+ * This file is distributed in the hope that it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS
+ * FOR A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more
+ * details.
+ *
+ * You should have received a copy of the GNU Lesser General Public License
+ * along with this file; 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 <errno.h>
+#include <fcntl.h>
+#include <sys/stat.h>
+#include <unistd.h>
+
+#include "wimlib/apply.h"
+#include "wimlib/assert.h"
+#include "wimlib/blob_table.h"
+#include "wimlib/dentry.h"
+#include "wimlib/encoding.h"
+#include "wimlib/endianness.h"
+#include "wimlib/error.h"
+#include "wimlib/metadata.h"
+#include "wimlib/pathlist.h"
+#include "wimlib/paths.h"
+#include "wimlib/pattern.h"
+#include "wimlib/reparse.h"
+#include "wimlib/resource.h"
+#include "wimlib/security.h"
+#include "wimlib/unix_data.h"
+#include "wimlib/wim.h"
+#include "wimlib/win32.h" /* for realpath() equivalent */
+#include "wimlib/xml.h"
+
+#define WIMLIB_EXTRACT_FLAG_FROM_PIPE 0x80000000
+#define WIMLIB_EXTRACT_FLAG_IMAGEMODE 0x40000000
+
+/* 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 | \
+ WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS4K | \
+ WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS8K | \
+ WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS16K | \
+ WIMLIB_EXTRACT_FLAG_COMPACT_LZX \
+ )
+
+/* Send WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE or
+ * WIMLIB_PROGRESS_MSG_EXTRACT_METADATA. */
+int
+do_file_extract_progress(struct apply_ctx *ctx, enum wimlib_progress_msg msg)
+{
+ ctx->count_until_file_progress = 500; /* Arbitrary value to limit calls */
+ return extract_progress(ctx, msg);
+}
+
+static int
+start_file_phase(struct apply_ctx *ctx, u64 end_file_count, enum wimlib_progress_msg msg)
+{
+ ctx->progress.extract.current_file_count = 0;
+ ctx->progress.extract.end_file_count = end_file_count;
+ return do_file_extract_progress(ctx, msg);
+}
+
+int
+start_file_structure_phase(struct apply_ctx *ctx, u64 end_file_count)
+{
+ return start_file_phase(ctx, end_file_count, WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE);
+}
+
+int
+start_file_metadata_phase(struct apply_ctx *ctx, u64 end_file_count)
+{
+ return start_file_phase(ctx, end_file_count, WIMLIB_PROGRESS_MSG_EXTRACT_METADATA);
+}
+
+static int
+end_file_phase(struct apply_ctx *ctx, enum wimlib_progress_msg msg)
+{
+ ctx->progress.extract.current_file_count = ctx->progress.extract.end_file_count;
+ return do_file_extract_progress(ctx, msg);
+}
+
+int
+end_file_structure_phase(struct apply_ctx *ctx)
+{
+ return end_file_phase(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE);
+}
+
+int
+end_file_metadata_phase(struct apply_ctx *ctx)
+{
+ return end_file_phase(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_METADATA);
+}
+
+#define PWM_FOUND_WIM_HDR (-1)
+
+/* Read the header for a blob in a pipable WIM. If @pwm_hdr_ret is not NULL,
+ * also look for a pipable WIM header and return PWM_FOUND_WIM_HDR if found. */
+static int
+read_pwm_blob_header(WIMStruct *pwm, u8 hash_ret[SHA1_HASH_SIZE],
+ struct wim_reshdr *reshdr_ret,
+ struct wim_header_disk *pwm_hdr_ret)
+{
+ int ret;
+ struct pwm_blob_hdr blob_hdr;
+ u64 magic;
+
+ ret = full_read(&pwm->in_fd, &blob_hdr, sizeof(blob_hdr));
+ if (unlikely(ret))
