X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Fextract.c;h=a4029ec23f39c718497e57d992d6de752917a63f;hp=d6cc9c45d75ee854382a6c9c4d4fcffc7a02bcd0;hb=ced16a28e197645a40fa04a54793d117a04526d7;hpb=ce6b853107173c96fb04754aeb9e250a07bd4a07 diff --git a/src/extract.c b/src/extract.c index d6cc9c45..a4029ec2 100644 --- a/src/extract.c +++ b/src/extract.c @@ -6,7 +6,7 @@ */ /* - * Copyright (C) 2012, 2013 Eric Biggers + * Copyright (C) 2012, 2013, 2014 Eric Biggers * * This file is part of wimlib, a library for working with WIM files. * @@ -24,437 +24,1750 @@ * 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 -#ifdef __WIN32__ -# include "wimlib/win32_common.h" /* For GetFullPathName() */ -#endif - #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/swm.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 -#ifdef WITH_NTFS_3G -# include /* for ntfs_mount(), ntfs_umount() */ -#endif +#include #include #include #include -#define MAX_EXTRACT_LONG_PATH_WARNINGS 5 +#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_HARDLINK | \ + WIMLIB_EXTRACT_FLAG_SYMLINK | \ + WIMLIB_EXTRACT_FLAG_VERBOSE | \ + WIMLIB_EXTRACT_FLAG_SEQUENTIAL | \ + WIMLIB_EXTRACT_FLAG_UNIX_DATA | \ + WIMLIB_EXTRACT_FLAG_NO_ACLS | \ + WIMLIB_EXTRACT_FLAG_STRICT_ACLS | \ + WIMLIB_EXTRACT_FLAG_RPFIX | \ + WIMLIB_EXTRACT_FLAG_NORPFIX | \ + WIMLIB_EXTRACT_FLAG_TO_STDOUT | \ + WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES | \ + WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS | \ + WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS | \ + WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES | \ + WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS | \ + WIMLIB_EXTRACT_FLAG_RESUME | \ + WIMLIB_EXTRACT_FLAG_FILE_ORDER | \ + WIMLIB_EXTRACT_FLAG_GLOB_PATHS | \ + WIMLIB_EXTRACT_FLAG_STRICT_GLOB | \ + WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES | \ + WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE | \ + WIMLIB_EXTRACT_FLAG_WIMBOOT) -static int -do_apply_op(struct wim_dentry *dentry, struct apply_args *args, - int (*apply_dentry_func)(const tchar *, size_t, - struct wim_dentry *, struct apply_args *)) +static bool +dentry_in_list(const struct wim_dentry *dentry) { - tchar *p; - size_t extraction_path_nchars; - struct wim_dentry *d; - LIST_HEAD(ancestor_list); - const tchar *target; - size_t target_nchars; + return dentry->extraction_list.next != NULL; +} -#ifdef __WIN32__ - if (args->target_lowlevel_path) { - target = args->target_lowlevel_path; - target_nchars = args->target_lowlevel_path_nchars; - } else -#endif - { - target = args->target; - target_nchars = args->target_nchars; - } +static inline bool +is_linked_extraction(const struct apply_ctx *ctx) +{ + return 0 != (ctx->extract_flags & (WIMLIB_EXTRACT_FLAG_HARDLINK | + WIMLIB_EXTRACT_FLAG_SYMLINK)); +} - extraction_path_nchars = target_nchars; +static inline bool +can_extract_named_data_streams(const struct apply_ctx *ctx) +{ + return ctx->supported_features.named_data_streams && + !is_linked_extraction(ctx); +} +/* Inform library user of progress of stream extraction following the successful + * extraction of a copy of the stream specified by @lte. */ +static void +update_extract_progress(struct apply_ctx *ctx, + const struct wim_lookup_table_entry *lte) +{ + wimlib_progress_func_t progress_func = ctx->progress_func; + union wimlib_progress_info *progress = &ctx->progress; - for (d = dentry; d != args->extract_root; d = d->parent) { - if (d->not_extracted) - return 0; - extraction_path_nchars += d->extraction_name_nchars + 1; - list_add(&d->tmp_list, &ancestor_list); + progress->extract.completed_bytes += lte->size; + if (progress_func && + progress->extract.completed_bytes >= ctx->next_progress) + { + progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS, progress); + if (progress->extract.completed_bytes >= + progress->extract.total_bytes) + { + ctx->next_progress = ~0ULL; + } else { + ctx->next_progress += progress->extract.total_bytes / 128; + if (ctx->next_progress > progress->extract.total_bytes) + ctx->next_progress = progress->extract.total_bytes; + } } +} - tchar extraction_path[extraction_path_nchars + 1]; - p = tmempcpy(extraction_path, target, target_nchars); +#ifndef __WIN32__ +/* Extract a symbolic link (not directly as reparse data), handling fixing up + * the target of absolute symbolic links and updating the extract progress. + * + * @inode must specify the WIM inode for a symbolic link or junction reparse + * point. + * + * @lte_override overrides the resource used as the reparse data for the + * symbolic link. */ +static int +extract_symlink(const tchar *path, struct apply_ctx *ctx, + struct wim_inode *inode, + struct wim_lookup_table_entry *lte_override) +{ + ssize_t bufsize = ctx->ops->path_max; + tchar target[bufsize]; + tchar *buf = target; + tchar *fixed_target; + ssize_t sret; + int ret; + /* If absolute symbolic link fixups requested, reserve space in the link + * target buffer for the absolute path of the target directory. */ + if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) + { + buf += ctx->realtarget_nchars; + bufsize -= ctx->realtarget_nchars; + } - list_for_each_entry(d, &ancestor_list, tmp_list) { - *p++ = OS_PREFERRED_PATH_SEPARATOR; - p = tmempcpy(p, d->extraction_name, d->extraction_name_nchars); + /* Translate the WIM inode's reparse data into the link target. */ + sret = wim_inode_readlink(inode, buf, bufsize - 1, lte_override); + if (sret < 0) { + errno = -sret; + return WIMLIB_ERR_READLINK; } - *p = T('\0'); + buf[sret] = '\0'; -#ifdef __WIN32__ - /* Warn the user if the path exceeds MAX_PATH */ + if ((ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) && + buf[0] == '/') + { + /* Fix absolute symbolic link target to point into the + * actual extraction destination. */ + tmemcpy(target, ctx->realtarget, ctx->realtarget_nchars); + fixed_target = target; + } else { + /* Keep same link target. */ + fixed_target = buf; + } - /* + 1 for '\0', -4 for \\?\. */ - if (extraction_path_nchars + 1 - 4 > MAX_PATH) { - if (dentry->needs_extraction && - args->num_long_paths < MAX_EXTRACT_LONG_PATH_WARNINGS) - { - WARNING("Path \"%ls\" exceeds MAX_PATH and will not be accessible " - "to most Windows software", extraction_path); - if (++args->num_long_paths == MAX_EXTRACT_LONG_PATH_WARNINGS) - WARNING("Suppressing further warnings about long paths"); - } + /* Call into the apply_operations to create the symbolic link. */ + DEBUG("Creating symlink \"%"TS"\" => \"%"TS"\"", + path, fixed_target); + ret = ctx->ops->create_symlink(fixed_target, path, ctx); + if (ret) { + ERROR_WITH_ERRNO("Failed to create symlink " + "\"%"TS"\" => \"%"TS"\"", path, fixed_target); + return ret; } -#endif - return (*apply_dentry_func)(extraction_path, extraction_path_nchars, - dentry, args); -} + /* Account for reparse data consumed. */ + update_extract_progress(ctx, + (lte_override ? lte_override : + inode_unnamed_lte_resolved(inode))); + return 0; +} +#endif /* !__WIN32__ */ -/* Extracts a file, directory, or symbolic link from the WIM archive. */ +/* Create a file, directory, or symbolic link. */ static int -apply_dentry_normal(struct wim_dentry *dentry, void *arg) +extract_inode(const tchar *path, struct apply_ctx *ctx, struct wim_inode *inode) { -#ifdef __WIN32__ - return do_apply_op(dentry, arg, win32_do_apply_dentry); -#else - return do_apply_op(dentry, arg, unix_do_apply_dentry); -#endif -} + int ret; +#ifndef __WIN32__ + if (ctx->supported_features.symlink_reparse_points && + !ctx->supported_features.reparse_points && + inode_is_symlink(inode)) + { + ret = extract_symlink(path, ctx, inode, NULL); + } else +#endif /* !__WIN32__ */ + if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY) { + ret = ctx->ops->create_directory(path, ctx, &inode->extract_cookie); + if (ret) { + ERROR_WITH_ERRNO("Failed to create the directory " + "\"%"TS"\"", path); + } + } else if ((inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) && + ctx->ops->extract_encrypted_stream_creates_file && + ctx->supported_features.encrypted_files) { + ret = ctx->ops->extract_encrypted_stream( + path, inode_unnamed_lte_resolved(inode), ctx); + if (ret) { + ERROR_WITH_ERRNO("Failed to create and extract " + "encrypted file \"%"TS"\"", path); + } + } else { + ret = ctx->ops->create_file(path, ctx, &inode->extract_cookie); + if (ret) { + ERROR_WITH_ERRNO("Failed to create the file " + "\"%"TS"\"", path); + } + } + return ret; +} -/* Apply timestamps to an extracted file or directory */ static int -apply_dentry_timestamps_normal(struct wim_dentry *dentry, void *arg) +extract_hardlink(const tchar *oldpath, const tchar *newpath, + struct apply_ctx *ctx) { -#ifdef __WIN32__ - return do_apply_op(dentry, arg, win32_do_apply_dentry_timestamps); -#else - return do_apply_op(dentry, arg, unix_do_apply_dentry_timestamps); -#endif -} + int ret; -static bool -dentry_is_dot_or_dotdot(const struct wim_dentry *dentry) -{ - const utf16lechar *file_name = dentry->file_name; - return file_name != NULL && - file_name[0] == cpu_to_le16('.') && - (file_name[1] == cpu_to_le16('\0') || - (file_name[1] == cpu_to_le16('.') && - file_name[2] == cpu_to_le16('\0'))); + DEBUG("Creating hardlink \"%"TS"\" => \"%"TS"\"", newpath, oldpath); + ret = ctx->ops->create_hardlink(oldpath, newpath, ctx); + if (ret) { + ERROR_WITH_ERRNO("Failed to create hardlink " + "\"%"TS"\" => \"%"TS"\"", + newpath, oldpath); + } + return ret; } -/* Extract a dentry if it hasn't already been extracted and either - * WIMLIB_EXTRACT_FLAG_NO_STREAMS is not specified, or the dentry is a directory - * and/or has no unnamed stream. */ +#ifdef __WIN32__ static int -maybe_apply_dentry(struct wim_dentry *dentry, void *arg) +try_extract_rpfix(u8 *rpbuf, + u16 *rpbuflen_p, + const wchar_t *extract_root_realpath, + unsigned extract_root_realpath_nchars) { - struct apply_args *args = arg; + struct reparse_data rpdata; + wchar_t *target; + size_t target_nchars; + size_t stripped_nchars; + wchar_t *stripped_target; + wchar_t stripped_target_nchars; int ret; - if (!dentry->needs_extraction) - return 0; + utf16lechar *new_target; + utf16lechar *new_print_name; + size_t new_target_nchars; + size_t new_print_name_nchars; + utf16lechar *p; + + ret = parse_reparse_data(rpbuf, *rpbuflen_p, &rpdata); + if (ret) + return ret; - if (args->extract_flags & WIMLIB_EXTRACT_FLAG_NO_STREAMS && - !dentry_is_directory(dentry) && - inode_unnamed_lte_resolved(dentry->d_inode) != NULL) + if (extract_root_realpath[0] == L'\0' || + extract_root_realpath[1] != L':' || + extract_root_realpath[2] != L'\\') + return WIMLIB_ERR_REPARSE_POINT_FIXUP_FAILED; + + ret = parse_substitute_name(rpdata.substitute_name, + rpdata.substitute_name_nbytes, + rpdata.rptag); + if (ret < 0) return 0; + stripped_nchars = ret; + target = rpdata.substitute_name; + target_nchars = rpdata.substitute_name_nbytes / sizeof(utf16lechar); + stripped_target = target + stripped_nchars; + stripped_target_nchars = target_nchars - stripped_nchars; + + new_target = alloca((6 + extract_root_realpath_nchars + + stripped_target_nchars) * sizeof(utf16lechar)); + + p = new_target; + if (stripped_nchars == 6) { + /* Include \??