X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Fresource.c;h=51bde70822c7ab33e5341ef33fdb63cd0e64fa83;hp=78adfafe6415ea85bfc46524f44cf5cbaa3868ba;hb=52cccded7706cd556f78ce7d2506937d2a216293;hpb=21a83b1b20f5ae138bc4bb9bdedd80a2a2d92c0a diff --git a/src/resource.c b/src/resource.c index 78adfafe..51bde708 100644 --- a/src/resource.c +++ b/src/resource.c @@ -5,47 +5,56 @@ */ /* - * Copyright (C) 2010 Carl Thijssen * Copyright (C) 2012 Eric Biggers * * This file is part of wimlib, a library for working with WIM files. * * wimlib 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 - * Software Foundation; either version 2.1 of the License, or (at your option) - * any later version. + * terms of the GNU General Public License as published by the Free Software + * Foundation; either version 3 of the License, or (at your option) any later + * version. * * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR - * A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more - * details. + * A PARTICULAR PURPOSE. See the GNU General Public License for more details. * - * You should have received a copy of the GNU Lesser General Public License - * along with wimlib; if not, see http://www.gnu.org/licenses/. + * You should have received a copy of the GNU General Public License along with + * wimlib; if not, see http://www.gnu.org/licenses/. */ +#include "config.h" + +#include +#include + +#include "dentry.h" + #include "wimlib_internal.h" #include "lookup_table.h" #include "io.h" #include "lzx.h" #include "xpress.h" #include "sha1.h" -#include "dentry.h" -#include "config.h" #include #include -#include +#ifdef WITH_NTFS_3G +#include +#include +#include +#include +#endif -/* + +/* * Reads all or part of a compressed resource into an in-memory buffer. * * @fp: The FILE* for the WIM file. - * @resource_compressed_size: The compressed size of the resource. + * @resource_compressed_size: The compressed size of the resource. * @resource_uncompressed_size: The uncompressed size of the resource. * @resource_offset: The offset of the start of the resource from * the start of the stream @fp. - * @resource_ctype: The compression type of the resource. + * @resource_ctype: The compression type of the resource. * @len: The number of bytes of uncompressed data to read from * the resource. * @offset: The offset of the bytes to read within the uncompressed @@ -55,9 +64,9 @@ * * Returns zero on success, nonzero on failure. */ -static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, - u64 resource_uncompressed_size, - u64 resource_offset, int resource_ctype, +static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, + u64 resource_uncompressed_size, + u64 resource_offset, int resource_ctype, u64 len, u64 offset, u8 contents_ret[]) { @@ -88,7 +97,7 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, * The chunk offsets are measured relative to the end of the chunk * table. The first chunk is omitted from the table in the WIM file * because its offset is implicitly given by the fact that it directly - * follows the chunk table and therefore must have an offset of 0. + * follows the chunk table and therefore must have an offset of 0. */ /* Calculate how many chunks the resource conists of in its entirety. */ @@ -136,7 +145,7 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, /* According to M$'s documentation, if the uncompressed size of * the file is greater than 4 GB, the chunk entries are 8-byte * integers. Otherwise, they are 4-byte integers. */ - u64 chunk_entry_size = (resource_uncompressed_size >= (u64)1 << 32) ? + u64 chunk_entry_size = (resource_uncompressed_size >= (u64)1 << 32) ? 8 : 4; /* Size of the full chunk table in the WIM file. */ @@ -150,11 +159,11 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, /* Number of entries we need to actually read from the chunk * table (excludes the implicit first chunk). */ - u64 num_needed_chunk_entries = (start_chunk == 0) ? + u64 num_needed_chunk_entries = (start_chunk == 0) ? num_needed_chunks - 1 : num_needed_chunks; /* Skip over unneeded chunk table entries. */ - u64 file_offset_of_needed_chunk_entries = resource_offset + + u64 file_offset_of_needed_chunk_entries = resource_offset + start_table_idx * chunk_entry_size; if (fseeko(fp, file_offset_of_needed_chunk_entries, SEEK_SET) != 0) { ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" to read " @@ -181,17 +190,17 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, if (chunk_entry_size == 4) { u32 *entries = (u32*)chunk_tab_buf; while (num_needed_chunk_entries--) - *chunk_tab_p++ = to_le32(*entries++); + *chunk_tab_p++ = le32_to_cpu(*entries++); } else { u64 *entries = (u64*)chunk_tab_buf; while (num_needed_chunk_entries--) - *chunk_tab_p++ = to_le64(*entries++); + *chunk_tab_p++ = le64_to_cpu(*entries++); } /* Done with the chunk table now. We must now seek to the first chunk * that is needed for the read. */ - u64 file_offset_of_first_needed_chunk = resource_offset + + u64 file_offset_of_first_needed_chunk = resource_offset + chunk_table_size + chunk_offsets[0]; if (fseeko(fp, file_offset_of_first_needed_chunk, SEEK_SET) != 0) { ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" to read " @@ -227,7 +236,7 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, * expand to WIM_CHUNK_SIZE uncompressed, and the amount * of compressed data for the chunk is given by the * difference of offsets in the chunk offset table. */ - compressed_chunk_size = chunk_offsets[i + 1 - start_chunk] - + compressed_chunk_size = chunk_offsets[i + 1 - start_chunk] - chunk_offsets[i - start_chunk]; uncompressed_chunk_size = WIM_CHUNK_SIZE; } else { @@ -235,15 +244,15 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, * bytes in the file resource, and the last uncompressed * chunk has size equal to however many bytes are left- * that is, the remainder of the uncompressed size when - * divided by WIM_CHUNK_SIZE. + * divided by WIM_CHUNK_SIZE. * * Note that the resource_compressed_size includes the * chunk table, so the size of it must be subtracted. */ - compressed_chunk_size = resource_compressed_size - + compressed_chunk_size = resource_compressed_size - chunk_table_size - chunk_offsets[i - start_chunk]; - uncompressed_chunk_size = resource_uncompressed_size % + uncompressed_chunk_size = resource_uncompressed_size % WIM_CHUNK_SIZE; /* If the remainder is 0, the last chunk actually @@ -270,7 +279,7 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, end_offset = WIM_CHUNK_SIZE - 1; u64 partial_chunk_size = end_offset + 1 - start_offset; - bool is_partial_chunk = (partial_chunk_size != + bool is_partial_chunk = (partial_chunk_size != uncompressed_chunk_size); DEBUG2("start_offset = %u, end_offset = %u", start_offset, @@ -290,7 +299,7 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, return WIMLIB_ERR_READ; } } - if (fread(out_p, 1, partial_chunk_size, fp) != + if (fread(out_p, 1, partial_chunk_size, fp) != partial_chunk_size) goto err; } else { @@ -298,7 +307,7 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, int ret; /* Read the compressed data into compressed_buf. */ - if (fread(compressed_buf, 1, compressed_chunk_size, + if (fread(compressed_buf, 1, compressed_chunk_size, fp) != compressed_chunk_size) goto err; @@ -309,7 +318,7 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, ret = decompress(compressed_buf, compressed_chunk_size, - uncompressed_buf, + uncompressed_buf, uncompressed_chunk_size); if (ret != 0) return WIMLIB_ERR_DECOMPRESSION; @@ -340,7 +349,7 @@ err: return WIMLIB_ERR_READ; } -/* +/* * Reads uncompressed data from an open file stream. */ int read_uncompressed_resource(FILE *fp, u64 offset, u64 len, @@ -380,8 +389,21 @@ const u8 *get_resource_entry(const u8 *p, struct resource_entry *entry) p = get_u8(p, &flags); entry->size = size; entry->flags = flags; + + /* offset and original_size are truncated to 62 bits to avoid possible + * overflows, when converting to a signed 64-bit integer (off_t) or when + * adding size or original_size. This is okay since no one would ever + * actually have a WIM bigger than 4611686018427387903 bytes... */ p = get_u64(p, &entry->offset); + if (entry->offset & 0xc000000000000000ULL) { + WARNING("Truncating offset in resource entry"); + entry->offset &= 0x3fffffffffffffffULL; + } p = get_u64(p, &entry->original_size); + if (entry->original_size & 0xc000000000000000ULL) { + WARNING("Truncating original_size in resource entry"); + entry->original_size &= 0x3fffffffffffffffULL; + } return p; } @@ -397,45 +419,136 @@ u8 *put_resource_entry(u8 *p, const struct resource_entry *entry) return p; } -static int __read_wim_resource(const struct lookup_table_entry *lte, - u8 buf[], size_t size, u64 offset, bool raw) +#ifdef WITH_FUSE +static FILE *wim_get_fp(WIMStruct *w) +{ + pthread_mutex_lock(&w->fp_tab_mutex); + FILE *fp; + + wimlib_assert(w->filename != NULL); + + for (size_t i = 0; i < w->num_allocated_fps; i++) { + if (w->fp_tab[i]) { + fp = w->fp_tab[i]; + w->fp_tab[i] = NULL; + goto out; + } + } + DEBUG("Opening extra file descriptor to `%s'", w->filename); + fp = fopen(w->filename, "rb"); + if (!fp) + ERROR_WITH_ERRNO("Failed to open `%s'", w->filename); +out: + pthread_mutex_unlock(&w->fp_tab_mutex); + return