include -#include -#include #include -#include +#include +#include -#ifdef HAVE_UTIME_H -#include -#endif +#define WIMLIB_EXTRACT_FLAG_MULTI_IMAGE 0x80000000 +#define WIMLIB_EXTRACT_FLAG_FROM_PIPE 0x40000000 +#define WIMLIB_EXTRACT_FLAG_IMAGEMODE 0x20000000 + +/* Keep in sync with wimlib.h */ +#define WIMLIB_EXTRACT_MASK_PUBLIC \ + (WIMLIB_EXTRACT_FLAG_NTFS | \ + 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_GLOB_PATHS | \ + WIMLIB_EXTRACT_FLAG_STRICT_GLOB | \ + WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES | \ + WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE | \ + WIMLIB_EXTRACT_FLAG_WIMBOOT) + +/* Check whether the extraction of a dentry should be skipped completely. */ +static bool +dentry_is_supported(struct wim_dentry *dentry, + const struct wim_features *supported_features) +{ + struct wim_inode *inode = dentry->d_inode; + if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) { + 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; +} -#include -#include "dentry.h" -#include "lookup_table.h" -#include "timestamp.h" -#include "wimlib_internal.h" -#include "xml.h" +#define PWM_ALLOW_WIM_HDR 0x00001 +#define PWM_SILENT_EOF 0x00002 -#ifdef WITH_NTFS_3G -#include -#endif +/* 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; -static int extract_regular_file_linked(struct dentry *dentry, - const char *output_path, - struct apply_args *args, - struct lookup_table_entry *lte) -{ - /* This mode overrides the normal hard-link extraction and - * instead either symlinks or hardlinks *all* identical files in - * the WIM, even if they are in a different image (in the case - * of a multi-image extraction) */ - - if (args->extract_flags & WIMLIB_EXTRACT_FLAG_HARDLINK) { - if (link(lte->extracted_file, output_path) != 0) { - ERROR_WITH_ERRNO("Failed to hard link " - "`%s' to `%s'", - output_path, lte->extracted_file); - return WIMLIB_ERR_LINK; - } - } else { - int num_path_components; - int num_output_dir_path_components; - size_t extracted_file_len; - char *p; - const char *p2; - size_t i; - - num_path_components = - get_num_path_components(dentry->full_path_utf8) - 1; - num_output_dir_path_components = - get_num_path_components(args->target); - - if (args->extract_flags & WIMLIB_EXTRACT_FLAG_MULTI_IMAGE) { - num_path_components++; - num_output_dir_path_components--; - } - extracted_file_len = strlen(lte->extracted_file); + ret = full_read(&pwm->in_fd, &buf.stream_hdr, sizeof(buf.stream_hdr)); + if (ret) + goto read_error; - char buf[extracted_file_len + 3 * num_path_components + 1]; - p = &buf[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)); - for (i = 0; i < num_path_components; i++) { - *p++ = '.'; - *p++ = '.'; - *p++ = '/'; - } - p2 = lte->extracted_file; - while (*p2 == '/') - p2++; - while (num_output_dir_path_components--) - p2 = path_next_part(p2, NULL); - strcpy(p, p2); - if (symlink(buf, output_path) != 0) { - ERROR_WITH_ERRNO("Failed to symlink `%s' to " - "`%s'", - buf, lte->extracted_file); - return WIMLIB_ERR_LINK; - } + 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_regular_file_unlinked(struct dentry *dentry, - struct apply_args *args, - const char *output_path, - struct lookup_table_entry *lte) +static int +load_streams_from_pipe(struct apply_ctx *ctx, + const struct read_stream_list_callbacks *cbs) { - /* Normal mode of extraction. Regular files and hard links are - * extracted in the way that they appear in the WIM. */ - - int out_fd; + struct wim_lookup_table_entry *found_lte = NULL; + struct wim_resource_spec *rspec = NULL; + struct wim_lookup_table *lookup_table; int ret; - struct inode *inode = dentry->d_inode; - if (!((args->extract_flags & WIMLIB_EXTRACT_FLAG_MULTI_IMAGE) - && (args->extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK | - WIMLIB_EXTRACT_FLAG_HARDLINK)))) - { - /* If the dentry is one of a hard link set of at least 2 - * dentries and one of the other dentries has already been - * extracted, make a hard link to the file corresponding to this - * already-extracted directory. Otherwise, extract the file, - * and set the inode->extracted_file field so that other - * dentries in the hard link group can link to it. */ - if (inode->link_count > 1) { - if (inode->extracted_file) { - DEBUG("Extracting hard link `%s' => `%s'", - output_path, inode->extracted_file); - if (link(inode->extracted_file, output_path) != 0) { - ERROR_WITH_ERRNO("Failed to hard link " - "`%s' to `%s'", - output_path, - inode->extracted_file); - return WIMLIB_ERR_LINK; - } - return 0; + ret = WIMLIB_ERR_NOMEM; + found_lte = new_lookup_table_entry(); + if (!found_lte) + goto out; + + rspec = MALLOC(sizeof(struct wim_resource_spec)); + if (!rspec) + goto out; + + lookup_table = ctx->wim->lookup_table; + memcpy(ctx->progress.extract.guid, ctx->wim->hdr.guid, WIM_GUID_LEN); + ctx->progress.extract.part_number = ctx->wim->hdr.part_number; + ctx->progress.extract.total_parts = ctx->wim->hdr.total_parts; + ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN); + if (ret) + goto out; + + while (ctx->num_streams_remaining) { + struct wim_header_disk pwm_hdr; + struct wim_lookup_table_entry *needed_lte; + + 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_ALLOW_WIM_HDR, &pwm_hdr); + if (ret) + goto out; + + if ((found_lte->resource_location != RESOURCE_NONEXISTENT) + && !(found_lte->flags & WIM_RESHDR_FLAG_METADATA) + && (needed_lte = lookup_stream(lookup_table, found_lte->hash)) + && (needed_lte->out_refcnt)) + { + 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); + + ret = (*cbs->begin_stream)(needed_lte, 0, + cbs->begin_stream_ctx); + if (ret) { + lte_unbind_wim_resource_spec(needed_lte); + goto out; } - FREE(inode->extracted_file); - inode->extracted_file = STRDUP(output_path); - if (!inode->extracted_file) { - ERROR("Failed to allocate memory for filename"); - return WIMLIB_ERR_NOMEM; + + ret = extract_stream(needed_lte, needed_lte->size, + cbs->consume_chunk, + cbs->consume_chunk_ctx); + + ret = (*cbs->end_stream)(needed_lte, ret, + cbs->end_stream_ctx); + lte_unbind_wim_resource_spec(needed_lte); + if (ret) + goto out; + ctx->num_streams_remaining--; + } else if (found_lte->resource_location != RESOURCE_NONEXISTENT) { + ret = skip_wim_stream(found_lte); + if (ret) + goto out; + } else { + u16 part_number = le16_to_cpu(pwm_hdr.part_number); + u16 total_parts = le16_to_cpu(pwm_hdr.total_parts); + + if (part_number != ctx->progress.extract.part_number || + total_parts != ctx->progress.extract.total_parts || + memcmp(pwm_hdr.guid, ctx->progress.extract.guid, + WIM_GUID_LEN)) + { + ctx->progress.extract.part_number = part_number; + ctx->progress.extract.total_parts = total_parts; + memcpy(ctx->progress.extract.guid, + pwm_hdr.guid, WIM_GUID_LEN); + ret = extract_progress(ctx, + WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN); + if (ret) + goto out; } } } + ret = 0; +out: + if (found_lte->resource_location != RESOURCE_IN_WIM) + FREE(rspec); + free_lookup_table_entry(found_lte); + return ret; +} - /* Extract the contents of the file to @output_path. */ +static int +begin_extract_stream_with_progress(struct wim_lookup_table_entry *lte, + u32 flags, void *_ctx) +{ + struct apply_ctx *ctx = _ctx; - out_fd = open(output_path, O_WRONLY | O_CREAT | O_TRUNC, 0644); - if (out_fd == -1) { - ERROR_WITH_ERRNO("Failed to open the file `%s' for writing", - output_path); - return WIMLIB_ERR_OPEN; - } + ctx->cur_stream = lte; - if (!lte) { - /* Empty file with no lookup table entry */ - DEBUG("Empty file `%s'.", output_path); - ret = 0; - goto out; - } + return (*ctx->saved_cbs->begin_stream)(lte, flags, + ctx->saved_cbs->begin_stream_ctx); +} - ret = extract_full_wim_resource_to_fd(lte, out_fd); - if (ret != 0) { - ERROR("Failed to extract resource to `%s'", output_path); - goto out; +static int +consume_chunk_with_progress(const void *chunk, size_t size, void *_ctx) +{ + struct apply_ctx *ctx = _ctx; + union wimlib_progress_info *progress = &ctx->progress; + int ret; + + if (likely(ctx->supported_features.hard_links)) { + progress->extract.completed_bytes += + (u64)size * ctx->cur_stream->out_refcnt; + } else { + const struct stream_owner *owners = stream_owners(ctx->cur_stream); + for (u32 i = 0; i < ctx->cur_stream->out_refcnt; i++) { + const struct wim_inode *inode = owners[i].inode; + const struct wim_dentry *dentry; + + list_for_each_entry(dentry, + &inode->i_extraction_aliases, + d_extraction_alias_node) + { + progress->extract.completed_bytes += size; + } + } } - args->progress.extract.completed_bytes += wim_resource_size(lte); -out: - if (close(out_fd) != 0) { - ERROR_WITH_ERRNO("Failed to close file `%s'", output_path); - ret = WIMLIB_ERR_WRITE; + if (progress->extract.completed_bytes >= ctx->next_progress) { + + ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS); + if (ret) + return ret; + + if (progress->extract.completed_bytes >= + progress->extract.total_bytes) + { + ctx->next_progress = UINT64_MAX; + } else { + ctx->next_progress += progress->extract.total_bytes / 128; + if (ctx->next_progress > progress->extract.total_bytes) + ctx->next_progress = progress->extract.total_bytes; + } } - return ret; + return (*ctx->saved_cbs->consume_chunk)(chunk, size, + ctx->saved_cbs->consume_chunk_ctx); } /* - * Extracts a regular file from the WIM archive. + * Read the list of single-instance streams to extract and feed their data into + * the specified callback functions. + * + * This handles checksumming each stream. + * + * This also handles sending WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS. + * + * This also works if the WIM is being read from a pipe, whereas attempting to + * read streams directly (e.g. with read_full_stream_into_buf()) will not. */ -static int extract_regular_file(struct dentry *dentry, - struct apply_args *args, - const char *output_path) +int +extract_stream_list(struct apply_ctx *ctx, + const struct read_stream_list_callbacks *cbs) { - struct lookup_table_entry *lte; - const struct inode *inode = dentry->d_inode; + struct read_stream_list_callbacks wrapper_cbs = { + .begin_stream = begin_extract_stream_with_progress, + .begin_stream_ctx = ctx, + .consume_chunk = consume_chunk_with_progress, + .consume_chunk_ctx = ctx, + .end_stream = cbs->end_stream, + .end_stream_ctx = cbs->end_stream_ctx, + }; + if (ctx->progfunc) { + ctx->saved_cbs = cbs; + cbs = &wrapper_cbs; + } + if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) { + return load_streams_from_pipe(ctx, cbs); + } else { + return read_stream_list(&ctx->stream_list, + offsetof(struct wim_lookup_table_entry, + extraction_list), + cbs, VERIFY_STREAM_HASHES); + } +} - lte = inode_unnamed_lte_resolved(inode); +/* 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 +extract_dentry_to_stdout(struct wim_dentry *dentry, + const struct wim_lookup_table *lookup_table) +{ + struct wim_inode *inode = dentry->d_inode; + struct wim_lookup_table_entry *lte; + struct filedes _stdout; - if (lte && (args->extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK | - WIMLIB_EXTRACT_FLAG_HARDLINK))) + if (inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT | + FILE_ATTRIBUTE_DIRECTORY)) { - if (lte->extracted_file) { - return extract_regular_file_linked(dentry, output_path, args, lte); - } else { - lte->extracted_file = STRDUP(output_path); - if (!lte->extracted_file) - return WIMLIB_ERR_NOMEM; - } + 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; } - return extract_regular_file_unlinked(dentry, args, output_path, lte); + + 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; + } + + filedes_init(&_stdout, STDOUT_FILENO); + return extract_full_stream_to_fd(lte, &_stdout); } -static int extract_symlink(struct dentry *dentry, - struct apply_args *args, - const char *output_path) +static int +extract_dentries_to_stdout(struct wim_dentry **dentries, size_t num_dentries, + const struct wim_lookup_table *lookup_table) { - char target[4096]; - ssize_t ret = inode_readlink(dentry->d_inode, target, - sizeof(target), args->w, 0); - struct lookup_table_entry *lte; + 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; +} - if (ret <= 0) { - ERROR("Could not read the symbolic link from dentry `%s'", - dentry->full_path_utf8); - return WIMLIB_ERR_INVALID_DENTRY; +/**********************************************************************/ + +/* + * Removes duplicate dentries from the array. + * + * 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]; + } } - ret = symlink(target, output_path); - if (ret != 0) { - ERROR_WITH_ERRNO("Failed to symlink `%s' to `%s'", - output_path, target); - return WIMLIB_ERR_LINK; + 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; } - lte = inode_unnamed_lte_resolved(dentry->d_inode); - args->progress.extract.completed_bytes += wim_resource_size(lte); + + for (i = 0; i < j; i++) + trees[i]->tmp_flag = 0; + return j; +} + +static int +dentry_append_to_list(struct wim_dentry *dentry, void *_dentry_list) +{ + struct list_head *dentry_list = _dentry_list; + list_add_tail(&dentry->d_extraction_list_node, dentry_list); + return 0; +} + +static void +dentry_reset_extraction_list_node(struct wim_dentry *dentry) +{ + dentry->d_extraction_list_node = (struct list_head){NULL, NULL}; +} + +static int +dentry_delete_from_list(struct wim_dentry *dentry, void *_ignore) +{ + list_del(&dentry->d_extraction_list_node); + dentry_reset_extraction_list_node(dentry); return 0; } /* - * Extracts a directory from the WIM archive. + * Build the preliminary list of dentries to be extracted. * - * @dentry: The directory entry for the directory. - * @output_path: The path to which the directory is to be extracted to. - * @return: True on success, false on failure. + * 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 int extract_directory(const char *output_path, bool is_root) +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 (will_extract_dentry(ancestor)) { + place_after = &ancestor->d_extraction_list_node; + break; + } + } while (!dentry_is_root(ancestor)); + + ancestor = dentry; + do { + ancestor = ancestor->parent; + if (will_extract_dentry(ancestor)) + break; + list_add(&ancestor->d_extraction_list_node, place_after); + } while (!dentry_is_root(ancestor)); + } + } +} + +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, d_extraction_list_node) { + inode = dentry->d_inode; + dentry_reset_extraction_list_node(dentry); + inode->i_visited = 0; + if ((void *)dentry->d_extraction_name != (void *)dentry->file_name) + FREE(dentry->d_extraction_name); + dentry->d_extraction_name = NULL; + dentry->d_extraction_name_nchars = 0; + } +} + +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_stream_owners)) + FREE(lte->stream_owners); +} + +#ifdef __WIN32__ +static const utf16lechar replacement_char = cpu_to_le16(0xfffd); +#else +static const utf16lechar replacement_char = cpu_to_le16('?'); +#endif + +static bool +file_name_valid(utf16lechar *name, size_t num_chars, bool fix) +{ + size_t i; + + if (num_chars == 0) + return true; + for (i = 0; i < num_chars; i++) { + switch (name[i]) { + #ifdef __WIN32__ + case cpu_to_le16('\\'): + case cpu_to_le16(':'): + case cpu_to_le16('*'): + case cpu_to_le16('?'): + case cpu_to_le16('"'): + case cpu_to_le16('<'): + case cpu_to_le16('>'): + case cpu_to_le16('|'): + #endif + case cpu_to_le16('/'): + case cpu_to_le16('\0'): + if (fix) + name[i] = replacement_char; + else + return false; + } + } + +#ifdef __WIN32__ + if (name[num_chars - 1] == cpu_to_le16(' ') || + name[num_chars - 1] == cpu_to_le16('.')) + { + if (fix) + name[num_chars - 1] = replacement_char; + else + return false; + } +#endif + return true; +} + +static int +dentry_calculate_extraction_name(struct wim_dentry *dentry, + struct apply_ctx *ctx) { int ret; - struct stat stbuf; - ret = stat(output_path, &stbuf); - if (ret == 0) { - if (S_ISDIR(stbuf.st_mode)) { - /*if (!is_root)*/ - /*WARNING("`%s' already exists", output_path);*/ - return 0; - } else { - ERROR("`%s' is not a directory", output_path); - return WIMLIB_ERR_MKDIR; + + if (!dentry_is_supported(dentry, &ctx->supported_features)) + goto skip_dentry; + + if (dentry_is_root(dentry)) + return 0; + + if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) { + dentry->d_extraction_name = dentry->file_name; + dentry->d_extraction_name_nchars = dentry->file_name_nbytes / + sizeof(utf16lechar); + return 0; + } + + if (!ctx->supported_features.case_sensitive_filenames) { + struct wim_dentry *other; + list_for_each_entry(other, &dentry->d_ci_conflict_list, + d_ci_conflict_list) + { + if (will_extract_dentry(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; + } + } } + } + + if (file_name_valid(dentry->file_name, dentry->file_name_nbytes / 2, false)) { + ret = utf16le_get_tstr(dentry->file_name, + dentry->file_name_nbytes, + (const tchar **)&dentry->d_extraction_name, + &dentry->d_extraction_name_nchars); + dentry->d_extraction_name_nchars /= sizeof(tchar); + return ret; } else { - if (errno != ENOENT) { - ERROR_WITH_ERRNO("Failed to stat `%s'", output_path); - return WIMLIB_ERR_STAT; + if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES) + { + WARNING("\"%"TS"\" has an invalid filename " + "that is not supported on this platform; " + "extracting dummy name instead", + dentry_full_path(dentry)); + goto out_replace; + } else { + WARNING("Not extracting \"%"TS"\": has an invalid filename " + "that is not supported on this platform", + dentry_full_path(dentry)); + goto skip_dentry; } } - /* Compute the output path directory to the directory. */ - if (mkdir(output_path, S_IRWXU | S_IRGRP | S_IXGRP | - S_IROTH | S_IXOTH) != 0) { - ERROR_WITH_ERRNO("Cannot create directory `%s'", - output_path); - return WIMLIB_ERR_MKDIR; + +out_replace: + { + utf16lechar utf16_name_copy[dentry->file_name_nbytes / 2]; + + memcpy(utf16_name_copy, dentry->file_name, dentry->file_name_nbytes); + file_name_valid(utf16_name_copy, dentry->file_name_nbytes / 2, true); + + const tchar *tchar_name; + size_t tchar_nchars; + + ret = utf16le_get_tstr(utf16_name_copy, + dentry->file_name_nbytes, + &tchar_name, &tchar_nchars); + if (ret) + return ret; + + tchar_nchars /= sizeof(tchar); + + size_t fixed_name_num_chars = tchar_nchars; + tchar fixed_name[tchar_nchars + 50]; + + tmemcpy(fixed_name, tchar_name, tchar_nchars); + fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars, + T(" (invalid filename #%lu)"), + ++ctx->invalid_sequence); + + utf16le_put_tstr(tchar_name); + + dentry->d_extraction_name = memdup(fixed_name, + 2 * fixed_name_num_chars + 2); + if (!dentry->d_extraction_name) + return WIMLIB_ERR_NOMEM; + dentry->d_extraction_name_nchars = fixed_name_num_chars; } return 0; + +skip_dentry: + for_dentry_in_tree(dentry, dentry_delete_from_list, NULL); + return 0; } /* - * Extracts a file, directory, or symbolic link from the WIM archive. For use - * in for_dentry_in_tree(). + * 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. + * + * ctx->supported_features must be filled in. + * + * Possible error codes: WIMLIB_ERR_NOMEM, WIMLIB_ERR_INVALID_UTF16_STRING */ -static int apply_dentry_normal(struct dentry *dentry, void *arg) +static int +dentry_list_calculate_extraction_names(struct list_head *dentry_list, + struct apply_ctx *ctx) { - struct apply_args *args = arg; - int extract_flags = args->extract_flags; - struct inode *inode = dentry->d_inode; - size_t len; - int ret; + struct list_head *prev, *cur; - if (dentry->is_extracted) - return 0; + /* 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. */ - if (extract_flags & WIMLIB_EXTRACT_FLAG_NO_STREAMS) - if (inode_unnamed_lte_resolved(inode)) - return 0; + prev = dentry_list; + for (;;) { + struct wim_dentry *dentry; + int ret; - if ((extract_flags & WIMLIB_EXTRACT_FLAG_VERBOSE) && - args->progress_func) - { - args->progress.extract.cur_path = dentry->full_path_utf8; - args->progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DENTRY, - &args->progress); - } + cur = prev->next; + if (cur == dentry_list) + break; - len = strlen(args->target); - char output_path[len + dentry->full_path_utf8_len + 1]; - memcpy(output_path, args->target, len); - memcpy(output_path + len, dentry->full_path_utf8, dentry->full_path_utf8_len); - output_path[len + dentry->full_path_utf8_len] = '\0'; + dentry = list_entry(cur, struct wim_dentry, d_extraction_list_node); - if (inode_is_symlink(inode)) - ret = extract_symlink(dentry, args, output_path); - else if (inode_is_directory(inode)) - ret = extract_directory(output_path, false); - else - ret = extract_regular_file(dentry, args, output_path); - if (ret == 0) - dentry->is_extracted = 1; - return ret; + 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; } -/* Apply timestamp to extracted file */ -static int apply_dentry_timestamps_normal(struct