\ prefix if it was present before */ + p = wmempcpy(p, L"\\??\\", 4); + } - if ((args->extract_flags & WIMLIB_EXTRACT_FLAG_VERBOSE) && - args->progress_func) { - ret = calculate_dentry_full_path(dentry); - if (ret) - return ret; - args->progress.extract.cur_path = dentry->_full_path; - args->progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DENTRY, - &args->progress); + /* Print name excludes the \??\ if present. */ + new_print_name = p; + if (stripped_nchars != 0) { + /* Get drive letter from real path to extract root, if a drive + * letter was present before. */ + *p++ = extract_root_realpath[0]; + *p++ = extract_root_realpath[1]; } - ret = args->apply_dentry(dentry, args); - if (ret == 0) - dentry->needs_extraction = 0; - return ret; + /* Copy the rest of the extract root */ + p = wmempcpy(p, extract_root_realpath + 2, extract_root_realpath_nchars - 2); + + /* Append the stripped target */ + p = wmempcpy(p, stripped_target, stripped_target_nchars); + new_target_nchars = p - new_target; + new_print_name_nchars = p - new_print_name; + + if (new_target_nchars * sizeof(utf16lechar) >= REPARSE_POINT_MAX_SIZE || + new_print_name_nchars * sizeof(utf16lechar) >= REPARSE_POINT_MAX_SIZE) + return WIMLIB_ERR_REPARSE_POINT_FIXUP_FAILED; + + rpdata.substitute_name = new_target; + rpdata.substitute_name_nbytes = new_target_nchars * sizeof(utf16lechar); + rpdata.print_name = new_print_name; + rpdata.print_name_nbytes = new_print_name_nchars * sizeof(utf16lechar); + return make_reparse_buffer(&rpdata, rpbuf, rpbuflen_p); } +#endif /* __WIN32__ */ -static void -calculate_bytes_to_extract(struct list_head *stream_list, - int extract_flags, - union wimlib_progress_info *progress) +/* Set reparse data on extracted file or directory that has + * FILE_ATTRIBUTE_REPARSE_POINT set. */ +static int +extract_reparse_data(const tchar *path, struct apply_ctx *ctx, + struct wim_inode *inode, + struct wim_lookup_table_entry *lte_override) { - struct wim_lookup_table_entry *lte; - u64 total_bytes = 0; - u64 num_streams = 0; + int ret; + u8 rpbuf[REPARSE_POINT_MAX_SIZE]; + u16 rpbuflen; + + ret = wim_inode_get_reparse_data(inode, rpbuf, &rpbuflen, lte_override); + if (ret) + goto error; - /* For each stream to be extracted... */ - list_for_each_entry(lte, stream_list, extraction_list) { - if (extract_flags & - (WIMLIB_EXTRACT_FLAG_SYMLINK | WIMLIB_EXTRACT_FLAG_HARDLINK)) +#ifdef __WIN32__ + /* Fix up target of absolute symbolic link or junction points so + * that they point into the actual extraction target. */ + if ((ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) && + (inode->i_reparse_tag == WIM_IO_REPARSE_TAG_SYMLINK || + inode->i_reparse_tag == WIM_IO_REPARSE_TAG_MOUNT_POINT) && + !inode->i_not_rpfixed) + { + ret = try_extract_rpfix(rpbuf, &rpbuflen, ctx->realtarget, + ctx->realtarget_nchars); + if (ret && !(ctx->extract_flags & + WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS)) { - /* 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); + WARNING("Reparse point fixup of \"%"TS"\" " + "failed", path); + ret = 0; } + if (ret) + goto error; } - progress->extract.num_streams = num_streams; - progress->extract.total_bytes = total_bytes; - progress->extract.completed_bytes = 0; -} +#endif -static void -maybe_add_stream_for_extraction(struct wim_lookup_table_entry *lte, - struct list_head *stream_list) -{ - if (++lte->out_refcnt == 1) { - INIT_LIST_HEAD(<e->lte_dentry_list); - list_add_tail(<e->extraction_list, stream_list); + ret = ctx->ops->set_reparse_data(path, rpbuf, rpbuflen, ctx); + + /* On Windows, the SeCreateSymbolicLink privilege is required to create + * symbolic links. To be more friendly towards non-Administrator users, + * we merely warn the user if symbolic links cannot be created due to + * insufficient permissions or privileges, unless + * WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS was provided. */ +#ifdef __WIN32__ + if (ret && inode_is_symlink(inode) && + (errno == EACCES || errno == EPERM) && + !(ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SYMLINKS)) + { + WARNING("Can't set reparse data on \"%"TS"\": " + "Access denied!\n" + " You may be trying to " + "extract a symbolic link without the\n" + " SeCreateSymbolicLink privilege, " + "which by default non-Administrator\n" + " accounts do not have.", + path); + ret = 0; } -} +#endif + if (ret) + goto error; -struct find_streams_ctx { - struct list_head stream_list; - int extract_flags; -}; + /* Account for reparse data consumed. */ + update_extract_progress(ctx, + (lte_override ? lte_override : + inode_unnamed_lte_resolved(inode))); + return 0; + +error: + ERROR_WITH_ERRNO("Failed to set reparse data on \"%"TS"\"", path); + return ret; +} +/* + * Extract zero or more streams to a file. + * + * This function operates slightly differently depending on whether @lte_spec is + * NULL or not. When @lte_spec is NULL, the behavior is to extract the default + * file contents (unnamed stream), and, if named data streams are supported in + * the extract mode and volume, any named data streams. When @lte_spec is not + * NULL, the behavior is to extract only all copies of the stream @lte_spec, and + * in addition use @lte_spec to set the reparse data or create the symbolic link + * if appropriate. + * + * @path + * Path to file to extract (as can be passed to apply_operations + * functions). + * @ctx + * Apply context. + * @dentry + * WIM dentry that corresponds to the file being extracted. + * @lte_spec + * If non-NULL, specifies the lookup table entry for a stream to extract, + * and only that stream will be extracted (although there may be more than + * one instance of it). + * @lte_override + * Used only if @lte_spec != NULL; it is passed to the extraction functions + * rather than @lte_spec, allowing the location of the stream to be + * overridden. (This is used when the WIM is being read from a nonseekable + * file, such as a pipe, when streams need to be used more than once; each + * such stream is extracted to a temporary file.) + */ static int -dentry_find_streams_to_extract(struct wim_dentry *dentry, void *_ctx) +extract_streams(const tchar *path, struct apply_ctx *ctx, + struct wim_dentry *dentry, + struct wim_lookup_table_entry *lte_spec, + struct wim_lookup_table_entry *lte_override) { - struct find_streams_ctx *ctx = _ctx; struct wim_inode *inode = dentry->d_inode; struct wim_lookup_table_entry *lte; - bool dentry_added = false; - struct list_head *stream_list = &ctx->stream_list; - int extract_flags = ctx->extract_flags; + file_spec_t file_spec; + int ret; - if (!dentry->needs_extraction) + if (dentry->was_linked) return 0; - lte = inode_unnamed_lte_resolved(inode); - if (lte) { - if (!inode->i_visited) - maybe_add_stream_for_extraction(lte, stream_list); - list_add_tail(&dentry->extraction_stream_list, <e->lte_dentry_list); - dentry_added = true; - } - - /* Determine whether to include alternate data stream entries or not. - * - * UNIX: Include them if extracting using NTFS-3g. - * - * Windows: Include them undconditionally, although if the filesystem is - * not NTFS we won't actually be able to extract them. */ -#if defined(WITH_NTFS_3G) - if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) -#elif defined(__WIN32__) - if (1) -#else - if (0) +#ifdef ENABLE_DEBUG + if (lte_spec) { + char sha1_str[100]; + char *p = sha1_str; + for (unsigned i = 0; i < SHA1_HASH_SIZE; i++) + p += sprintf(p, "%02x", lte_override->hash[i]); + DEBUG("Extracting stream SHA1=%s to \"%"TS"\"", + sha1_str, path, inode->i_ino); + } else { + DEBUG("Extracting streams to \"%"TS"\"", path, inode->i_ino); + } #endif - { - for (unsigned i = 0; i < inode->i_num_ads; i++) { - if (inode->i_ads_entries[i].stream_name_nbytes != 0) { - lte = inode->i_ads_entries[i].lte; - if (lte) { - if (!inode->i_visited) { - maybe_add_stream_for_extraction(lte, - stream_list); - } - if (!dentry_added) { - list_add_tail(&dentry->extraction_stream_list, - <e->lte_dentry_list); - dentry_added = true; - } + + if (ctx->ops->uses_cookies) + file_spec.cookie = inode->extract_cookie; + else + file_spec.path = path; + + /* Unnamed data stream. */ + lte = inode_unnamed_lte_resolved(inode); + if (lte && (!lte_spec || lte == lte_spec)) { + if (lte_spec) + lte = lte_override; + if (!(inode->i_attributes & (FILE_ATTRIBUTE_DIRECTORY | + FILE_ATTRIBUTE_REPARSE_POINT))) + { + if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED && + ctx->supported_features.encrypted_files) { + if (!ctx->ops->extract_encrypted_stream_creates_file) { + ret = ctx->ops->extract_encrypted_stream( + path, lte, ctx); + if (ret) + goto error; } + } else { + ret = ctx->ops->extract_unnamed_stream( + file_spec, lte, ctx, + dentry); + if (ret) + goto error; } + update_extract_progress(ctx, lte); + } + else if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) + { + ret = 0; + if (ctx->supported_features.reparse_points) + ret = extract_reparse_data(path, ctx, inode, lte); + #ifndef __WIN32__ + else if ((inode_is_symlink(inode) && + ctx->supported_features.symlink_reparse_points)) + ret = extract_symlink(path, ctx, inode, lte); + #endif + if (ret) + return ret; + } + } + + /* Named data streams. */ + if (can_extract_named_data_streams(ctx)) { + for (u16 i = 0; i < inode->i_num_ads; i++) { + struct wim_ads_entry *entry = &inode->i_ads_entries[i]; + + if (!ads_entry_is_named_stream(entry)) + continue; + lte = entry->lte; + if (!lte) + continue; + if (lte_spec && lte_spec != lte) + continue; + if (lte_spec) + lte = lte_override; + ret = ctx->ops->extract_named_stream(file_spec, entry->stream_name, + entry->stream_name_nbytes / 2, + lte, ctx); + if (ret) + goto error; + update_extract_progress(ctx, lte); } } - inode->i_visited = 1; return 0; + +error: + ERROR_WITH_ERRNO("Failed to extract data of \"%"TS"\"", path); + return ret; } +/* Set attributes on an extracted file or directory if supported by the + * extraction mode. */ static int -dentry_resolve_and_zero_lte_refcnt(struct wim_dentry *dentry, void *_lookup_table) +extract_file_attributes(const tchar *path, struct apply_ctx *ctx, + struct wim_dentry *dentry, unsigned pass) { - struct wim_inode *inode = dentry->d_inode; - struct wim_lookup_table *lookup_table = _lookup_table; - struct wim_lookup_table_entry *lte; int ret; - ret = inode_resolve_ltes(inode, lookup_table); - if (ret) - return ret; - for (unsigned i = 0; i <= inode->i_num_ads; i++) { - lte = inode_stream_lte_resolved(inode, i); - if (lte) - lte->out_refcnt = 0; + if (ctx->ops->set_file_attributes && + !(ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES) && + !(dentry == ctx->target_dentry && ctx->root_dentry_is_special)) { + u32 attributes = dentry->d_inode->i_attributes; + + /* Clear unsupported attributes. */ + attributes &= ctx->supported_attributes_mask; + + if ((attributes & FILE_ATTRIBUTE_DIRECTORY && + !ctx->supported_features.encrypted_directories) || + (!(attributes & FILE_ATTRIBUTE_DIRECTORY) && + !ctx->supported_features.encrypted_files)) + { + attributes &= ~FILE_ATTRIBUTE_ENCRYPTED; + } + + if (attributes == 0) + attributes = FILE_ATTRIBUTE_NORMAL; + + ret = ctx->ops->set_file_attributes(path, attributes, ctx, pass); + if (ret) { + ERROR_WITH_ERRNO("Failed to set attributes on " + "\"%"TS"\"", path); + return ret; + } } return 0; } + +/* Set or remove the short (DOS) name on an extracted file or directory if + * supported by the extraction mode. Since DOS names are unimportant and it's + * easy to run into problems setting them on Windows (SetFileShortName() + * requires SE_RESTORE privilege, which only the Administrator can request, and + * also requires DELETE access to the file), failure is ignored unless + * WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES is set. */ static int -find_streams_for_extraction(struct wim_dentry *root, - struct list_head *stream_list, - struct wim_lookup_table *lookup_table, - int extract_flags) +extract_short_name(const tchar *path, struct apply_ctx *ctx, + struct wim_dentry *dentry) { - struct find_streams_ctx ctx; int ret; - INIT_LIST_HEAD(&ctx.stream_list); - ctx.extract_flags = extract_flags; - ret = for_dentry_in_tree(root, dentry_resolve_and_zero_lte_refcnt, lookup_table); - if (ret) - return ret; - for_dentry_in_tree(root, dentry_find_streams_to_extract, &ctx); - list_transfer(&ctx.stream_list, stream_list); + /* The root of the dentry tree being extracted may not be extracted to + * its original name, so its short name should be ignored. */ + if (dentry == ctx->target_dentry) + return 0; + + if (ctx->supported_features.short_names) { + ret = ctx->ops->set_short_name(path, + dentry->short_name, + dentry->short_name_nbytes / 2, + ctx); + if (ret && (ctx->extract_flags & + WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES)) + { + ERROR_WITH_ERRNO("Failed to set short name of " + "\"%"TS"\"", path); + return ret; + } + } return 0; } -struct apply_operations { - int (*apply_dentry)(struct wim_dentry *dentry, void *arg); - int (*apply_dentry_timestamps)(struct wim_dentry *dentry, void *arg); -}; +/* Set security descriptor, UNIX data, or neither on an extracted file, taking + * into account the current extraction mode and flags. */ +static int +extract_security(const tchar *path, struct apply_ctx *ctx, + struct wim_dentry *dentry) +{ + int ret; + struct wim_inode *inode = dentry->d_inode; -static const struct apply_operations normal_apply_operations = { - .apply_dentry = apply_dentry_normal, - .apply_dentry_timestamps = apply_dentry_timestamps_normal, -}; + if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS) + return 0; -#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 ((ctx->target_dentry == dentry) && ctx->root_dentry_is_special) + return 0; + +#ifndef __WIN32__ + if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) { + struct wimlib_unix_data data; + + ret = inode_get_unix_data(inode, &data, NULL); + if (ret < 0) + ret = 0; + else if (ret == 0) + ret = ctx->ops->set_unix_data(path, &data, ctx); + if (ret) { + if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_ACLS) { + ERROR_WITH_ERRNO("Failed to set UNIX owner, " + "group, and/or mode on " + "\"%"TS"\"", path); + return ret; + } else { + WARNING_WITH_ERRNO("Failed to set UNIX owner, " + "group, and/or/mode on " + "\"%"TS"\"", path); + } + } + } + else +#endif /* __WIN32__ */ + if (ctx->supported_features.security_descriptors && + inode->i_security_id != -1) + { + const struct wim_security_data *sd; + const u8 *desc; + size_t desc_size; + + sd = wim_const_security_data(ctx->wim); + desc = sd->descriptors[inode->i_security_id]; + desc_size = sd->sizes[inode->i_security_id]; + + ret = ctx->ops->set_security_descriptor(path, desc, + desc_size, ctx); + if (ret) { + if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_ACLS) { + ERROR_WITH_ERRNO("Failed to set security " + "descriptor on \"%"TS"\"", path); + return ret; + } else { + #if 0 + if (errno != EACCES) { + WARNING_WITH_ERRNO("Failed to set " + "security descriptor " + "on \"%"TS"\"", path); + } + #endif + ctx->no_security_descriptors++; + } + } + } + return 0; +} +/* Set timestamps on an extracted file. Failure is warning-only unless + * WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS is set. */ 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) +extract_timestamps(const tchar *path, struct apply_ctx *ctx, + struct wim_dentry *dentry) { - uint64_t bytes_per_progress = args->progress.extract.total_bytes / 100; - uint64_t next_progress = bytes_per_progress; - struct wim_lookup_table_entry *lte; - struct wim_dentry *dentry; + struct wim_inode *inode = dentry->d_inode; int ret; - /* This complicated loop is essentially looping through the dentries, - * although dentries may be visited more than once (if a dentry contains - * two different nonempty streams) or not at all (if a dentry contains - * no non-empty streams). - * - * The outer loop is 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, extraction_list) { - /* For each dentry to be extracted that is a name for an inode - * containing the stream */ - list_for_each_entry(dentry, <e->lte_dentry_list, extraction_stream_list) { - /* Extract the dentry if it was not already - * extracted */ - ret = maybe_apply_dentry(dentry, args); - if (ret) + if ((ctx->target_dentry == dentry) && ctx->root_dentry_is_special) + return 0; + + if (ctx->ops->set_timestamps) { + ret = ctx->ops->set_timestamps(path, + inode->i_creation_time, + inode->i_last_write_time, + inode->i_last_access_time, + ctx); + if (ret) { + if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) { + ERROR_WITH_ERRNO("Failed to set timestamps " + "on \"%"TS"\"", path); return ret; - if (progress_func && - args->progress.extract.completed_bytes >= next_progress) - { - progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS, - &args->progress); - if (args->progress.extract.completed_bytes >= - args->progress.extract.total_bytes) - { - next_progress = ~0ULL; - } else { - next_progress = - min (args->progress.extract.completed_bytes + - bytes_per_progress, - args->progress.extract.total_bytes); - } + } else { + WARNING_WITH_ERRNO("Failed to set timestamps " + "on \"%"TS"\"", path); } } } return 0; } +/* 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; +} + +/* Given a WIM dentry to extract, build the path to which to extract it, in the + * format understood by the callbacks in the apply_operations being used. + * + * Write the resulting path into @path, which must have room for at least + * ctx->ops->path_max characters. + * + * Return %true if successful; %false if this WIM dentry doesn't actually need + * to be extracted or if the calculated path exceeds ctx->ops->max_path + * characters. + * + * This function clobbers the tmp_list member of @dentry and its ancestors up + * until the extraction root. */ +static bool +build_extraction_path(tchar path[], struct wim_dentry *dentry, + const struct apply_ctx *ctx) +{ + size_t path_nchars; + LIST_HEAD(ancestor_list); + tchar *p = path; + const tchar *target_prefix; + size_t target_prefix_nchars; + struct wim_dentry *d; + + path_nchars = ctx->ops->path_prefix_nchars; + + if (ctx->ops->requires_realtarget_in_paths) { + target_prefix = ctx->realtarget; + target_prefix_nchars = ctx->realtarget_nchars; + } else if (ctx->ops->requires_target_in_paths) { + target_prefix = ctx->target; + target_prefix_nchars = ctx->target_nchars; + } else { + target_prefix = NULL; + target_prefix_nchars = 0; + } + path_nchars += target_prefix_nchars; + + for (d = dentry; d != ctx->target_dentry; d = d->parent) { + if (!dentry_in_list(d)) + break; + + path_nchars += d->extraction_name_nchars + 1; + list_add(&d->tmp_list, &ancestor_list); + } + + path_nchars++; /* null terminator */ + + if (path_nchars > ctx->ops->path_max) { + WARNING("\"%"TS"\": Path too long to extract", + dentry_full_path(dentry)); + return false; + } + + p = tmempcpy(p, ctx->ops->path_prefix, ctx->ops->path_prefix_nchars); + p = tmempcpy(p, target_prefix, target_prefix_nchars); + list_for_each_entry(d, &ancestor_list, tmp_list) { + *p++ = ctx->ops->path_separator; + p = tmempcpy(p, d->extraction_name, d->extraction_name_nchars); + } + *p++ = T('\0'); + wimlib_assert(p - path == path_nchars); + return true; +} + +static unsigned +get_num_path_components(const tchar *path, tchar path_separator) +{ + unsigned num_components = 0; +#ifdef __WIN32__ + /* Ignore drive letter. */ + if (path[0] != L'\0' && path[1] == L':') + path += 2; +#endif + + while (*path) { + while (*path == path_separator) + path++; + if (*path) + num_components++; + while (*path && *path != path_separator) + path++; + } + return num_components; +} + static int -sort_stream_list_by_wim_position(struct list_head *stream_list) +extract_multiimage_symlink(const tchar *oldpath, const tchar *newpath, + struct apply_ctx *ctx, struct wim_dentry *dentry) { - struct list_head *cur; - size_t num_streams; - struct wim_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; + size_t num_raw_path_components; + const struct wim_dentry *d; + size_t num_target_path_components; + tchar *p; + const tchar *p_old; + int ret; + + num_raw_path_components = 0; + for (d = dentry; d != ctx->target_dentry; d = d->parent) + num_raw_path_components++; + + if (ctx->ops->requires_realtarget_in_paths) + num_target_path_components = get_num_path_components(ctx->realtarget, + ctx->ops->path_separator); + else if (ctx->ops->requires_target_in_paths) + num_target_path_components = get_num_path_components(ctx->target, + ctx->ops->path_separator); + else + num_target_path_components = 0; + + if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_MULTI_IMAGE) { + wimlib_assert(num_target_path_components > 0); + num_raw_path_components++; + num_target_path_components--; } - cur = stream_list->next; - for (i = 0; i < num_streams; i++) { - array[i] = container_of(cur, struct wim_lookup_table_entry, extraction_list); - cur = cur->next; + + p_old = oldpath + ctx->ops->path_prefix_nchars; +#ifdef __WIN32__ + if (p_old[0] != L'\0' && p_old[1] == ':') + p_old += 2; +#endif + while (*p_old == ctx->ops->path_separator) + p_old++; + while (--num_target_path_components) { + while (*p_old != ctx->ops->path_separator) + p_old++; + while (*p_old == ctx->ops->path_separator) + p_old++; } - qsort(array, num_streams, sizeof(array[0]), cmp_streams_by_wim_position); + tchar symlink_target[tstrlen(p_old) + 3 * num_raw_path_components + 1]; + + p = &symlink_target[0]; + while (num_raw_path_components--) { + *p++ = '.'; + *p++ = '.'; + *p++ = ctx->ops->path_separator; + } + tstrcpy(p, p_old); + DEBUG("Creating symlink \"%"TS"\" => \"%"TS"\"", + newpath, symlink_target); + ret = ctx->ops->create_symlink(symlink_target, newpath, ctx); + if (ret) { + ERROR_WITH_ERRNO("Failed to create symlink " + "\"%"TS"\" => \"%"TS"\"", + newpath, symlink_target); + } + return ret; +} + +/* Create the "skeleton" of an extracted file or directory. Don't yet extract + * data streams, reparse data (including symbolic links), timestamps, and + * security descriptors. Basically, everything that doesn't require reading + * non-metadata resources from the WIM file and isn't delayed until the final + * pass. */ +static int +do_dentry_extract_skeleton(tchar path[], struct wim_dentry *dentry, + struct apply_ctx *ctx) +{ + struct wim_inode *inode = dentry->d_inode; + int ret; + const tchar *oldpath; + + if (unlikely(is_linked_extraction(ctx))) { + struct wim_lookup_table_entry *unnamed_lte; + + unnamed_lte = inode_unnamed_lte_resolved(dentry->d_inode); + if (unnamed_lte && unnamed_lte->extracted_file) { + oldpath = unnamed_lte->extracted_file; + if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_HARDLINK) + goto hardlink; + else + goto symlink; + } + } + + /* Create hard link if this dentry corresponds to an already-extracted + * inode. */ + if (inode->i_extracted_file) { + oldpath = inode->i_extracted_file; + goto hardlink; + } + + /* Skip symlinks unless they can be extracted as reparse points rather + * than created directly. */ + if (inode_is_symlink(inode) && !ctx->supported_features.reparse_points) + return 0; + + /* Create this file or directory unless it's the extraction root, which + * was already created if necessary. */ + if (dentry != ctx->target_dentry) { + ret = extract_inode(path, ctx, inode); + if (ret) + return ret; + } + + /* Create empty named data streams. */ + if (can_extract_named_data_streams(ctx)) { + for (u16 i = 0; i < inode->i_num_ads; i++) { + file_spec_t file_spec; + struct wim_ads_entry *entry = &inode->i_ads_entries[i]; + + if (!ads_entry_is_named_stream(entry)) + continue; + if (entry->lte) + continue; + if (ctx->ops->uses_cookies) + file_spec.cookie = inode->extract_cookie; + else + file_spec.path = path; + ret = ctx->ops->extract_named_stream(file_spec, + entry->stream_name, + entry->stream_name_nbytes / 2, + entry->lte, ctx); + if (ret) { + ERROR_WITH_ERRNO("\"%"TS"\": failed to create " + "empty named data stream", + path); + return ret; + } + } + } + + /* Set file attributes (if supported). */ + ret = extract_file_attributes(path, ctx, dentry, 0); + if (ret) + return ret; + + /* Set or remove file short name (if supported). */ + ret = extract_short_name(path, ctx, dentry); + if (ret) + return ret; + + /* If inode has multiple links and hard links are supported in this + * extraction mode and volume, save the path to the extracted file in + * case it's needed to create a hard link. */ + if (unlikely(is_linked_extraction(ctx))) { + struct wim_lookup_table_entry *unnamed_lte; + + unnamed_lte = inode_unnamed_lte_resolved(dentry->d_inode); + if (unnamed_lte) { + unnamed_lte->extracted_file = TSTRDUP(path); + if (!unnamed_lte->extracted_file) + return WIMLIB_ERR_NOMEM; + } + } else if (inode->i_nlink > 1 && ctx->supported_features.hard_links) { + inode->i_extracted_file = TSTRDUP(path); + if (!inode->i_extracted_file) + return WIMLIB_ERR_NOMEM; + } + return 0; + +symlink: + ret = extract_multiimage_symlink(oldpath, path, ctx, dentry); + if (ret) + return ret; + dentry->was_linked = 1; + return 0; + +hardlink: + ret = extract_hardlink(oldpath, path, ctx); + if (ret) + return ret; + dentry->was_linked = 1; + return 0; +} + +/* This is a wrapper around do_dentry_extract_skeleton() that handles building + * the path, doing short name reordering. This is also idempotent; dentries + * already processed have skeleton_extracted set and no action is taken. See + * apply_operations.requires_short_name_reordering for more details about short + * name reordering. */ +static int +dentry_extract_skeleton(struct wim_dentry *dentry, struct apply_ctx *ctx) +{ + tchar path[ctx->ops->path_max]; + struct wim_dentry *orig_dentry; + struct wim_dentry *other_dentry; + int ret; + + if (dentry->skeleton_extracted) + return 0; + + orig_dentry = NULL; + if (ctx->supported_features.short_names + && ctx->ops->requires_short_name_reordering + && !dentry_has_short_name(dentry) + && !dentry->d_inode->i_dos_name_extracted) + { + inode_for_each_dentry(other_dentry, dentry->d_inode) { + if (dentry_has_short_name(other_dentry) + && !other_dentry->skeleton_extracted + && dentry_in_list(other_dentry)) + { + DEBUG("Creating %"TS" before %"TS" " + "to guarantee correct DOS name extraction", + dentry_full_path(other_dentry), + dentry_full_path(dentry)); + orig_dentry = dentry; + dentry = other_dentry; + break; + } + } + } +again: + if (!build_extraction_path(path, dentry, ctx)) + return 0; + ret = do_dentry_extract_skeleton(path, dentry, ctx); + if (ret) + return ret; + + dentry->skeleton_extracted = 1; + + if (orig_dentry) { + dentry = orig_dentry; + orig_dentry = NULL; + goto again; + } + dentry->d_inode->i_dos_name_extracted = 1; + return 0; +} + +/* Create a file or directory, then immediately extract all streams. The WIM + * may not be read sequentially by this function. */ +static int +dentry_extract(struct wim_dentry *dentry, struct apply_ctx *ctx) +{ + tchar path[ctx->ops->path_max]; + int ret; + + ret = dentry_extract_skeleton(dentry, ctx); + if (ret) + return ret; + + 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; + int open_flags; + +retry: + name = ttempnam(NULL, T("wimlib")); + if (name == NULL) { + 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 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(struct wim_lookup_table_entry *lte, + u32 flags, void *_ctx) +{ + struct apply_ctx *ctx = _ctx; + int ret; + + if (flags & BEGIN_STREAM_FLAG_WHOLE_STREAM) { + DEBUG("Whole stream (size=%"PRIu64") will be read into memory", + lte->size); + ctx->cur_stream = lte; + filedes_invalidate(&ctx->tmpfile_fd); + return 0; + } + + if (!need_tmpfile_to_extract(lte, + (flags & BEGIN_STREAM_FLAG_PARTIAL_RESOURCE))) + { + 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 +extract_chunk(const void *chunk, size_t size, void *_ctx) +{ + struct apply_ctx *ctx = _ctx; + int ret; + + if (filedes_valid(&ctx->tmpfile_fd)) { + ret = full_write(&ctx->tmpfile_fd, chunk, size); + if (ret) + ERROR_WITH_ERRNO("Error writing to file descriptor"); + } else { + struct wim_lookup_table_entry lte_override; + + memcpy(<e_override, ctx->cur_stream, + sizeof(struct wim_lookup_table_entry)); + + lte_override.resource_location = RESOURCE_IN_ATTACHED_BUFFER; + lte_override.size = size; + lte_override.attached_buffer = (void *)chunk; + + ret = extract_stream_instances(ctx->cur_stream, <e_override, ctx); + } + return ret; +} + +static int +end_extract_stream(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; + + if (!filedes_valid(&ctx->tmpfile_fd)) + return status; + + 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, + .begin_stream_ctx = ctx, + .consume_chunk = extract_chunk, + .consume_chunk_ctx = ctx, + .end_stream = end_extract_stream, + .