+ goto read_error;
+
+ magic = le64_to_cpu(blob_hdr.magic);
+
+ if (magic == PWM_MAGIC && pwm_hdr_ret != NULL) {
+ memcpy(pwm_hdr_ret, &blob_hdr, sizeof(blob_hdr));
+ ret = full_read(&pwm->in_fd,
+ (u8 *)pwm_hdr_ret + sizeof(blob_hdr),
+ sizeof(*pwm_hdr_ret) - sizeof(blob_hdr));
+ if (unlikely(ret))
+ goto read_error;
+ return PWM_FOUND_WIM_HDR;
+ }
+
+ if (unlikely(magic != PWM_BLOB_MAGIC)) {
+ ERROR("Data read on pipe is invalid (expected blob header)");
+ return WIMLIB_ERR_INVALID_PIPABLE_WIM;
+ }
+
+ copy_hash(hash_ret, blob_hdr.hash);
+
+ reshdr_ret->size_in_wim = 0; /* Not available */
+ reshdr_ret->flags = le32_to_cpu(blob_hdr.flags);
+ reshdr_ret->offset_in_wim = pwm->in_fd.offset;
+ reshdr_ret->uncompressed_size = le64_to_cpu(blob_hdr.uncompressed_size);
+
+ if (unlikely(reshdr_ret->uncompressed_size == 0)) {
+ ERROR("Data read on pipe is invalid (resource is of 0 size)");
+ return WIMLIB_ERR_INVALID_PIPABLE_WIM;
+ }
+
+ return 0;
+
+read_error:
+ if (ret == WIMLIB_ERR_UNEXPECTED_END_OF_FILE)
+ ERROR("The pipe ended before all needed data was sent!");
+ else
+ ERROR_WITH_ERRNO("Error reading pipable WIM from pipe");
+ return ret;
+}
+
+static int
+read_blobs_from_pipe(struct apply_ctx *ctx, const struct read_blob_callbacks *cbs)
+{
+ int ret;
+ u8 hash[SHA1_HASH_SIZE];
+ struct wim_reshdr reshdr;
+ struct wim_header_disk pwm_hdr;
+ struct wim_resource_descriptor rdesc;
+ struct blob_descriptor *blob;
+
+ copy_guid(ctx->progress.extract.guid, ctx->wim->hdr.guid);
+ 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)
+ return ret;
+
+ while (ctx->num_blobs_remaining) {
+
+ ret = read_pwm_blob_header(ctx->wim, hash, &reshdr, &pwm_hdr);
+
+ if (ret == PWM_FOUND_WIM_HDR) {
+ 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 &&
+ guids_equal(pwm_hdr.guid, ctx->progress.extract.guid))
+ continue;
+
+ copy_guid(ctx->progress.extract.guid, pwm_hdr.guid);
+ ctx->progress.extract.part_number = part_number;
+ ctx->progress.extract.total_parts = total_parts;
+ ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN);
+ if (ret)
+ return ret;
+
+ continue;
+ }
+
+ if (ret)
+ return ret;
+
+ if (!(reshdr.flags & WIM_RESHDR_FLAG_METADATA)
+ && (blob = lookup_blob(ctx->wim->blob_table, hash))
+ && (blob->out_refcnt))
+ {
+ wim_reshdr_to_desc_and_blob(&reshdr, ctx->wim, &rdesc, blob);
+ ret = read_blob_with_sha1(blob, cbs);
+ blob_unset_is_located_in_wim_resource(blob);
+ if (ret)
+ return ret;
+ ctx->num_blobs_remaining--;
+ } else {
+ wim_reshdr_to_desc(&reshdr, ctx->wim, &rdesc);
+ ret = skip_wim_resource(&rdesc);
+ 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 open_flags;
+ int raw_fd;
+
+retry:
+ name = ttempnam(NULL, T("wimlib"));
+ if (!name) {
+ ERROR_WITH_ERRNO("Failed to create temporary filename");
+ return WIMLIB_ERR_NOMEM;
+ }
+
+ open_flags = O_WRONLY | O_CREAT | O_EXCL | O_BINARY;
+#ifdef __WIN32__
+ open_flags |= _O_SHORT_LIVED;
+#endif
+ raw_fd = topen(name, open_flags, 0600);
+
+ if (raw_fd < 0) {
+ if (errno == EEXIST) {
+ FREE(name);
+ goto retry;
+ }
+ ERROR_WITH_ERRNO("Failed to create temporary file "
+ "\"%"TS"\"", name);
+ FREE(name);
+ return WIMLIB_ERR_OPEN;
+ }
+
+ filedes_init(fd_ret, raw_fd);
+ *name_ret = name;
+ return 0;
+}
+
+static int
+begin_extract_blob_wrapper(struct blob_descriptor *blob, void *_ctx)
