+ if (!build_extraction_path(path, dentry, ctx))
+ return 0;
+
+ return extract_streams(path, ctx, dentry, NULL, NULL);
+}
+
+/* Finish extracting a file, directory, or symbolic link by setting file
+ * security and timestamps. */
+static int
+dentry_extract_final(struct wim_dentry *dentry, struct apply_ctx *ctx)
+{
+ int ret;
+ tchar path[ctx->ops->path_max];
+
+ if (!build_extraction_path(path, dentry, ctx))
+ return 0;
+
+ ret = extract_security(path, ctx, dentry);
+ if (ret)
+ return ret;
+
+ if (ctx->ops->requires_final_set_attributes_pass) {
+ /* Set file attributes (if supported). */
+ ret = extract_file_attributes(path, ctx, dentry, 1);
+ if (ret)
+ return ret;
+ }
+
+ return extract_timestamps(path, ctx, dentry);
+}
+
+static int
+extract_structure(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_skeleton(dentry, ctx);
+ if (ret)
+ return ret;
+ }
+ return 0;
+}
+
+static int
+extract_dir_structure(struct list_head *dentry_list, struct apply_ctx *ctx)
+{
+ struct wim_dentry *dentry;
+ int ret;
+
+ 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;
+}
+
+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)
+ return ret;
+ }
+ return 0;
+}
+
+static int
+extract_final_metadata(struct list_head *dentry_list, struct apply_ctx *ctx)
+{
+ struct wim_dentry *dentry;
+ int ret;
+
+ list_for_each_entry_reverse(dentry, dentry_list, extraction_list) {
+ ret = dentry_extract_final(dentry, ctx);
+ if (ret)
+ return ret;
+ }
+ return 0;
+}
+
+/* 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
+create_temporary_file(struct filedes *fd_ret, tchar **name_ret)
+{
+ 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_stream_instances(struct wim_lookup_table_entry *lte,
+ struct wim_lookup_table_entry *lte_override,
+ struct apply_ctx *ctx)
+{
+ 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 == 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))
+ pwm_flags |= PWM_SILENT_EOF;
+ memcpy(ctx->progress.extract.guid, ctx->wim->hdr.guid, WIM_GID_LEN);
+ ctx->progress.extract.part_number = ctx->wim->hdr.part_number;
+ ctx->progress.extract.total_parts = ctx->wim->hdr.total_parts;
+ if (ctx->progress_func)
+ ctx->progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN,
+ &ctx->progress);
+ while (ctx->num_streams_remaining) {
+ 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))
+ {
+ goto resume_done;
+ }
+ goto out_free_found_lte;
+ }
+
+ 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))
+ {
+ 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_wim_stream(found_lte);
+ if (ret)
+ goto out_free_found_lte;
+ } 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_GID_LEN))
+ {
+ ctx->progress.extract.part_number = part_number;
+ ctx->progress.extract.total_parts = total_parts;
+ memcpy(ctx->progress.extract.guid,
+ pwm_hdr.guid, WIM_GID_LEN);
+ if (ctx->progress_func) {
+ ctx->progress_func(
+ WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN,
+ &ctx->progress);
+ }
+
+ }
+ }
+ }
+ 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;
+
+resume_done:
+ /* TODO */
+ return 0;
+}
+
+/* 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 (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;
+ }
+
+ filedes_init(&_stdout, STDOUT_FILENO);
+ return extract_full_stream_to_fd(lte, &_stdout);
+}
+
+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;
+}
+
+/**********************************************************************/
+
+/*
+ * 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];
+ }
+ }
+ 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
+dentry_append_to_list(struct wim_dentry *dentry, void *_dentry_list)
+{
+ 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));
+ }
+ }
+}
+
+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__