fp; +} + +static int wim_release_fp(WIMStruct *w, FILE *fp) +{ + int ret = 0; + FILE **fp_tab; + + pthread_mutex_lock(&w->fp_tab_mutex); + + for (size_t i = 0; i < w->num_allocated_fps; i++) { + if (w->fp_tab[i] == NULL) { + w->fp_tab[i] = fp; + goto out; + } + } + + fp_tab = REALLOC(w->fp_tab, sizeof(FILE*) * (w->num_allocated_fps + 4)); + if (!fp_tab) { + ret = WIMLIB_ERR_NOMEM; + goto out; + } + w->fp_tab = fp_tab; + memset(&w->fp_tab[w->num_allocated_fps], 0, 4 * sizeof(FILE*)); + w->fp_tab[w->num_allocated_fps] = fp; + w->num_allocated_fps += 4; +out: + pthread_mutex_unlock(&w->fp_tab_mutex); + return ret; +} +#endif + +/* + * Reads some data from the resource corresponding to a WIM lookup table entry. + * + * @lte: The WIM lookup table entry for the resource. + * @buf: Buffer into which to write the data. + * @size: Number of bytes to read. + * @offset: Offset at which to start reading the resource. + * + * Returns zero on success, nonzero on failure. + */ +int read_wim_resource(const struct lookup_table_entry *lte, u8 buf[], + size_t size, u64 offset, int flags) { - wimlib_assert(offset + size <= wim_resource_size(lte)); int ctype; - int ret; + int ret = 0; FILE *fp; + + /* We shouldn't be allowing read over-runs in any part of the library. + * */ + if (flags & WIMLIB_RESOURCE_FLAG_RAW) + wimlib_assert(offset + size <= lte->resource_entry.size); + else + wimlib_assert(offset + size <= lte->resource_entry.original_size); + switch (lte->resource_location) { case RESOURCE_IN_WIM: - wimlib_assert(lte->wim); - wimlib_assert(lte->wim->fp); - ctype = wim_resource_compression_type(lte); - if (ctype == WIM_COMPRESSION_TYPE_NONE && - lte->resource_entry.original_size != - lte->resource_entry.size) { - ERROR("WIM resource at offset %"PRIu64", size %"PRIu64 - "has an original size of %"PRIu64", but is " - "uncompressed", - lte->resource_entry.offset, - lte->resource_entry.size, - lte->resource_entry.original_size); - return WIMLIB_ERR_INVALID_RESOURCE_SIZE; + /* The resource is in a WIM file, and its WIMStruct is given by + * the lte->wim member. The resource may be either compressed + * or uncompressed. */ + wimlib_assert(lte->wim != NULL); + + #ifdef WITH_FUSE + if (flags & WIMLIB_RESOURCE_FLAG_MULTITHREADED) { + fp = wim_get_fp(lte->wim); + if (!fp) + return WIMLIB_ERR_OPEN; + } else + #endif + { + wimlib_assert(!(flags & WIMLIB_RESOURCE_FLAG_MULTITHREADED)); + wimlib_assert(lte->wim->fp != NULL); + fp = lte->wim->fp; } - if (raw || ctype == WIM_COMPRESSION_TYPE_NONE) - return read_uncompressed_resource(lte->wim->fp, - offset, size, buf); + + ctype = wim_resource_compression_type(lte); + + wimlib_assert(ctype != WIM_COMPRESSION_TYPE_NONE || + (lte->resource_entry.original_size == + lte->resource_entry.size)); + + if ((flags & WIMLIB_RESOURCE_FLAG_RAW) + || ctype == WIM_COMPRESSION_TYPE_NONE) + ret = read_uncompressed_resource(fp, + lte->resource_entry.offset + offset, + size, buf); else - return read_compressed_resource(lte->wim->fp, - lte->resource_entry.size, - lte->resource_entry.original_size, - lte->resource_entry.offset, - ctype, size, offset, buf); + ret = read_compressed_resource(fp, + lte->resource_entry.size, + lte->resource_entry.original_size, + lte->resource_entry.offset, + ctype, size, offset, buf); + #ifdef WITH_FUSE + if (flags & WIMLIB_RESOURCE_FLAG_MULTITHREADED) { + int ret2 = wim_release_fp(lte->wim, fp); + if (ret == 0) + ret = ret2; + } + #endif break; case RESOURCE_IN_STAGING_FILE: - wimlib_assert(lte->staging_file_name); - wimlib_assert(0); - break; case RESOURCE_IN_FILE_ON_DISK: + /* The resource is in some file on the external filesystem and + * needs to be read uncompressed */ wimlib_assert(lte->file_on_disk); + wimlib_assert(<e->file_on_disk == <e->staging_file_name); + /* Use existing file pointer if available; otherwise open one + * temporarily */ if (lte->file_on_disk_fp) { fp = lte->file_on_disk_fp; } else { @@ -443,265 +556,141 @@ static int __read_wim_resource(const struct lookup_table_entry *lte, if (!fp) { ERROR_WITH_ERRNO("Failed to open the file " "`%s'", lte->file_on_disk); + ret = WIMLIB_ERR_OPEN; + break; } } - ret = read_uncompressed_resource(lte->file_on_disk_fp, - offset, size, buf); + ret = read_uncompressed_resource(fp, offset, size, buf); if (fp != lte->file_on_disk_fp) fclose(fp); - return ret; break; case RESOURCE_IN_ATTACHED_BUFFER: - wimlib_assert(lte->attached_buffer); + /* The resource is directly attached uncompressed in an + * in-memory buffer. */ + wimlib_assert(lte->attached_buffer != NULL); memcpy(buf, lte->attached_buffer + offset, size); - return 0; break; +#ifdef WITH_NTFS_3G + case