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}; +} - INIT_LIST_HEAD(stream_list); - for (i = 0; i < num_streams; i++) - list_add_tail(&array[i]->extraction_list, stream_list); - FREE(array); +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; } /* - * Extract a dentry to standard output. + * Build the preliminary list of dentries to be extracted. * - * 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. + * 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_dentry_to_stdout(struct wim_dentry *dentry) +static void +build_dentry_list(struct list_head *dentry_list, struct wim_dentry **trees, + size_t num_trees, bool add_ancestors) { - int ret = 0; - if (dentry->d_inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT | - FILE_ATTRIBUTE_DIRECTORY)) - { - ERROR("\"%"TS"\" is not a regular file and therefore cannot be " - "extracted to standard output", dentry_full_path(dentry)); - ret = WIMLIB_ERR_NOT_A_REGULAR_FILE; - } else { - struct wim_lookup_table_entry *lte; - - lte = inode_unnamed_lte_resolved(dentry->d_inode); - if (lte) { - ret = extract_wim_resource_to_fd(lte, STDOUT_FILENO, - wim_resource_size(lte)); + INIT_LIST_HEAD(dentry_list); + + /* Add the trees recursively. */ + for (size_t i = 0; i < num_trees; i++) + for_dentry_in_tree(trees[i], dentry_append_to_list, dentry_list); + + /* If requested, add ancestors of the trees. */ + if (add_ancestors) { + for (size_t i = 0; i < num_trees; i++) { + struct wim_dentry *dentry = trees[i]; + struct wim_dentry *ancestor; + struct list_head *place_after; + + if (dentry_is_root(dentry)) + continue; + + place_after = dentry_list; + ancestor = dentry; + do { + ancestor = ancestor->parent; + if (dentry_in_list(ancestor)) { + place_after = &ancestor->extraction_list; + break; + } + } while (!dentry_is_root(ancestor)); + + ancestor = dentry; + do { + ancestor = ancestor->parent; + if (dentry_in_list(ancestor)) + break; + list_add(&ancestor->extraction_list, place_after); + } while (!dentry_is_root(ancestor)); } } - return ret; +} + +static const struct apply_operations * +select_apply_operations(int extract_flags) +{ +#ifdef WITH_NTFS_3G + if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) + return &ntfs_3g_apply_ops; +#endif +#ifdef __WIN32__ + return &win32_apply_ops; +#else + return &unix_apply_ops; +#endif } #ifdef __WIN32__ @@ -504,80 +1817,55 @@ file_name_valid(utf16lechar *name, size_t num_chars, bool fix) return true; } -/* - * dentry_calculate_extraction_path- - * - * Calculate the actual filename component at which a WIM dentry will be - * extracted, handling invalid filenames "properly". - * - * dentry->extraction_name usually will be set the same as dentry->file_name (on - * UNIX, converted into the platform's multibyte encoding). However, if the - * file name contains characters that are not valid on the current platform or - * has some other format that is not valid, leave dentry->extraction_name as - * NULL and clear dentry->needs_extraction to indicate that this dentry should - * not be extracted, unless the appropriate flag - * WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES is set in the extract flags, in - * which case a substitute filename will be created and set instead. - * - * Conflicts with case-insensitive names on Windows are handled similarly; see - * below. - */ static int -dentry_calculate_extraction_path(struct wim_dentry *dentry, void *_args) +dentry_calculate_extraction_name(struct wim_dentry *dentry, + struct apply_ctx *ctx) { - struct apply_args *args = _args; int ret; - dentry->needs_extraction = 1; - - if (dentry == args->extract_root) + if (dentry == ctx->target_dentry) return 0; - if (dentry_is_dot_or_dotdot(dentry)) { - /* WIM files shouldn't contain . or .. entries. But if they are - * there, don't attempt to extract them. */ - WARNING("Skipping extraction of unexpected . or .. file \"%"TS"\"", - dentry_full_path(dentry)); + if (!dentry_is_supported(dentry, &ctx->supported_features)) goto skip_dentry; - } -#ifdef __WIN32__ - struct wim_dentry *other; - list_for_each_entry(other, &dentry->case_insensitive_conflict_list, - case_insensitive_conflict_list) + if (!ctx->ops->supports_case_sensitive_filenames) { - if (other->needs_extraction) { - if (args->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; + struct wim_dentry *other; + list_for_each_entry(other, &dentry->d_ci_conflict_list, + d_ci_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; + } } } } -#endif if (file_name_valid(dentry->file_name, dentry->file_name_nbytes / 2, false)) { -#ifdef __WIN32__ - 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, + ret = utf16le_get_tstr(dentry->file_name, dentry->file_name_nbytes, - &dentry->extraction_name, + (const tchar **)&dentry->extraction_name, &dentry->extraction_name_nchars); -#endif + dentry->extraction_name_nchars /= sizeof(tchar); + return ret; } else { - if (args->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES) + if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES) { WARNING("\"%"TS"\" has an invalid filename " "that is not supported on this platform; " @@ -599,539 +1887,1096 @@ out_replace: 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; + const tchar *tchar_name; size_t tchar_nchars; - #ifdef __WIN32__ - 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); + + ret = utf16le_get_tstr(utf16_name_copy, + dentry->file_name_nbytes, + &tchar_name, &tchar_nchars); if (ret) return ret; - #endif + + 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)"), - ++args->invalid_sequence); - #ifndef __WIN32__ - FREE(tchar_name); - #endif - dentry->extraction_name = memdup(fixed_name, 2 * fixed_name_num_chars + 2); + ++ctx->invalid_sequence); + + utf16le_put_tstr(tchar_name); + + 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: - dentry->needs_extraction = 0; - dentry->not_extracted = 1; + 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_reset_needs_extraction(struct wim_dentry *dentry, void *_ignore) +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; + } - dentry->needs_extraction = 0; - dentry->not_extracted = 0; - inode->i_visited = 0; - inode->i_dos_name_extracted = 0; - FREE(inode->i_extracted_file); - inode->i_extracted_file = NULL; - if ((void*)dentry->extraction_name != (void*)dentry->file_name) - FREE(dentry->extraction_name); - dentry->extraction_name = NULL; + 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; } -#define WINDOWS_NT_MAX_PATH 32768 +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; +} /* - * extract_tree - Extract a file or directory tree from the currently selected - * WIM image. + * 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. * - * @wim: WIMStruct for the WIM file, with the desired image selected - * (as wim->current_image). - * @wim_source_path: - * "Canonical" (i.e. no leading or trailing slashes, path - * separators forwald slashes) path inside the WIM image to - * extract. An empty string means the full image. - * @target: - * Filesystem path to extract the file or directory tree to. + * Also, if doing a sequential extraction, build a mapping from each stream to + * the dentries referencing it. * - * @extract_flags: - * WIMLIB_EXTRACT_FLAG_*. Also, the private flag - * WIMLIB_EXTRACT_FLAG_MULTI_IMAGE will be set if this is being - * called through wimlib_extract_image() with WIMLIB_ALL_IMAGES as - * the image. + * This also initializes the extract progress info with byte and stream + * information. * - * @progress_func: - * If non-NULL, progress function for the extraction. The messages - * we may in this function are: + * 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. * - * WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN or - * WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN; - * WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN; - * WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END; - * WIMLIB_PROGRESS_MSG_EXTRACT_DENTRY; - * WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS; - * WIMLIB_PROGRESS_MSG_APPLY_TIMESTAMPS; - * WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END or - * WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END. + * Note: this uses the i_visited member of the inodes (assumed to be 0 + * initially), but does not reset it. * - * Returns 0 on success; nonzero on failure. + * Possible error codes: WIMLIB_ERR_NOMEM. */ static int -extract_tree(WIMStruct *wim, const tchar *wim_source_path, const tchar *target, - int extract_flags, wimlib_progress_func_t progress_func) +dentry_list_ref_streams(struct list_head *dentry_list, struct apply_ctx *ctx) { + struct wim_dentry *dentry; int ret; - struct list_head stream_list; - struct apply_args args; - const struct apply_operations *ops; - struct wim_dentry *root; - memset(&args, 0, sizeof(args)); + 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); - args.w = wim; - args.target = target; - args.target_nchars = tstrlen(target); - args.extract_flags = extract_flags; - args.progress_func = progress_func; + list_for_each_entry(dentry, dentry_list, extraction_list) + dentry->d_inode->i_visited = 0; +} -#ifdef __WIN32__ - /* Work around defective behavior in Windows where paths longer than 260 - * characters are not supported by default; instead they need to be - * turned into absolute paths and prefixed with "\\?\". */ - args.target_lowlevel_path = MALLOC(WINDOWS_NT_MAX_PATH * sizeof(wchar_t)); - if (!args.target_lowlevel_path) +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) { - ret = WIMLIB_ERR_NOMEM; - goto out; + ERROR("Extraction backend does not support hard links!"); + return WIMLIB_ERR_UNSUPPORTED; } - args.target_lowlevel_path_nchars = - GetFullPathName(args.target, WINDOWS_NT_MAX_PATH - 4, - &args.target_lowlevel_path[4], NULL); + if (required_features->hard_links && !supported_features->hard_links) + WARNING("Extracting %lu hard links as independent files", + required_features->hard_links); - if (args.target_lowlevel_path_nchars == 0 || - args.target_lowlevel_path_nchars >= WINDOWS_NT_MAX_PATH - 4) + /* Symbolic links and reparse points. */ + if ((extract_flags & WIMLIB_EXTRACT_FLAG_SYMLINK) && + !supported_features->symlink_reparse_points) { - WARNING("Can't get full path name for \"%ls\"", args.target); - FREE(args.target_lowlevel_path); - args.target_lowlevel_path = NULL; - } else { - wmemcpy(args.target_lowlevel_path, L"\\\\?