X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Fextract.c;h=446b55551b29337b6e2d5fdd08e027b58fa2162b;hp=df144785c92ce76e43d0b8937c92365a8cb89290;hb=d1301021c8edf915004faf291f31e4c99b0bcfde;hpb=b6f6a919c8291da9cf2a9ea72ec7f8e47fbd79cf diff --git a/src/extract.c b/src/extract.c index df144785..446b5555 100644 --- a/src/extract.c +++ b/src/extract.c @@ -24,405 +24,1389 @@ * along with wimlib; if not, see http://www.gnu.org/licenses/. */ +/* + * This file provides the API functions wimlib_extract_image(), + * wimlib_extract_files(), and wimlib_extract_image_from_pipe(). Internally, + * all end up calling extract_tree() zero or more times to extract a tree of + * files from the currently selected WIM image to the specified target directory + * or NTFS volume. + * + * 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/paths.h" +#include "wimlib/reparse.h" #include "wimlib/resource.h" +#include "wimlib/security.h" #include "wimlib/swm.h" #ifdef __WIN32__ # include "wimlib/win32.h" /* for realpath() equivalent */ #endif #include "wimlib/xml.h" +#include "wimlib/wim.h" #include -#include -#ifdef WITH_NTFS_3G -# include /* for ntfs_mount(), ntfs_umount() */ -#endif +#include #include #include #include -#define MAX_LONG_PATH_WARNINGS 5 +#define WIMLIB_EXTRACT_FLAG_MULTI_IMAGE 0x80000000 +#define WIMLIB_EXTRACT_FLAG_FROM_PIPE 0x40000000 +#define WIMLIB_EXTRACT_MASK_PUBLIC 0x3fffffff +/* Given a WIM dentry in the tree to be extracted, resolve all streams in the + * corresponding inode and set 'out_refcnt' in each to 0. */ static int -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 *)) +dentry_resolve_and_zero_lte_refcnt(struct wim_dentry *dentry, void *_ctx) { - tchar *p; - size_t extraction_path_nchars; - struct wim_dentry *d; - LIST_HEAD(ancestor_list); - const tchar *target; - size_t target_nchars; + struct apply_ctx *ctx = _ctx; + struct wim_inode *inode = dentry->d_inode; + struct wim_lookup_table_entry *lte; + int ret; + bool force = false; -#ifdef __WIN32__ - if (args->target_lowlevel_path) { - target = args->target_lowlevel_path; - target_nchars = args->target_lowlevel_path_nchars; - } else -#endif + if (dentry->extraction_skipped) + return 0; + + /* Special case: when extracting from a pipe, the WIM lookup table is + * initially empty, so "resolving" an inode's streams is initially not + * possible. However, we still need to keep track of which streams, + * identified by SHA1 message digests, need to be extracted, so we + * "resolve" the inode's streams anyway by allocating new entries. */ + if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) + force = true; + ret = inode_resolve_ltes(inode, ctx->wim->lookup_table, force); + if (ret) + return ret; + for (unsigned i = 0; i <= inode->i_num_ads; i++) { + lte = inode_stream_lte_resolved(inode, i); + if (lte) + lte->out_refcnt = 0; + } + return 0; +} + +static inline bool +is_linked_extraction(const struct apply_ctx *ctx) +{ + return 0 != (ctx->extract_flags & (WIMLIB_EXTRACT_FLAG_HARDLINK | + WIMLIB_EXTRACT_FLAG_SYMLINK)); +} + +static inline bool +can_extract_named_data_streams(const struct apply_ctx *ctx) +{ + return ctx->supported_features.named_data_streams && + !is_linked_extraction(ctx); +} + +static int +ref_stream_to_extract(struct wim_lookup_table_entry *lte, + struct wim_dentry *dentry, struct apply_ctx *ctx) +{ + if (!lte) + return 0; + + if (likely(!is_linked_extraction(ctx)) || (lte->out_refcnt == 0 && + lte->extracted_file == NULL)) { - target = args->target; - target_nchars = args->target_nchars; + ctx->progress.extract.total_bytes += wim_resource_size(lte); + ctx->progress.extract.num_streams++; } - extraction_path_nchars = target_nchars; + if (lte->out_refcnt == 0) { + list_add_tail(<e->extraction_list, &ctx->stream_list); + ctx->num_streams_remaining++; + } - 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); + if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_SEQUENTIAL) { + struct wim_dentry **lte_dentries; + + /* Append dentry to this stream's array of dentries referencing + * it. Use inline array to avoid memory allocation until the + * number of dentries becomes too large. */ + if (lte->out_refcnt < ARRAY_LEN(lte->inline_lte_dentries)) { + lte_dentries = lte->inline_lte_dentries; + } else { + struct wim_dentry **prev_lte_dentries; + size_t alloc_lte_dentries; + + if (lte->out_refcnt == ARRAY_LEN(lte->inline_lte_dentries)) { + prev_lte_dentries = NULL; + alloc_lte_dentries = ARRAY_LEN(lte->inline_lte_dentries); + } else { + prev_lte_dentries = lte->lte_dentries; + alloc_lte_dentries = lte->alloc_lte_dentries; + } + + if (lte->out_refcnt == alloc_lte_dentries) { + alloc_lte_dentries *= 2; + lte_dentries = REALLOC(prev_lte_dentries, + alloc_lte_dentries * + sizeof(lte_dentries[0])); + if (!lte_dentries) + return WIMLIB_ERR_NOMEM; + if (prev_lte_dentries == NULL) { + memcpy(lte_dentries, + lte->inline_lte_dentries, + sizeof(lte->inline_lte_dentries)); + } + lte->lte_dentries = lte_dentries; + lte->alloc_lte_dentries = alloc_lte_dentries; + } + lte_dentries = lte->lte_dentries; + } + lte_dentries[lte->out_refcnt] = dentry; } + lte->out_refcnt++; + return 0; +} - tchar extraction_path[extraction_path_nchars + 1]; - p = tmempcpy(extraction_path, target, target_nchars); +/* Given a WIM dentry in the tree to be extracted, iterate through streams that + * need to be extracted. For each one, add it to the list of streams to be + * extracted (ctx->stream_list) if not already done so, and also update the + * progress information (ctx->progress) with the stream. Furthermore, if doing + * a sequential extraction, build a mapping from each the stream to the dentries + * referencing it. */ +static int +dentry_add_streams_to_extract(struct wim_dentry *dentry, void *_ctx) +{ + struct apply_ctx *ctx = _ctx; + struct wim_inode *inode = dentry->d_inode; + int ret; + /* Don't process dentries marked as skipped. */ + if (dentry->extraction_skipped) + return 0; - list_for_each_entry(d, &ancestor_list, tmp_list) { - *p++ = OS_PREFERRED_PATH_SEPARATOR; - p = tmempcpy(p, d->extraction_name, d->extraction_name_nchars); + /* Don't process additional hard links. */ + if (inode->i_visited && ctx->supported_features.hard_links) + return 0; + + /* The unnamed data stream will always be extracted, except in an + * unlikely case. */ + if (!inode_is_encrypted_directory(inode)) { + ret = ref_stream_to_extract(inode_unnamed_lte_resolved(inode), + dentry, ctx); + if (ret) + return ret; } - *p = T('\0'); -#ifdef __WIN32__ - /* Warn the user if the path exceeds MAX_PATH */ + /* Named data streams will be extracted only if supported in the current + * extraction mode and volume, and to avoid complications, if not doing + * a linked extraction. */ + if (can_extract_named_data_streams(ctx)) { + for (u16 i = 0; i < inode->i_num_ads; i++) { + if (!ads_entry_is_named_stream(&inode->i_ads_entries[i])) + continue; + ret = ref_stream_to_extract(inode->i_ads_entries[i].lte, + dentry, ctx); + if (ret) + return ret; + } + } + inode->i_visited = 1; + return 0; +} + +/* Inform library user of progress of stream extraction following the successful + * extraction of a copy of the stream specified by @lte. */ +static void +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; - /* + 1 for '\0', -4 for \\?