RESOURCE_IN_NTFS_VOLUME: + wimlib_assert(lte->ntfs_loc != NULL); + wimlib_assert(lte->attr != NULL); + if (lte->ntfs_loc->is_reparse_point) + offset += 8; + if (ntfs_attr_pread(lte->attr, offset, size, buf) != size) { + ERROR_WITH_ERRNO("Error reading NTFS attribute " + "at `%s'", + lte->ntfs_loc->path_utf8); + ret = WIMLIB_ERR_NTFS_3G; + } + break; +#endif default: - assert(0); - } -} - -int read_wim_resource(const struct lookup_table_entry *lte, u8 buf[], - size_t size, u64 offset) -{ - return __read_wim_resource(lte, buf, size, offset, false); -} - -int read_full_wim_resource(const struct lookup_table_entry *lte, u8 buf[]) -{ - return __read_wim_resource(lte, buf, lte->resource_entry.original_size, - 0, false); -} - -struct chunk_table { - off_t file_offset; - u64 num_chunks; - u64 original_resource_size; - u64 bytes_per_chunk_entry; - u64 table_disk_size; - u64 cur_offset; - u64 *cur_offset_p; - u64 offsets[0]; -}; - -static int -begin_wim_resource_chunk_tab(const struct lookup_table_entry *lte, - FILE *out_fp, - struct chunk_table **chunk_tab_ret) -{ - u64 size = wim_resource_size(lte); - u64 num_chunks = (size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE; - struct chunk_table *chunk_tab = MALLOC(sizeof(struct chunk_table) + - num_chunks * sizeof(u64)); - int ret = 0; - - if (!chunk_tab) { - ERROR("Failed to allocate chunk table for %"PRIu64" byte " - "resource", size); - ret = WIMLIB_ERR_NOMEM; - goto out; - } - chunk_tab->file_offset = ftello(out_fp); - if (chunk_tab->file_offset == -1) { - ERROR_WITH_ERRNO("Failed to get file offset in output WIM"); - ret = WIMLIB_ERR_WRITE; - goto out; - } - chunk_tab->num_chunks = num_chunks; - chunk_tab->cur_offset_p = chunk_tab->offsets; - chunk_tab->original_resource_size = lte->resource_entry.original_size; - chunk_tab->bytes_per_chunk_entry = - (lte->resource_entry.original_size >= (1ULL << 32)) - ? 8 : 4; - chunk_tab->table_disk_size = chunk_tab->bytes_per_chunk_entry * - (num_chunks - 1); - - if (fwrite(chunk_tab, 1, chunk_tab->table_disk_size, out_fp) != - chunk_tab->table_disk_size) - { - ERROR_WITH_ERRNO("Failed to write chunk table in compressed " - "file resource"); - ret = WIMLIB_ERR_WRITE; - goto out; + wimlib_assert(0); + ret = -1; + break; } - - *chunk_tab_ret = chunk_tab; -out: return ret; } -static int compress_chunk(const u8 chunk[], unsigned chunk_sz, - u8 compressed_chunk[], - unsigned *compressed_chunk_len_ret, - int ctype) -{ - unsigned compressed_chunk_sz; - int (*compress)(const void *, unsigned, void *, unsigned *); - if (ctype == WIM_COMPRESSION_TYPE_LZX) - compress = lzx_compress; - else - compress = xpress_compress; - return (*compress)(chunk, chunk_sz, compressed_chunk, - compressed_chunk_len_ret); -} - -static int write_wim_resource_chunk(const u8 chunk[], unsigned chunk_size, - FILE *out_fp, int out_ctype, - struct chunk_table *chunk_tab) -{ - const u8 *out_chunk; - unsigned out_chunk_size; - - if (out_ctype == WIM_COMPRESSION_TYPE_NONE) { - out_chunk = chunk; - out_chunk_size = chunk_size; - } else { - u8 *compressed_chunk = alloca(chunk_size); - int ret; - unsigned compressed_chunk_len; - - wimlib_assert(chunk_tab != NULL); - - ret = compress_chunk(chunk, chunk_size, compressed_chunk, - &out_chunk_size, out_ctype); - if (ret > 0) - return ret; - else if (ret < 0) { - out_chunk = chunk; - out_chunk_size = chunk_size; - } else { - out_chunk = compressed_chunk; - } - *chunk_tab->cur_offset_p++ = chunk_tab->cur_offset; - chunk_tab->cur_offset += out_chunk_size; - } - - if (fwrite(out_chunk, 1, out_chunk_size, out_fp) != out_chunk_size) { - ERROR_WITH_ERRNO("Failed to write WIM resource chunk"); - return WIMLIB_ERR_WRITE; - } - return 0; -} - -static int -finish_wim_resource_chunk_tab(const struct chunk_table *chunk_tab, - FILE *out_fp) -{ - if (fseeko(out_fp, chunk_tab->file_offset, SEEK_SET) != 0) { - ERROR_WITH_ERRNO("Failed to seet to byte "PRIu64" of output " - "WIM file", chunk_tab->file_offset); - return WIMLIB_ERR_WRITE; - } - - if (chunk_tab->bytes_per_chunk_entry == 8) { - array_to_le64(chunk_tab->offsets, chunk_tab->num_chunks - 1); - } else { - for (u64 i = 0; i < chunk_tab->num_chunks - 1; i++) - ((u32*)chunk_tab->offsets)[i] = - to_le32(chunk_tab->offsets[i]); - } - if (fwrite(chunk_tab->offsets, 1, chunk_tab->table_disk_size, out_fp) != - chunk_tab->table_disk_size) - { - ERROR_WITH_ERRNO("Failed to write chunk table in compressed " - "file resource"); - return WIMLIB_ERR_WRITE; - } - if (fseeko(out_fp, chunk_tab->file_offset, SEEK_SET) != 0) { - ERROR_WITH_ERRNO("Failed to seet to byte "PRIu64" of output " - "WIM file", chunk_tab->file_offset); - return