\\", 4); - args.target_lowlevel_path_nchars += 4; + 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); + } } -#endif - if (progress_func) { - args.progress.extract.wimfile_name = wim->filename; - args.progress.extract.image = wim->current_image; - args.progress.extract.extract_flags = (extract_flags & - WIMLIB_EXTRACT_MASK_PUBLIC); - args.progress.extract.image_name = wimlib_get_image_name(wim, - wim->current_image); - args.progress.extract.extract_root_wim_source_path = wim_source_path; - args.progress.extract.target = target; + /* 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; } -#ifdef 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 `%"TS"'", - target); - ret = WIMLIB_ERR_NTFS_3G; - goto out_free_target_lowlevel_path; + /* 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; } - ops = &ntfs_apply_operations; - } else + 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; +} + +static void +do_extract_warnings(struct apply_ctx *ctx) +{ + if (ctx->partial_security_descriptors == 0 && + ctx->no_security_descriptors == 0) + return; + + WARNING("Extraction to \"%"TS"\" complete, but with one or more warnings:", + ctx->target); + if (ctx->partial_security_descriptors != 0) { + WARNING("- Could only partially set the security descriptor\n" + " on %lu files or directories.", + ctx->partial_security_descriptors); + } + if (ctx->no_security_descriptors != 0) { + WARNING("- Could not set security descriptor at all\n" + " on %lu files or directories.", + ctx->no_security_descriptors); + } +#ifdef __WIN32__ + WARNING("To fully restore all security descriptors, run the program\n" + " with Administrator rights."); #endif - ops = &normal_apply_operations; +} + +static void +destroy_dentry_list(struct list_head *dentry_list) +{ + struct wim_dentry *dentry, *tmp; + struct wim_inode *inode; + + list_for_each_entry_safe(dentry, tmp, dentry_list, extraction_list) { + inode = dentry->d_inode; + dentry_reset_extraction_list_node(dentry); + dentry->was_linked = 0; + dentry->skeleton_extracted = 0; + inode->i_visited = 0; + FREE(inode->i_extracted_file); + inode->i_extracted_file = NULL; + inode->i_dos_name_extracted = 0; + if ((void*)dentry->extraction_name != (void*)dentry->file_name) + FREE(dentry->extraction_name); + dentry->extraction_name = NULL; + } +} + +static void +destroy_stream_list(struct list_head *stream_list) +{ + struct wim_lookup_table_entry *lte; + + list_for_each_entry(lte, stream_list, extraction_list) + if (lte->out_refcnt > ARRAY_LEN(lte->inline_lte_dentries)) + FREE(lte->lte_dentries); +} - root = get_dentry(wim, wim_source_path); - if (!root) { - ERROR("Path \"%"TS"\" does not exist in WIM image %d", - wim_source_path, wim->current_image); - ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST; - goto out_ntfs_umount; +static int +extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees, + const tchar *target, int extract_flags, + wimlib_progress_func_t progress_func) +{ + int ret; + struct apply_ctx ctx; + struct list_head dentry_list; + struct wim_features required_features; + + /* Handle stdout extraction as a separate case. */ + if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) + return extract_dentries_to_stdout(trees, num_trees, + wim->lookup_table); + + /* Start initializing the apply_ctx. */ + memset(&ctx, 0, sizeof(struct apply_ctx)); + ctx.wim = wim; + ctx.extract_flags = extract_flags; + ctx.target = target; + ctx.target_nchars = tstrlen(target); + ctx.progress_func = progress_func; + if (progress_func) { + ctx.progress.extract.wimfile_name = wim->filename; + ctx.progress.extract.image = wim->current_image; + ctx.progress.extract.extract_flags = (extract_flags & + WIMLIB_EXTRACT_MASK_PUBLIC); + ctx.progress.extract.image_name = wimlib_get_image_name(wim, + wim->current_image); + ctx.progress.extract.target = target; } - args.extract_root = root; - /* Calculate the actual filename component of each extracted dentry, and - * in the process set the dentry->needs_extraction flag on dentries that - * will be extracted. */ - ret = for_dentry_in_tree(root, dentry_calculate_extraction_path, &args); + ctx.target_dentry = wim_root_dentry(wim); + /* Note: ctx.target_dentry represents the dentry that gets extracted to + * @target. There may be none, in which case it gets set to the image + * root and never matches any of the dentries actually being extracted. + */ + + num_trees = remove_duplicate_trees(trees, num_trees); + + /* All trees are now distinct. */ + + num_trees = remove_contained_trees(trees, num_trees); + + /* All trees are now distinct and non-overlapping. */ + + /* Build list of dentries to be extracted. */ + build_dentry_list(&dentry_list, trees, num_trees, + !(extract_flags & WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE)); + + /* Select the appropriate apply_operations based on the platform and + * extract_flags. */ + ctx.ops = select_apply_operations(extract_flags); + + /* Figure out whether the root dentry is being extracted to the root of + * a volume and therefore needs to be treated "specially", for example + * not being explicitly created and not having attributes set. */ + if (ctx.ops->target_is_root && ctx.ops->root_directory_is_special) + ctx.root_dentry_is_special = ctx.ops->target_is_root(target); + + /* Call the start_extract() callback. This gives the apply_operations + * implementation a chance to do any setup needed to access the volume. + * Furthermore, start_extract() is expected to set the supported + * features of this extraction mode (ctx.supported_features), which are + * determined at runtime as they may vary depending on the actual + * volume. */ + ret = ctx.ops->start_extract(target, &ctx); + if (ret) + goto out_destroy_dentry_list; + + /* Get and check the features required to extract the dentries. */ + dentry_list_get_features(&dentry_list, &required_features); + ret = do_feature_check(&required_features, &ctx.supported_features, + extract_flags, ctx.ops); if (ret) - goto out_dentry_reset_needs_extraction; + goto out_finish_or_abort_extract; + + ctx.supported_attributes_mask = + compute_supported_attributes_mask(&ctx.supported_features); - /* Build a list of the streams that need to be extracted */ - ret = find_streams_for_extraction(root, - &stream_list, - wim->lookup_table, extract_flags); + /* Calculate extraction name for each dentry and remove subtrees that + * can't be extracted due to naming problems. */ + ret = dentry_list_calculate_extraction_names(&dentry_list, &ctx); if (ret) - goto out_dentry_reset_needs_extraction; + goto out_finish_or_abort_extract; - /* Calculate the number of bytes of data that will be extracted */ - calculate_bytes_to_extract(&stream_list, extract_flags, - &args.progress); + /* Build list of streams to extract. */ + ret = dentry_list_resolve_streams(&dentry_list, &ctx); + if (ret) + goto out_finish_or_abort_extract; + INIT_LIST_HEAD(&ctx.stream_list); + ret = dentry_list_ref_streams(&dentry_list, &ctx); + if (ret) + goto out_destroy_stream_list; + + if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) { + /* When extracting from a pipe, the number of bytes of data to + * 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); + } + } - if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) { - ret = extract_dentry_to_stdout(root); - goto out_dentry_reset_needs_extraction; + if (ctx.ops->realpath_works_on_nonexisting_files && + ((extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) || + ctx.ops->requires_realtarget_in_paths)) + { + ctx.realtarget = realpath(target, NULL); + if (!ctx.realtarget) { + ret = WIMLIB_ERR_NOMEM; + goto out_destroy_stream_list; + } + ctx.realtarget_nchars = tstrlen(ctx.realtarget); + #ifdef __WIN32__ + /* Strip trailing slashes. If we don't do this, we may create a + * path with multiple consecutive backslashes, which for some + * reason causes Windows to report that the file cannot be found. + */ + while (ctx.realtarget_nchars >= 2 + && ctx.realtarget[ctx.realtarget_nchars - 1] == L'\\' + && ctx.realtarget[ctx.realtarget_nchars - 2] != L':') + { + ctx.realtarget[--ctx.realtarget_nchars] = L'\0'; + } + #endif } if (progress_func) { - progress_func(*wim_source_path ? WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN : - WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN, - &args.progress); - } - - /* If a sequential extraction was specified, sort the streams to be - * extracted by their position in the WIM file, so that the WIM file can - * be read sequentially. */ - if (extract_flags & WIMLIB_EXTRACT_FLAG_SEQUENTIAL) { - 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; + int msg; + if (extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) + msg = WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN; + else + msg = WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN; + progress_func(msg, &ctx.progress); + } + + if (!ctx.root_dentry_is_special) { + tchar path[ctx.ops->path_max]; + if (build_extraction_path(path, ctx.target_dentry, &ctx)) { + ret = extract_inode(path, &ctx, ctx.target_dentry->d_inode); + if (ret) + goto out_free_realtarget; + } + } + + /* If we need to fix up the targets of absolute symbolic links + * (WIMLIB_EXTRACT_FLAG_RPFIX) or the extraction mode requires paths to + * be absolute, use realpath() (or its replacement on Windows) to get + * the absolute path to the extraction target. Note that this requires + * the target directory to exist, unless + * realpath_works_on_nonexisting_files is set in the apply_operations. + * */ + if (!ctx.realtarget && + (((extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) && + required_features.symlink_reparse_points) || + ctx.ops->requires_realtarget_in_paths)) + { + ctx.realtarget = realpath(target, NULL); + if (!ctx.realtarget) { + ret = WIMLIB_ERR_NOMEM; + goto out_free_realtarget; } + ctx.realtarget_nchars = tstrlen(ctx.realtarget); } - if (progress_func) { - progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN, - &args.progress); + if (ctx.ops->requires_short_name_reordering) { + if (progress_func) + progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN, + &ctx.progress); + ret = extract_dir_structure(&dentry_list, &ctx); + if (ret) + goto out_free_realtarget; } - /* Make the directory structure and extract empty files */ - args.extract_flags |= WIMLIB_EXTRACT_FLAG_NO_STREAMS; - args.apply_dentry = ops->apply_dentry; - ret = for_dentry_in_tree(root, maybe_apply_dentry, &args); - args.extract_flags &= ~WIMLIB_EXTRACT_FLAG_NO_STREAMS; - if (ret) - goto out_dentry_reset_needs_extraction; - - if (progress_func) { - progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END, - &args.progress); - } + /* Finally, the important part: extract the tree of files. */ + if (!(extract_flags & WIMLIB_EXTRACT_FLAG_FILE_ORDER)) { + /* Sequential extraction requested, so two passes are needed + * (one for file structure, one for streams.) */ + if (progress_func && !ctx.ops->requires_short_name_reordering) + progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN, + &ctx.progress); - if (extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) { - args.target_realpath = realpath(target, NULL); - if (!args.target_realpath) { - ret = WIMLIB_ERR_NOMEM; - goto out_dentry_reset_needs_extraction; + if (!