+{
+ struct apply_ctx *ctx = _ctx;
+
+ ctx->cur_blob = blob;
+ ctx->cur_blob_offset = 0;
+
+ if (unlikely(blob->out_refcnt > MAX_OPEN_FILES))
+ return create_temporary_file(&ctx->tmpfile_fd, &ctx->tmpfile_name);
+
+ return call_begin_blob(blob, ctx->saved_cbs);
+}
+
+static int
+extract_chunk_wrapper(const void *chunk, size_t size, void *_ctx)
+{
+ struct apply_ctx *ctx = _ctx;
+ union wimlib_progress_info *progress = &ctx->progress;
+ int ret;
+
+ ctx->cur_blob_offset += size;
+
+ if (likely(ctx->supported_features.hard_links)) {
+ progress->extract.completed_bytes +=
+ (u64)size * ctx->cur_blob->out_refcnt;
+ if (ctx->cur_blob_offset == ctx->cur_blob->size)
+ progress->extract.completed_streams += ctx->cur_blob->out_refcnt;
+ } else {
+ const struct blob_extraction_target *targets =
+ blob_extraction_targets(ctx->cur_blob);
+ for (u32 i = 0; i < ctx->cur_blob->out_refcnt; i++) {
+ const struct wim_inode *inode = targets[i].inode;
+ const struct wim_dentry *dentry;
+
+ inode_for_each_extraction_alias(dentry, inode) {
+ progress->extract.completed_bytes += size;
+ if (ctx->cur_blob_offset == ctx->cur_blob->size)
+ progress->extract.completed_streams++;
+ }
+ }
+ }
+ if (progress->extract.completed_bytes >= ctx->next_progress) {
+
+ ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS);
+ if (ret)
+ return ret;
+
+ set_next_progress(progress->extract.completed_bytes,
+ progress->extract.total_bytes,
+ &ctx->next_progress);
+ }
+
+ if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
+ /* Just extracting to temporary file for now. */
+ ret = full_write(&ctx->tmpfile_fd, chunk, size);
+ if (ret) {
+ ERROR_WITH_ERRNO("Error writing data to "
+ "temporary file \"%"TS"\"",
+ ctx->tmpfile_name);
+ }
+ return ret;
+ }
+
+ return call_consume_chunk(chunk, size, ctx->saved_cbs);
+}
+
+/* Copy the blob's data from the temporary file to each of its targets.
+ *
+ * This is executed only in the very uncommon case that a blob is being
+ * extracted to more than MAX_OPEN_FILES targets! */
+static int
+extract_from_tmpfile(const tchar *tmpfile_name,
+ const struct blob_descriptor *orig_blob,
+ const struct read_blob_callbacks *cbs)
+{
+ struct blob_descriptor tmpfile_blob;
+ const struct blob_extraction_target *targets = blob_extraction_targets(orig_blob);
+ int ret;
+
+ memcpy(&tmpfile_blob, orig_blob, sizeof(struct blob_descriptor));
+ tmpfile_blob.blob_location = BLOB_IN_FILE_ON_DISK;
+ tmpfile_blob.file_on_disk = (tchar *)tmpfile_name;
+ tmpfile_blob.out_refcnt = 1;
+
+ for (u32 i = 0; i < orig_blob->out_refcnt; i++) {
+ tmpfile_blob.inline_blob_extraction_targets[0] = targets[i];
+ ret = read_blob_with_cbs(&tmpfile_blob, cbs);
+ if (ret)
+ return ret;
+ }
+ return 0;
+}
+
+static int
+end_extract_blob_wrapper(struct blob_descriptor *blob, int status, void *_ctx)
+{
+ struct apply_ctx *ctx = _ctx;
+
+ if (unlikely(filedes_valid(&ctx->tmpfile_fd))) {
+ filedes_close(&ctx->tmpfile_fd);
+ if (!status)
+ status = extract_from_tmpfile(ctx->tmpfile_name, blob,
+ ctx->saved_cbs);
+ filedes_invalidate(&ctx->tmpfile_fd);
+ tunlink(ctx->tmpfile_name);
+ FREE(ctx->tmpfile_name);
+ return status;
+ }
+
+ return call_end_blob(blob, status, ctx->saved_cbs);
+}
+
+/*
+ * Read the list of blobs to extract and feed their data into the specified
+ * callback functions.
