X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Fextract.c;h=ea9c25228d8c8a4877a59223b6daa78e2777d749;hp=b0df25d0719245c3dbab0d3580ea524019a607cb;hb=846d3ae386f6c12953816108c04de244533e3a4a;hpb=29ac4319aa9c75811cd5629cd3471a681fbeb552 diff --git a/src/extract.c b/src/extract.c index b0df25d0..ea9c2522 100644 --- a/src/extract.c +++ b/src/extract.c @@ -6,7 +6,7 @@ */ /* - * Copyright (C) 2012, 2013, 2014 Eric Biggers + * Copyright (C) 2012, 2013, 2014, 2015 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 @@ -39,12 +39,18 @@ # include "config.h" #endif +#include +#include +#include +#include + #include "wimlib/apply.h" +#include "wimlib/assert.h" +#include "wimlib/blob_table.h" #include "wimlib/dentry.h" #include "wimlib/encoding.h" #include "wimlib/endianness.h" #include "wimlib/error.h" -#include "wimlib/lookup_table.h" #include "wimlib/metadata.h" #include "wimlib/pathlist.h" #include "wimlib/paths.h" @@ -52,18 +58,10 @@ #include "wimlib/resource.h" #include "wimlib/security.h" #include "wimlib/unix_data.h" -#ifdef __WIN32__ -# include "wimlib/win32.h" /* for realpath() equivalent */ -#endif -#include "wimlib/xml.h" #include "wimlib/wildcard.h" #include "wimlib/wim.h" - -#include -#include -#include -#include -#include +#include "wimlib/win32.h" /* for realpath() equivalent */ +#include "wimlib/xml.h" #define WIMLIB_EXTRACT_FLAG_FROM_PIPE 0x80000000 #define WIMLIB_EXTRACT_FLAG_IMAGEMODE 0x40000000 @@ -98,7 +96,7 @@ do_file_extract_progress(struct apply_ctx *ctx, enum wimlib_progress_msg msg) } static int -start_file_phase(struct apply_ctx *ctx, uint64_t end_file_count, enum wimlib_progress_msg msg) +start_file_phase(struct apply_ctx *ctx, u64 end_file_count, enum wimlib_progress_msg msg) { ctx->progress.extract.current_file_count = 0; ctx->progress.extract.end_file_count = end_file_count; @@ -106,13 +104,13 @@ start_file_phase(struct apply_ctx *ctx, uint64_t end_file_count, enum wimlib_pro } int -start_file_structure_phase(struct apply_ctx *ctx, uint64_t end_file_count) +start_file_structure_phase(struct apply_ctx *ctx, u64 end_file_count) { return start_file_phase(ctx, end_file_count, WIMLIB_PROGRESS_MSG_EXTRACT_FILE_STRUCTURE); } int -start_file_metadata_phase(struct apply_ctx *ctx, uint64_t end_file_count) +start_file_metadata_phase(struct apply_ctx *ctx, u64 end_file_count) { return start_file_phase(ctx, end_file_count, WIMLIB_PROGRESS_MSG_EXTRACT_METADATA); } @@ -136,180 +134,124 @@ end_file_metadata_phase(struct apply_ctx *ctx) return end_file_phase(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_METADATA); } -/* Check whether the extraction of a dentry should be skipped completely. */ -static bool -dentry_is_supported(struct wim_dentry *dentry, - const struct wim_features *supported_features) -{ - struct wim_inode *inode = dentry->d_inode; - - if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) { - return supported_features->reparse_points || - (inode_is_symlink(inode) && - supported_features->symlink_reparse_points); - } - if (inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED) { - if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY) - return supported_features->encrypted_directories != 0; - else - return supported_features->encrypted_files != 0; - } - return true; -} - - -#define PWM_ALLOW_WIM_HDR 0x00001 +#define PWM_FOUND_WIM_HDR (-1) -/* Read the header from a stream in a pipable WIM. */ +/* Read the header for a blob in a pipable WIM. If @pwm_hdr_ret is not NULL, + * also look for a pipable WIM header and return PWM_FOUND_WIM_HDR if found. */ static int -read_pwm_stream_header(WIMStruct *pwm, struct wim_lookup_table_entry *lte, - struct wim_resource_spec *rspec, - int flags, struct wim_header_disk *hdr_ret) +read_pwm_blob_header(WIMStruct *pwm, u8 hash_ret[SHA1_HASH_SIZE], + struct wim_reshdr *reshdr_ret, + struct wim_header_disk *pwm_hdr_ret) { - union { - struct pwm_stream_hdr stream_hdr; - struct wim_header_disk pwm_hdr; - } buf; - struct wim_reshdr reshdr; int ret; + struct pwm_blob_hdr blob_hdr; + u64 magic; - ret = full_read(&pwm->in_fd, &buf.stream_hdr, sizeof(buf.stream_hdr)); - if (ret) + ret = full_read(&pwm->in_fd, &blob_hdr, sizeof(blob_hdr)); + if (unlikely(ret)) goto read_error; - if ((flags & PWM_ALLOW_WIM_HDR) && - le64_to_cpu(buf.stream_hdr.magic) == PWM_MAGIC) - { - BUILD_BUG_ON(sizeof(buf.pwm_hdr) < sizeof(buf.stream_hdr)); - ret = full_read(&pwm->in_fd, &buf.stream_hdr + 1, - sizeof(buf.pwm_hdr) - sizeof(buf.stream_hdr)); + magic = le64_to_cpu(blob_hdr.magic); - if (ret) + if (magic == PWM_MAGIC && pwm_hdr_ret != NULL) { + memcpy(pwm_hdr_ret, &blob_hdr, sizeof(blob_hdr)); + ret = full_read(&pwm->in_fd, + (u8 *)pwm_hdr_ret + sizeof(blob_hdr), + sizeof(*pwm_hdr_ret) - sizeof(blob_hdr)); + if (unlikely(ret)) goto read_error; - lte->resource_location = RESOURCE_NONEXISTENT; - memcpy(hdr_ret, &buf.pwm_hdr, sizeof(buf.pwm_hdr)); - return 0; + return PWM_FOUND_WIM_HDR; } - if (le64_to_cpu(buf.stream_hdr.magic) != PWM_STREAM_MAGIC) { - ERROR("Data read on pipe is invalid (expected stream header)."); + if (unlikely(magic != PWM_BLOB_MAGIC)) { + ERROR("Data read on pipe is invalid (expected blob header)"); return WIMLIB_ERR_INVALID_PIPABLE_WIM; } - copy_hash(lte->hash, buf.stream_hdr.hash); + copy_hash(hash_ret, blob_hdr.hash); + + reshdr_ret->size_in_wim = 0; /* Not available */ + reshdr_ret->flags = le32_to_cpu(blob_hdr.flags); + reshdr_ret->offset_in_wim = pwm->in_fd.offset; + reshdr_ret->uncompressed_size = le64_to_cpu(blob_hdr.uncompressed_size); + + if (unlikely(reshdr_ret->uncompressed_size == 0)) { + ERROR("Data read on pipe is invalid (resource is of 0 size)"); + return WIMLIB_ERR_INVALID_PIPABLE_WIM; + } - reshdr.size_in_wim = 0; - reshdr.flags = le32_to_cpu(buf.stream_hdr.flags); - reshdr.offset_in_wim = pwm->in_fd.offset; - reshdr.uncompressed_size = le64_to_cpu(buf.stream_hdr.uncompressed_size); - wim_res_hdr_to_spec(&reshdr, pwm, rspec); - lte_bind_wim_resource_spec(lte, rspec); - lte->flags = rspec->flags; - lte->size = rspec->uncompressed_size; - lte->offset_in_res = 0; return 0; read_error: - ERROR_WITH_ERRNO("Error reading pipable WIM from pipe"); + if (ret == WIMLIB_ERR_UNEXPECTED_END_OF_FILE) + ERROR("The pipe ended before all needed data was sent!"); + else + ERROR_WITH_ERRNO("Error reading pipable WIM from pipe"); return ret; } static int -load_streams_from_pipe(struct apply_ctx *ctx, - const struct read_stream_list_callbacks *cbs) +read_blobs_from_pipe(struct apply_ctx *ctx, const struct read_blob_callbacks *cbs) { - struct wim_lookup_table_entry *found_lte = NULL; - struct wim_resource_spec *rspec = NULL; - struct wim_lookup_table *lookup_table; int ret; + u8 hash[SHA1_HASH_SIZE]; + struct wim_reshdr reshdr; + struct wim_header_disk pwm_hdr; + struct wim_resource_descriptor rdesc; + struct blob_descriptor *blob; - ret = WIMLIB_ERR_NOMEM; - found_lte = new_lookup_table_entry(); - if (!found_lte) - goto out; - - rspec = MALLOC(sizeof(struct wim_resource_spec)); - if (!rspec) - goto out; - - lookup_table = ctx->wim->lookup_table; - memcpy(ctx->progress.extract.guid, ctx->wim->hdr.guid, WIM_GUID_LEN); + copy_guid(ctx->progress.extract.guid, ctx->wim->hdr.guid); ctx->progress.extract.part_number = ctx->wim->hdr.part_number; ctx->progress.extract.total_parts = ctx->wim->hdr.total_parts; ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN); if (ret) - goto out; + return ret; - while (ctx->num_streams_remaining) { - struct wim_header_disk pwm_hdr; - struct wim_lookup_table_entry *needed_lte; + while (ctx->num_blobs_remaining) { - if (found_lte->resource_location != RESOURCE_NONEXISTENT) - lte_unbind_wim_resource_spec(found_lte); - ret = read_pwm_stream_header(ctx->wim, found_lte, rspec, - PWM_ALLOW_WIM_HDR, &pwm_hdr); - if (ret) - goto out; + ret = read_pwm_blob_header(ctx->wim, hash, &reshdr, &pwm_hdr); - if ((found_lte->resource_location != RESOURCE_NONEXISTENT) - && !(found_lte->flags & WIM_RESHDR_FLAG_METADATA) - && (needed_lte = lookup_stream(lookup_table, found_lte->hash)) - && (needed_lte->out_refcnt)) - { - needed_lte->offset_in_res = found_lte->offset_in_res; - needed_lte->flags = found_lte->flags; - needed_lte->size = found_lte->size; + if (ret == PWM_FOUND_WIM_HDR) { + u16 part_number = le16_to_cpu(pwm_hdr.part_number); + u16 total_parts = le16_to_cpu(pwm_hdr.total_parts); + + if (part_number == ctx->progress.extract.part_number && + total_parts == ctx->progress.extract.total_parts && + guids_equal(pwm_hdr.guid, ctx->progress.extract.guid)) + continue; - lte_unbind_wim_resource_spec(found_lte); - lte_bind_wim_resource_spec(needed_lte, rspec); + copy_guid(ctx->progress.extract.guid, pwm_hdr.guid); + ctx->progress.extract.part_number = part_number; + ctx->progress.extract.total_parts = total_parts; + ret = extract_progress(ctx, WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN); + if (ret) + return ret; - ret = (*cbs->begin_stream)(needed_lte, - cbs->begin_stream_ctx); - if (ret) { - lte_unbind_wim_resource_spec(needed_lte); - goto out; - } + continue; + } - ret = extract_stream(needed_lte, needed_lte->size, - cbs->consume_chunk, - cbs->consume_chunk_ctx); + if (ret) + return ret; - ret = (*cbs->end_stream)(needed_lte, ret, - cbs->end_stream_ctx); - lte_unbind_wim_resource_spec(needed_lte); - if (ret) - goto out; - ctx->num_streams_remaining--; - } else if (found_lte->resource_location != RESOURCE_NONEXISTENT) { - ret = skip_wim_stream(found_lte); + if (!