WIMLIB_ERR_WRITE; - } - return 0; -} - -static int write_wim_resource(const struct lookup_table_entry *lte, - FILE *out_fp, int out_ctype) +/* + * Reads all the data from the resource corresponding to a WIM lookup table + * entry. + * + * @lte: The WIM lookup table entry for the resource. + * @buf: Buffer into which to write the data. It must be at least + * wim_resource_size(lte) bytes long. + * + * Returns 0 on success; nonzero on failure. + */ +int read_full_wim_resource(const struct lookup_table_entry *lte, u8 buf[], + int flags) { - u64 bytes_remaining = wim_resource_size(lte); - char buf[min(WIM_CHUNK_SIZE, bytes_remaining)]; - u64 offset = 0; - int ret = 0; - bool raw = wim_resource_compression_type(lte) == out_ctype; - struct chunk_table *chunk_tab = NULL; - - if (raw) - out_ctype = WIM_COMPRESSION_TYPE_NONE; - - if (out_ctype != WIM_COMPRESSION_TYPE_NONE) { - ret = begin_wim_resource_chunk_tab(lte, out_fp, - &chunk_tab); - if (ret != 0) - return 0; - } - - while (bytes_remaining) { - u64 to_read = min(bytes_remaining, WIM_CHUNK_SIZE); - ret = __read_wim_resource(lte, buf, to_read, offset, raw); - if (ret != 0) - break; - ret = write_wim_resource_chunk(buf, to_read, out_fp, - out_ctype, chunk_tab); - if (ret != 0) - break; - bytes_remaining -= to_read; - offset += to_read; - } - if (out_ctype != WIM_COMPRESSION_TYPE_NONE) - ret = finish_wim_resource_chunk_tab(chunk_tab, out_fp); - return ret; + return read_wim_resource(lte, buf, wim_resource_size(lte), 0, flags); } +/* Like write_wim_resource(), but the resource is specified by a buffer of + * uncompressed data rather a lookup table entry; also writes the SHA1 hash of + * the buffer to @hash. */ static int write_wim_resource_from_buffer(const u8 *buf, u64 buf_size, - FILE *out_fp, int out_ctype) + FILE *out_fp, int out_ctype, + struct resource_entry *out_res_entry, + u8 hash[SHA1_HASH_SIZE]) { + /* Set up a temporary lookup table entry to provide to + * write_wim_resource(). */ struct lookup_table_entry lte; - lte.resource_entry.flags = 0; + int ret; + lte.resource_entry.flags = 0; lte.resource_entry.original_size = buf_size; - lte.resource_entry.size = buf_size; - lte.resource_entry.offset = 0; - lte.resource_location = RESOURCE_IN_ATTACHED_BUFFER; - lte.attached_buffer = (u8*)buf; - return write_wim_resource(<e, out_fp, out_ctype); + lte.resource_entry.size = buf_size; + lte.resource_entry.offset = 0; + lte.resource_location = RESOURCE_IN_ATTACHED_BUFFER; + lte.attached_buffer = (u8*)buf; + + zero_out_hash(lte.hash); + ret = write_wim_resource(<e, out_fp, out_ctype, out_res_entry, 0); + if (ret != 0) + return ret; + copy_hash(hash, lte.hash); + return 0; } -/* - * Extracts the first @size bytes of the resource specified by @lte to the open - * file @fd. Returns nonzero on error. +/* + * Extracts the first @size bytes of the WIM resource specified by @lte to the + * open file descriptor @fd. + * + * Returns 0 on success; nonzero on failure. */ int extract_wim_resource_to_fd(const struct lookup_table_entry *lte, int fd, u64 size) { u64 bytes_remaining = size; - char buf[min(WIM_CHUNK_SIZE, bytes_remaining)]; + u8 buf[min(WIM_CHUNK_SIZE, bytes_remaining)]; u64 offset = 0; int ret = 0; + u8 hash[SHA1_HASH_SIZE]; + + SHA_CTX ctx; + sha1_init(&ctx); while (bytes_remaining) { u64 to_read = min(bytes_remaining, WIM_CHUNK_SIZE); - ret = read_wim_resource(lte, buf, to_read, offset); + ret = read_wim_resource(lte, buf, to_read, offset, 0); if (ret != 0) break; - if (full_write(fd, buf, to_read) < 0) { + sha1_update(&ctx, buf, to_read); + if (full_write(fd, buf, to_read) < to_read) { ERROR_WITH_ERRNO("Error extracting WIM resource"); return WIMLIB_ERR_WRITE; } bytes_remaining -= to_read; offset += to_read; } + sha1_final(hash, &ctx); + if (!hashes_equal(hash, lte->hash)) { + ERROR("Invalid checksum on a WIM resource " + "(detected when extracting to external file)"); + ERROR("The following WIM resource is invalid:"); + print_lookup_table_entry(lte); + return WIMLIB_ERR_INVALID_RESOURCE_HASH; + } return 0; } +/* + * Extracts the WIM resource specified by @lte to the open file descriptor @fd. + * + * Returns 0 on success; nonzero on failure. + */ int extract_full_wim_resource_to_fd(const struct lookup_table_entry *lte, int fd) { - return extract_wim_resource_to_fd(lte, fd, - lte->resource_entry.original_size); + return extract_wim_resource_to_fd(lte, fd, wim_resource_size(lte)); } -/* +/* * Copies the file resource specified by the lookup table entry @lte from the - * input WIM, pointed to by the fp field of the WIMStruct, to the output WIM, - * pointed to by the out_fp field