+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;
+
+ if (dentry == ctx->target_dentry)
+ return 0;
+
+ if (!dentry_is_supported(dentry, &ctx->supported_features))
+ goto skip_dentry;
+
+ 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 (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;
+ }
+ }
+ }
+ }
+
+ if (file_name_valid(dentry->file_name, dentry->file_name_nbytes / 2, false)) {
+#if TCHAR_IS_UTF16LE
+ dentry->extraction_name = dentry->file_name;
+ dentry->extraction_name_nchars = dentry->file_name_nbytes / 2;
+ return 0;
+#else
+ return utf16le_to_tstr(dentry->file_name,
+ dentry->file_name_nbytes,
+ &dentry->extraction_name,
+ &dentry->extraction_name_nchars);
+#endif
+ } else {
+ 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;
+ }
+ }
+
+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);
+
+ tchar *tchar_name;
+ size_t tchar_nchars;
+ #if TCHAR_IS_UTF16LE
+ tchar_name = utf16_name_copy;
+ tchar_nchars = dentry->file_name_nbytes / 2;
+ #else
+ ret = utf16le_to_tstr(utf16_name_copy,
+ dentry->file_name_nbytes,
+ &tchar_name, &tchar_nchars);
+ if (ret)
+ return ret;
+ #endif
+ 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);
+ #if !TCHAR_IS_UTF16LE
+ FREE(tchar_name);
+ #endif
+ dentry->extraction_name = memdup(fixed_name,
+ 2 * fixed_name_num_chars + 2);
+ if (!dentry->extraction_name)
+ return WIMLIB_ERR_NOMEM;
+ dentry->extraction_name_nchars = fixed_name_num_chars;
+ }
+ return 0;
+
+skip_dentry:
+ for_dentry_in_tree(dentry, dentry_delete_from_list, NULL);
+ return 0;
+}
+
+/*
+ * 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;
+
+ /* 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;
+}
+
+/*
+ * 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_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_inode *inode = dentry->d_inode;
+
+ 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_visited)
+ features->hard_links++;
+ 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++;
+ }
+ if (inode->i_security_id != -1)
+ features->security_descriptors++;
+ if (dentry->short_name_nbytes)
+ features->short_names++;
+ if (inode_has_unix_data(inode))
+ features->unix_data++;
+ inode->i_visited = 1;
+}
+
+/* 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 wim_dentry *dentry;
+
+ memset(features, 0, sizeof(struct wim_features));
+
+ 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 u32
+compute_supported_attributes_mask(const struct wim_features *supported_features)
+{
+ u32 mask = (u32)~0UL;
+
+ if (!supported_features->archive_files)
+ mask &= ~FILE_ATTRIBUTE_ARCHIVE;
+
+ if (!supported_features->hidden_files)
+ mask &= ~FILE_ATTRIBUTE_HIDDEN;
+
+ if (!supported_features->system_files)
+ mask &= ~FILE_ATTRIBUTE_SYSTEM;
+
+ if (!supported_features->not_context_indexed_files)
+ mask &= ~FILE_ATTRIBUTE_NOT_CONTENT_INDEXED;
+
+ if (!supported_features->compressed_files)
+ mask &= ~FILE_ATTRIBUTE_COMPRESSED;
+
+ if (!supported_features->sparse_files)
+ mask &= ~FILE_ATTRIBUTE_SPARSE_FILE;
+
+ if (!supported_features->reparse_points)
+ mask &= ~FILE_ATTRIBUTE_REPARSE_POINT;
+
+ 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 ||
+ (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)
+ {
+ 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);
+
+ /* Symbolic links and reparse points. */
+ if ((extract_flags & WIMLIB_EXTRACT_FLAG_SYMLINK) &&
+ !supported_features->symlink_reparse_points)
+ {
+ ERROR("Extraction backend does not support symbolic links!");
+ return WIMLIB_ERR_UNSUPPORTED;
+ }
+ 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) &&
+ !ops->set_timestamps)
+ {
+ ERROR("Extraction backend does not support timestamps!");
+ return WIMLIB_ERR_UNSUPPORTED;
+ }
+
+ return 0;
+}