\. */ - if (extraction_path_nchars + 1 - 4 > MAX_PATH) { - if (dentry->needs_extraction && - args->num_long_paths < MAX_LONG_PATH_WARNINGS) + progress->extract.completed_bytes += wim_resource_size(lte); + 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) { - WARNING("Path \"%ls\" exceeds MAX_PATH and will not be accessible " - "to most Windows software", extraction_path); - if (++args->num_long_paths == MAX_LONG_PATH_WARNINGS) - WARNING("Suppressing further warnings about long paths"); + 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; } } -#endif - return (*apply_dentry_func)(extraction_path, extraction_path_nchars, - dentry, args); } +#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; + } -/* Extracts a file, directory, or symbolic link from the WIM archive. */ + /* 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; + } + buf[sret] = '\0'; + + 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; + } + + /* 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; + } + + /* Account for reparse data consumed. */ + update_extract_progress(ctx, + (lte_override ? lte_override : + inode_unnamed_lte_resolved(inode))); + return 0; +} +#endif /* !__WIN32__ */ + +/* 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); + if (ret) { + ERROR_WITH_ERRNO("Failed to create the directory " + "\"%"TS"\"", path); + } + } else { + ret = ctx->ops->create_file(path, ctx); + if (ret) { + ERROR_WITH_ERRNO("Failed to create the file " + "\"%"TS"\"", path); + } + } + return ret; +} + +static int +extract_hardlink(const tchar *oldpath, const tchar *newpath, + struct apply_ctx *ctx) +{ + int ret; + + 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; } +#ifdef __WIN32__ +static int +try_extract_rpfix(u8 *rpbuf, + u16 *rpbuflen_p, + const wchar_t *extract_root_realpath, + unsigned extract_root_realpath_nchars) +{ + 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; + + utf16lechar *new_target; + utf16lechar *new_print_name; + size_t new_target_nchars; + size_t new_print_name_nchars; + utf16lechar *p; -/* Apply timestamps to an extracted file or directory */ + ret = parse_reparse_data(rpbuf, *rpbuflen_p, &rpdata); + if (ret) + return ret; + + 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); + } + + /* 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]; + } + /* 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__ */ + +/* Set reparse data on extracted file or directory that has + * FILE_ATTRIBUTE_REPARSE_POINT set. */ static int -apply_dentry_timestamps_normal(struct wim_dentry *dentry, void *arg) +extract_reparse_data(const tchar *path, struct apply_ctx *ctx, + struct wim_inode *inode, + struct wim_lookup_table_entry *lte_override) { + 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; + #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); + /* 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)) + { + WARNING("Reparse point fixup of \"%"TS"\" " + "failed", path); + ret = 0; + } + if (ret) + goto error; + } +#endif + + 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; + + /* 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 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 +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 wim_inode *inode = dentry->d_inode; + struct wim_lookup_table_entry *lte; + int ret; + + if (dentry->was_hardlinked) + return 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 + + /* 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) + ret = ctx->ops->extract_encrypted_stream(path, lte, ctx); + else + ret = ctx->ops->extract_unnamed_stream(path, lte, ctx); + 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(path, entry->stream_name, + entry->stream_name_nbytes / 2, + lte, ctx); + if (ret) + goto error; + update_extract_progress(ctx, lte); + } + } + 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 +extract_file_attributes(const tchar *path, struct apply_ctx *ctx, + struct wim_dentry *dentry) +{ + int ret; + + if (ctx->ops->set_file_attributes) { + if (dentry == ctx->extract_root && ctx->root_dentry_is_special) + return 0; + ret = ctx->ops->set_file_attributes(path, + dentry->d_inode->i_attributes, + ctx); + if (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 +extract_short_name(const tchar *path, struct apply_ctx *ctx, + struct wim_dentry *dentry) +{ + int ret; + + /* The root of the dentry tree being extracted may not be extracted to + * its original name, so its short name should be ignored. */ + if (dentry == ctx->extract_root) + 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; +} + +/* 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; + + if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_NO_ACLS) + return 0; + + if ((ctx->extract_root == 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 (errno != EACCES) { + WARNING_WITH_ERRNO("Failed to set " + "security descriptor " + "on \"%"TS"\"", path); + } + } + } + } + return 0; +} + +/* Set timestamps on an extracted file. Failure is warning-only unless + * WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS is set. */ +static int +extract_timestamps(const tchar *path, struct apply_ctx *ctx, + struct wim_dentry *dentry) +{ + struct wim_inode *inode = dentry->d_inode; + int ret; + + if ((ctx->extract_root == 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; + } 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_dot_or_dotdot(const struct wim_dentry *dentry) +dentry_is_supported(struct wim_dentry *dentry, + const struct wim_features *supported_features) { - 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'))); + struct wim_inode *inode = dentry->d_inode; + + if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) { + if (supported_features->reparse_points) + return true; + if (supported_features->symlink_reparse_points && + inode_is_symlink(inode)) + return true; + return false; + } + return 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->max_path characters including the null-terminator. + * + * 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, + 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; + + if (dentry->extraction_skipped) + return false; + + 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->extract_root; d = d->parent) { + 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; + + while (*path) { + while (*path == path_separator) + path++; + if (*path) + num_components++; + while (*path && *path != path_separator) + path++; + } + return num_components; +} + +static int +extract_multiimage_symlink(const tchar *oldpath, const tchar *newpath, + struct apply_ctx *ctx, struct wim_dentry *dentry) +{ + 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->extract_root; 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--; + } + + p_old = oldpath; + 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++; + } + + 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->extract_root) { + 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++) { + struct wim_ads_entry *entry = &inode->i_ads_entries[i]; + + if (!ads_entry_is_named_stream(entry)) + continue; + if (entry->lte) + continue; + ret = ctx->ops->extract_named_stream(path, + 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); + 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_hardlinked = 1; + return 0; + +hardlink: + ret = extract_hardlink(oldpath, path, ctx); + if (ret) + return ret; + dentry->was_hardlinked = 1; + return 0; +} + +static int +dentry_extract_skeleton(struct wim_dentry *dentry, void *_ctx) +{ + struct apply_ctx *ctx = _ctx; + tchar path[ctx->ops->path_max]; + struct wim_dentry *orig_dentry; + struct wim_dentry *other_dentry; + int ret; + + /* Here we may re-order the extraction of multiple names (hard links) + * for the same file in the same directory in order to ensure the short + * (DOS) name is set correctly. A short name is always associated with + * exactly one long name, and at least on NTFS, only one long name for a + * file can have a short name associated with it. (More specifically, + * there can be unlimited names in the POSIX namespace, but only one + * name can be in the Win32+DOS namespace, or one name in the Win32 + * namespace with a corresponding name in the DOS namespace.) To ensure + * the short name of a file is associated with the correct long name in + * a directory, we extract the long name with a corresponding short name + * before any additional names. This can affect NTFS-3g extraction + * (which uses ntfs_set_ntfs_dos_name(), which doesn't allow specifying + * the long name to associate with a short name) and may affect Win32 + * extraction as well (which uses SetFileShortName()). */ + + if (dentry->skeleton_extracted) + return 0; + orig_dentry = NULL; + if (ctx->supported_features.short_names + && !dentry_has_short_name(dentry) + && !dentry->d_inode->i_dos_name_extracted) + { + inode_for_each_dentry(other_dentry, dentry->d_inode) { + if (dentry_has_short_name(other_dentry) + && !other_dentry->skeleton_extracted + && other_dentry->parent == dentry->parent) + { + 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; } -/* 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. */ +/* Create a file or directory, then immediately extract all streams. This + * assumes that WIMLIB_EXTRACT_FLAG_SEQUENTIAL is not specified, since the WIM + * may not be read sequentially by this function. */ static int -maybe_apply_dentry(struct wim_dentry *dentry, void *arg) +dentry_extract(struct wim_dentry *dentry, void *_ctx) { - struct apply_args *args = arg; + struct apply_ctx *ctx = _ctx; + tchar path[ctx->ops->path_max]; int ret; - if (!dentry->needs_extraction) - return 0; + ret = dentry_extract_skeleton(dentry, ctx); + 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 (!build_extraction_path(path, dentry, ctx)) return 0; - 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); - } - ret = args->apply_dentry(dentry, args); - if (ret == 0) - dentry->needs_extraction = 0; - return ret; + return extract_streams(path, ctx, dentry, NULL, NULL); } -static void -calculate_bytes_to_extract(struct list_head *stream_list, - int extract_flags, - union wimlib_progress_info *progress) +/* Extract all instances of the stream @lte that are being extracted in this + * call of extract_tree(). @can_seek specifies whether the WIM file descriptor + * is seekable or not (e.g. is a pipe). If not and the stream needs to be + * extracted multiple times, it is extracted to a temporary file first. + * + * This is intended for use with sequential extraction of a WIM image + * (WIMLIB_EXTRACT_FLAG_SEQUENTIAL specified). */ +static int +extract_stream_instances(struct wim_lookup_table_entry *lte, + struct apply_ctx *ctx, bool can_seek) { - struct wim_lookup_table_entry *lte; - u64 total_bytes = 0; - u64 num_streams = 0; + struct wim_dentry **lte_dentries; + struct wim_lookup_table_entry *lte_tmp = NULL; + struct wim_lookup_table_entry *lte_override; + tchar *stream_tmp_filename = NULL; + tchar path[ctx->ops->path_max]; + unsigned i; + int ret; - /* 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)) - { - /* 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); + if (lte->out_refcnt <= ARRAY_LEN(lte->inline_lte_dentries)) + lte_dentries = lte->inline_lte_dentries; + else + lte_dentries = lte->lte_dentries; + + if (likely(can_seek || lte->out_refcnt < 2)) { + lte_override = lte; + } else { + /* Need to extract stream to temporary file. */ + struct filedes fd; + int raw_fd; + + stream_tmp_filename = ttempnam(NULL, T("wimlib")); + if (!stream_tmp_filename) { + ERROR_WITH_ERRNO("Failed to create temporary filename"); + ret = WIMLIB_ERR_OPEN; + goto out; + } + + lte_tmp = memdup(lte, sizeof(struct wim_lookup_table_entry)); + if (!lte_tmp) { + ret = WIMLIB_ERR_NOMEM; + goto out_free_stream_tmp_filename; } + lte_tmp->resource_location = RESOURCE_IN_FILE_ON_DISK; + lte_tmp->file_on_disk = stream_tmp_filename; + lte_override = lte_tmp; + + raw_fd = topen(stream_tmp_filename, + O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, 0600); + if (raw_fd < 0) { + ERROR_WITH_ERRNO("Failed to open temporary file"); + ret = WIMLIB_ERR_OPEN; + goto out_free_lte_tmp; + } + filedes_init(&fd, raw_fd); + ret = extract_wim_resource_to_fd(lte, &fd, + wim_resource_size(lte)); + if (filedes_close(&fd) && !ret) + ret = WIMLIB_ERR_WRITE; + if (ret) + goto out_unlink_stream_tmp_file; } - progress->extract.num_streams = num_streams; - progress->extract.total_bytes = total_bytes; - progress->extract.completed_bytes = 0; -} -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); + /* Extract all instances of the stream, reading either from the stream + * in the WIM file or from the temporary file containing the stream. + * dentry->tmp_flag is used to ensure that each dentry is processed only + * once regardless of how many times this stream appears in the streams + * of the corresponding inode. */ + for (i = 0; i < lte->out_refcnt; i++) { + struct wim_dentry *dentry = lte_dentries[i]; + + if (dentry->tmp_flag) + continue; + if (!build_extraction_path(path, dentry, ctx)) + continue; + ret = extract_streams(path, ctx, dentry, + lte, lte_override); + if (ret) + goto out_clear_tmp_flags; + dentry->tmp_flag = 1; } + ret = 0; +out_clear_tmp_flags: + for (i = 0; i < lte->out_refcnt; i++) + lte_dentries[i]->tmp_flag = 0; +out_unlink_stream_tmp_file: + if (stream_tmp_filename) + tunlink(stream_tmp_filename); +out_free_lte_tmp: + FREE(lte_tmp); +out_free_stream_tmp_filename: + FREE(stream_tmp_filename); +out: + return ret; } -struct find_streams_ctx { - struct list_head stream_list; - int extract_flags; -}; - +/* Extracts a list of streams (ctx.stream_list), assuming that the directory + * structure and empty files were already created. This relies on the + * per-`struct wim_lookup_table_entry' list of dentries that reference each + * stream that was constructed earlier. Streams are extracted exactly in the + * order of the stream list; however, unless the WIM's file descriptor is + * detected to be non-seekable, streams may be read from the WIM file more than + * one time if multiple copies need to be extracted. */ static int -dentry_find_streams_to_extract(struct wim_dentry *dentry, void *_ctx) +extract_stream_list(struct apply_ctx *ctx) { - 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; - - if (!dentry->needs_extraction) - return 0; + bool can_seek; + int ret; - 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) -#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; - } - } - } - } + can_seek = (lseek(ctx->wim->in_fd.fd, 0, SEEK_CUR) != -1); + list_for_each_entry(lte, &ctx->stream_list, extraction_list) { + ret = extract_stream_instances(lte, ctx, can_seek); + if (ret) + return ret; } - inode->i_visited = 1; return 0; } +#define PWM_ALLOW_WIM_HDR 0x00001 +#define PWM_SILENT_EOF 0x00002 + +/* Read