(extract_flags & WIMLIB_EXTRACT_FLAG_RESUME)) { + ret = extract_structure(&dentry_list, &ctx); + if (ret) + goto out_free_realtarget; } - args.target_realpath_len = tstrlen(args.target_realpath); + if (progress_func) + progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END, + &ctx.progress); + if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) + ret = extract_streams_from_pipe(&ctx); + else + ret = extract_stream_list(&ctx); + if (ret) + goto out_free_realtarget; + } else { + /* Sequential extraction was not requested, so we can make do + * with one pass where we both create the files and extract + * streams. */ + if (progress_func && !ctx.ops->requires_short_name_reordering) + progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN, + &ctx.progress); + ret = extract_dentries(&dentry_list, &ctx); + if (ret) + goto out_free_realtarget; + if (progress_func) + progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END, + &ctx.progress); } - /* Extract non-empty files */ - ret = apply_stream_list(&stream_list, &args, ops, progress_func); - if (ret) - goto out_free_target_realpath; - - if (progress_func) { - progress_func(WIMLIB_PROGRESS_MSG_APPLY_TIMESTAMPS, - &args.progress); + /* If the total number of bytes to extract was miscalculated, just jump + * to the calculated number in order to avoid confusing the progress + * function. This should only occur when extracting from a pipe. */ + if (ctx.progress.extract.completed_bytes != ctx.progress.extract.total_bytes) + { + DEBUG("Calculated %"PRIu64" bytes to extract, but actually " + "extracted %"PRIu64, + ctx.progress.extract.total_bytes, + ctx.progress.extract.completed_bytes); + } + if (progress_func && + ctx.progress.extract.completed_bytes < ctx.progress.extract.total_bytes) + { + ctx.progress.extract.completed_bytes = ctx.progress.extract.total_bytes; + progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS, &ctx.progress); } - /* Apply timestamps */ - ret = for_dentry_in_tree_depth(root, - ops->apply_dentry_timestamps, &args); + /* Apply security descriptors and timestamps. This is done at the end, + * and in a depth-first manner, to prevent timestamps from getting + * changed by subsequent extract operations and to minimize the chance + * of the restored security descriptors getting in our way. */ + if (progress_func) + progress_func(WIMLIB_PROGRESS_MSG_APPLY_TIMESTAMPS, + &ctx.progress); + ret = extract_final_metadata(&dentry_list, &ctx); if (ret) - goto out_free_target_realpath; + goto out_free_realtarget; if (progress_func) { - progress_func(*wim_source_path ? WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END : - WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END, - &args.progress); - } -out_free_target_realpath: - FREE(args.target_realpath); -out_dentry_reset_needs_extraction: - for_dentry_in_tree(root, dentry_reset_needs_extraction, NULL); -out_ntfs_umount: -#ifdef WITH_NTFS_3G - /* Unmount the NTFS volume */ - if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) { - if (ntfs_umount(args.vol, FALSE) != 0) { - ERROR_WITH_ERRNO("Failed to unmount NTFS volume `%"TS"'", - args.target); - if (ret == 0) - ret = WIMLIB_ERR_NTFS_3G; - } + int msg; + if (extract_flags & WIMLIB_EXTRACT_FLAG_IMAGEMODE) + msg = WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END; + else + msg = WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END; + progress_func(msg, &ctx.progress); } -#endif -out_free_target_lowlevel_path: -#ifdef __WIN32__ - FREE(args.target_lowlevel_path); -#endif -out: + do_extract_warnings(&ctx); + ret = 0; +out_free_realtarget: + FREE(ctx.realtarget); +out_destroy_stream_list: + if (!(ctx.extract_flags & WIMLIB_EXTRACT_FLAG_FILE_ORDER)) + destroy_stream_list(&ctx.stream_list); +out_finish_or_abort_extract: + if (ret) { + if (ctx.ops->abort_extract) + ctx.ops->abort_extract(&ctx); + } else { + if (ctx.ops->finish_extract) + ret = ctx.ops->finish_extract(&ctx); + } +out_destroy_dentry_list: + destroy_dentry_list(&dentry_list); return ret; } -/* Validates a single wimlib_extract_command, mostly checking to make sure the - * extract flags make sense. */ static int -check_extract_command(struct wimlib_extract_command *cmd, int wim_header_flags) +mkdir_if_needed(const tchar *target) { - int extract_flags; - bool is_entire_image = (cmd->wim_source_path[0] == T('\0')); - - /* Empty destination path? */ - if (cmd->fs_dest_path[0] == T('\0')) - return WIMLIB_ERR_INVALID_PARAM; + struct stat stbuf; + if (tstat(target, &stbuf)) { + if (errno == ENOENT) { + if (tmkdir(target, 0755)) { + ERROR_WITH_ERRNO("Failed to create directory " + "\"%"TS"\"", target); + return WIMLIB_ERR_MKDIR; + } + } else { + ERROR_WITH_ERRNO("Failed to stat \"%"TS"\"", target); + return WIMLIB_ERR_STAT; + } + } else if (!S_ISDIR(stbuf.st_mode)) { + ERROR("\"%"TS"\" is not a directory", target); + return WIMLIB_ERR_NOTDIR; + } + return 0; +} - extract_flags = cmd->extract_flags; +/* Make sure the extraction flags make sense, and update them if needed. */ +static int +check_extract_flags(const WIMStruct *wim, int *extract_flags_p) +{ + int extract_flags = *extract_flags_p; - /* Specified both symlink and hardlink modes? */ + /* Check for invalid flag combinations */ if ((extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK | WIMLIB_EXTRACT_FLAG_HARDLINK)) == (WIMLIB_EXTRACT_FLAG_SYMLINK | WIMLIB_EXTRACT_FLAG_HARDLINK)) return WIMLIB_ERR_INVALID_PARAM; -#ifdef __WIN32__ - /* Wanted UNIX data on Windows? */ - if (extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) { - ERROR("Extracting UNIX data is not supported on Windows"); + 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; - } - /* Wanted linked extraction on Windows? (XXX This is possible, just not - * implemented yet.) */ - if (extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK | - WIMLIB_EXTRACT_FLAG_HARDLINK)) - { - ERROR("Linked extraction modes are not supported on Windows"); + + if ((extract_flags & + (WIMLIB_EXTRACT_FLAG_RPFIX | + WIMLIB_EXTRACT_FLAG_NORPFIX)) == (WIMLIB_EXTRACT_FLAG_RPFIX | + WIMLIB_EXTRACT_FLAG_NORPFIX)) + return WIMLIB_ERR_INVALID_PARAM; + + if ((extract_flags & + (WIMLIB_EXTRACT_FLAG_RESUME | + WIMLIB_EXTRACT_FLAG_FROM_PIPE)) == WIMLIB_EXTRACT_FLAG_RESUME) return WIMLIB_ERR_INVALID_PARAM; - } -#endif +#ifndef WITH_NTFS_3G if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) { - /* NTFS-3g extraction mode requested */ -#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 (!is_entire_image && - (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS)) - { - ERROR("When applying directly to a NTFS volume you can " - "only extract a full image, not part of one"); - return WIMLIB_ERR_INVALID_PARAM; - } - if (extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) { - ERROR("Cannot restore UNIX-specific data in " - "the NTFS extraction mode"); - 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"); + ERROR("wimlib was compiled without support for NTFS-3g, so\n" + " it cannot apply a WIM image directly to a NTFS volume."); return WIMLIB_ERR_UNSUPPORTED; -#endif } +#endif - if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX | - WIMLIB_EXTRACT_FLAG_NORPFIX)) == - (WIMLIB_EXTRACT_FLAG_RPFIX | WIMLIB_EXTRACT_FLAG_NORPFIX)) - { - ERROR("Cannot specify RPFIX and NORPFIX flags at the same time!"); - return WIMLIB_ERR_INVALID_PARAM; +#ifndef __WIN32__ + if (extract_flags & WIMLIB_EXTRACT_FLAG_WIMBOOT) { + ERROR("WIMBoot extraction is only supported on Windows!"); + return WIMLIB_ERR_UNSUPPORTED; } +#endif if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX | - WIMLIB_EXTRACT_FLAG_NORPFIX)) == 0) + WIMLIB_EXTRACT_FLAG_NORPFIX | + WIMLIB_EXTRACT_FLAG_IMAGEMODE)) == + WIMLIB_EXTRACT_FLAG_IMAGEMODE) { - /* Do reparse point fixups by default if the WIM header says - * they are enabled and we are extracting a full image. */ - if ((wim_header_flags & WIM_HDR_FLAG_RP_FIX) && is_entire_image) + /* 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; } - if (!is_entire_image && (extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX)) { - ERROR("Cannot specify --rpfix when not extracting entire image"); - return WIMLIB_ERR_INVALID_PARAM; + /* TODO: Since UNIX data entries are stored in the file resources, in a + * completely sequential extraction they may come up before the + * corresponding file or symbolic link data. This needs to be handled + * better. */ + if ((extract_flags & (WIMLIB_EXTRACT_FLAG_UNIX_DATA | + WIMLIB_EXTRACT_FLAG_FILE_ORDER)) + == WIMLIB_EXTRACT_FLAG_UNIX_DATA) + { + if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) { + WARNING("Setting UNIX file/owner group may " + "be impossible on some\n" + " symbolic links " + "when applying from a pipe."); + } else { + extract_flags |= WIMLIB_EXTRACT_FLAG_FILE_ORDER; + WARNING("Disabling sequential extraction for " + "UNIX data mode"); + } } - cmd->extract_flags = extract_flags; + *extract_flags_p = extract_flags; return 0; } +static u32 +get_wildcard_flags(int extract_flags) +{ + u32 wildcard_flags = 0; + + if (extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_GLOB) + wildcard_flags |= WILDCARD_FLAG_ERROR_IF_NO_MATCH; + else + wildcard_flags |= WILDCARD_FLAG_WARN_IF_NO_MATCH; + + if (default_ignore_case) + wildcard_flags |= WILDCARD_FLAG_CASE_INSENSITIVE; + + return wildcard_flags; +} + +struct append_dentry_ctx { + struct wim_dentry **dentries; + size_t num_dentries; + size_t num_alloc_dentries; +}; -/* Internal function to execute extraction commands for a WIM image. */ static int -do_wimlib_extract_files(WIMStruct *wim, - int image, - struct wimlib_extract_command *cmds, - size_t num_cmds, - wimlib_progress_func_t progress_func) +append_dentry_cb(struct wim_dentry *dentry, void *_ctx) +{ + struct append_dentry_ctx *ctx = _ctx; + + if (ctx->num_dentries == ctx->num_alloc_dentries) { + struct wim_dentry **new_dentries; + size_t new_length; + + new_length = max(ctx->num_alloc_dentries + 8, + ctx->num_alloc_dentries * 3 / 2); + new_dentries = REALLOC(ctx->dentries, + new_length * sizeof(ctx->dentries[0])); + if (new_dentries == NULL) + return WIMLIB_ERR_NOMEM; + ctx->dentries = new_dentries; + ctx->num_alloc_dentries = new_length; + } + ctx->dentries[ctx->num_dentries++] = dentry; + return 0; +} + +static int +do_wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target, + const tchar * const *paths, size_t num_paths, + int extract_flags, wimlib_progress_func_t progress_func) { int ret; - bool found_link_cmd = false; - bool found_nolink_cmd = false; + struct wim_dentry **trees; + size_t num_trees; + + if (wim == NULL || target == NULL || target[0] == T('\0') || + (num_paths != 0 && paths == NULL)) + return WIMLIB_ERR_INVALID_PARAM; + + ret = check_extract_flags(wim, &extract_flags); + if (ret) + return ret; - /* Select the image from which we are extracting files */ ret = select_wim_image(wim, image); if (ret) return ret; - /* Make sure there