(reshdr.flags & WIM_RESHDR_FLAG_METADATA) + && (blob = lookup_blob(ctx->wim->blob_table, hash)) + && (blob->out_refcnt)) + { + wim_reshdr_to_desc_and_blob(&reshdr, ctx->wim, &rdesc, blob); + ret = read_blob_with_sha1(blob, cbs); + blob_unset_is_located_in_wim_resource(blob); if (ret) - goto out; + return ret; + ctx->num_blobs_remaining--; } else { - u16 part_number = le16_to_cpu(pwm_hdr.part_number); - u16 total_parts = le16_to_cpu(pwm_hdr.total_parts); - - if (part_number != ctx->progress.extract.part_number || - total_parts != ctx->progress.extract.total_parts || - memcmp(pwm_hdr.guid, ctx->progress.extract.guid, - WIM_GUID_LEN)) - { - ctx->progress.extract.part_number = part_number; - ctx->progress.extract.total_parts = total_parts; - memcpy(ctx->progress.extract.guid, - pwm_hdr.guid, WIM_GUID_LEN); - ret = extract_progress(ctx, - WIMLIB_PROGRESS_MSG_EXTRACT_SPWM_PART_BEGIN); - if (ret) - goto out; - } + wim_reshdr_to_desc(&reshdr, ctx->wim, &rdesc); + ret = skip_wim_resource(&rdesc); + if (ret) + return ret; } } - ret = 0; -out: - if (found_lte && found_lte->resource_location != RESOURCE_IN_WIM) - FREE(rspec); - free_lookup_table_entry(found_lte); - return ret; + + return 0; } /* Creates a temporary file opened for writing. The open file descriptor is @@ -352,17 +294,17 @@ retry: } static int -begin_extract_stream_wrapper(struct wim_lookup_table_entry *lte, void *_ctx) +begin_extract_blob_wrapper(struct blob_descriptor *blob, void *_ctx) { struct apply_ctx *ctx = _ctx; - ctx->cur_stream = lte; - ctx->cur_stream_offset = 0; + ctx->cur_blob = blob; + ctx->cur_blob_offset = 0; - if (unlikely(lte->out_refcnt > MAX_OPEN_STREAMS)) + if (unlikely(blob->out_refcnt > MAX_OPEN_FILES)) return create_temporary_file(&ctx->tmpfile_fd, &ctx->tmpfile_name); - else - return (*ctx->saved_cbs->begin_stream)(lte, ctx->saved_cbs->begin_stream_ctx); + + return call_begin_blob(blob, ctx->saved_cbs); } static int @@ -372,25 +314,23 @@ extract_chunk_wrapper(const void *chunk, size_t size, void *_ctx) union wimlib_progress_info *progress = &ctx->progress; int ret; - ctx->cur_stream_offset += size; + ctx->cur_blob_offset += size; if (likely(ctx->supported_features.hard_links)) { progress->extract.completed_bytes += - (u64)size * ctx->cur_stream->out_refcnt; - if (ctx->cur_stream_offset == ctx->cur_stream->size) - progress->extract.completed_streams += ctx->cur_stream->out_refcnt; + (u64)size * ctx->cur_blob->out_refcnt; + if (ctx->cur_blob_offset == ctx->cur_blob->size) + progress->extract.completed_streams += ctx->cur_blob->out_refcnt; } else { - const struct stream_owner *owners = stream_owners(ctx->cur_stream); - for (u32 i = 0; i < ctx->cur_stream->out_refcnt; i++) { - const struct wim_inode *inode = owners[i].inode; + const struct blob_extraction_target *targets = + blob_extraction_targets(ctx->cur_blob); + for (u32 i = 0; i < ctx->cur_blob->out_refcnt; i++) { + const struct wim_inode *inode = targets[i].inode; const struct wim_dentry *dentry; - list_for_each_entry(dentry, - &inode->i_extraction_aliases, - d_extraction_alias_node) - { + inode_for_each_extraction_alias(dentry, inode) { progress->extract.completed_bytes += size; - if (ctx->cur_stream_offset == ctx->cur_stream->size) + if (ctx->cur_blob_offset == ctx->cur_blob->size) progress->extract.completed_streams++; } } @@ -436,127 +376,91 @@ extract_chunk_wrapper(const void *chunk, size_t size, void *_ctx) ctx->tmpfile_name); } return ret; - } else { - return (*ctx->saved_cbs->consume_chunk)(chunk, size, - ctx->saved_cbs->consume_chunk_ctx); } + + return call_consume_chunk(chunk, size, ctx->saved_cbs); } +/* Copy the blob's data from the temporary file to each of its targets. + * + * This is executed only in the very uncommon case that a blob is being + * extracted to more than MAX_OPEN_FILES targets! */ static int -extract_from_tmpfile(const tchar *tmpfile_name, struct apply_ctx *ctx) +extract_from_tmpfile(const tchar *tmpfile_name, + const struct blob_descriptor *orig_blob, + const struct read_blob_callbacks *cbs) { - struct wim_lookup_table_entry tmpfile_lte; - struct wim_lookup_table_entry *orig_lte = ctx->cur_stream; - const struct read_stream_list_callbacks *cbs = ctx->saved_cbs; + struct blob_descriptor tmpfile_blob; + const struct blob_extraction_target *targets = blob_extraction_targets(orig_blob); int ret; - const u32 orig_refcnt = orig_lte->out_refcnt; - - BUILD_BUG_ON(MAX_OPEN_STREAMS < ARRAY_LEN(orig_lte->inline_stream_owners)); - - struct stream_owner *owners = orig_lte->stream_owners; - - /* Copy the stream's data from the temporary file to each of its - * destinations. - * - * This is executed only in the very uncommon case that a - * single-instance stream is being extracted to more than - * MAX_OPEN_STREAMS locations! */ - - memcpy(&tmpfile_lte, orig_lte, sizeof(struct wim_lookup_table_entry)); - tmpfile_lte.resource_location = RESOURCE_IN_FILE_ON_DISK; - tmpfile_lte.file_on_disk = ctx->tmpfile_name; - ret = 0; - for (u32 i = 0; i < orig_refcnt; i++) { - - /* Note: it usually doesn't matter whether we pass the original - * stream entry to callbacks provided by the extraction backend - * as opposed to the tmpfile stream entry, since they shouldn't - * actually read data from the stream other than through the - * read_stream_prefix() call below. But for - * WIMLIB_EXTRACT_FLAG_WIMBOOT mode on Windows it does matter - * because it needs the original stream location in order to - * create the