+ *
+ * This handles checksumming each blob.
+ *
+ * This also handles sending WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS.
+ *
+ * This also works if the WIM is being read from a pipe.
+ *
+ * This also will split up blobs that will need to be extracted to more than
+ * MAX_OPEN_FILES locations, as measured by the 'out_refcnt' of each blob.
+ * Therefore, the apply_operations implementation need not worry about running
+ * out of file descriptors, unless it might open more than one file descriptor
+ * per 'blob_extraction_target' (e.g. Win32 currently might because the
+ * destination file system might not support hard links).
+ */
+int
+extract_blob_list(struct apply_ctx *ctx, const struct read_blob_callbacks *cbs)
+{
+ struct read_blob_callbacks wrapper_cbs = {
+ .begin_blob = begin_extract_blob_wrapper,
+ .consume_chunk = extract_chunk_wrapper,
+ .end_blob = end_extract_blob_wrapper,
+ .ctx = ctx,
+ };
+ ctx->saved_cbs = cbs;
+ if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) {
+ return read_blobs_from_pipe(ctx, &wrapper_cbs);
+ } else {
+ return read_blob_list(&ctx->blob_list,
+ offsetof(struct blob_descriptor,
+ extraction_list),
+ &wrapper_cbs, VERIFY_BLOB_HASHES);
+ }
+}
+
+/* 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 blob_table *blob_table)
+{
+ struct wim_inode *inode = dentry->d_inode;
+ struct blob_descriptor *blob;
+ struct filedes _stdout;
+
+ if (inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT |
+ FILE_ATTRIBUTE_DIRECTORY |
+ FILE_ATTRIBUTE_ENCRYPTED))
+ {
+ 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;
+ }
+
+ blob = inode_get_blob_for_unnamed_data_stream(inode, blob_table);
+ if (!blob) {
+ const u8 *hash = inode_get_hash_of_unnamed_data_stream(inode);
+ if (!is_zero_hash(hash))
+ return blob_not_found_error(inode, hash);
+ return 0;
+ }
+
+ filedes_init(&_stdout, STDOUT_FILENO);
+ return extract_blob_to_fd(blob, &_stdout);
+}
+
+static int
+extract_dentries_to_stdout(struct wim_dentry **dentries, size_t num_dentries,
+ const struct blob_table *blob_table)
+{
+ for (size_t i = 0; i < num_dentries; i++) {
+ int ret = extract_dentry_to_stdout(dentries[i], blob_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]->d_tmp_flag) {
+ /* Found distinct dentry. */
+ trees[i]->d_tmp_flag = 1;
+ trees[j++] = trees[i];
+ }
+ }
+ for (i = 0; i < j; i++)
+ trees[i]->d_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]->d_tmp_flag = 1;
+ for (i = 0; i < num_trees; i++) {
+ struct wim_dentry *d = trees[i];
+ while (!dentry_is_root(d)) {
+ d = d->d_parent;
+ if (d->d_tmp_flag)
+ goto tree_contained;
+ }
+ trees[j++] = trees[i];
+ continue;
+
+ tree_contained:
+ trees[i]->d_tmp_flag = 0;
+ }
+
+ for (i = 0; i < j; i++)
+ trees[i]->d_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)
+{
+ if (will_extract_dentry(dentry)) {
+ list_del(&dentry->d_extraction_list_node);
+ 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->d_parent;
+ if (will_extract_dentry(ancestor)) {
+ place_after = &ancestor->d_extraction_list_node;
+ break;
+ }
+ } while (!dentry_is_root(ancestor));
+
+ ancestor = dentry;
+ do {
+ ancestor = ancestor->d_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;
+ inode->i_can_externally_back = 0;
+ if ((void *)dentry->d_extraction_name != (void *)dentry->d_name)
+ FREE(dentry->d_extraction_name);
+ dentry->d_extraction_name = NULL;
+ dentry->d_extraction_name_nchars = 0;
+ }
+}
+
+static void
+destroy_blob_list(struct list_head *blob_list)
+{
+ struct blob_descriptor *blob;
+
+ list_for_each_entry(blob, blob_list, extraction_list)
+ if (blob->out_refcnt > ARRAY_LEN(blob->inline_blob_extraction_targets))
