X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Fextract.c;h=f59cffb55347acc014da0feb0aad58f42d2150b8;hp=4f1c54a00df805cef6ff092e009ec32bc35e603a;hb=088dff37aa334c218e1cac96cc847f5dd14f7124;hpb=f03f71aeaedbe90772338635cc221bf065f356c6 diff --git a/src/extract.c b/src/extract.c index 4f1c54a0..f59cffb5 100644 --- a/src/extract.c +++ b/src/extract.c @@ -6,7 +6,7 @@ */ /* - * Copyright (C) 2012, 2013, 2014, 2015 Eric Biggers + * Copyright (C) 2012-2017 Eric Biggers * * This file is free software; you can redistribute it and/or modify it under * the terms of the GNU Lesser General Public License as published by the Free @@ -28,7 +28,7 @@ * wimlib_extract_pathlist(). Internally, all end up calling * do_wimlib_extract_paths() and extract_trees(). * - * Although wimlib supports multiple extraction modes/backends (NTFS-3g, UNIX, + * 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 @@ -41,6 +41,7 @@ #include #include +#include #include #include @@ -52,6 +53,7 @@ #include "wimlib/endianness.h" #include "wimlib/error.h" #include "wimlib/metadata.h" +#include "wimlib/object_id.h" #include "wimlib/pathlist.h" #include "wimlib/paths.h" #include "wimlib/pattern.h" @@ -61,6 +63,7 @@ #include "wimlib/unix_data.h" #include "wimlib/wim.h" #include "wimlib/win32.h" /* for realpath() equivalent */ +#include "wimlib/xattr.h" #include "wimlib/xml.h" #define WIMLIB_EXTRACT_FLAG_FROM_PIPE 0x80000000 @@ -84,7 +87,12 @@ WIMLIB_EXTRACT_FLAG_STRICT_GLOB | \ WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES | \ WIMLIB_EXTRACT_FLAG_NO_PRESERVE_DIR_STRUCTURE | \ - WIMLIB_EXTRACT_FLAG_WIMBOOT) + WIMLIB_EXTRACT_FLAG_WIMBOOT | \ + WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS4K | \ + WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS8K | \ + WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS16K | \ + WIMLIB_EXTRACT_FLAG_COMPACT_LZX \ + ) /* Send WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE or * WIMLIB_PROGRESS_MSG_EXTRACT_METADATA. */ @@ -134,6 +142,70 @@ end_file_metadata_phase(struct apply_ctx *ctx) return end_file_phase(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_METADATA); } +/* Are all bytes in the specified buffer zero? */ +static bool +is_all_zeroes(const u8 *p, const size_t size) +{ + const u8 * const end = p + size; + + for (; (uintptr_t)p % WORDBYTES && p != end; p++) + if (*p) + return false; + + for (; end - p >= WORDBYTES; p += WORDBYTES) + if (*(const machine_word_t *)p) + return false; + + for (; p != end; p++) + if (*p) + return false; + + return true; +} + +/* + * Sparse regions should be detected at the granularity of the filesystem block + * size. For now just assume 4096 bytes, which is the default block size on + * NTFS and most Linux filesystems. + */ +#define SPARSE_UNIT 4096 + +/* + * Detect whether the specified buffer begins with a region of all zero bytes. + * Return %true if a zero region was found or %false if a nonzero region was + * found, and sets *len_ret to the length of the region. This operates at a + * granularity of SPARSE_UNIT bytes, meaning that to extend a zero region, there + * must be SPARSE_UNIT zero bytes with no interruption, but to extend a nonzero + * region, just one nonzero byte in the next SPARSE_UNIT bytes is sufficient. + * + * Note: besides compression, the WIM format doesn't yet have a way to + * efficiently represent zero regions, so that's why we need to detect them + * ourselves. Things will still fall apart badly on extremely large sparse + * files, but this is a start... + */ +bool +detect_sparse_region(const void *data, size_t size, size_t *len_ret) +{ + const void *p = data; + const void * const end = data + size; + size_t len = 0; + bool zeroes = false; + + while (p != end) { + size_t n = min(end - p, SPARSE_UNIT); + bool z = is_all_zeroes(p, n); + + if (len != 0 && z != zeroes) + break; + zeroes = z; + len += n; + p += n; + } + + *len_ret = len; + return zeroes; +} + #define PWM_FOUND_WIM_HDR (-1) /* Read the header for a blob in a pipable WIM. If @pwm_hdr_ret is not NULL, @@ -254,6 +326,55 @@ read_blobs_from_pipe(struct apply_ctx *ctx, const struct read_blob_callbacks *cb return 0; } +static int +handle_pwm_metadata_resource(WIMStruct *pwm, int image, bool is_needed) +{ + struct blob_descriptor *blob; + struct wim_reshdr reshdr; + struct wim_resource_descriptor *rdesc; + int ret; + + ret = WIMLIB_ERR_NOMEM; + blob = new_blob_descriptor(); + if (!blob) + goto out; + + ret = read_pwm_blob_header(pwm, blob->hash, &reshdr, NULL); + if (ret) + goto out; + + ret = WIMLIB_ERR_INVALID_PIPABLE_WIM; + if (!(reshdr.flags & WIM_RESHDR_FLAG_METADATA)) { + ERROR("Expected metadata resource, but found non-metadata " + "resource"); + goto out; + } + + ret = WIMLIB_ERR_NOMEM; + rdesc = MALLOC(sizeof(*rdesc)); + if (!rdesc) + goto out; + + wim_reshdr_to_desc_and_blob(&reshdr, pwm, rdesc, blob); + pwm->refcnt++; + + ret = WIMLIB_ERR_NOMEM; + pwm->image_metadata[image - 1] = new_unloaded_image_metadata(blob); + if (!pwm->image_metadata[image - 1]) + goto out; + blob = NULL; + + /* If the metadata resource is for the image being extracted, then parse + * it and save the metadata in memory. Otherwise, skip over it. */ + if (is_needed) + ret = select_wim_image(pwm, image); + else + ret = skip_wim_resource(rdesc); +out: + free_blob_descriptor(blob); + return ret; +} + /* Creates a temporary file opened for writing. The open file descriptor is * returned in @fd_ret and its name is returned in @name_ret (dynamically * allocated). */ @@ -261,27 +382,33 @@ static int create_temporary_file(struct filedes *fd_ret, tchar **name_ret) { tchar *name; - int open_flags; int raw_fd; +#ifdef __WIN32__ retry: - name = ttempnam(NULL, T("wimlib")); + name = _wtempnam(NULL, L"wimlib"); if (!name) { ERROR_WITH_ERRNO("Failed to create temporary filename"); return WIMLIB_ERR_NOMEM; } - - open_flags = O_WRONLY | O_CREAT | O_EXCL | O_BINARY; -#ifdef __WIN32__ - open_flags |= _O_SHORT_LIVED; -#endif - raw_fd = topen(name, open_flags, 0600); + raw_fd = _wopen(name, O_WRONLY | O_CREAT | O_EXCL | O_BINARY | + _O_SHORT_LIVED, 0600); + if (raw_fd < 0 && errno == EEXIST) { + FREE(name); + goto retry; + } +#else /* __WIN32__ */ + const char *tmpdir = getenv("TMPDIR"); + if (!tmpdir) + tmpdir = P_tmpdir; + name = MALLOC(strlen(tmpdir) + 1 + 6 + 6 + 1); + if (!name) + return WIMLIB_ERR_NOMEM; + sprintf(name, "%s/wimlibXXXXXX", tmpdir); + raw_fd = mkstemp(name); +#endif /* !__WIN32__ */ if (raw_fd < 0) { - if (errno == EEXIST) { - FREE(name); - goto retry; - } ERROR_WITH_ERRNO("Failed to create temporary file " "\"%"TS"\"", name); FREE(name); @@ -294,13 +421,10 @@ retry: } static int -begin_extract_blob_wrapper(struct blob_descriptor *blob, void *_ctx) +begin_extract_blob(struct blob_descriptor *blob, void *_ctx) { struct apply_ctx *ctx = _ctx; - ctx->cur_blob = blob; - ctx->cur_blob_offset = 0; - if (unlikely(blob->out_refcnt > MAX_OPEN_FILES)) return create_temporary_file(&ctx->tmpfile_fd, &ctx->tmpfile_name); @@ -308,29 +432,29 @@ begin_extract_blob_wrapper(struct blob_descriptor *blob, void *_ctx) } static int -extract_chunk_wrapper(const void *chunk, size_t size, void *_ctx) +extract_chunk(const struct blob_descriptor *blob, u64 offset, + const void *chunk, size_t size, void *_ctx) { struct apply_ctx *ctx = _ctx; union wimlib_progress_info *progress = &ctx->progress; + bool last = (offset + size == blob->size); int ret; - ctx->cur_blob_offset += size; - if (likely(ctx->supported_features.hard_links)) { progress->extract.completed_bytes += - (u64)size * ctx->cur_blob->out_refcnt; - if (ctx->cur_blob_offset == ctx->cur_blob->size) - progress->extract.completed_streams += ctx->cur_blob->out_refcnt; + (u64)size * blob->out_refcnt; + if (last) + progress->extract.completed_streams += blob->out_refcnt; } else { const struct blob_extraction_target *targets = - blob_extraction_targets(ctx->cur_blob); - for (u32 i = 0; i < ctx->cur_blob->out_refcnt; i++) { + blob_extraction_targets(blob); + for (u32 i = 0; i < blob->out_refcnt; i++) { const struct wim_inode *inode = targets[i].inode; const struct wim_dentry *dentry; inode_for_each_extraction_alias(dentry, inode) { progress->extract.completed_bytes += size; - if (ctx->cur_blob_offset == ctx->cur_blob->size) + if (last) progress->extract.completed_streams++; } } @@ -357,7 +481,7 @@ extract_chunk_wrapper(const void *chunk, size_t size, void *_ctx) return ret; } - return call_consume_chunk(chunk, size, ctx->saved_cbs); + return call_continue_blob(blob, offset, chunk, size, ctx->saved_cbs); } /* Copy the blob's data from the temporary file to each of its targets. @@ -388,7 +512,7 @@ extract_from_tmpfile(const tchar *tmpfile_name, } static int -end_extract_blob_wrapper(struct blob_descriptor *blob, int status, void *_ctx) +end_extract_blob(struct blob_descriptor *blob, int status, void *_ctx) { struct apply_ctx *ctx = _ctx; @@ -427,9 +551,9 @@ int extract_blob_list(struct apply_ctx *ctx, const struct read_blob_callbacks *cbs) { struct read_blob_callbacks wrapper_cbs = { - .begin_blob = begin_extract_blob_wrapper, - .consume_chunk = extract_chunk_wrapper, - .end_blob = end_extract_blob_wrapper, + .begin_blob = begin_extract_blob, + .continue_blob = extract_chunk, + .end_blob = end_extract_blob, .ctx = ctx, }; ctx->saved_cbs = cbs; @@ -699,9 +823,7 @@ dentry_calculate_extraction_name(struct wim_dentry *dentry, if (!ctx->supported_features.case_sensitive_filenames) { struct wim_dentry *other; - list_for_each_entry(other, &dentry->d_ci_conflict_list, - d_ci_conflict_list) - { + dentry_for_each_ci_match(other, dentry) { if (will_extract_dentry(other)) { if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_ALL_CASE_CONFLICTS) { @@ -973,10 +1095,11 @@ ref_stream_if_needed(struct wim_dentry *dentry, struct wim_inode *inode, * - file is a directory * - file is encrypted * - backend needs to create the file as UNIX symlink - * - backend will extract the stream as externally backed + * - backend will extract the stream as externally + * backed from the WIM archive itself */ - if (ctx->apply_ops->will_externally_back) { - int ret = (*ctx->apply_ops->will_externally_back)(dentry, ctx); + if (ctx->apply_ops->will_back_from_wim) { + int ret = (*ctx->apply_ops->will_back_from_wim)(dentry, ctx); if (ret > 0) /* Error? */ return ret; if (ret < 0) /* Won't externally back? */ @@ -1067,12 +1190,14 @@ static void inode_tally_features(const struct wim_inode *inode, struct wim_features *features) { - if (inode->i_attributes & FILE_ATTRIBUTE_ARCHIVE) - features->archive_files++; + if (inode->i_attributes & FILE_ATTRIBUTE_READONLY) + features->readonly_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_ARCHIVE) + features->archive_files++; if (inode->i_attributes & FILE_ATTRIBUTE_COMPRESSED) features->compressed_files++; if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) { @@ -1098,6 +1223,10 @@ inode_tally_features(const struct wim_inode *inode, features->security_descriptors++; if (inode_has_unix_data(inode)) features->unix_data++; + if (inode_has_object_id(inode)) + features->object_ids++; + if (inode_has_linux_xattrs(inode)) + features->linux_xattrs++; } /* Tally features necessary to extract a dentry and the corresponding inode. */ @@ -1150,9 +1279,11 @@ do_feature_check(const struct wim_features *required_features, /* File attributes. */ if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES)) { - /* Note: Don't bother the user about FILE_ATTRIBUTE_ARCHIVE. - * We're an archive program, so theoretically we can do what we - * want with it. */ + + if (required_features->readonly_files && + !supported_features->readonly_files) + WARNING("Ignoring FILE_ATTRIBUTE_READONLY of %lu files", + required_features->readonly_files); if (required_features->hidden_files && !supported_features->hidden_files) @@ -1164,6 +1295,10 @@ do_feature_check(const struct wim_features *required_features, WARNING("Ignoring FILE_ATTRIBUTE_SYSTEM of %lu files", required_features->system_files); + /* Note: Don't bother the user about FILE_ATTRIBUTE_ARCHIVE. + * We're an archive program, so theoretically we can do what we + * want with it. */ + if (required_features->compressed_files && !supported_features->compressed_files) WARNING("Ignoring FILE_ATTRIBUTE_COMPRESSED of %lu files", @@ -1229,19 +1364,37 @@ do_feature_check(const struct wim_features *required_features, WARNING("Ignoring Windows NT security descriptors of %lu files", required_features->security_descriptors); - /* UNIX data. */ - if ((extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) && - required_features->unix_data && !supported_features->unix_data) + /* Standard UNIX metadata */ + if (required_features->unix_data && + (!supported_features->unix_data || + !(extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA))) { - ERROR("Extraction backend does not support UNIX data!"); - return WIMLIB_ERR_UNSUPPORTED; + if (extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA) { + ERROR("Requested UNIX metadata extraction, but " + "extraction backend does not support it!"); + return WIMLIB_ERR_UNSUPPORTED; + } + WARNING("Ignoring UNIX metadata (uid/gid/mode/rdev) of %lu files%"TS, + required_features->unix_data, + (supported_features->unix_data ? + T("\n (use --unix-data mode to extract these)") : T(""))); } - if (required_features->unix_data && - !(extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA)) + /* Linux-style extended attributes */ + if (required_features->linux_xattrs && + (!supported_features->linux_xattrs || + !(extract_flags & WIMLIB_EXTRACT_FLAG_UNIX_DATA))) { - WARNING("Ignoring UNIX metadata of %lu files", - required_features->unix_data); + WARNING("Ignoring Linux-style extended attributes of %lu files%"TS, + required_features->linux_xattrs, + (supported_features->linux_xattrs ? + T("\n (use --unix-data mode to extract these)") : T(""))); + } + + /* Object IDs. */ + if (required_features->object_ids && !supported_features->object_ids) { + WARNING("Ignoring object IDs of %lu files", + required_features->object_ids); } /* DOS Names. */ @@ -1380,12 +1533,12 @@ extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees, * subtract from this if hard links are * supported by the extraction mode.) */ ctx->progress.extract.total_bytes = - wim_info_get_image_total_bytes(wim->wim_info, - wim->current_image); + xml_get_image_total_bytes(wim->xml_info, + wim->current_image); if (ctx->supported_features.hard_links) { ctx->progress.extract.total_bytes -= - wim_info_get_image_hard_link_bytes(wim->wim_info, - wim->current_image); + xml_get_image_hard_link_bytes(wim->xml_info, + wim->current_image); } } @@ -1463,7 +1616,7 @@ check_extract_flags(const WIMStruct *wim, int *extract_flags_p) #ifndef WITH_NTFS_3G if (extract_flags & WIMLIB_EXTRACT_FLAG_NTFS) { - ERROR("wimlib was compiled without support for NTFS-3g, so\n" + ERROR("wimlib was compiled without support for NTFS-3G, so\n" " it cannot apply a WIM image directly to an NTFS volume."); return WIMLIB_ERR_UNSUPPORTED; } @@ -1479,6 +1632,34 @@ check_extract_flags(const WIMStruct *wim, int *extract_flags_p) #endif } + if (extract_flags & (WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS4K | + WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS8K | + WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS16K | + WIMLIB_EXTRACT_FLAG_COMPACT_LZX)) + { + #ifdef __WIN32__ + int count = 0; + count += ((extract_flags & WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS4K) != 0); + count += ((extract_flags & WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS8K) != 0); + count += ((extract_flags & WIMLIB_EXTRACT_FLAG_COMPACT_XPRESS16K) != 0); + count += ((extract_flags & WIMLIB_EXTRACT_FLAG_COMPACT_LZX) != 0); + if (count != 1) { + ERROR("Only one compression format can be specified " + "for compact-mode extraction!"); + return WIMLIB_ERR_INVALID_PARAM; + } + if (extract_flags & WIMLIB_EXTRACT_FLAG_WIMBOOT) { + ERROR("Compact-mode extraction and WIMBoot-mode " + "extraction are mutually exclusive!"); + return WIMLIB_ERR_INVALID_PARAM; + } + #else + ERROR("Compact-mode extraction (System Compression) " + "is only supported on Windows!"); + return WIMLIB_ERR_UNSUPPORTED; + #endif + } + if ((extract_flags & (WIMLIB_EXTRACT_FLAG_RPFIX | WIMLIB_EXTRACT_FLAG_NORPFIX | @@ -1649,13 +1830,11 @@ extract_single_image(WIMStruct *wim, int image, } static const tchar * const filename_forbidden_chars = -T( #ifdef __WIN32__ -"<>:\"/\\|?*" +T("<>:\"/\\|?*"); #else -"/" +T("/"); #endif -); /* This function checks if it is okay to use a WIM image's name as a directory * name. */ @@ -1665,7 +1844,8 @@ image_name_ok_as_dir(const tchar *image_name) return image_name && *image_name && !tstrpbrk(image_name, filename_forbidden_chars) && tstrcmp(image_name, T(".")) && - tstrcmp(image_name, T("..")); + tstrcmp(image_name, T("..")) && + tstrlen(image_name) <= 128; } /* Extracts all images from the WIM to the directory @target, with the images @@ -1673,9 +1853,8 @@ image_name_ok_as_dir(const tchar *image_name) static int extract_all_images(WIMStruct *wim, const tchar *target, int extract_flags) { - size_t image_name_max_len = max(xml_get_max_image_name_len(wim), 20); size_t output_path_len = tstrlen(target); - tchar buf[output_path_len + 1 + image_name_max_len + 1]; + tchar buf[output_path_len + 1 + 128 + 1]; int ret; int image; const tchar *image_name; @@ -1831,7 +2010,7 @@ wimlib_extract_image_from_pipe_with_progress(int pipe_fd, if (ret) goto out_wimlib_free; - if (wim_info_get_num_images(pwm->wim_info) != pwm->hdr.image_count) { + if (xml_get_image_count(pwm->xml_info) != pwm->hdr.image_count) { ERROR("Image count in XML data is not the same as in WIM header."); ret = WIMLIB_ERR_IMAGE_COUNT; goto out_wimlib_free; @@ -1864,52 +2043,9 @@ wimlib_extract_image_from_pipe_with_progress(int pipe_fd, /* Load the needed metadata resource. */ for (i = 1; i <= pwm->hdr.image_count; i++) { - struct wim_image_metadata *imd; - struct wim_reshdr reshdr; - struct wim_resource_descriptor *metadata_rdesc; - - imd = pwm->image_metadata[i - 1]; - - ret = WIMLIB_ERR_NOMEM; - imd->metadata_blob = new_blob_descriptor(); - if (!imd->metadata_blob) - goto out_wimlib_free; - - imd->metadata_blob->is_metadata = 1; - - ret = read_pwm_blob_header(pwm, imd->metadata_blob->hash, - &reshdr, NULL); + ret = handle_pwm_metadata_resource(pwm, i, i == image); if (ret) goto out_wimlib_free; - - if (!(reshdr.flags & WIM_RESHDR_FLAG_METADATA)) { - ERROR("Expected metadata resource, but found " - "non-metadata resource"); - ret = WIMLIB_ERR_INVALID_PIPABLE_WIM; - goto out_wimlib_free; - } - - ret = WIMLIB_ERR_NOMEM; - metadata_rdesc = MALLOC(sizeof(struct wim_resource_descriptor)); - if (!metadata_rdesc) - goto out_wimlib_free; - wim_reshdr_to_desc_and_blob(&reshdr, pwm, metadata_rdesc, - imd->metadata_blob); - - if (i == image) { - /* Metadata resource is for the image being extracted. - * Parse it and save the metadata in memory. */ - ret = read_metadata_resource(imd); - if (ret) - goto out_wimlib_free; - imd->modified = 1; - } else { - /* Metadata resource is not for the image being - * extracted. Skip over it. */ - ret = skip_wim_resource(metadata_rdesc); - if (ret) - goto out_wimlib_free; - } } /* Extract the image. */ extract_flags |= WIMLIB_EXTRACT_FLAG_FROM_PIPE;