external backing reference. */ - - orig_lte->out_refcnt = 1; - orig_lte->inline_stream_owners[0] = owners[i]; - - ret = (*cbs->begin_stream)(orig_lte, cbs->begin_stream_ctx); - if (ret) - break; - /* Extra SHA-1 isn't necessary here, but it shouldn't hurt as - * this case is very rare anyway. */ - ret = extract_stream(&tmpfile_lte, tmpfile_lte.size, - cbs->consume_chunk, - cbs->consume_chunk_ctx); + memcpy(&tmpfile_blob, orig_blob, sizeof(struct blob_descriptor)); + tmpfile_blob.blob_location = BLOB_IN_FILE_ON_DISK; + tmpfile_blob.file_on_disk = (tchar *)tmpfile_name; + tmpfile_blob.out_refcnt = 1; - ret = (*cbs->end_stream)(orig_lte, ret, cbs->end_stream_ctx); + for (u32 i = 0; i < orig_blob->out_refcnt; i++) { + tmpfile_blob.inline_blob_extraction_targets[0] = targets[i]; + ret = read_blob_with_cbs(&tmpfile_blob, cbs); if (ret) - break; + return ret; } - FREE(owners); - orig_lte->out_refcnt = 0; - return ret; + return 0; } static int -end_extract_stream_wrapper(struct wim_lookup_table_entry *stream, - int status, void *_ctx) +end_extract_blob_wrapper(struct blob_descriptor *blob, int status, void *_ctx) { struct apply_ctx *ctx = _ctx; if (unlikely(filedes_valid(&ctx->tmpfile_fd))) { filedes_close(&ctx->tmpfile_fd); if (!status) - status = extract_from_tmpfile(ctx->tmpfile_name, ctx); + status = extract_from_tmpfile(ctx->tmpfile_name, blob, + ctx->saved_cbs); filedes_invalidate(&ctx->tmpfile_fd); tunlink(ctx->tmpfile_name); FREE(ctx->tmpfile_name); return status; - } else { - return (*ctx->saved_cbs->end_stream)(stream, status, - ctx->saved_cbs->end_stream_ctx); } + + return call_end_blob(blob, status, ctx->saved_cbs); } /* - * Read the list of single-instance streams to extract and feed their data into - * the specified callback functions. + * Read the list of blobs to extract and feed their data into the specified + * callback functions. * - * This handles checksumming each stream. + * This handles checksumming each blob. * * This also handles sending WIMLIB_PROGRESS_MSG_EXTRACT_STREAMS. * - * This also works if the WIM is being read from a pipe, whereas attempting to - * read streams directly (e.g. with read_full_stream_into_buf()) will not. + * This also works if the WIM is being read from a pipe. * - * This also will split up streams that will need to be extracted to more than - * MAX_OPEN_STREAMS locations, as measured by the 'out_refcnt' of each stream. + * This also will split up blobs that will need to be extracted to more than + * MAX_OPEN_FILES locations, as measured by the 'out_refcnt' of each blob. * Therefore, the apply_operations implementation need not worry about running * out of file descriptors, unless it might open more than one file descriptor - * per nominal destination (e.g. Win32 currently might because the destination - * file system might not support hard links). + * per 'blob_extraction_target' (e.g. Win32 currently might because the + * destination file system might not support hard links). */ int -extract_stream_list(struct apply_ctx *ctx, - const struct read_stream_list_callbacks *cbs) +extract_blob_list(struct apply_ctx *ctx, const struct read_blob_callbacks *cbs) { - struct read_stream_list_callbacks wrapper_cbs = { - .begin_stream = begin_extract_stream_wrapper, - .begin_stream_ctx = ctx, - .consume_chunk = extract_chunk_wrapper, - .consume_chunk_ctx = ctx, - .end_stream = end_extract_stream_wrapper, - .end_stream_ctx = ctx, + struct read_blob_callbacks wrapper_cbs = { + .begin_blob = begin_extract_blob_wrapper, + .consume_chunk = extract_chunk_wrapper, + .end_blob = end_extract_blob_wrapper, + .ctx = ctx, }; ctx->saved_cbs = cbs; if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) { - return load_streams_from_pipe(ctx, &wrapper_cbs); + return read_blobs_from_pipe(ctx, &wrapper_cbs); } else { - return read_stream_list(&ctx->stream_list, - offsetof(struct wim_lookup_table_entry, - extraction_list), - &wrapper_cbs, VERIFY_STREAM_HASHES); + return read_blob_list(&ctx->blob_list, + offsetof(struct blob_descriptor, + extraction_list), + &wrapper_cbs, VERIFY_BLOB_HASHES); } } @@ -567,38 +471,39 @@ extract_stream_list(struct apply_ctx *ctx, * unnamed data stream only. */ static int extract_dentry_to_stdout(struct wim_dentry *dentry, - const struct wim_lookup_table *lookup_table) + const struct blob_table *blob_table) { struct wim_inode *inode = dentry->d_inode; - struct wim_lookup_table_entry *lte; + struct blob_descriptor *blob; struct filedes _stdout; if (inode->i_attributes & (FILE_ATTRIBUTE_REPARSE_POINT | - FILE_ATTRIBUTE_DIRECTORY)) + FILE_ATTRIBUTE_DIRECTORY | + FILE_ATTRIBUTE_ENCRYPTED)) { ERROR("\"%"TS"\" is not a regular file and therefore cannot be " "extracted to standard output", dentry_full_path(dentry)); return WIMLIB_ERR_NOT_A_REGULAR_FILE; } - lte = inode_unnamed_lte(inode, lookup_table); - if (!lte) { - const u8 *hash = inode_unnamed_stream_hash(inode); + blob = inode_get_blob_for_unnamed_data_stream(inode, blob_table); + if (!blob) { + const u8 *hash = inode_get_hash_of_unnamed_data_stream(inode); if (!is_zero_hash(hash)) - return stream_not_found_error(inode, hash); + return blob_not_found_error(inode, hash); return 0; } filedes_init(&_stdout, STDOUT_FILENO); - return extract_full_stream_to_fd(lte, &_stdout); + return extract_blob_to_fd(blob, &_stdout); } static int extract_dentries_to_stdout(struct wim_dentry **dentries, size_t num_dentries, - const struct wim_lookup_table *lookup_table) + const struct blob_table *blob_table) { for (size_t i = 0; i < num_dentries; i++) { - int ret = extract_dentry_to_stdout(dentries[i], lookup_table); + int ret = extract_dentry_to_stdout(dentries[i], blob_table); if (ret) return ret; } @@ -746,13 +651,13 @@ destroy_dentry_list(struct list_head *dentry_list) } static void -destroy_stream_list(struct list_head *stream_list) +destroy_blob_list(struct list_head *blob_list) { - struct wim_lookup_table_entry *lte; + struct blob_descriptor *blob; - list_for_each_entry(lte, stream_list, extraction_list) - if (lte->out_refcnt > ARRAY_LEN(lte->inline_stream_owners)) - FREE(lte->stream_owners); + list_for_each_entry(blob, blob_list, extraction_list) + if (blob->out_refcnt > ARRAY_LEN(blob->inline_blob_extraction_targets)) + FREE(blob->blob_extraction_targets); } #ifdef __WIN32__ @@ -808,9 +713,6 @@ dentry_calculate_extraction_name(struct wim_dentry *dentry, { int ret; - if (!dentry_is_supported(dentry, &ctx->supported_features)) - goto skip_dentry; - if (dentry_is_root(dentry)) return 0; @@ -901,8 +803,7 @@ out_replace: utf16le_put_tstr(tchar_name); - dentry->d_extraction_name = memdup(fixed_name, - 2 * fixed_name_num_chars + 2); + dentry->d_extraction_name = TSTRDUP(fixed_name); if (!dentry->d_extraction_name) return WIMLIB_ERR_NOMEM; dentry->d_extraction_name_nchars = fixed_name_num_chars; @@ -959,34 +860,35 @@ dentry_list_calculate_extraction_names(struct list_head *dentry_list, static int dentry_resolve_streams(struct wim_dentry *dentry, int extract_flags, - struct wim_lookup_table *lookup_table) + struct blob_table *blob_table) { struct wim_inode *inode = dentry->d_inode; - struct wim_lookup_table_entry *lte; + struct blob_descriptor *blob; int ret; bool force = false; - /* Special case: when extracting from a pipe, the WIM lookup table is + /* Special case: when extracting from a pipe, the WIM blob table is * initially empty, so "resolving" an inode's streams is initially not - * possible. However, we still need to keep track of which streams, - * identified by SHA1 message digests, need to be extracted, so we - * "resolve" the inode's streams anyway by allocating new entries. */ + * possible. However, we still need to keep track of which blobs, + * identified by SHA-1 message digests, need to be extracted, so we + * "resolve" the inode's streams anyway by allocating a 'struct + * blob_descriptor' for each one. */ if (extract_flags & WIMLIB_EXTRACT_FLAG_FROM_PIPE) force = true; - ret = inode_resolve_streams(inode, lookup_table, force); + ret = inode_resolve_streams(inode, blob_table, force); if (ret) return ret; - for (u32 i = 0; i <= inode->i_num_ads; i++) { - lte = inode_stream_lte_resolved(inode, i); - if (lte) - lte->out_refcnt = 0; + for (unsigned i = 0; i < inode->i_num_streams; i++) { + blob = stream_blob_resolved(&inode->i_streams[i]); + if (blob) + blob->out_refcnt = 0; } return 0; } /* * For each dentry to be extracted, resolve all streams in the corresponding - * inode and set 'out_refcnt' in each to 0. + * inode and set 'out_refcnt' in all referenced blob_descriptors to 0. * * Possible error codes: WIMLIB_ERR_RESOURCE_NOT_FOUND, WIMLIB_ERR_NOMEM. */ @@ -1000,7 +902,7 @@ dentry_list_resolve_streams(struct list_head *dentry_list, list_for_each_entry(dentry, dentry_list, d_extraction_list_node) { ret = dentry_resolve_streams(dentry, ctx->extract_flags, - ctx->wim->lookup_table); + ctx->wim->blob_table); if (ret) return ret; } @@ -1008,145 +910,155 @@ dentry_list_resolve_streams(struct list_head *dentry_list, } static int -ref_stream(struct wim_lookup_table_entry *lte, u32 stream_idx, - struct wim_dentry *dentry, struct apply_ctx *ctx) +ref_stream(struct wim_inode_stream *strm, struct wim_dentry *dentry, + struct apply_ctx *ctx) { struct wim_inode *inode = dentry->d_inode; - struct stream_owner *stream_owners; + struct blob_descriptor *blob = stream_blob_resolved(strm); + struct blob_extraction_target *targets; - if (!lte) + if (!blob) return 0; - /* Tally the size only for each extraction of the stream (not hard - * links). */ + /* Tally the size only for each actual extraction of the stream (not + * additional hard links to the inode). */ if (inode->i_visited && ctx->supported_features.hard_links) return 0; - ctx->progress.extract.total_bytes += lte->size; + ctx->progress.extract.total_bytes += blob->size; ctx->progress.extract.total_streams++; if (inode->i_visited) return 0; - /* Add stream to the dentry_list only one time, even if