+ FREE(blob->blob_extraction_targets);
+}
+
+#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 (le16_to_cpu(name[i])) {
+ #ifdef __WIN32__
+ case '\x01'...'\x1F':
+ case '\\':
+ case ':':
+ case '*':
+ case '?':
+ case '"':
+ case '<':
+ case '>':
+ case '|':
+ #endif
+ case '/':
+ case '\0':
+ if (fix)
+ name[i] = replacement_char;
+ else
+ return false;
+ }
+ }
+
+ return true;
+}
+
+static int
+dentry_calculate_extraction_name(struct wim_dentry *dentry,
+ struct apply_ctx *ctx)
+{
+ int ret;
+
+ if (dentry_is_root(dentry))
+ return 0;
+
+#ifdef WITH_NTFS_3G
+ if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) {
+ dentry->d_extraction_name = dentry->d_name;
+ dentry->d_extraction_name_nchars = dentry->d_name_nbytes /
+ sizeof(utf16lechar);
+ return 0;
+ }
+#endif
+
+ 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->d_name, dentry->d_name_nbytes / 2, false)) {
+ size_t nbytes = 0;
+ ret = utf16le_get_tstr(dentry->d_name,
+ dentry->d_name_nbytes,
+ (const tchar **)&dentry->d_extraction_name,
+ &nbytes);
+ dentry->d_extraction_name_nchars = nbytes / sizeof(tchar);
+ return ret;
+ } 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->d_name_nbytes / 2];
+
+ memcpy(utf16_name_copy, dentry->d_name, dentry->d_name_nbytes);
+ file_name_valid(utf16_name_copy, dentry->d_name_nbytes / 2, true);
+
+ const tchar *tchar_name;
+ size_t tchar_nchars;
+
+ ret = utf16le_get_tstr(utf16_name_copy,
+ dentry->d_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 = TSTRDUP(fixed_name);
+ 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;
+}
+
+/*
+ * 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
+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, d_extraction_list_node);
+
+ 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 blob_table *blob_table)
+{
+ struct wim_inode *inode = dentry->d_inode;
+ struct blob_descriptor *blob;
+ int ret;
+ bool force = false;
+
+ /* Special case: when extracting from a pipe, the WIM blob table is
+ * initially empty, so "resolving" an inode's streams is initially not
+ * possible. However, we still need to keep track of which blobs,
+ * identified by SHA-1 message digests, need to be extracted, so we
+ * "resolve" the inode's streams anyway by allocating a 'struct
+ * blob_descriptor' for each one. */
+ if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE)
+ force = true;
+ ret = inode_resolve_streams(inode, blob_table, force);
+ if (ret)
+ return ret;
+ for (unsigned i = 0; i < inode->i_num_streams; i++) {
+ blob = stream_blob_resolved(&inode->i_streams[i]);
+ if (blob)
+ blob->out_refcnt = 0;
+ }
+ return 0;
+}
+
+/*
+ * For each dentry to be extracted, resolve all streams in the corresponding
+ * inode and set 'out_refcnt' in all referenced blob_descriptors 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->blob_table);
+ if (ret)
+ return ret;
+ }
+ return 0;
+}
+
+static int
+ref_stream(struct wim_inode_stream *strm, struct wim_dentry *dentry,
+ struct apply_ctx *ctx)
+{
+ struct wim_inode *inode = dentry->d_inode;
+ struct blob_descriptor *blob = stream_blob_resolved(strm);
+ struct blob_extraction_target *targets;
+
+ if (!blob)
+ return 0;
+
+ /* Tally the size only for each actual extraction of the stream (not
+ * additional hard links to the inode). */
+ if (inode->i_visited && ctx->supported_features.hard_links)
+ return 0;
+
+ ctx->progress.extract.total_bytes += blob->size;
+ ctx->progress.extract.total_streams++;
+
+ if (inode->i_visited)
+ return 0;
+
+ /* Add each blob to 'ctx->blob_list' only one time, regardless of how
+ * many extraction targets it will have. */
+ if (blob->out_refcnt == 0) {