dentry *dentry, void *arg) +static int +dentry_resolve_streams(struct wim_dentry *dentry, int extract_flags, + struct wim_lookup_table *lookup_table) { - struct apply_args *args = arg; - size_t len = strlen(args->target); - char output_path[len + dentry->full_path_utf8_len + 1]; - const struct inode *inode = dentry->d_inode; + struct wim_inode *inode = dentry->d_inode; + struct wim_lookup_table_entry *lte; int ret; + bool force = false; + + /* 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; +} - memcpy(output_path, args->target, len); - memcpy(output_path + len, dentry->full_path_utf8, dentry->full_path_utf8_len); - output_path[len + dentry->full_path_utf8_len] = '\0'; - - struct timeval tv[2]; - wim_timestamp_to_timeval(inode->last_access_time, &tv[0]); - wim_timestamp_to_timeval(inode->last_write_time, &tv[1]); - #ifdef HAVE_LUTIMES - ret = lutimes(output_path, tv); - #else - ret = -1; - errno = ENOSYS; - #endif - if (ret != 0) { - #ifdef HAVE_UTIME - if (errno == ENOSYS) { - struct utimbuf buf; - buf.actime = wim_timestamp_to_unix(inode->last_access_time); - buf.modtime = wim_timestamp_to_unix(inode->last_write_time); - if (utime(output_path, &buf) == 0) - return 0; +/* + * 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, d_extraction_list_node) { + 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, u32 stream_idx, + struct wim_dentry *dentry, struct apply_ctx *ctx) +{ + struct wim_inode *inode = dentry->d_inode; + struct stream_owner *stream_owners; + + if (!lte) + return 0; + + /* Tally the size only for each extraction of the stream (not hard + * links). */ + if (inode->i_visited && ctx->supported_features.hard_links) + return 0; + + ctx->progress.extract.total_bytes += lte->size; + ctx->progress.extract.num_streams++; + + if (inode->i_visited) + return 0; + + /* Add stream to the dentry_list only one time, even if it's going + * to be extracted to multiple inodes. */ + if (lte->out_refcnt == 0) { + list_add_tail(<e->extraction_list, &ctx->stream_list); + ctx->num_streams_remaining++; + } + + /* If inode not yet been visited, append it to the stream_owners array. */ + if (lte->out_refcnt < ARRAY_LEN(lte->inline_stream_owners)) { + stream_owners = lte->inline_stream_owners; + } else { + struct stream_owner *prev_stream_owners; + size_t alloc_stream_owners; + + if (lte->out_refcnt == ARRAY_LEN(lte->inline_stream_owners)) { + prev_stream_owners = NULL; + alloc_stream_owners = ARRAY_LEN(lte->inline_stream_owners); + } else { + prev_stream_owners = lte->stream_owners; + alloc_stream_owners = lte->alloc_stream_owners; } - #endif - if (errno != ENOSYS || args->num_lutimes_warnings < 10) { - /*WARNING("Failed to set timestamp on file `%s': %s",*/ - /*output_path, strerror(errno));*/ - args->num_lutimes_warnings++; + + if (lte->out_refcnt == alloc_stream_owners) { + alloc_stream_owners *= 2; + stream_owners = REALLOC(prev_stream_owners, + alloc_stream_owners * + sizeof(stream_owners[0])); + if (!stream_owners) + return WIMLIB_ERR_NOMEM; + if (!prev_stream_owners) { + memcpy(stream_owners, + lte->inline_stream_owners, + sizeof(lte->inline_stream_owners)); + } + lte->stream_owners = stream_owners; + lte->alloc_stream_owners = alloc_stream_owners; } + stream_owners = lte->stream_owners; } + stream_owners[lte->out_refcnt].inode = inode; + if (stream_idx == 0) { + stream_owners[lte->out_refcnt].stream_name = NULL; + } else { + stream_owners[lte->out_refcnt].stream_name = + inode->i_ads_entries[stream_idx - 1].stream_name; + } + lte->out_refcnt++; return 0; } -static int cmp_streams_by_wim_position(const void *p1, const void *p2) +static int +dentry_ref_streams(struct wim_dentry *dentry, struct apply_ctx *ctx) { - const struct lookup_table_entry *lte1, *lte2; - lte1 = *(const struct lookup_table_entry**)p1; - lte2 = *(const struct lookup_table_entry**)p2; - if (lte1->resource_entry.offset < lte2->resource_entry.offset) - return -1; - else if (lte1->resource_entry.offset > lte2->resource_entry.offset) - return 1; - else - return 0; + 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)) { + u16 stream_idx; + struct wim_lookup_table_entry *stream; + + stream = inode_unnamed_stream_resolved(inode, &stream_idx); + ret = ref_stream(stream, stream_idx, 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 (ctx->supported_features.named_data_streams) { + 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, i + 1, + dentry, ctx); + if (ret) + return ret; + } + } + inode->i_visited = 1; + return 0; } -static int sort_stream_list_by_wim_position(struct list_head *stream_list) +/* + * 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 builds a mapping from each stream to the inodes referencing it. + * + * This also initializes the extract progress info with byte and stream + * information. + * + * ctx->supported_features must be filled in. + * + * Possible error codes: WIMLIB_ERR_NOMEM. + */ +static int +dentry_list_ref_streams(struct list_head *dentry_list, struct apply_ctx *ctx) { - struct list_head *cur; - size_t num_streams; - struct lookup_table_entry **array; - size_t i; - size_t array_size; - - num_streams = 0; - list_for_each(cur, stream_list) - num_streams++; - array_size = num_streams * sizeof(array[0]); - array = MALLOC(array_size); - if (!array) { - ERROR("Failed to allocate %zu bytes to sort stream entries", - array_size); - return WIMLIB_ERR_NOMEM; - } - cur = stream_list->next; - for (i = 0; i < num_streams; i++) { - array[i] = container_of(cur, struct lookup_table_entry, staging_list); - cur = cur->next; - } - - qsort(array, num_streams, sizeof(array[0]), cmp_streams_by_wim_position); - - INIT_LIST_HEAD(stream_list); - for (i = 0; i < num_streams; i++) - list_add_tail(&array[i]->staging_list, stream_list); - FREE(array); + struct wim_dentry *dentry; + int ret; + + list_for_each_entry(dentry, dentry_list, d_extraction_list_node) { + ret = dentry_ref_streams(dentry, ctx); + if (ret) + return ret; + } + list_for_each_entry(dentry, dentry_list, d_extraction_list_node) + dentry->d_inode->i_visited = 0; return 0; } -static void calculate_bytes_to_extract(struct list_head *stream_list, - int extract_flags, - union wimlib_progress_info *progress) +static void +dentry_list_build_inode_alias_lists(struct list_head *dentry_list) { - struct lookup_table_entry *lte; - struct inode *inode; - u64 total_bytes = 0; - u64 num_streams = 0; + struct wim_dentry *dentry; + struct wim_inode *inode; + + list_for_each_entry(dentry, dentry_list, d_extraction_list_node) { + inode = dentry->d_inode; + if (!inode->i_visited) + INIT_LIST_HEAD(&inode->i_extraction_aliases); + list_add_tail(&dentry->d_extraction_alias_node, + &inode->i_extraction_aliases); + inode->i_visited = 1; + } + list_for_each_entry(dentry, dentry_list, d_extraction_list_node) + dentry->d_inode->i_visited = 0; +} - /* For each stream to be extracted... */ - list_for_each_entry(lte, stream_list, staging_list) { - if (extract_flags & - (WIMLIB_EXTRACT_FLAG_SYMLINK | WIMLIB_EXTRACT_FLAG_HARDLINK)) - { - /* In the symlink or hard link extraction mode, each - * stream will be extracted one time regardless of how - * many dentries share the stream. */ - wimlib_assert(!(extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)); - if (!lte->extracted_file) { - num_streams++; - total_bytes += wim_resource_size(lte); - } - } else { - num_streams += lte->out_refcnt; - total_bytes += lte->out_refcnt * wim_resource_size(lte); - } +static void +inode_tally_features(const struct wim_inode *inode, + struct wim_features *features) +{ + if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE) + features->archive_files++; + if (inode->i_attributes & FILE_ATTRIBUTE_HIDDEN) + features->hidden_files++; + if (inode->i_attributes & FILE_ATTRIBUTE_SYSTEM) + features->system_files++; + if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED) + features->compressed_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) + features->sparse_files++; + if (inode_has_named_stream(inode)) + features->named_data_streams++; + if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) { + features->reparse_points++; + if (inode_is_symlink(inode)) + features->symlink_reparse_points++; + else + features->other_reparse_points++; } - progress->extract.num_streams = num_streams; - progress->extract.total_bytes = total_bytes; - progress->extract.completed_bytes = 0; + if (inode->i_security_id != -1) + features->security_descriptors++; + if (inode_has_unix_data(inode)) + features->unix_data++; } -static void maybe_add_stream_for_extraction(struct lookup_table_entry *lte, - struct list_head *stream_list) +/* Tally features necessary to extract a dentry and the corresponding inode. */ +static void +dentry_tally_features(struct wim_dentry *dentry, struct wim_features *features) { - if (++lte->out_refcnt == 1) { - INIT_LIST_HEAD(<e->inode_list); - list_add_tail(<e->staging_list, stream_list); + struct wim_inode *inode = dentry->d_inode; + + if (dentry_has_short_name(dentry)) + features->short_names++; + + if (inode->i_visited) { + features->hard_links++; + } else { + inode_tally_features(inode, features); + inode->i_visited = 1; } } -static void inode_find_streams_for_extraction(struct inode *inode, - struct list_head *stream_list, - int extract_flags) +/* Tally the features necessary to extract the specified dentries. */ +static void +dentry_list_get_features(struct list_head *dentry_list, + struct wim_features *features) { - struct lookup_table_entry *lte; - bool inode_added = false; + struct wim_dentry *dentry; - lte = inode_unnamed_lte_resolved(inode); + list_for_each_entry(dentry, dentry_list, d_extraction_list_node) + dentry_tally_features(dentry, features); - if (lte) { - maybe_add_stream_for_extraction(lte, stream_list); - list_add_tail(&inode->lte_inode_list, <e->inode_list); - inode_added = true; + list_for_each_entry(dentry, dentry_list, d_extraction_list_node) + 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) +{ + /* 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); } -#ifdef WITH_NTFS_3G - if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) { - for (unsigned i = 0; i < inode->num_ads; i++) { - if (inode->ads_entries[i].stream_name_len != 0) { - lte = inode->ads_entries[i].lte; - if (lte) { - maybe_add_stream_for_extraction(lte, - stream_list); - if (!inode_added) { - list_add_tail(&inode->lte_inode_list, - <e->inode_list); - inode_added = true; - } - } + + /* 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)) + WARNING("Ignoring named data streams of %lu files", + required_features->named_data_streams); + + /* Hard links. */ + if (required_features->hard_links && !supported_features->hard_links) + WARNING("Extracting %lu hard links as independent files", + required_features->hard_links); + + /* Symbolic links and reparse points. */ + if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS) && + required_features->symlink_reparse_points && + !supported_features->symlink_reparse_points && + !supported_features->reparse_points) + { + ERROR("Extraction backend does not support symbolic links!"); + return WIMLIB_ERR_UNSUPPORTED; + } + if (required_features->reparse_points && + !supported_features->reparse_points) + { + if (supported_features->symlink_reparse_points) { + if (required_features->other_reparse_points) { + WARNING("Ignoring %lu non-symlink/junction " + "reparse point files", + required_features->other_reparse_points); } + } else { + WARNING("Ignoring %lu reparse point files", + required_features->reparse_points); } } + + /* 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) + { + 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("Extraction backend does not support UNIX data!"); + return WIMLIB_ERR_UNSUPPORTED; + } + + /* DOS Names. */ + if (required_features->short_names && + !supported_features->short_names) + { + 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) && + !supported_features->timestamps) + { + ERROR("Extraction backend does not support timestamps!"); + return WIMLIB_ERR_UNSUPPORTED; + } + + return 0; +} + +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 } -static void find_streams_for_extraction(struct hlist_head *inode_list, - struct list_head *stream_list, - struct lookup_table *lookup_table, - int extract_flags) +static int +extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees, + const tchar *target, int extract_flags) { - struct inode *inode; - struct hlist_node *cur; - struct dentry *dentry; + const struct apply_operations *ops; + struct apply_ctx *ctx; + int ret; + LIST_HEAD(dentry_list); - for_lookup_table_entry(lookup_table, lte_zero_out_refcnt, NULL); - INIT_LIST_HEAD(stream_list); - hlist_for_each_entry(inode, cur, inode_list, hlist) { - if (!inode->resolved) - inode_resolve_ltes(inode, lookup_table); - inode_for_each_dentry(dentry, inode) - dentry->is_extracted = 0; - inode_find_streams_for_extraction(inode, stream_list, - extract_flags); + if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) { + ret = extract_dentries_to_stdout(trees, num_trees, + wim->lookup_table); + goto out; } -} -struct apply_operations { - int (*apply_dentry)(struct dentry *dentry, void *arg); - int (*apply_dentry_timestamps)(struct dentry *dentry, void *arg); -}; + num_trees = remove_duplicate_trees(trees, num_trees); + num_trees = remove_contained_trees(trees, num_trees); -static const struct apply_operations normal_apply_operations = { - .apply_dentry = apply_dentry_normal, - .apply_dentry_timestamps = apply_dentry_timestamps_normal, -}; + ops = select_apply_operations(extract_flags); -#ifdef WITH_NTFS_3G -static const struct apply_operations ntfs_apply_operations = { - .apply_dentry = apply_dentry_ntfs, - .apply_dentry_timestamps = apply_dentry_timestamps_ntfs, -}; -#endif + if (num_trees > 1 && ops->single_tree_only) { + ERROR("Extracting multiple directory trees " + "at once is not supported in %s extraction mode!", + ops->name); + ret = WIMLIB_ERR_UNSUPPORTED; + goto out; + } -static int apply_stream_list(struct list_head *stream_list, - struct apply_args *args, - const struct apply_operations *ops, - wimlib_progress_func_t progress_func) -{ - uint64_t bytes_per_progress = args->progress.extract.total_bytes / 100; - uint64_t next_progress = bytes_per_progress; - struct lookup_table_entry *lte; - struct inode *inode; - struct dentry *dentry; - int ret = 0; - - /* This complicated loop is actually just looping through the dentries - * (as for_dentry_in_tree() does), but the outer loop is actually over - * the distinct streams to be extracted so that sequential reading of - * the WIM can be implemented. */ - - /* For each distinct stream to be extracted */ - list_for_each_entry(lte, stream_list, staging_list) { - /* For each inode that contains the stream */ - list_for_each_entry(inode, <e->inode_list, lte_inode_list) { - /* For each dentry that points to the inode */ - inode_for_each_dentry(dentry, inode) { - ret = ops->apply_dentry(dentry, args); - if (ret != 0) - goto out; - if (args->progress.extract.completed_bytes >= next_progress - && args->progress.extract.total_bytes != 0) - { - progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS, - &args->progress); - next_progress += bytes_per_progress; - } - } + ctx = CALLOC(1, ops->context_size); + if (!ctx) { + ret = WIMLIB_ERR_NOMEM; + goto out; + } + + ctx->wim = wim; + ctx->target = target; + ctx->target_nchars = tstrlen(target); + ctx->extract_flags = extract_flags; + if (ctx->wim->progfunc) { + ctx->progfunc = ctx->wim->progfunc; + ctx->progctx = ctx->wim->progctx; + ctx->progress.extract.image = wim->current_image; + ctx->progress.extract.extract_flags = (extract_flags & + WIMLIB_EXTRACT_MASK_PUBLIC); + ctx->progress.extract.wimfile_name = wim->filename; + ctx->progress.extract.image_name = wimlib_get_image_name(wim, + wim->current_image); + ctx->progress.extract.target = target; + } + INIT_LIST_HEAD(&ctx->stream_list); + + ret = (*ops->get_supported_features)(target, &ctx->supported_features); + if (ret) + goto out_cleanup; + + build_dentry_list(&dentry_list, trees, num_trees, + !(extract_flags & + WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE)); + + dentry_list_get_features(&dentry_list, &ctx->required_features); + + ret = do_feature_check(&ctx->required_features, &ctx->supported_features, + ctx->extract_flags); + if (ret) + goto out_cleanup; + + ret = dentry_list_calculate_extraction_names(&dentry_list, ctx); + if (ret) + goto out_cleanup; + + ret = dentry_list_resolve_streams(&dentry_list, ctx); + if (ret) + goto out_cleanup; + + ret = dentry_list_ref_streams(&dentry_list, ctx); + if (ret) + goto out_cleanup; + + dentry_list_build_inode_alias_lists(&dentry_list); + + if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) { + /* When extracting from a pipe, the number of bytes of data to + * extract can't be determined in the normal way (examining the + * lookup table), since at this point all we have is a set of + * SHA1 message digests of streams that need to be extracted. + * However, we can get a reasonably accurate estimate by taking + * from the corresponding in the WIM XML + * data. This does assume that a full image is being extracted, + * but currently there is no API for doing otherwise. (Also, + * subtract from this if hard links are + * supported by the extraction mode.) */ + ctx->progress.extract.total_bytes = + wim_info_get_image_total_bytes(wim->wim_info, + wim->current_image); + if (ctx->supported_features.hard_links) { + ctx->progress.extract.total_bytes -= + wim_info_get_image_hard_link_bytes(wim->wim_info, + wim->current_image); } } + + ret = extract_progress(ctx, + ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ? + WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN : + WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN)); + if (ret) + goto out_cleanup; + + ret = (*ops->extract)(&dentry_list, ctx); + if (ret) + goto out_cleanup; + + if (ctx->progress.extract.completed_bytes < + ctx->progress.extract.total_bytes) + { + ctx->progress.extract.completed_bytes = + ctx->progress.extract.total_bytes; + ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS); + if (ret) + goto out_cleanup; + } + + ret = extract_progress(ctx, + ((extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) ? + WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END : + WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END)); +out_cleanup: + destroy_stream_list(&ctx->stream_list); + destroy_dentry_list(&dentry_list); + FREE(ctx); out: return ret; } -static int extract_single_image(WIMStruct *w, int image, - const char *target, int extract_flags, - wimlib_progress_func_t progress_func) +static int +mkdir_if_needed(const tchar *target) { - int ret; - struct list_head stream_list; - struct hlist_head *inode_list; + 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; +} - struct apply_args args; - const struct apply_operations *ops; +/* 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; - args.w = w; - args.target = target; - args.extract_flags = extract_flags; - args.num_lutimes_warnings = 0; - args.target = target; - args.stream_list = &stream_list; - args.progress_func = progress_func; + /* Check for invalid flag combinations */ - if (progress_func) { - args.progress.extract.image = image; - args.progress.extract.image_name = wimlib_get_image_name(w, image); - args.progress.extract.target = target; - } + if ((extract_flags & + (WIMLIB_EXTRACT_FLAG_NO_ACLS | + WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) == (WIMLIB_EXTRACT_FLAG_NO_ACLS | + WIMLIB_EXTRACT_FLAG_STRICT_ACLS)) + return WIMLIB_ERR_INVALID_PARAM; -#ifdef WITH_NTFS_3G + if ((extract_flags & + (WIMLIB_EXTRACT_FLAG_RPFIX | + WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX | + WIMLIB_EXTRACT_FLAG_NORPFIX)) + return WIMLIB_ERR_INVALID_PARAM; + +#ifndef WITH_NTFS_3G if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) { - args.vol = ntfs_mount(target, 0); - if (!args.vol) { - ERROR_WITH_ERRNO("Failed to mount NTFS volume `%s'", target); - return WIMLIB_ERR_NTFS_3G; - } - ops = &ntfs_apply_operations; - } else + ERROR("wimlib was compiled without support for NTFS-3g, so\n" + " it cannot apply a WIM image directly to an NTFS volume."); + return WIMLIB_ERR_UNSUPPORTED; + } #endif - ops = &normal_apply_operations; - ret = select_wim_image(w, image); - if (ret != 0) - goto out; +#ifndef __WIN32__ + if (extract_flags & WIMLIB_EXTRACT_FLAG_WIMBOOT) { + ERROR("WIMBoot extraction is only supported on Windows!"); + return WIMLIB_ERR_UNSUPPORTED; + } +#endif - inode_list = &w->image_metadata[image - 1].inode_list; + if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX | + WIMLIB_EXTRACT_FLAG_NORPFIX | + WIMLIB_EXTRACT_FLAG_IMAGEMODE)) == + WIMLIB_EXTRACT_FLAG_IMAGEMODE) + { + /* 