the header from a stream in a pipable WIM. */ static int -dentry_resolve_and_zero_lte_refcnt(struct wim_dentry *dentry, void *_lookup_table) +read_pwm_stream_header(WIMStruct *pwm, struct wim_lookup_table_entry *lte, + int flags) { - struct wim_inode *inode = dentry->d_inode; - struct wim_lookup_table *lookup_table = _lookup_table; - struct wim_lookup_table_entry *lte; + struct pwm_stream_hdr stream_hdr; + int ret; - inode_resolve_ltes(inode, lookup_table); - for (unsigned i = 0; i <= inode->i_num_ads; i++) { - lte = inode_stream_lte_resolved(inode, i); - if (lte) - lte->out_refcnt = 0; + ret = full_read(&pwm->in_fd, &stream_hdr, sizeof(stream_hdr)); + if (ret) + goto read_error; + + if ((flags & PWM_ALLOW_WIM_HDR) && stream_hdr.magic == PWM_MAGIC) { + u8 buf[WIM_HEADER_DISK_SIZE - sizeof(stream_hdr)]; + ret = full_read(&pwm->in_fd, buf, sizeof(buf)); + if (ret) + goto read_error; + lte->resource_location = RESOURCE_NONEXISTENT; + return 0; + } + + if (stream_hdr.magic != PWM_STREAM_MAGIC) { + ERROR("Data read on pipe is invalid (expected stream header)."); + return WIMLIB_ERR_INVALID_PIPABLE_WIM; + } + + lte->resource_entry.original_size = le64_to_cpu(stream_hdr.uncompressed_size); + copy_hash(lte->hash, stream_hdr.hash); + lte->resource_entry.flags = le32_to_cpu(stream_hdr.flags); + lte->resource_entry.offset = pwm->in_fd.offset; + lte->resource_location = RESOURCE_IN_WIM; + lte->wim = pwm; + if (lte->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED) { + lte->compression_type = pwm->compression_type; + lte->resource_entry.size = 0; + } else { + lte->compression_type = WIMLIB_COMPRESSION_TYPE_NONE; + lte->resource_entry.size = lte->resource_entry.original_size; } + lte->is_pipable = 1; return 0; + +read_error: + if (ret != WIMLIB_ERR_UNEXPECTED_END_OF_FILE || !(flags & PWM_SILENT_EOF)) + ERROR_WITH_ERRNO("Error reading pipable WIM from pipe"); + return ret; } -static void -find_streams_for_extraction(struct wim_dentry *root, - struct list_head *stream_list, - struct wim_lookup_table *lookup_table, - int extract_flags) +/* Skip over an unneeded stream in a pipable WIM being read from a pipe. */ +static int +skip_pwm_stream(struct wim_lookup_table_entry *lte) { - struct find_streams_ctx ctx; - - INIT_LIST_HEAD(&ctx.stream_list); - ctx.extract_flags = extract_flags; - for_dentry_in_tree(root, dentry_resolve_and_zero_lte_refcnt, lookup_table); - for_dentry_in_tree(root, dentry_find_streams_to_extract, &ctx); - list_transfer(&ctx.stream_list, stream_list); + return read_partial_wim_resource(lte, wim_resource_size(lte), + NULL, NULL, + WIMLIB_READ_RESOURCE_FLAG_SEEK_ONLY, + 0); } -struct apply_operations { - int (*apply_dentry)(struct wim_dentry *dentry, void *arg); - int (*apply_dentry_timestamps)(struct wim_dentry *dentry, void *arg); -}; - -static const struct apply_operations normal_apply_operations = { - .apply_dentry = apply_dentry_normal, - .apply_dentry_timestamps = apply_dentry_timestamps_normal, -}; - -#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 - 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_streams_from_pipe(struct apply_ctx *ctx) { - 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_lookup_table_entry *found_lte; + struct wim_lookup_table_entry *needed_lte; + struct wim_lookup_table *lookup_table; int ret; + int pwm_flags; - /* 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) - return ret; - if (progress_func && - args->progress.extract.completed_bytes >= next_progress) + ret = WIMLIB_ERR_NOMEM; + found_lte = new_lookup_table_entry(); + if (!found_lte) + goto out; + + 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; + while (ctx->num_streams_remaining) { + ret = read_pwm_stream_header(ctx->wim, found_lte, pwm_flags); + if (ret) { + if (ret == WIMLIB_ERR_UNEXPECTED_END_OF_FILE && + (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_RESUME)) { - 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); - } + goto resume_done; } + goto out_free_found_lte; + } + + if ((found_lte->resource_location != RESOURCE_NONEXISTENT) + && !(found_lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) + && (needed_lte = __lookup_resource(lookup_table, found_lte->hash)) + && (needed_lte->out_refcnt)) + { + copy_resource_entry(&needed_lte->resource_entry, + &found_lte->resource_entry); + needed_lte->resource_location = found_lte->resource_location; + needed_lte->wim = found_lte->wim; + needed_lte->compression_type = found_lte->compression_type; + needed_lte->is_pipable = found_lte->is_pipable; + + ret = extract_stream_instances(needed_lte, ctx, false); + if (ret) + goto out_free_found_lte; + ctx->num_streams_remaining--; + } else if (found_lte->resource_location != RESOURCE_NONEXISTENT) { + ret = skip_pwm_stream(found_lte); + if (ret) + goto out_free_found_lte; } } + ret = 0; +out_free_found_lte: + free_lookup_table_entry(found_lte); +out: + return ret; + +resume_done: + /* TODO */ return 0; } +/* Finish extracting a file, directory, or symbolic link by setting file + * security and timestamps. */ static int -sort_stream_list_by_wim_position(struct list_head *stream_list) +dentry_extract_final(struct wim_dentry *dentry, void *_ctx) { - 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; - } - 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; - } + struct apply_ctx *ctx = _ctx; + int ret; + tchar path[ctx->ops->path_max]; - qsort(array, num_streams, sizeof(array[0]), cmp_streams_by_wim_position); + if (!build_extraction_path(path, dentry, ctx)) + return 0; - INIT_LIST_HEAD(stream_list); - for (i = 0; i < num_streams; i++) - list_add_tail(&array[i]->extraction_list, stream_list); - FREE(array); - return 0; + ret = extract_security(path, ctx, dentry); + if (ret) + return ret; + + return extract_timestamps(path, ctx, dentry); } /* - * Extract a dentry to standard output. + * 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 @@ -432,16 +1416,20 @@ static int extract_dentry_to_stdout(struct wim_dentry *dentry) { int ret = 0; - if (!dentry_is_regular_file(dentry)) { + 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); + "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, + struct filedes _stdout; + filedes_init(&_stdout, STDOUT_FILENO); + ret = extract_wim_resource_to_fd(lte, &_stdout, wim_resource_size(lte)); } } @@ -495,6 +1483,24 @@ file_name_valid(utf16lechar *name, size_t num_chars, bool fix) return true; } +static bool +dentry_is_dot_or_dotdot(const struct wim_dentry *dentry) +{ + const utf16lechar *file_name = dentry->file_name; + return file_name != NULL && + file_name[0] == cpu_to_le16('.') && + (file_name[1] == cpu_to_le16('\0') || + (file_name[1] == cpu_to_le16('.') && + file_name[2] == cpu_to_le16('\0'))); +} + +static int +dentry_mark_skipped(struct wim_dentry *dentry, void *_ignore) +{ + dentry->extraction_skipped = 1; + return 0; +} + /* * dentry_calculate_extraction_path- * @@ -505,7 +1511,7 @@ file_name_valid(utf16lechar *name, size_t num_chars, bool fix) * 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 + * NULL and set dentry->extraction_skipped to indicate that this dentry should * not be extracted, unless the appropriate flag * WIMLIB_EXTRACT_FLAG_REPLACE_INVALID_FILENAMES is set in the extract flags, in * which case a substitute filename will be created and set instead. @@ -516,37 +1522,42 @@ file_name_valid(utf16lechar *name, size_t num_chars, bool