are no streams in the WIM that have not been - * checksummed yet. */ ret = wim_checksum_unhashed_streams(wim); if (ret) return ret; - /* Check for problems with the extraction commands */ - for (size_t i = 0; i < num_cmds; i++) { - ret = check_extract_command(&cmds[i], wim->hdr.flags); - if (ret) - return ret; - if (cmds[i].extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK | - WIMLIB_EXTRACT_FLAG_HARDLINK)) { - found_link_cmd = true; - } else { - found_nolink_cmd = true; - } - if (found_link_cmd && found_nolink_cmd) { - ERROR("Symlink or hardlink extraction mode must " - "be set on all extraction commands"); - return WIMLIB_ERR_INVALID_PARAM; - } - } - - /* Execute the extraction commands */ - for (size_t i = 0; i < num_cmds; i++) { - ret = extract_tree(wim, - cmds[i].wim_source_path, - cmds[i].fs_dest_path, - cmds[i].extract_flags, - progress_func); + if ((extract_flags & (WIMLIB_EXTRACT_FLAG_NTFS | + WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE)) == + (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE)) + { + ret = mkdir_if_needed(target); if (ret) return ret; } - return 0; -} - -/* Extract files or directories from a WIM image. */ -WIMLIBAPI int -wimlib_extract_files(WIMStruct *wim, - int image, - const struct wimlib_extract_command *cmds, - size_t num_cmds, - int default_extract_flags, - WIMStruct **additional_swms, - unsigned num_additional_swms, - wimlib_progress_func_t progress_func) -{ - int ret; - struct wimlib_extract_command *cmds_copy; - int all_flags = 0; - - default_extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC; - ret = verify_swm_set(wim, additional_swms, num_additional_swms); - if (ret) - goto out; + if (extract_flags & WIMLIB_EXTRACT_FLAG_GLOB_PATHS) { - if (num_cmds == 0) - goto out; + struct append_dentry_ctx append_dentry_ctx = { + .dentries = NULL, + .num_dentries = 0, + .num_alloc_dentries = 0, + }; - if (num_additional_swms) - merge_lookup_tables(wim, additional_swms, num_additional_swms); + u32 wildcard_flags = get_wildcard_flags(extract_flags); - cmds_copy = CALLOC(num_cmds, sizeof(cmds[0])); - if (!cmds_copy) { - ret = WIMLIB_ERR_NOMEM; - goto out_restore_lookup_table; - } + for (size_t i = 0; i < num_paths; i++) { + tchar *path = canonicalize_wim_path(paths[i]); + if (path == NULL) { + ret = WIMLIB_ERR_NOMEM; + trees = append_dentry_ctx.dentries; + goto out_free_trees; + } + ret = expand_wildcard(wim, path, + append_dentry_cb, + &append_dentry_ctx, + wildcard_flags); + FREE(path); + if (ret) { + trees = append_dentry_ctx.dentries; + goto out_free_trees; + } + } + trees = append_dentry_ctx.dentries; + num_trees = append_dentry_ctx.num_dentries; + } else { + trees = MALLOC(num_paths * sizeof(trees[0])); + if (trees == NULL) + return WIMLIB_ERR_NOMEM; - for (size_t i = 0; i < num_cmds; i++) { - cmds_copy[i].extract_flags = (default_extract_flags | - cmds[i].extract_flags) - & WIMLIB_EXTRACT_MASK_PUBLIC; - all_flags |= cmds_copy[i].extract_flags; + for (size_t i = 0; i < num_paths; i++) { - cmds_copy[i].wim_source_path = canonicalize_wim_path(cmds[i].wim_source_path); - if (!cmds_copy[i].wim_source_path) { - ret = WIMLIB_ERR_NOMEM; - goto out_free_cmds_copy; - } + tchar *path = canonicalize_wim_path(paths[i]); + if (path == NULL) { + ret = WIMLIB_ERR_NOMEM; + goto out_free_trees; + } - cmds_copy[i].fs_dest_path = canonicalize_fs_path(cmds[i].fs_dest_path); - if (!cmds_copy[i].fs_dest_path) { - ret = WIMLIB_ERR_NOMEM; - goto out_free_cmds_copy; + trees[i] = get_dentry(wim, path, + WIMLIB_CASE_PLATFORM_DEFAULT); + FREE(path); + if (trees[i] == NULL) { + ERROR("Path \"%"TS"\" does not exist " + "in WIM image %d", + paths[i], wim->current_image); + ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST; + goto out_free_trees; + } } - + num_trees = num_paths; } - ret = do_wimlib_extract_files(wim, image, - cmds_copy, num_cmds, - progress_func); - if (all_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK | - WIMLIB_EXTRACT_FLAG_HARDLINK)) - { - for_lookup_table_entry(wim->lookup_table, - lte_free_extracted_file, NULL); + if (num_trees == 0) { + ret = 0; + goto out_free_trees; } -out_free_cmds_copy: - for (size_t i = 0; i < num_cmds; i++) { - FREE(cmds_copy[i].wim_source_path); - FREE(cmds_copy[i].fs_dest_path); - } - FREE(cmds_copy); -out_restore_lookup_table: - if (num_additional_swms) - unmerge_lookup_table(wim); -out: + + ret = extract_trees(wim, trees, num_trees, + target, extract_flags, progress_func); +out_free_trees: + FREE(trees); return ret; } -/* - * Extracts an image from a WIM file. - * - * @wim: WIMStruct for the WIM file. - * - * @image: Number of the single image to extract. - * - * @target: Directory or NTFS volume to extract the image to. - * - * @extract_flags: Bitwise or of WIMLIB_EXTRACT_FLAG_*. - * - * @progress_func: If non-NULL, a progress function to be called - * periodically. - * - * Returns 0 on success; nonzero on failure. - */ static int extract_single_image(WIMStruct *wim, int image, const tchar *target, int extract_flags, wimlib_progress_func_t progress_func) { - int ret; - tchar *target_copy = canonicalize_fs_path(target); - if (!target_copy) - return WIMLIB_ERR_NOMEM; - struct wimlib_extract_command cmd = { - .wim_source_path = T(""), - .fs_dest_path = target_copy, - .extract_flags = extract_flags, - }; - ret = do_wimlib_extract_files(wim, image, &cmd, 1, progress_func); - FREE(target_copy); - return ret; + const tchar *path = WIMLIB_WIM_ROOT_PATH; + extract_flags |= WIMLIB_EXTRACT_FLAG_IMAGEMODE; + return do_wimlib_extract_paths(wim, image, target, &path, 1, + extract_flags, progress_func); } static const tchar * const filename_forbidden_chars = @@ -1168,26 +3013,17 @@ extract_all_images(WIMStruct *wim, int ret; int image; const tchar *image_name; - struct stat stbuf; - if (tstat(target, &stbuf)) { - if (errno == ENOENT) - { - if (tmkdir(target, S_IRWXU | S_IRGRP | S_IXGRP | - S_IROTH | S_IXOTH)) - { - 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; + 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; } + 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++) { @@ -1207,45 +3043,254 @@ extract_all_images(WIMStruct *wim, return 0; } -/* Extracts a single image or all images from a WIM file to a directory or NTFS - * volume. */ +static void +clear_lte_extracted_file(WIMStruct *wim, int extract_flags) +{ + if (unlikely(extract_flags & (WIMLIB_EXTRACT_FLAG_SYMLINK | + WIMLIB_EXTRACT_FLAG_HARDLINK))) + for_lookup_table_entry(wim->lookup_table, + lte_free_extracted_file, NULL); +} + +static int +do_wimlib_extract_image(WIMStruct *wim, + int image, + const tchar *target, + int extract_flags, + wimlib_progress_func_t progress_func) +{ + int ret; + + if (extract_flags & (WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE | + WIMLIB_EXTRACT_FLAG_TO_STDOUT | + WIMLIB_EXTRACT_FLAG_GLOB_PATHS)) + return WIMLIB_ERR_INVALID_PARAM; + + if (image == WIMLIB_ALL_IMAGES) + ret = extract_all_images(wim, target, extract_flags, + progress_func); + else + ret = extract_single_image(wim, image, target, extract_flags, + progress_func); + + clear_lte_extracted_file(wim, extract_flags); + return ret; +} + + +/**************************************************************************** + * Extraction API * + ****************************************************************************/ + WIMLIBAPI int -wimlib_extract_image(WIMStruct *wim, - int image, - const tchar *target, - int extract_flags, - WIMStruct **additional_swms, - unsigned num_additional_swms, - wimlib_progress_func_t progress_func) +wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target, + const tchar * const *paths, size_t num_paths, + int extract_flags, wimlib_progress_func_t progress_func) +{ + int ret; + + if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC) + return WIMLIB_ERR_INVALID_PARAM; + + ret = do_wimlib_extract_paths(wim, image, target, paths, num_paths, + extract_flags, progress_func); + clear_lte_extracted_file(wim, extract_flags); + return ret; +} + +WIMLIBAPI int +wimlib_extract_pathlist(WIMStruct *wim, int image, const tchar *target, + const tchar *path_list_file, int extract_flags, + wimlib_progress_func_t progress_func) +{ + int ret; + tchar **paths; + size_t num_paths; + void *mem; + + ret = read_path_list_file(path_list_file, &paths, &num_paths, &mem); + if (ret) { + ERROR("Failed to read path list file \"%"TS"\"", + path_list_file); + return ret; + } + + ret = wimlib_extract_paths(wim, image, target, + (const tchar * const *)paths, num_paths, + extract_flags, progress_func); + FREE(paths); + FREE(mem); + return ret; +} + +WIMLIBAPI int +wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name, + const tchar *target, int extract_flags, + wimlib_progress_func_t progress_func) { int ret; + WIMStruct *pwm; + struct filedes *in_fd; + int image; + unsigned i; + + if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC) + return WIMLIB_ERR_INVALID_PARAM; - extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC; + if (extract_flags & WIMLIB_EXTRACT_FLAG_FILE_ORDER) + return WIMLIB_ERR_INVALID_PARAM; - ret = verify_swm_set(wim, additional_swms, num_additional_swms); + /* 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, progress_func); if (ret) return ret; - if (num_additional_swms) - merge_lookup_tables(wim, additional_swms, num_additional_swms); + /* 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) { - ret = extract_all_images(wim, target, - extract_flags | WIMLIB_EXTRACT_FLAG_MULTI_IMAGE, - progress_func); - } else { - ret = extract_single_image(wim, 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(wim->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; + } + + /* 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; + } } - if (num_additional_swms) - unmerge_lookup_table(wim); + /* Extract the image. */ + extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE; + ret = do_wimlib_extract_image(pwm, image, target, + extract_flags, progress_func); + /* Clean up and return. */ +out_wimlib_free: + wimlib_free(pwm); return ret; } + +WIMLIBAPI int +wimlib_extract_image(WIMStruct *wim, int image, const tchar *target, + int extract_flags, wimlib_progress_func_t progress_func) +{ + if (extract_flags & ~WIMLIB_EXTRACT_MASK_PUBLIC) + return WIMLIB_ERR_INVALID_PARAM; + return do_wimlib_extract_image(wim, image, target, extract_flags, + progress_func); +}