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; + } - find_streams_for_extraction(inode_list, &stream_list, - w->lookup_table, extract_flags); + *extract_flags_p = extract_flags; + return 0; +} - calculate_bytes_to_extract(&stream_list, extract_flags, - &args.progress); +static u32 +get_wildcard_flags(int extract_flags) +{ + u32 wildcard_flags = 0; - if (progress_func) { - progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN, - &args.progress); - } + 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 (extract_flags & WIMLIB_EXTRACT_FLAG_SEQUENTIAL) { - ret = sort_stream_list_by_wim_position(&stream_list); - if (ret != 0) { - WARNING("Falling back to non-sequential extraction"); - extract_flags &= ~WIMLIB_EXTRACT_FLAG_SEQUENTIAL; - } - } + if (default_ignore_case) + wildcard_flags |= WILDCARD_FLAG_CASE_INSENSITIVE; - if (progress_func) { - progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN, - &args.progress); - } + return wildcard_flags; +} - args.extract_flags |= WIMLIB_EXTRACT_FLAG_NO_STREAMS; - ret = for_dentry_in_tree(wim_root_dentry(w), ops->apply_dentry, &args); - args.extract_flags &= ~WIMLIB_EXTRACT_FLAG_NO_STREAMS; - if (ret != 0) - goto out; +struct append_dentry_ctx { + struct wim_dentry **dentries; + size_t num_dentries; + size_t num_alloc_dentries; +}; - if (progress_func) { - progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END, - &args.progress); +static int +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; +} - ret = apply_stream_list(&stream_list, &args, ops, progress_func); - if (ret != 0) - goto out; +static int +do_wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target, + const tchar * const *paths, size_t num_paths, + int extract_flags) +{ + int ret; + struct wim_dentry **trees; + size_t num_trees; - if (progress_func) - progress_func(WIMLIB_PROGRESS_MSG_APPLY_TIMESTAMPS, NULL); + if (wim == NULL || target == NULL || target[0] == T('\0') || + (num_paths != 0 && paths == NULL)) + return WIMLIB_ERR_INVALID_PARAM; - ret = for_dentry_in_tree_depth(wim_root_dentry(w), - ops->apply_dentry_timestamps, &args); - if (ret != 0) - goto out; + ret = check_extract_flags(wim, &extract_flags); + if (ret) + return ret; + + ret = select_wim_image(wim, image); + if (ret) + return ret; - if (progress_func) { - progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END, - &args.progress); + ret = wim_checksum_unhashed_streams(wim); + if (ret) + return ret; + + if ((extract_flags & (WIMLIB_EXTRACT_FLAG_NTFS | + WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE)) == + (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE)) + { + ret = mkdir_if_needed(target); + if (ret) + return ret; } -out: -#ifdef WITH_NTFS_3G - if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) { - if (ntfs_umount(args.vol, FALSE) != 0) { - ERROR_WITH_ERRNO("Failed to unmount NTFS volume `%s'", args.target); - if (ret == 0) - ret = WIMLIB_ERR_NTFS_3G; + + if (extract_flags & WIMLIB_EXTRACT_FLAG_GLOB_PATHS) { + + struct append_dentry_ctx append_dentry_ctx = { + .dentries = NULL, + .num_dentries = 0, + .num_alloc_dentries = 0, + }; + + u32 wildcard_flags = get_wildcard_flags(extract_flags); + + 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_paths; i++) { + + tchar *path = canonicalize_wim_path(paths[i]); + if (path == NULL) { + ret = WIMLIB_ERR_NOMEM; + goto out_free_trees; + } + + 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; } -#endif + + if (num_trees == 0) { + ret = 0; + goto out_free_trees; + } + + ret = extract_trees(wim, trees, num_trees, target, extract_flags); +out_free_trees: + FREE(trees); return ret; } +static int +extract_single_image(WIMStruct *wim, int image, + const tchar *target, int extract_flags) +{ + const tchar *path = WIMLIB_WIM_ROOT_PATH; + extract_flags |= WIMLIB_EXTRACT_FLAG_IMAGEMODE; + return do_wimlib_extract_paths(wim, image, target, &path, 1, extract_flags); +} -/* Extracts all images from the WIM to @output_dir, with the images placed in - * subdirectories named by their image names. */ -static int extract_all_images(WIMStruct *w, const char *target, - int extract_flags, - wimlib_progress_func_t progress_func) +static const tchar * const filename_forbidden_chars = +T( +#ifdef __WIN32__ +"<>:\"/\\|?*" +#else +"/" +#endif +); + +/* This function checks if it is okay to use a WIM image's name as a directory + * name. */ +static bool +image_name_ok_as_dir(const tchar *image_name) +{ + return image_name && *image_name && + !tstrpbrk(image_name, filename_forbidden_chars) && + tstrcmp(image_name, T(".")) && + tstrcmp(image_name, T("..")); +} + +/* Extracts all images from the WIM to the directory @target, with the images + * placed in subdirectories named by their image names. */ +static int +extract_all_images(WIMStruct *wim, const tchar *target, int extract_flags) { - size_t image_name_max_len = max(xml_get_max_image_name_len(w), 20); - size_t output_path_len = strlen(target); - char buf[output_path_len + 1 + image_name_max_len + 1]; + size_t image_name_max_len = max(xml_get_max_image_name_len(wim), 20); + size_t output_path_len = tstrlen(target); + tchar buf[output_path_len + 1 + image_name_max_len + 1]; int ret; int image; - const char *image_name; + const tchar *image_name; - ret = extract_directory(target, true); - if (ret != 0) - return ret; + extract_flags |= WIMLIB_EXTRACT_FLAG_MULTI_IMAGE; + + if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) { + ERROR("Cannot extract multiple images in NTFS extraction mode."); + return WIMLIB_ERR_INVALID_PARAM; + } - memcpy(buf, target, output_path_len); - buf[output_path_len] = '/'; - for (image = 1; image <= w->hdr.image_count; image++) { - image_name = wimlib_get_image_name(w, image); - if (image_name && *image_name) { - strcpy(buf + output_path_len + 1, image_name); + 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++) { + image_name = wimlib_get_image_name(wim, image); + if (image_name_ok_as_dir(image_name)) { + tstrcpy(buf + output_path_len + 1, image_name); } else { - /* Image name is empty. Use image number instead */ - sprintf(buf + output_path_len + 1, "%d", image); + /* Image name is empty or contains forbidden characters. + * Use image number instead. */ + tsprintf(buf + output_path_len + 1, T("%d"), image); } - ret = extract_single_image(w, image, buf, extract_flags, - progress_func); - if (ret != 0) + ret = extract_single_image(wim, image, buf, extract_flags); + if (ret) return ret; } return 0; } -/* Extracts a single image or all images from a WIM file. */ -WIMLIBAPI int wimlib_extract_image(WIMStruct *w, int image, - const char *target, - int extract_flags, - WIMStruct **additional_swms, - unsigned num_additional_swms, - wimlib_progress_func_t progress_func) +static int +do_wimlib_extract_image(WIMStruct *wim, int image, const tchar *target, + int extract_flags) { - struct lookup_table *joined_tab, *w_tab_save; - int ret; - - if (!w || !target) + 