of the WIMStruct. + * input WIM to the output WIM that has its FILE * given by + * ((WIMStruct*)wim)->out_fp. * * The output_resource_entry, out_refcnt, and part_number fields of @lte are * updated. @@ -713,63 +702,22 @@ int copy_resource(struct lookup_table_entry *lte, void *wim) { WIMStruct *w = wim; int ret; - off_t new_offset; if ((lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) && !w->write_metadata) return 0; - new_offset = ftello(w->out_fp); - if (new_offset == -1) { - ERROR_WITH_ERRNO("Could not get offset in output WIM"); - return WIMLIB_ERR_WRITE; - } - - ret = write_wim_resource(lte, w->out_fp, - wimlib_get_compression_type((WIMStruct*)w)); + wim_resource_compression_type(lte), + <e->output_resource_entry, 0); if (ret != 0) return ret; - - memcpy(<e->output_resource_entry, <e->resource_entry, - sizeof(struct resource_entry)); - - lte->output_resource_entry.offset = new_offset; lte->out_refcnt = lte->refcnt; lte->part_number = w->hdr.part_number; return 0; } -/* - * Writes a dentry's resources, including the main file resource as well as all - * alternate data streams, to the output file. - * - * @dentry: The dentry for the file. - * @wim_p: A pointer to the WIMStruct. The fields of interest to this - * function are the input and output file streams and the lookup - * table. - * - * @return zero on success, nonzero on failure. - */ -int write_dentry_resources(struct dentry *dentry, void *wim_p) -{ - WIMStruct *w = wim_p; - int ret = 0; - struct lookup_table_entry *lte; - int ctype = wimlib_get_compression_type(w); - - for (unsigned i = 0; i <= dentry->num_ads; i++) { - lte = dentry_stream_lte(dentry, i, w->lookup_table); - if (lte && ++lte->out_refcnt == 1) { - ret = write_wim_resource(lte, w->out_fp, ctype); - if (ret != 0) - break; - } - } - return ret; -} - -/* +/* * Reads the metadata metadata resource from the WIM file. The metadata * resource consists of the security data, followed by the directory entry for * the root directory, followed by all the other directory entries in the @@ -780,66 +728,89 @@ int write_dentry_resources(struct dentry *dentry, void *wim_p) * * @fp: The FILE* for the input WIM file. * @wim_ctype: The compression type of the WIM file. - * @imd: Pointer to the image metadata structure. Its - * `lookup_table_entry' member specifies the lookup table entry for - * the metadata resource. The rest of the image metadata entry - * will be filled in by this function. + * @imd: Pointer to the image metadata structure. Its `metadata_lte' + * member specifies the lookup table entry for the metadata + * resource. The rest of the image metadata entry will be filled + * in by this function. * * @return: Zero on success, nonzero on failure. */ -int read_metadata_resource(FILE *fp, int wim_ctype, struct image_metadata *imd) +int read_metadata_resource(WIMStruct *w, struct image_metadata *imd) { u8 *buf; - int ctype; u32 dentry_offset; int ret; struct dentry *dentry; - struct wim_security_data *sd; - struct link_group_table *lgt; + struct inode_table inode_tab; const struct lookup_table_entry *metadata_lte; - const struct resource_entry *res_entry; + u64 metadata_len; + u64 metadata_offset; + struct hlist_head inode_list; metadata_lte = imd->metadata_lte; - res_entry = &metadata_lte->resource_entry; + metadata_len = wim_resource_size(metadata_lte); + metadata_offset = metadata_lte->resource_entry.offset; DEBUG("Reading metadata resource: length = %"PRIu64", " - "offset = %"PRIu64"", - res_entry->original_size, res_entry->offset); + "offset = %"PRIu64"", metadata_len, metadata_offset); + + /* There is no way the metadata resource could possibly be less than (8 + * + WIM_DENTRY_DISK_SIZE) bytes, where the 8 is for security data (with + * no security descriptors) and WIM_DENTRY_DISK_SIZE is for the root + * dentry. */ + if (metadata_len < 8 + WIM_DENTRY_DISK_SIZE) { + ERROR("Expected at least %u bytes for the metadata resource", + 8 + WIM_DENTRY_DISK_SIZE); + return WIMLIB_ERR_INVALID_RESOURCE_SIZE; + } - if (res_entry->original_size < 8) { - ERROR("Expected at least 8 bytes for the metadata resource"); + if (sizeof(size_t) < 8 && metadata_len > 0xffffffff) { + ERROR("Metadata resource is too large (%"PRIu64" bytes", + metadata_len); return WIMLIB_ERR_INVALID_RESOURCE_SIZE; } /* Allocate memory for the uncompressed metadata resource. */ - buf = MALLOC(res_entry->original_size); + buf = MALLOC(metadata_len); if (!buf) { ERROR("Failed to allocate %"PRIu64" bytes for uncompressed " - "metadata resource", res_entry->original_size); + "metadata resource", metadata_len); return WIMLIB_ERR_NOMEM; } - /* Determine the