fix) static int dentry_calculate_extraction_path(struct wim_dentry *dentry, void *_args) { - struct apply_args *args = _args; + struct apply_ctx *ctx = _args; int ret; - dentry->needs_extraction = 1; - - if (dentry == args->extract_root) + if (dentry == ctx->extract_root || dentry->extraction_skipped) return 0; + if (!dentry_is_supported(dentry, &ctx->supported_features)) + goto skip_dentry; + if (dentry_is_dot_or_dotdot(dentry)) { /* WIM files shouldn't contain . or .. entries. But if they are * there, don't attempt to extract them. */ - WARNING("Skipping extraction of unexpected . or .. file \"%"TS"\"", - dentry_full_path(dentry)); + WARNING("Skipping extraction of unexpected . or .. file " + "\"%"TS"\"", dentry_full_path(dentry)); 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", + struct wim_dentry *other; + list_for_each_entry(other, &dentry->case_insensitive_conflict_list, + case_insensitive_conflict_list) + { + if (ctx->extract_flags & + WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) { + WARNING("\"%"TS"\" has the same " + "case-insensitive name as " + "\"%"TS"\"; extracting " + "dummy name instead", dentry_full_path(dentry), dentry_full_path(other)); goto out_replace; } else { - WARNING("Not extracting \"%"TS"\": has same case-insensitive " + WARNING("Not extracting \"%"TS"\": " + "has same case-insensitive " "name as \"%"TS"\"", dentry_full_path(dentry), dentry_full_path(other)); @@ -554,7 +1565,9 @@ dentry_calculate_extraction_path(struct wim_dentry *dentry, void *_args) } } } -#endif +#else /* __WIN32__ */ + wimlib_assert(ctx->ops->supports_case_sensitive_filenames); +#endif /* !__WIN32__ */ if (file_name_valid(dentry->file_name, dentry->file_name_nbytes / 2, false)) { #ifdef __WIN32__ @@ -568,7 +1581,7 @@ dentry_calculate_extraction_path(struct wim_dentry *dentry, void *_args) &dentry->extraction_name_nchars); #endif } 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; " @@ -583,61 +1596,319 @@ dentry_calculate_extraction_path(struct wim_dentry *dentry, void *_args) } } -out_replace: +out_replace: + { + utf16lechar utf16_name_copy[dentry->file_name_nbytes / 2]; + + memcpy(utf16_name_copy, dentry->file_name, dentry->file_name_nbytes); + file_name_valid(utf16_name_copy, dentry->file_name_nbytes / 2, true); + + tchar *tchar_name; + size_t tchar_nchars; + #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); + if (ret) + return ret; + #endif + size_t fixed_name_num_chars = tchar_nchars; + tchar fixed_name[tchar_nchars + 50]; + + tmemcpy(fixed_name, tchar_name, tchar_nchars); + fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars, + T(" (invalid filename #%lu)"), + ++ctx->invalid_sequence); + #ifndef __WIN32__ + FREE(tchar_name); + #endif + dentry->extraction_name = memdup(fixed_name, + 2 * fixed_name_num_chars + 2); + if (!dentry->extraction_name) + return WIMLIB_ERR_NOMEM; + dentry->extraction_name_nchars = fixed_name_num_chars; + } + return 0; + +skip_dentry: + for_dentry_in_tree(dentry, dentry_mark_skipped, NULL); + return 0; +} + +/* Clean up dentry and inode structure after extraction. */ +static int +dentry_reset_needs_extraction(struct wim_dentry *dentry, void *_ignore) +{ + struct wim_inode *inode = dentry->d_inode; + + dentry->extraction_skipped = 0; + dentry->was_hardlinked = 0; + dentry->skeleton_extracted = 0; + inode->i_visited = 0; + FREE(inode->i_extracted_file); + inode->i_extracted_file = NULL; + inode->i_dos_name_extracted = 0; + if ((void*)dentry->extraction_name != (void*)dentry->file_name) + FREE(dentry->extraction_name); + dentry->extraction_name = NULL; + return 0; +} + +/* Tally features necessary to extract a dentry and the corresponding inode. */ +static int +dentry_tally_features(struct wim_dentry *dentry, void *_features) +{ + struct wim_features *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) + 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; + return 0; +} + +static int +dentry_clear_inode_visited(struct wim_dentry *dentry, void *_ignore) +{ + dentry->d_inode->i_visited = 0; + return 0; +} + +/* Tally the features necessary to extract a dentry tree. */ +static void +dentry_tree_get_features(struct wim_dentry *root, struct wim_features *features) +{ + memset(features, 0, sizeof(struct wim_features)); + for_dentry_in_tree(root, dentry_tally_features, features); + for_dentry_in_tree(root, dentry_clear_inode_visited, NULL); +} + +static int +do_feature_check(const struct wim_features *required_features, + const struct wim_features *supported_features, + int extract_flags, + const struct apply_operations *ops, + const tchar *wim_source_path) +{ + const tchar *loc; + const tchar *mode = "this extraction mode"; + + if (wim_source_path[0] == '\0') + loc = "the WIM image"; + else + loc = wim_source_path; + + /* We're an archive program, so theoretically we can do what we want + * with FILE_ATTRIBUTE_ARCHIVE (which is a dumb flag anyway). Don't + * bother the user about it. */ +#if 0 + if (required_features->archive_files && !supported_features->archive_files) + { + WARNING( + "%lu files in %"TS" are marked as archived, but this attribute\n" +" is not supported in %"TS".", + required_features->archive_files, loc, mode); + } +#endif + + if (required_features->hidden_files && !supported_features->hidden_files) + { + WARNING( + "%lu files in %"TS" are marked as hidden, but this\n" +" attribute is not supported in %"TS".", + required_features->hidden_files, loc, mode); + } + + if (required_features->system_files && !supported_features->system_files) + { + WARNING( + "%lu files in %"TS" are marked as system files,\n" +" but this attribute is not supported in %"TS".", + required_features->system_files, loc, mode); + } + + if (required_features->compressed_files && !supported_features->compressed_files) + { + WARNING( + "%lu files in %"TS" are marked as being transparently\n" +" compressed, but transparent compression is not supported in\n" +" %"TS". These files will be extracted as uncompressed.", + required_features->compressed_files, loc, mode); + } + + if (required_features->encrypted_files && !supported_features->encrypted_files) + { + WARNING( + "%lu files in %"TS" are marked as being encrypted,\n" +" but encryption is not supported in %"TS". These files\n" +" will be extracted as raw encrypted data instead.", + required_features->encrypted_files, loc, mode); + } + + if (required_features->not_context_indexed_files && + !supported_features->not_context_indexed_files) + { + WARNING( + "%lu files in %"TS" are marked as not content indexed,\n" +" but this attribute is not supported in %"TS".", + required_features->not_context_indexed_files, loc, mode); + } + + if (required_features->sparse_files && !supported_features->sparse_files) + { + WARNING( + "%lu files in %"TS" are marked as sparse, but creating\n" +" sparse files is not supported in %"TS". These files\n" +" will be extracted as non-sparse.", + required_features->sparse_files, loc, mode); + } + + if (required_features->named_data_streams && + !supported_features->named_data_streams) + { + WARNING( + "%lu files in %"TS" contain one or more alternate (named)\n" +" data streams, which are not supported in %"TS".