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; - extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC; + if (image == WIMLIB_ALL_IMAGES) + return extract_all_images(wim, target, extract_flags); + else + return extract_single_image(wim, image, target, extract_flags); +} - if ((extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK | WIMLIB_EXTRACT_FLAG_HARDLINK)) - == (WIMLIB_EXTRACT_FLAG_SYMLINK | WIMLIB_EXTRACT_FLAG_HARDLINK)) + +/**************************************************************************** + * Extraction API * + ****************************************************************************/ + +WIMLIBAPI int +wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target, + const tchar * const *paths, size_t num_paths, + int extract_flags) +{ + if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC) return WIMLIB_ERR_INVALID_PARAM; - if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) { -#ifdef WITH_NTFS_3G - if ((extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK | WIMLIB_EXTRACT_FLAG_HARDLINK))) { - ERROR("Cannot specify symlink or hardlink flags when applying\n" - " directly to a NTFS volume"); - return WIMLIB_ERR_INVALID_PARAM; - } - if (image == WIMLIB_ALL_IMAGES) { - ERROR("Can only apply a single image when applying " - "directly to a NTFS volume"); - return WIMLIB_ERR_INVALID_PARAM; - } -#else - ERROR("wimlib was compiled without support for NTFS-3g, so"); - ERROR("we cannot apply a WIM image directly to a NTFS volume"); - return WIMLIB_ERR_UNSUPPORTED; -#endif + return do_wimlib_extract_paths(wim, image, target, paths, num_paths, + extract_flags); +} + +WIMLIBAPI int +wimlib_extract_pathlist(WIMStruct *wim, int image, const tchar *target, + const tchar *path_list_file, int extract_flags) +{ + 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; } - ret = verify_swm_set(w, additional_swms, num_additional_swms); - if (ret != 0) + ret = wimlib_extract_paths(wim, image, target, + (const tchar * const *)paths, num_paths, + extract_flags); + FREE(paths); + FREE(mem); + return ret; +} + +WIMLIBAPI int +wimlib_extract_image_from_pipe_with_progress(int pipe_fd, + const tchar *image_num_or_name, + const tchar *target, + int extract_flags, + wimlib_progress_func_t progfunc, + void *progctx) +{ + int ret; + WIMStruct *pwm; + struct filedes *in_fd; + int image; + unsigned i; + + if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC) + 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, &pwm, + progfunc, progctx); + if (ret) return ret; - if (num_additional_swms) { - ret = new_joined_lookup_table(w, additional_swms, - num_additional_swms, &joined_tab); - if (ret != 0) - return ret; - w_tab_save = w->lookup_table; - w->lookup_table = joined_tab; + /* Sanity check to make sure this is a pipable WIM. */ + if (pwm->hdr.magic != PWM_MAGIC) { + ERROR("The WIM being read from file descriptor %d " + "is not pipable!", pipe_fd); + ret = WIMLIB_ERR_NOT_PIPABLE; + goto out_wimlib_free; } - if (image == WIMLIB_ALL_IMAGES) { - extract_flags |= WIMLIB_EXTRACT_FLAG_MULTI_IMAGE; - ret = extract_all_images(w, target, extract_flags, - progress_func); - } else { - extract_flags &= ~WIMLIB_EXTRACT_FLAG_MULTI_IMAGE; - ret = extract_single_image(w, image, target, extract_flags, - progress_func); + /* Sanity check to make sure the first part of a pipable split WIM is + * sent over the pipe first. */ + if (pwm->hdr.part_number != 1) { + ERROR("The first part of the split WIM must be " + "sent over the pipe first."); + ret = WIMLIB_ERR_INVALID_PIPABLE_WIM; + goto out_wimlib_free; } - if (extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK | - WIMLIB_EXTRACT_FLAG_HARDLINK)) + in_fd = &pwm->in_fd; + wimlib_assert(in_fd->offset == WIM_HEADER_DISK_SIZE); + + /* As mentioned, the WIMStruct we created from the pipe does not have + * XML data yet. Fix this by reading the extra copy of the XML data + * that directly follows the header in pipable WIMs. (Note: see + * write_pipable_wim() for more details about the format of pipable + * WIMs.) */ { - for_lookup_table_entry(w->lookup_table, - lte_free_extracted_file, - NULL); + struct wim_lookup_table_entry xml_lte; + 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.flags & WIM_RESHDR_FLAG_METADATA)) + { + ERROR("Expected XML data, but found non-metadata " + "stream."); + ret = WIMLIB_ERR_INVALID_PIPABLE_WIM; + goto out_wimlib_free; + } + + 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_IMAGE_COUNT; + goto out_wimlib_free; + } + } + + /* Get image index (this may use the XML data that was just read to + * resolve an image name). */ + if (image_num_or_name) { + image = wimlib_resolve_image(pwm, image_num_or_name); + if (image == WIMLIB_NO_IMAGE) { + ERROR("\"%"TS"\" is not a valid image in the pipable WIM!", + 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!"); + ret = WIMLIB_ERR_INVALID_IMAGE; + goto out_wimlib_free; + } + } else { + if (pwm->hdr.image_count != 1) { + ERROR("No image was specified, but the pipable WIM " + "did not contain exactly 1 image"); + ret = WIMLIB_ERR_INVALID_IMAGE; + goto out_wimlib_free; + } + image = 1; } - if (num_additional_swms) { - free_lookup_table(w->lookup_table); - w->lookup_table = w_tab_save; + /* Load the needed metadata resource. */ + 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 == 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, metadata_rspec, 0, NULL); + imd = pwm->image_metadata[i - 1]; + imd->metadata_lte = metadata_lte; + if (ret) { + FREE(metadata_rspec); + goto out_wimlib_free; + } + + if (!(metadata_lte->flags & WIM_RESHDR_FLAG_METADATA)) { + ERROR("Expected metadata resource, but found " + "non-metadata stream."); + ret = WIMLIB_ERR_INVALID_PIPABLE_WIM; + goto out_wimlib_free; + } + + if (i == image) { + /* Metadata resource is for the image being extracted. + * Parse it and save the metadata in memory. */ + ret = read_metadata_resource(pwm, imd); + if (ret) + goto out_wimlib_free; + imd->modified = 1; + } else { + /* Metadata resource is not for the image being + * extracted. Skip over it. */ + ret = skip_wim_stream(metadata_lte); + if (ret) + goto out_wimlib_free; + } } + /* Extract the image. */ + extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE; + ret = do_wimlib_extract_image(pwm, image, target, extract_flags); + /* Clean up and return. */ +out_wimlib_free: + wimlib_free(pwm); return ret; } + + +WIMLIBAPI int +wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name, + const tchar *target, int extract_flags) +{ + return wimlib_extract_image_from_pipe_with_progress(pipe_fd, + image_num_or_name, + target, + extract_flags, + NULL, + NULL); +} + +WIMLIBAPI int +wimlib_extract_image(WIMStruct *wim, int image, const tchar *target, + int extract_flags) +{ + if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC) + return WIMLIB_ERR_INVALID_PARAM; + return do_wimlib_extract_image(wim, image, target, extract_flags); +}