compression type of the metadata resource. */ - /* Read the metadata resource into memory. (It may be compressed.) */ - ret = read_full_wim_resource(metadata_lte, buf); + ret = read_full_wim_resource(metadata_lte, buf, 0); if (ret != 0) goto out_free_buf; DEBUG("Finished reading metadata resource into memory."); - /* The root directory entry starts after security data, on an 8-byte - * aligned address. + /* The root directory entry starts after security data, aligned on an + * 8-byte boundary within the metadata resource. * * The security data starts with a 4-byte integer giving its total - * length. */ + * length, so if we round that up to an 8-byte boundary that gives us + * the offset of the root dentry. + * + * Here we read the security data into a wim_security_data structure, + * and if successful, go ahead and calculate the offset in the metadata + * resource of the root dentry. */ - /* Read the security data into a wim_security_data structure. */ - ret = read_security_data(buf, res_entry->original_size, &sd); + wimlib_assert(imd->security_data == NULL); + ret = read_security_data(buf, metadata_len, &imd->security_data); if (ret != 0) goto out_free_buf; + dentry_offset = (imd->security_data->total_length + 7) & ~7; + + if (dentry_offset == 0) { + ERROR("Integer overflow while reading metadata resource"); + ret = WIMLIB_ERR_INVALID_SECURITY_DATA; + goto out_free_security_data; + } + + /* Allocate memory for the root dentry and read it into memory */ dentry = MALLOC(sizeof(struct dentry)); if (!dentry) { ERROR("Failed to allocate %zu bytes for root dentry", @@ -848,93 +819,101 @@ int read_metadata_resource(FILE *fp, int wim_ctype, struct image_metadata *imd) goto out_free_security_data; } - get_u32(buf, &dentry_offset); - if (dentry_offset == 0) - dentry_offset = 8; - dentry_offset = (dentry_offset + 7) & ~7; - - ret = read_dentry(buf, res_entry->original_size, dentry_offset, dentry); - /* This is the root dentry, so set its pointers correctly. */ + ret = read_dentry(buf, metadata_len, dentry_offset, dentry); + + /* This is the root dentry, so set its parent to itself. */ dentry->parent = dentry; - dentry->next = dentry; - dentry->prev = dentry; + if (ret != 0) goto out_free_dentry_tree; + inode_add_dentry(dentry, dentry->d_inode); + /* Now read the entire directory entry tree into memory. */ DEBUG("Reading dentry tree"); - /* Now read the entire directory entry tree. */ - ret = read_dentry_tree(buf, res_entry->original_size, dentry); + ret = read_dentry_tree(buf, metadata_len, dentry); if (ret != 0) goto out_free_dentry_tree; - DEBUG("Calculating dentry full paths"); /* Calculate the full paths in the dentry tree. */ + DEBUG("Calculating dentry full paths"); ret = for_dentry_in_tree(dentry, calculate_dentry_full_path, NULL); if (ret != 0) goto out_free_dentry_tree; - DEBUG("Building link group table"); /* Build hash table that maps hard link group IDs to dentry sets */ - lgt = new_link_group_table(9001); - if (!lgt) + DEBUG("Building link group table"); + ret = init_inode_table(&inode_tab, 9001); + if (ret != 0) goto out_free_dentry_tree; - ret = for_dentry_in_tree(dentry, link_group_table_insert, lgt); + + for_dentry_in_tree(dentry, inode_table_insert, &inode_tab); + + DEBUG("Fixing inconsistencies in the hard link groups"); + ret = fix_inodes(&inode_tab, &inode_list); + destroy_inode_table(&inode_tab); if (ret != 0) - goto out_free_lgt; + goto out_free_dentry_tree; - DEBUG("Freeing duplicate ADS entries in link group table"); - ret = link_groups_free_duplicate_data(lgt); + DEBUG("Running miscellaneous verifications on the dentry tree"); + for_lookup_table_entry(w->lookup_table, lte_zero_real_refcnt, NULL); + ret = for_dentry_in_tree(dentry, verify_dentry, w); if (ret != 0) - goto out_free_lgt; + goto out_free_dentry_tree; + DEBUG("Done reading image metadata"); - imd->lgt = lgt; - imd->security_data = sd; - imd->root_dentry = dentry; + imd->root_dentry = dentry; + imd->inode_list = inode_list; goto out_free_buf; -out_free_lgt: - free_link_group_table(lgt); out_free_dentry_tree: free_dentry_tree(dentry, NULL); out_free_security_data: - free_security_data(sd); + free_security_data(imd->security_data); + imd->security_data = NULL; out_free_buf: FREE(buf); return ret; } -/* Write the metadata resource for the current image. */ +/* Write the metadata resource for the current WIM image. */ int write_metadata_resource(WIMStruct *w) { - FILE *out; u8 *buf; u8 *p; int ret; u64 subdir_offset; struct dentry *root; struct lookup_table_entry *lte; - off_t metadata_offset; u64 metadata_original_size; - u64 metadata_compressed_size; - int