\n" +" Alternate data streams will NOT be extracted.", + required_features->named_data_streams, loc, mode); + } + + if (unlikely(extract_flags & (WIMLIB_EXTRACT_FLAG_HARDLINK | + WIMLIB_EXTRACT_FLAG_SYMLINK)) && + required_features->named_data_streams && + supported_features->named_data_streams) + { + WARNING( + "%lu files in %"TS" contain one or more alternate (named)\n" +" data streams, which are not supported in linked extraction mode.\n" +" Alternate data streams will NOT be extracted.", + required_features->named_data_streams, loc); + } + + if (required_features->hard_links && !supported_features->hard_links) + { + WARNING( + "%lu files in %"TS" are hard links, but hard links are\n" +" not supported in %"TS". Hard links will be extracted as\n" +" duplicate copies of the linked files.", + required_features->hard_links, loc, mode); + } + + if (required_features->reparse_points && !supported_features->reparse_points) + { + if (supported_features->symlink_reparse_points) { + if (required_features->other_reparse_points) { + WARNING( + "%lu files in %"TS" are reparse points that are neither\n" +" symbolic links nor junction points and are not supported in\n" +" %"TS". These reparse points will not be extracted.", + required_features->other_reparse_points, loc, + mode); + } + } else { + WARNING( + "%lu files in %"TS" are reparse points, which are\n" +" not supported in %"TS" and will not be extracted.", + required_features->reparse_points, loc, mode); + } + } + + if (required_features->security_descriptors && + !supported_features->security_descriptors) + { + WARNING( + "%lu files in %"TS" have Windows NT security descriptors,\n" +" but extracting security descriptors is not supported in\n" +" %"TS". No security descriptors will be extracted.", + required_features->security_descriptors, loc, mode); + } + + if (required_features->short_names && !supported_features->short_names) + { + WARNING( + "%lu files in %"TS" have short (DOS) names, but\n" +" extracting short names is not supported in %"TS".\n" +" Short names will not be extracted.\n", + required_features->short_names, loc, mode); + } + + if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) && + required_features->unix_data && !supported_features->unix_data) + { + ERROR("Extracting UNIX data is not supported in %"TS, mode); + return WIMLIB_ERR_UNSUPPORTED; + } + if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_SHORT_NAMES) && + required_features->short_names && !supported_features->short_names) + { + ERROR("Extracting short names is not supported in %"TS"", mode); + return WIMLIB_ERR_UNSUPPORTED; + } + if ((extract_flags & WIMLIB_EXTRACT_FLAG_STRICT_TIMESTAMPS) && + !ops->set_timestamps) + { + ERROR("Extracting timestamps is not supported in %"TS"", mode); + return WIMLIB_ERR_UNSUPPORTED; + } + if (((extract_flags & (WIMLIB_EXTRACT_FLAG_STRICT_ACLS | + WIMLIB_EXTRACT_FLAG_UNIX_DATA)) + == WIMLIB_EXTRACT_FLAG_STRICT_ACLS) && + required_features->security_descriptors && + !supported_features->security_descriptors) + { + ERROR("Extracting security descriptors is not supported in %"TS, mode); + return WIMLIB_ERR_UNSUPPORTED; + } + + if ((extract_flags & WIMLIB_EXTRACT_FLAG_HARDLINK) && + !supported_features->hard_links) + { + ERROR("Hard link extraction mode requested, but " + "%"TS" does not support hard links!", mode); + return WIMLIB_ERR_UNSUPPORTED; + } + + if ((extract_flags & WIMLIB_EXTRACT_FLAG_SYMLINK) && + !supported_features->symlink_reparse_points) { - utf16lechar utf16_name_copy[dentry->file_name_nbytes / 2]; - - memcpy(utf16_name_copy, dentry->file_name, dentry->file_name_nbytes); - file_name_valid(utf16_name_copy, dentry->file_name_nbytes / 2, true); - - tchar *tchar_name; - size_t tchar_nchars; - #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); - if (ret) - return ret; - #endif - size_t fixed_name_num_chars = tchar_nchars; - tchar fixed_name[tchar_nchars + 50]; - - tmemcpy(fixed_name, tchar_name, tchar_nchars); - fixed_name_num_chars += tsprintf(fixed_name + tchar_nchars, - T(" (invalid filename #%lu)"), - ++args->invalid_sequence); - #ifndef __WIN32__ - FREE(tchar_name); - #endif - dentry->extraction_name = memdup(fixed_name, 2 * fixed_name_num_chars + 2); - if (!dentry->extraction_name) - return WIMLIB_ERR_NOMEM; - dentry->extraction_name_nchars = fixed_name_num_chars; + ERROR("Symbolic link extraction mode requested, but " + "%"TS" does not support symbolic " + "links!", mode); + return WIMLIB_ERR_UNSUPPORTED; } return 0; -skip_dentry: - dentry->needs_extraction = 0; - dentry->not_extracted = 1; - return 0; -} - -static int -dentry_reset_needs_extraction(struct wim_dentry *dentry, void *_ignore) -{ - dentry->needs_extraction = 0; - dentry->not_extracted = 0; - dentry->is_win32_name = 0; - dentry->d_inode->i_visited = 0; - dentry->d_inode->i_dos_name_extracted = 0; - FREE(dentry->d_inode->i_extracted_file); - dentry->d_inode->i_extracted_file = NULL; - if ((void*)dentry->extraction_name != (void*)dentry->file_name) - FREE(dentry->extraction_name); - dentry->extraction_name = NULL; - return 0; } /* @@ -646,12 +1917,15 @@ dentry_reset_needs_extraction(struct wim_dentry *dentry, void *_ignore) * * @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 + * separators WIM_PATH_SEPARATOR) path inside the WIM image to * extract. An empty string means the full image. + * * @target: * Filesystem path to extract the file or directory tree to. + * (Or, with WIMLIB_EXTRACT_FLAG_NTFS: the name of a NTFS volume.) * * @extract_flags: * WIMLIB_EXTRACT_FLAG_*. Also, the private flag @@ -661,7 +1935,7 @@ dentry_reset_needs_extraction(struct wim_dentry *dentry, void *_ignore) * * @progress_func: * If non-NULL, progress function for the extraction. The messages - * we may in this function are: + * that may be sent in this function are: * * WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN or * WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN; @@ -673,196 +1947,286 @@ dentry_reset_needs_extraction(struct wim_dentry *dentry, void *_ignore) * WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END or * WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END. * - * Returns 0 on success; nonzero on failure. + * Returns 0 on success; a positive WIMLIB_ERR_* code on failure. */ static int extract_tree(WIMStruct *wim, const tchar *wim_source_path, const tchar *target, int extract_flags, wimlib_progress_func_t progress_func) { - int ret; - struct list_head stream_list; - struct apply_args args; - const struct apply_operations *ops; struct wim_dentry *root; + struct wim_features required_features; + struct apply_ctx ctx; + int ret; + struct wim_lookup_table_entry *lte; - memset(&args, 0, sizeof(args)); - - - args.w = wim; - args.target = target; - args.target_nchars = tstrlen(target); - args.extract_flags = extract_flags; - args.progress_func = progress_func; - -#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(32768 * sizeof(wchar_t)); - if (!args.target_lowlevel_path) - { - ret = WIMLIB_ERR_NOMEM; - goto out; - } - args.target_lowlevel_path[0] = L'\\'; - args.target_lowlevel_path[1] = L'\\'; - args.target_lowlevel_path[2] = L'?'; - args.target_lowlevel_path[3] = L'\\'; - args.target_lowlevel_path_nchars = - GetFullPathName(args.target, 32768 - 4, - &args.target_lowlevel_path[4], NULL); - - if (args.target_lowlevel_path_nchars == 0 || - args.target_lowlevel_path_nchars >= 32768 - 4) - { - WARNING("Can't get full path name for \"%ls\"", args.target); - FREE(args.target_lowlevel_path); - args.target_lowlevel_path = NULL; - } else { - args.target_lowlevel_path_nchars += 4; - } -#endif - + /* 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) { - 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; + 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.extract_root_wim_source_path = wim_source_path; + ctx.progress.extract.target = target; } + INIT_LIST_HEAD(&ctx.stream_list); -#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; - } - ops = &ntfs_apply_operations; - } else -#endif - ops = &normal_apply_operations; - + /* Translate the path to extract into the corresponding + * `struct wim_dentry', which will be the root of the + * "dentry tree" to extract. */ root = get_dentry(wim, wim_source_path); if (!root) { ERROR("Path \"%"TS"\" does not exist in WIM image %d", wim_source_path, wim->current_image); ret = WIMLIB_ERR_PATH_DOES_NOT_EXIST; - goto out_ntfs_umount; + goto out; } - 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.extract_root = root; + + /* Select the appropriate apply_operations based on the + * platform and extract_flags. */ +#ifdef __WIN32__ + ctx.ops = &win32_apply_ops; +#else + ctx.ops = &unix_apply_ops; +#endif + +#ifdef WITH_NTFS_3G + if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) + ctx.ops = &ntfs_3g_apply_ops; +#endif + + /* Call the start_extract() callback. This gives the apply_operations + * implementation a chance to do any setup needed to access the volume. + * Furthermore, it's expected to set the supported features of this + * extraction mode (ctx.supported_features), which are determined at + * runtime as they may vary depending on the actual volume. These + * features are then compared with the actual features extracting this + * dentry tree requires. Some mismatches will merely produce warnings + * and the unsupported data will be ignored; others will produce errors. + */ + ret = ctx.ops->start_extract(target, &ctx); + if (ret) + goto out; + + dentry_tree_get_features(root, &required_features); + ret = do_feature_check(&required_features, &ctx.supported_features, + extract_flags, ctx.ops, wim_source_path); + if (ret) + goto out_finish_or_abort_extract; + + /* Figure out whether the root dentry is being extracted to the root of + * a volume and therefore needs to be treated "specially", for example + * not being explicitly created and not having attributes set. */ + if (ctx.ops->target_is_root && ctx.ops->root_directory_is_special) + ctx.root_dentry_is_special = ctx.ops->target_is_root(target); + + /* Calculate the actual filename component of each extracted dentry. In + * the process, set the dentry->extraction_skipped flag on dentries that + * are being skipped for some reason (e.g. invalid filename). */ + ret = for_dentry_in_tree(root, dentry_calculate_extraction_path, &ctx); if (ret) goto out_dentry_reset_needs_extraction; - /* Build a list of the streams that need to be extracted */ - find_streams_for_extraction(root, - &stream_list, - wim->lookup_table, extract_flags); + /* Build the list of the streams that need to be extracted and + * initialize ctx.progress.extract with stream information. */ + ret = for_dentry_in_tree(ctx.extract_root, + dentry_resolve_and_zero_lte_refcnt, &ctx); + if (ret) + goto out_dentry_reset_needs_extraction; - /* Calculate the number of bytes of data that will be extracted */ - calculate_bytes_to_extract(&stream_list, extract_flags, - &args.progress); + ret = for_dentry_in_tree(ctx.extract_root, + dentry_add_streams_to_extract, &ctx); + if (ret) + goto out_teardown_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. */ + ctx.progress.extract.total_bytes = + wim_info_get_image_total_bytes(wim->wim_info, + wim->current_image); + } + /* Handle the special case of extracting a file to standard + * output. In that case, "root" should be a single file, not a + * directory tree. (If not, extract_dentry_to_stdout() will + * return an error.) */ if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) { ret = extract_dentry_to_stdout(root); - goto out_dentry_reset_needs_extraction; + goto out_teardown_stream_list; } - 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 | + WIMLIB_EXTRACT_FLAG_FROM_PIPE)) + == WIMLIB_EXTRACT_FLAG_SEQUENTIAL) + { + ret = sort_stream_list_by_sequential_order( + &ctx.stream_list, + offsetof(struct wim_lookup_table_entry, + extraction_list)); + if (ret) + goto out_teardown_stream_list; } - /* If 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; + 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_teardown_stream_list; } + ctx.realtarget_nchars = tstrlen(ctx.realtarget); } if (progress_func) { - progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN, - &args.progress); + progress_func(*wim_source_path ? WIMLIB_PROGRESS_MSG_EXTRACT_TREE_BEGIN : + WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_BEGIN, + &ctx.progress); } - /* 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); + if (!ctx.root_dentry_is_special) + { + tchar path[ctx.ops->path_max]; + if (build_extraction_path(path, root, &ctx)) + { + ret = extract_inode(path, &ctx, root->d_inode); + if (ret) + goto out_free_realtarget; + } } - if (extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) { - args.target_realpath = realpath(target, NULL); - if (!args.target_realpath) { + /* 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_dentry_reset_needs_extraction; + goto out_free_realtarget; } - args.target_realpath_len = tstrlen(args.target_realpath); + ctx.realtarget_nchars = tstrlen(ctx.realtarget); } - /* Extract non-empty files */ - ret = apply_stream_list(&stream_list, &args, ops, progress_func); - if (ret) - goto out_free_target_realpath; + /* Finally, the important part: extract the tree of files. */ + if (extract_flags & (WIMLIB_EXTRACT_FLAG_SEQUENTIAL | + WIMLIB_EXTRACT_FLAG_FROM_PIPE)) { + /* Sequential extraction requested, so two passes are needed + * (one for directory structure, one for streams.) */ + if (progress_func) + progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN, + &ctx.progress); + + if (!(extract_flags & WIMLIB_EXTRACT_FLAG_RESUME)) { + ret = for_dentry_in_tree(root, dentry_extract_skeleton, &ctx); + if (ret) + goto out_free_realtarget; + } + 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) + progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_BEGIN, + &ctx.progress); + ret = for_dentry_in_tree(root, dentry_extract, &ctx); + if (ret) + goto out_free_realtarget; + if (progress_func) + progress_func(WIMLIB_PROGRESS_MSG_EXTRACT_DIR_STRUCTURE_END, + &ctx.progress); + } - 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 = for_dentry_in_tree_depth(root, dentry_extract_final, &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); + &ctx.progress); } -out_free_target_realpath: - FREE(args.target_realpath); + + ret = 0; +out_free_realtarget: + FREE(ctx.realtarget); +out_teardown_stream_list: + /* Free memory allocated as part of the mapping from each + * wim_lookup_table_entry to the dentries that reference it. */ + if (ctx.extract_flags & WIMLIB_EXTRACT_FLAG_SEQUENTIAL) + list_for_each_entry(lte, &ctx.stream_list, extraction_list) + if (lte->out_refcnt > ARRAY_LEN(lte->inline_lte_dentries)) + FREE(lte->lte_dentries); out_dentry_reset_needs_extraction: for_dentry_in_tree(root, dentry_reset_needs_extraction, NULL); -out_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; - } +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); } -#endif -out_free_target_lowlevel_path: -#ifdef __WIN32__ - FREE(args.target_lowlevel_path); -#endif out: return ret; } @@ -873,7 +2237,6 @@ static int check_extract_command(struct wimlib_extract_command *cmd, int wim_header_flags) { 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')) @@ -881,77 +2244,66 @@ check_extract_command(struct wimlib_extract_command *cmd, int wim_header_flags) extract_flags = cmd->extract_flags; - /* 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 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"); +#ifndef WITH_NTFS_3G + ERROR("wimlib was compiled without support for NTFS-3g, so\n" + " we cannot apply a WIM image directly to a NTFS volume."); return WIMLIB_ERR_UNSUPPORTED; #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; - } - if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX | WIMLIB_EXTRACT_FLAG_NORPFIX)) == 0) { /* 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) + if (wim_header_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_SEQUENTIAL)) + == (WIMLIB_EXTRACT_FLAG_UNIX_DATA | + WIMLIB_EXTRACT_FLAG_SEQUENTIAL)) + { + 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_SEQUENTIAL; + WARNING("Disabling sequential extraction for " + "UNIX data mode"); + } } cmd->extract_flags = extract_flags; @@ -959,7 +2311,8 @@ check_extract_command(struct wimlib_extract_command *cmd, int wim_header_flags) } -/* Internal function to execute extraction commands for a WIM image. */ +/* Internal function to execute extraction commands for a WIM image. The paths + * in the extract commands are expected to be already "canonicalized". */ static int do_wimlib_extract_files(WIMStruct *wim, int image, @@ -977,7 +2330,7 @@ do_wimlib_extract_files(WIMStruct *wim, return ret; /* Make sure there are no streams in the WIM that have not been - * checksummed yet. */ + * checksummed yet. */ ret = wim_checksum_unhashed_streams(wim); if (ret) return ret; @@ -1013,7 +2366,7 @@ do_wimlib_extract_files(WIMStruct *wim, return 0; } -/* Extract files or directories from a WIM image. */ +/* API function documented in wimlib.h */ WIMLIBAPI int wimlib_extract_files(WIMStruct *wim, int image, @@ -1159,12 +2512,16 @@ extract_all_images(WIMStruct *wim, const tchar *image_name; struct stat stbuf; + 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; + } + if (tstat(target, &stbuf)) { - if (errno == ENOENT) - { - if (tmkdir(target, S_IRWXU | S_IRGRP | S_IXGRP | - S_IROTH | S_IXOTH)) - { + if (errno == ENOENT) { + if (tmkdir(target, 0755)) { ERROR_WITH_ERRNO("Failed to create directory \"%"TS"\"", target); return WIMLIB_ERR_MKDIR; } @@ -1178,7 +2535,7 @@ extract_all_images(WIMStruct *wim, } tmemcpy(buf, target, output_path_len); - buf[output_path_len] = T('/'); + buf[output_path_len] = OS_PREFERRED_PATH_SEPARATOR; for (image = 1; image <= wim->hdr.image_count; image++) { image_name = wimlib_get_image_name(wim, image); if (image_name_ok_as_dir(image_name)) { @@ -1196,31 +2553,31 @@ extract_all_images(WIMStruct *wim, return 0; } -/* Extracts a single image or all images from a WIM file to a directory or NTFS - * volume. */ -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) +static int +do_wimlib_extract_image(WIMStruct *wim, + int image, + const tchar *target, + int extract_flags, + WIMStruct **additional_swms, + unsigned num_additional_swms, + wimlib_progress_func_t progress_func) { int ret; - extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC; - - ret = verify_swm_set(wim, additional_swms, num_additional_swms); - if (ret) - return ret; + if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) { + wimlib_assert(wim->hdr.part_number == 1); + wimlib_assert(num_additional_swms == 0); + } else { + ret = verify_swm_set(wim, additional_swms, num_additional_swms); + if (ret) + return ret; - if (num_additional_swms) - merge_lookup_tables(wim, additional_swms, num_additional_swms); + if (num_additional_swms) + merge_lookup_tables(wim, additional_swms, num_additional_swms); + } if (image == WIMLIB_ALL_IMAGES) { - ret = extract_all_images(wim, target, - extract_flags | WIMLIB_EXTRACT_FLAG_MULTI_IMAGE, + ret = extract_all_images(wim, target, extract_flags, progress_func); } else { ret = extract_single_image(wim, image, target, extract_flags, @@ -1238,3 +2595,176 @@ wimlib_extract_image(WIMStruct *wim, unmerge_lookup_table(wim); return ret; } + +/* API function documented in wimlib.h */ +WIMLIBAPI int +wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name, + const tchar *target, int extract_flags, + wimlib_progress_func_t progress_func) +{ + int ret; + WIMStruct *pwm; + struct filedes *in_fd; + int image; + unsigned i; + + extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC; + + if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) + return WIMLIB_ERR_INVALID_PARAM; + + extract_flags |= WIMLIB_EXTRACT_FLAG_SEQUENTIAL; + + /* Read the WIM header from the pipe and get a WIMStruct to represent + * the pipable WIM. Caveats: Unlike getting a WIMStruct with + * wimlib_open_wim(), getting a WIMStruct in this way will result in + * an empty lookup table, no XML data read, and no filename set. */ + ret = open_wim_as_WIMStruct(&pipe_fd, + WIMLIB_OPEN_FLAG_FROM_PIPE | + WIMLIB_OPEN_FLAG_SPLIT_OK, + &pwm, progress_func); + if (ret) + return ret; + + /* 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; + } + + /* 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; + } + + 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.) */ + { + struct wim_lookup_table_entry xml_lte; + ret = read_pwm_stream_header(pwm, &xml_lte, 0); + if (ret) + goto out_wimlib_free; + + if (!(xml_lte.resource_entry.flags & WIM_RESHDR_FLAG_METADATA)) + { + ERROR("Expected XML data, but found non-metadata " + "stream."); + ret = WIMLIB_ERR_INVALID_PIPABLE_WIM; + goto out_wimlib_free; + } + + copy_resource_entry(&pwm->hdr.xml_res_entry, + &xml_lte.resource_entry); + + ret = read_wim_xml_data(pwm); + if (ret) + goto out_wimlib_free; + if (wim_info_get_num_images(pwm->wim_info) != pwm->hdr.image_count) { + ERROR("Image count in XML data is not the same as in WIM header."); + ret = WIMLIB_ERR_XML; + 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; + + metadata_lte = new_lookup_table_entry(); + if (!metadata_lte) { + ret = WIMLIB_ERR_NOMEM; + goto out_wimlib_free; + } + + ret = read_pwm_stream_header(pwm, metadata_lte, 0); + imd = pwm->image_metadata[i - 1]; + imd->metadata_lte = metadata_lte; + if (ret) + goto out_wimlib_free; + + if (!(metadata_lte->resource_entry.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 images 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_pwm_stream(metadata_lte); + if (ret) + goto out_wimlib_free; + } + } + /* Extract the image. */ + extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE; + ret = do_wimlib_extract_image(pwm, image, target, + extract_flags, NULL, 0, progress_func); + /* Clean up and return. */ +out_wimlib_free: + wimlib_free(pwm); + return ret; +} + +/* API function documented in wimlib.h */ +WIMLIBAPI int +wimlib_extract_image(WIMStruct *wim, + int image, + const tchar *target, + int extract_flags, + WIMStruct **additional_swms, + unsigned num_additional_swms, + wimlib_progress_func_t progress_func) +{ + extract_flags &= WIMLIB_EXTRACT_MASK_PUBLIC; + return do_wimlib_extract_image(wim, image, target, extract_flags, + additional_swms, num_additional_swms, + progress_func); +}