it's going - * to be extracted to multiple inodes. */ - if (lte->out_refcnt == 0) { - list_add_tail(<e->extraction_list, &ctx->stream_list); - ctx->num_streams_remaining++; + /* Add each blob to 'ctx->blob_list' only one time, regardless of how + * many extraction targets it will have. */ + if (blob->out_refcnt == 0) { + list_add_tail(&blob->extraction_list, &ctx->blob_list); + ctx->num_blobs_remaining++; } - /* If inode not yet been visited, append it to the stream_owners array. */ - if (lte->out_refcnt < ARRAY_LEN(lte->inline_stream_owners)) { - stream_owners = lte->inline_stream_owners; + /* Set this stream as an extraction target of 'blob'. */ + + if (blob->out_refcnt < ARRAY_LEN(blob->inline_blob_extraction_targets)) { + targets = blob->inline_blob_extraction_targets; } else { - struct stream_owner *prev_stream_owners; - size_t alloc_stream_owners; + struct blob_extraction_target *prev_targets; + size_t alloc_blob_extraction_targets; - if (lte->out_refcnt == ARRAY_LEN(lte->inline_stream_owners)) { - prev_stream_owners = NULL; - alloc_stream_owners = ARRAY_LEN(lte->inline_stream_owners); + if (blob->out_refcnt == ARRAY_LEN(blob->inline_blob_extraction_targets)) { + prev_targets = NULL; + alloc_blob_extraction_targets = ARRAY_LEN(blob->inline_blob_extraction_targets); } else { - prev_stream_owners = lte->stream_owners; - alloc_stream_owners = lte->alloc_stream_owners; + prev_targets = blob->blob_extraction_targets; + alloc_blob_extraction_targets = blob->alloc_blob_extraction_targets; } - if (lte->out_refcnt == alloc_stream_owners) { - alloc_stream_owners *= 2; - stream_owners = REALLOC(prev_stream_owners, - alloc_stream_owners * - sizeof(stream_owners[0])); - if (!stream_owners) + if (blob->out_refcnt == alloc_blob_extraction_targets) { + alloc_blob_extraction_targets *= 2; + targets = REALLOC(prev_targets, + alloc_blob_extraction_targets * + sizeof(targets[0])); + if (!targets) return WIMLIB_ERR_NOMEM; - if (!prev_stream_owners) { - memcpy(stream_owners, - lte->inline_stream_owners, - sizeof(lte->inline_stream_owners)); + if (!prev_targets) { + memcpy(targets, + blob->inline_blob_extraction_targets, + sizeof(blob->inline_blob_extraction_targets)); } - lte->stream_owners = stream_owners; - lte->alloc_stream_owners = alloc_stream_owners; + blob->blob_extraction_targets = targets; + blob->alloc_blob_extraction_targets = alloc_blob_extraction_targets; } - stream_owners = lte->stream_owners; - } - stream_owners[lte->out_refcnt].inode = inode; - if (stream_idx == 0) { - stream_owners[lte->out_refcnt].stream_name = NULL; - } else { - stream_owners[lte->out_refcnt].stream_name = - inode->i_ads_entries[stream_idx - 1].stream_name; + targets = blob->blob_extraction_targets; } - lte->out_refcnt++; + targets[blob->out_refcnt].inode = inode; + targets[blob->out_refcnt].stream = strm; + blob->out_refcnt++; return 0; } static int -ref_unnamed_stream(struct wim_dentry *dentry, struct apply_ctx *ctx) +ref_stream_if_needed(struct wim_dentry *dentry, struct wim_inode *inode, + struct wim_inode_stream *strm, struct apply_ctx *ctx) { - struct wim_inode *inode = dentry->d_inode; - int ret; - u16 stream_idx; - struct wim_lookup_table_entry *stream; - - if (unlikely(inode_is_encrypted_directory(inode))) - return 0; - - if (unlikely(ctx->apply_ops->will_externally_back)) { - ret = (*ctx->apply_ops->will_externally_back)(dentry, ctx); - if (ret >= 0) { - if (ret) /* Error */ - return ret; - /* Will externally back */ - return 0; + bool need_stream = false; + switch (strm->stream_type) { + case STREAM_TYPE_DATA: + if (stream_is_named(strm)) { + /* Named data stream */ + if (ctx->supported_features.named_data_streams) + need_stream = true; + } else if (!(inode->i_attributes & (FILE_ATTRIBUTE_DIRECTORY | + FILE_ATTRIBUTE_ENCRYPTED)) + && !(inode_is_symlink(inode) + && !ctx->supported_features.reparse_points + && ctx->supported_features.symlink_reparse_points)) + { + /* + * Unnamed data stream. Skip if any of the following is true: + * + * - file is a directory + * - file is encrypted + * - backend needs to create the file as UNIX symlink + * - backend will extract the stream as externally backed + */ + if (ctx->apply_ops->will_externally_back) { + int ret = (*ctx->apply_ops->will_externally_back)(dentry, ctx); + if (ret > 0) /* Error? */ + return ret; + if (ret < 0) /* Won't externally back? */ + need_stream = true; + } else { + need_stream = true; + } + } + break; + case STREAM_TYPE_REPARSE_POINT: + wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT); + if (ctx->supported_features.reparse_points || + (inode_is_symlink(inode) && + ctx->supported_features.symlink_reparse_points)) + need_stream = true; + break; + case STREAM_TYPE_EFSRPC_RAW_DATA: + wimlib_assert(inode->i_attributes & FILE_ATTRIBUTE_ENCRYPTED); + if (inode->i_attributes & FILE_ATTRIBUTE_DIRECTORY) { + if (ctx->supported_features.encrypted_directories) + need_stream = true; + } else { + if (ctx->supported_features.encrypted_files) + need_stream = true; } - /* Won't externally back */ + break; } - - stream = inode_unnamed_stream_resolved(inode, &stream_idx); - return ref_stream(stream, stream_idx, dentry, ctx); + if (need_stream) + return ref_stream(strm, dentry, ctx); + return 0; } static int dentry_ref_streams(struct wim_dentry *dentry, struct apply_ctx *ctx) { struct wim_inode *inode = dentry->d_inode; - int ret; - - /* The unnamed data stream will almost always be extracted, but there - * exist cases in which it won't be. */ - ret = ref_unnamed_stream(dentry, ctx); - if (ret) - return ret; - - /* Named data streams will be extracted only if supported in the current - * extraction mode and volume, and to avoid complications, if not doing - * a linked extraction. */ - if (ctx->supported_features.named_data_streams) { - for (u16 i = 0; i < inode->i_num_ads; i++) { - if (!ads_entry_is_named_stream(&inode->i_ads_entries[i])) - continue; - ret = ref_stream(inode->i_ads_entries[i].lte, i + 1, - dentry, ctx); - if (ret) - return ret; - } + for (unsigned i = 0; i < inode->i_num_streams; i++) { + int ret = ref_stream_if_needed(dentry, inode, + &inode->i_streams[i], ctx); + if (ret) + return ret; } inode->i_visited = 1; return 0; } /* - * For each dentry to be extracted, iterate through the data streams of the - * corresponding inode. For each such stream that is not to be ignored due to - * the supported features or extraction flags, add it to the list of streams to - * be extracted (ctx->stream_list) if not already done so. - * - * Also builds a mapping from each stream to the inodes referencing it. + * Given a list of dentries to be extracted, build the list of blobs that need + * to be extracted, and for each blob determine the streams to which that blob + * will be extracted. * - * This also initializes the extract progress info with byte and stream + * This also initializes the extract progress info with byte and blob * information. * * ctx->supported_features must be filled in. - * - * Possible error codes: WIMLIB_ERR_NOMEM. */ static int dentry_list_ref_streams(struct list_head *dentry_list, struct apply_ctx *ctx) @@ -1168,18 +1080,14 @@ static void dentry_list_build_inode_alias_lists(struct list_head *dentry_list) { struct wim_dentry *dentry; - struct wim_inode *inode; + + list_for_each_entry(dentry, dentry_list, d_extraction_list_node) + dentry->d_inode->i_first_extraction_alias = NULL; list_for_each_entry(dentry, dentry_list, d_extraction_list_node) { - inode = dentry->d_inode; - if (!inode->i_visited) - INIT_LIST_HEAD(&inode->i_extraction_aliases); - list_add_tail(&dentry->d_extraction_alias_node, - &inode->i_extraction_aliases); - inode->i_visited = 1; + dentry->d_next_extraction_alias = dentry->d_inode->i_first_extraction_alias; + dentry->d_inode->i_first_extraction_alias = dentry; } - list_for_each_entry(dentry, dentry_list, d_extraction_list_node) - dentry->d_inode->i_visited = 0; } static void @@ -1204,7 +1112,7 @@ inode_tally_features(const struct wim_inode *inode, features->not_context_indexed_files++; if (inode->i_attributes & FILE_ATTRIBUTE_SPARSE_FILE) features->sparse_files++; - if (inode_has_named_stream(inode)) + if (inode_has_named_data_stream(inode)) features->named_data_streams++; if (inode->i_attributes & FILE_ATTRIBUTE_REPARSE_POINT) { features->reparse_points++; @@ -1213,7 +1121,7 @@ inode_tally_features(const struct wim_inode *inode, else features->other_reparse_points++; } - if (inode->i_security_id != -1) + if (inode_has_security_descriptor(inode)) features->security_descriptors++; if (inode_has_unix_data(inode)) features->unix_data++; @@ -1255,6 +1163,18 @@ do_feature_check(const struct wim_features *required_features, const struct wim_features *supported_features, int extract_flags) { + /* Encrypted files. */ + if (required_features->encrypted_files && + !supported_features->encrypted_files) + WARNING("Ignoring EFS-encrypted data of %lu files", + required_features->encrypted_files); + + /* Named data streams. */ + if (required_features->named_data_streams && + !supported_features->named_data_streams) + WARNING("Ignoring named data streams of %lu files", + required_features->named_data_streams); + /* File attributes. */ if (!(extract_flags & WIMLIB_EXTRACT_FLAG_NO_ATTRIBUTES)) { /* Note: Don't bother the user about FILE_ATTRIBUTE_ARCHIVE. @@ -1292,18 +1212,6 @@ do_feature_check(const struct wim_features *required_features, required_features->encrypted_directories); } - /* Encrypted files. */ - if (required_features->encrypted_files && - !supported_features->encrypted_files) - WARNING("Ignoring %lu encrypted files", - required_features->encrypted_files); - - /* Named data streams. */ - if (required_features->named_data_streams && - (!supported_features->named_data_streams)) - WARNING("Ignoring named data streams of %lu files", - required_features->named_data_streams); - /* Hard links. */ if (required_features->hard_links && !supported_features->hard_links) WARNING("Extracting %lu hard links as independent files", @@ -1323,12 +1231,11 @@ do_feature_check(const struct wim_features *required_features, { if (supported_features->symlink_reparse_points) { if (required_features->other_reparse_points) { - WARNING("Ignoring %lu non-symlink/junction " - "reparse point files", + WARNING("Ignoring reparse data of %lu non-symlink/junction files", required_features->other_reparse_points); } } else { - WARNING("Ignoring %lu reparse point files", + WARNING("Ignoring reparse data of %lu files", required_features->reparse_points); } } @@ -1412,7 +1319,7 @@ extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees, if (extract_flags & WIMLIB_EXTRACT_FLAG_TO_STDOUT) { ret = extract_dentries_to_stdout(trees, num_trees, - wim->lookup_table); + wim->blob_table); goto out; } @@ -1450,7 +1357,7 @@ extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees, wim->current_image); ctx->progress.extract.target = target; } - INIT_LIST_HEAD(&ctx->stream_list); + INIT_LIST_HEAD(&ctx->blob_list); filedes_invalidate(&ctx->tmpfile_fd); ctx->apply_ops = ops; @@ -1473,6 +1380,11 @@ extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees, if (ret) goto out_cleanup; + if (unlikely(list_empty(&dentry_list))) { + WARNING("There is nothing to extract!"); + goto out_cleanup; + } + ret = dentry_list_resolve_streams(&dentry_list, ctx); if (ret) goto out_cleanup; @@ -1486,8 +1398,8 @@ extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees, 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. + * blob table), since at this point all we have is a set of + * SHA-1 message digests of blobs 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, @@ -1530,7 +1442,7 @@ extract_trees(WIMStruct *wim, struct wim_dentry **trees, size_t num_trees, WIMLIB_PROGRESS_MSG_EXTRACT_IMAGE_END : WIMLIB_PROGRESS_MSG_EXTRACT_TREE_END)); out_cleanup: - destroy_stream_list(&ctx->stream_list); + destroy_blob_list(&ctx->blob_list); destroy_dentry_list(&dentry_list); FREE(ctx); out: @@ -1675,7 +1587,7 @@ do_wimlib_extract_paths(WIMStruct *wim, int image, const tchar *target, if (ret) return ret; - ret = wim_checksum_unhashed_streams(wim); + ret = wim_checksum_unhashed_blobs(wim); if (ret) return ret; @@ -1897,8 +1809,8 @@ wimlib_extract_image_from_pipe_with_progress(int pipe_fd, /* 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. */ + * wimlib_open_wim(), getting a WIMStruct in this way will result in an + * empty blob table, no XML data read, and no filename set. */ ret = open_wim_as_WIMStruct(&pipe_fd, WIMLIB_OPEN_FLAG_FROM_PIPE, &pwm, progfunc, progctx); if (ret) @@ -1930,22 +1842,19 @@ wimlib_extract_image_from_pipe_with_progress(int pipe_fd, * write_pipable_wim() for more details about the format of pipable * WIMs.) */ { - struct wim_lookup_table_entry xml_lte; - struct wim_resource_spec xml_rspec; - ret = read_pwm_stream_header(pwm, &xml_lte, &xml_rspec, 0, NULL); + u8 hash[SHA1_HASH_SIZE]; + + ret = read_pwm_blob_header(pwm, hash, + &pwm->hdr.xml_data_reshdr, NULL); if (ret) goto out_wimlib_free; - if (!(xml_lte.flags & WIM_RESHDR_FLAG_METADATA)) - { - ERROR("Expected XML data, but found non-metadata " - "stream."); + if (!(pwm->hdr.xml_data_reshdr.flags & WIM_RESHDR_FLAG_METADATA)) { + ERROR("Expected XML data, but found non-metadata resource."); ret = WIMLIB_ERR_INVALID_PIPABLE_WIM; goto out_wimlib_free; } - wim_res_spec_to_hdr(&xml_rspec, &pwm->hdr.xml_data_reshdr); - ret = read_wim_xml_data(pwm); if (ret) goto out_wimlib_free; @@ -1983,48 +1892,49 @@ 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_lookup_table_entry *metadata_lte; struct wim_image_metadata *imd; - struct wim_resource_spec *metadata_rspec; + struct wim_reshdr reshdr; + struct wim_resource_descriptor *metadata_rdesc; - metadata_lte = new_lookup_table_entry(); - if (metadata_lte == NULL) { - ret = WIMLIB_ERR_NOMEM; - goto out_wimlib_free; - } - metadata_rspec = MALLOC(sizeof(struct wim_resource_spec)); - if (metadata_rspec == NULL) { - ret = WIMLIB_ERR_NOMEM; - free_lookup_table_entry(metadata_lte); + imd = pwm->image_metadata[i - 1]; + + ret = WIMLIB_ERR_NOMEM; + imd->metadata_blob = new_blob_descriptor(); + if (!imd->metadata_blob) goto out_wimlib_free; - } - ret = read_pwm_stream_header(pwm, metadata_lte, metadata_rspec, 0, NULL); - imd = pwm->image_metadata[i - 1]; - imd->metadata_lte = metadata_lte; - if (ret) { - FREE(metadata_rspec); + imd->metadata_blob->is_metadata = 1; + + ret = read_pwm_blob_header(pwm, imd->metadata_blob->hash, + &reshdr, NULL); + if (ret) goto out_wimlib_free; - } - if (!(metadata_lte->flags & WIM_RESHDR_FLAG_METADATA)) { + if (!(reshdr.flags & WIM_RESHDR_FLAG_METADATA)) { ERROR("Expected metadata resource, but found " - "non-metadata stream."); + "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(pwm, imd); + 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_stream(metadata_lte); + ret = skip_wim_resource(metadata_rdesc); if (ret) goto out_wimlib_free; }