+ list_add_tail(&blob->extraction_list, &ctx->blob_list);
+ ctx->num_blobs_remaining++;
+ }
+
+ /* Set this stream as an extraction target of 'blob'. */
+
+ if (blob->out_refcnt < ARRAY_LEN(blob->inline_blob_extraction_targets)) {
+ targets = blob->inline_blob_extraction_targets;
+ } else {
+ struct blob_extraction_target *prev_targets;
+ size_t alloc_blob_extraction_targets;
+
+ if (blob->out_refcnt == ARRAY_LEN(blob->inline_blob_extraction_targets)) {
+ prev_targets = NULL;
+ alloc_blob_extraction_targets = ARRAY_LEN(blob->inline_blob_extraction_targets);
+ } else {
+ prev_targets = blob->blob_extraction_targets;
+ alloc_blob_extraction_targets = blob->alloc_blob_extraction_targets;
+ }
+
+ if (blob->out_refcnt == alloc_blob_extraction_targets) {
+ alloc_blob_extraction_targets *= 2;
+ targets = REALLOC(prev_targets,
+ alloc_blob_extraction_targets *
+ sizeof(targets[0]));
+ if (!targets)
+ return WIMLIB_ERR_NOMEM;
+ if (!prev_targets) {
+ memcpy(targets,
+ blob->inline_blob_extraction_targets,
+ sizeof(blob->inline_blob_extraction_targets));
+ }
+ blob->blob_extraction_targets = targets;
+ blob->alloc_blob_extraction_targets = alloc_blob_extraction_targets;
+ }
+ targets = blob->blob_extraction_targets;
+ }
+ targets[blob->out_refcnt].inode = inode;
+ targets[blob->out_refcnt].stream = strm;
+ blob->out_refcnt++;
+ return 0;
+}
+
+static int
+ref_stream_if_needed(struct wim_dentry *dentry, struct wim_inode *inode,
+ struct wim_inode_stream *strm, struct apply_ctx *ctx)
+{
+ bool need_stream = false;
+ switch (strm->stream_type) {
+ case STREAM_TYPE_DATA:
+ if (stream_is_named(strm)) {
+ /* Named data stream */
+ if (ctx->supported_features.named_data_streams)
+ need_stream = true;
+ } else if (!(inode->i_attributes & (FILE_ATTRIBUTE_DIRECTORY |
+ FILE_ATTRIBUTE_ENCRYPTED))
+ && !(inode_is_symlink(inode)
+ && !ctx->supported_features.reparse_points
+ && ctx->supported_features.symlink_reparse_points))
+ {
+ /*
+ * Unnamed data stream. Skip if any of the following is true:
+ *
+ * - file is a directory
+ * - file is encrypted
+ * - backend needs to create the file as UNIX symlink
+ * - backend will extract the stream as externally
+ * backed from the WIM archive itself
+ */
+ if (ctx->apply_ops->will_back_from_wim) {
+ int ret = (*ctx->apply_ops->will_back_from_wim)(dentry, ctx);
+ if (ret > 0) /* Error? */
+ return ret;
+ if (ret < 0) /* Won't externally back? */
+ need_stream = true;
+ } else {
+ need_stream = true;
+ }
+ }
+ break;
+ case STREAM_TYPE_REPARSE_POINT:
+ wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT);
+ if (ctx->supported_features.reparse_points ||
+ (inode_is_symlink(inode) &&
+ ctx->supported_features.symlink_reparse_points))
+ need_stream = true;
+ break;
+ case STREAM_TYPE_EFSRPC_RAW_DATA:
+ wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED);
+ if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY) {
+ if (ctx->supported_features.encrypted_directories)
+ need_stream = true;
+ } else {
+ if (ctx->supported_features.encrypted_files)
+ need_stream = true;
+ }
+ break;
+ }
+ if (need_stream)
+ return ref_stream(strm, dentry, ctx);
+ return 0;
+}
+
+static int
+dentry_ref_streams(struct wim_dentry *dentry, struct apply_ctx *ctx)
+{
+ struct wim_inode *inode = dentry->d_inode;
+ for (unsigned i = 0; i < inode->i_num_streams; i++) {
+ int ret = ref_stream_if_needed(dentry, inode,
+ &inode->i_streams[i], ctx);
+ if (ret)
+ return ret;
+ }
+ inode->i_visited = 1;
+ return 0;
+}
+
+/*
+ * Given a list of dentries to be extracted, build the list of blobs that need
+ * to be extracted, and for each blob determine the streams to which that blob
+ * will be extracted.