metadata_ctype; - u8 hash[SHA1_HASH_SIZE]; + const struct wim_security_data *sd; DEBUG("Writing metadata resource for image %d", w->current_image); - out = w->out_fp; root = wim_root_dentry(w); - metadata_ctype = wimlib_get_compression_type(w); - metadata_offset = ftello(out); - if (metadata_offset == -1) - return WIMLIB_ERR_WRITE; - - struct wim_security_data *sd = wim_security_data(w); - if (sd) - subdir_offset = sd->total_length + root->length + 8; - else - subdir_offset = 8 + root->length + 8; + sd = wim_security_data(w); + + /* We do not allow the security data pointer to be NULL, although it may + * point to an empty security data with no entries. */ + wimlib_assert(root != NULL); + wimlib_assert(sd != NULL); + + /* Offset of first child of the root dentry. It's equal to: + * - The total length of the security data, rounded to the next 8-byte + * boundary, + * - plus the total length of the root dentry, + * - plus 8 bytes for an end-of-directory entry following the root + * dentry (shouldn't really be needed, but just in case...) + */ + subdir_offset = ((sd->total_length + 7) & ~7) + + dentry_correct_total_length(root) + 8; + + /* Calculate the subdirectory offsets for the entire dentry tree. */ calculate_subdir_offsets(root, &subdir_offset); + + /* Total length of the metadata resource (uncompressed) */ metadata_original_size = subdir_offset; + + /* Allocate a buffer to contain the uncompressed metadata resource */ buf = MALLOC(metadata_original_size); if (!buf) { ERROR("Failed to allocate %"PRIu64" bytes for " @@ -942,40 +921,55 @@ int write_metadata_resource(WIMStruct *w) return WIMLIB_ERR_NOMEM; } + /* Write the security data into the resource buffer */ p = write_security_data(sd, buf); - DEBUG("Writing dentry tree."); + /* Write the dentry tree into the resource buffer */ p = write_dentry_tree(root, p); - /* Like file resources, the lookup table entry for a metadata resource - * uses for the hash code a SHA1 message digest of its uncompressed - * contents. */ - sha1_buffer(buf, metadata_original_size, hash); + /* We MUST have exactly filled the buffer; otherwise we calculated its + * size incorrectly or wrote the data incorrectly. */ + wimlib_assert(p - buf == metadata_original_size); + + /* Get the lookup table entry for the metadata resource so we can update + * it. */ + lte = wim_metadata_lookup_table_entry(w); + wimlib_assert(lte != NULL); + /* Write the metadata resource to the output WIM using the proper + * compression type. The lookup table entry for the metadata resource + * is updated. */ ret = write_wim_resource_from_buffer(buf, metadata_original_size, - out, metadata_ctype); - FREE(buf); + w->out_fp, + wimlib_get_compression_type(w), + <e->output_resource_entry, + lte->hash); if (ret != 0) - return ret; - - DEBUG("Updating metadata lookup table entry (size %zu)", - metadata_original_size); + goto out; - /* Update the lookup table entry, including the hash and output resource - * entry fields, for this image's metadata resource. */ - lte = wim_metadata_lookup_table_entry(w); - lte->out_refcnt++; - if (!hashes_equal(hash, lte->hash)) { - lookup_table_unlink(w->lookup_table, lte); - copy_hash(lte->hash, hash); - lookup_table_insert(w->lookup_table, lte); - } - lte->output_resource_entry.original_size = metadata_original_size; - lte->output_resource_entry.offset = metadata_offset; - lte->output_resource_entry.size = metadata_compressed_size; - lte->output_resource_entry.flags = WIM_RESHDR_FLAG_METADATA; - if (metadata_ctype != WIM_COMPRESSION_TYPE_NONE) - lte->output_resource_entry.flags |= WIM_RESHDR_FLAG_COMPRESSED; - return 0; + /* It's very likely the SHA1 message digest of the metadata resource + * changed, so re-insert the lookup table entry into the lookup table. + * + * We do not check for other lookup table entries having the same SHA1 + * message digest. It's possible for 2 absolutely identical images to + * be added, therefore causing 2 identical metadata resources to be in + * the WIM. However, in this case, it's expected for 2 separate lookup + * table entries to be created, even though this doesn't make a whole + * lot of sense since they will share the same SHA1 message digest. + * */ + lookup_table_unlink(w->lookup_table, lte); + lookup_table_insert(w->lookup_table, lte); + + wimlib_assert(lte->out_refcnt == 0); + lte->out_refcnt = 1; + + /* Make sure that the resource entry is written marked with the metadata + * flag. */ + lte->output_resource_entry.flags |= WIM_RESHDR_FLAG_METADATA; +out: + /* All the data has been written to the new WIM; no need for the buffer + * anymore */ + FREE(buf); + return ret; }