X-Git-Url: https://wimlib.net/git/?a=blobdiff_plain;f=src%2Fresource.c;h=86a2be6a18922c720f1f5b2e4b3be95bb0d25a19;hb=1cf5ed19520ae061e8b9b694100897864cbcf538;hp=600bf5ba1321bb772f4d788ccfe33aad456c661b;hpb=885632f08c75c1d7bb5d25436231c78f6ad7e0c0;p=wimlib diff --git a/src/resource.c b/src/resource.c index 600bf5ba..86a2be6a 100644 --- a/src/resource.c +++ b/src/resource.c @@ -2,43 +2,59 @@ * resource.c * * Read uncompressed and compressed metadata and file resources. - * - * Copyright (C) 2010 Carl Thijssen + */ + +/* * Copyright (C) 2012 Eric Biggers * - * wimlib - Library for working with WIM files + * This file is part of wimlib, a library for working with WIM files. * - * This library 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. + * wimlib is free software; you can redistribute it and/or modify it under the + * 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. * - * This library 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. + * 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 General Public License for more details. * - * You should have received a copy of the GNU Lesser General Public License along - * with this library; if not, write to the Free Software Foundation, Inc., 59 - * Temple Place, Suite 330, Boston, MA 02111-1307 USA + * 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 "dentry.h" +#include "sha1.h" #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 @@ -48,28 +64,26 @@ * * 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[]) { - DEBUG2("comp size = %"PRIu64", " - "uncomp size = %"PRIu64", " - "res offset = %"PRIu64"\n", - resource_compressed_size, - resource_uncompressed_size, - resource_offset); - DEBUG2("resource_ctype = %s, len = %"PRIu64", offset = %"PRIu64"\n", - wimlib_get_compression_type_string(resource_ctype), - len, offset); + DEBUG2("comp size = %"PRIu64", uncomp size = %"PRIu64", " + "res offset = %"PRIu64"", + resource_compressed_size, + resource_uncompressed_size, + resource_offset); + DEBUG2("resource_ctype = %s, len = %"PRIu64", offset = %"PRIu64"", + wimlib_get_compression_type_string(resource_ctype), len, offset); /* Trivial case */ if (len == 0) return 0; int (*decompress)(const void *, uint, void *, uint); /* Set the appropriate decompress function. */ - if (resource_ctype == WIM_COMPRESSION_TYPE_LZX) + if (resource_ctype == WIMLIB_COMPRESSION_TYPE_LZX) decompress = lzx_decompress; else decompress = xpress_decompress; @@ -83,7 +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. */ @@ -131,7 +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. */ @@ -145,17 +159,16 @@ 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("Failed to seek to byte %"PRIu64" " - "to read chunk table of compressed " - "resource: %m\n", - file_offset_of_needed_chunk_entries); + ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" to read " + "chunk table of compressed resource", + file_offset_of_needed_chunk_entries); return WIMLIB_ERR_READ; } @@ -177,23 +190,22 @@ 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("Failed to seek to byte %"PRIu64" " - "to read first chunk of compressed " - "resource: %m\n", - file_offset_of_first_needed_chunk); + ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" to read " + "first chunk of compressed resource", + file_offset_of_first_needed_chunk); return WIMLIB_ERR_READ; } @@ -213,8 +225,8 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, /* Decompress all the chunks. */ for (u64 i = start_chunk; i <= end_chunk; i++) { - DEBUG2("Chunk %"PRIu64" (start %"PRIu64", end %"PRIu64")\n", - i, start_chunk, end_chunk); + DEBUG2("Chunk %"PRIu64" (start %"PRIu64", end %"PRIu64").", + i, start_chunk, end_chunk); /* Calculate the sizes of the compressed chunk and of the * uncompressed chunk. */ @@ -224,7 +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 { @@ -232,15 +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 @@ -249,8 +261,9 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, uncompressed_chunk_size = WIM_CHUNK_SIZE; } - DEBUG2("compressed_chunk_size = %u, uncompressed_chunk_size = %u\n", - compressed_chunk_size, uncompressed_chunk_size); + DEBUG2("compressed_chunk_size = %u, " + "uncompressed_chunk_size = %u", + compressed_chunk_size, uncompressed_chunk_size); /* Figure out how much of this chunk we actually need to read */ @@ -266,12 +279,12 @@ 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\n", start_offset, + DEBUG2("start_offset = %u, end_offset = %u", start_offset, end_offset); - DEBUG2("partial_chunk_size = %u\n", partial_chunk_size); + DEBUG2("partial_chunk_size = %u", partial_chunk_size); /* This is undocumented, but chunks can be uncompressed. This * appears to always be the case when the compressed chunk size @@ -281,12 +294,12 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, if (start_offset != 0) { if (fseeko(fp, start_offset, SEEK_CUR) != 0) { - ERROR("Uncompressed partial chunk " - "fseek() error: %m\n"); + ERROR_WITH_ERRNO("Uncompressed partial " + "chunk fseek() error"); 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 { @@ -294,7 +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; @@ -305,14 +318,13 @@ 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; memcpy(out_p, uncompressed_buf + start_offset, partial_chunk_size); } else { - DEBUG2("out_p = %p\n"); ret = decompress(compressed_buf, compressed_chunk_size, out_p, @@ -331,136 +343,39 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, err: if (feof(fp)) - ERROR("Unexpected EOF in compressed file resource\n"); + ERROR("Unexpected EOF in compressed file resource"); else - ERROR("Error reading compressed file resource: %m\n"); + ERROR_WITH_ERRNO("Error reading compressed file resource"); return WIMLIB_ERR_READ; } -/* +/* * Reads uncompressed data from an open file stream. */ -int read_uncompressed_resource(FILE *fp, u64 offset, u64 len, - u8 contents_ret[]) +int read_uncompressed_resource(FILE *fp, u64 offset, u64 len, + u8 contents_ret[]) { if (fseeko(fp, offset, SEEK_SET) != 0) { ERROR("Failed to seek to byte %"PRIu64" of input file " - "to read uncompressed resource " - "(len = %"PRIu64")!\n", offset, len); + "to read uncompressed resource (len = %"PRIu64")", + offset, len); return WIMLIB_ERR_READ; } if (fread(contents_ret, 1, len, fp) != len) { if (feof(fp)) { - ERROR("Unexpected EOF in uncompressed file resource!\n"); + ERROR("Unexpected EOF in uncompressed file resource"); } else { ERROR("Failed to read %"PRIu64" bytes from " - "uncompressed resource at offset " - "%"PRIu64"\n", len, offset); + "uncompressed resource at offset %"PRIu64, + len, offset); } return WIMLIB_ERR_READ; } return 0; } -/* - * Reads a WIM resource. - * - * @fp: The FILE* for the WIM file. - * @resource_size: The compressed size of the resource. - * @resource_original_size: The uncompressed size of the resource. - * @resource_offset: The offset of the resource in the stream @fp. - * @resource_ctype: The compression type of the resource. - * (WIM_COMPRESSION_TYPE_*) - * @len: How many bytes of the resource should be read. - * @offset: The offset within the resource at which the read - * will occur. - * - * To read the whole file resource, specify offset = - * 0 and len = resource_original_size, or call - * read_full_resource(). - * - * @contents_ret: An array, that must have length at least @len, - * into which the uncompressed contents of - * the file resource starting at @offset and - * continuing for @len bytes will be written. - * - * @return: Zero on success, nonzero on failure. Failure may be due to - * being unable to read the data from the WIM file at the - * specified length and offset, or it may be due to the - * compressed data (if the data is compressed) being - * invalid. - */ -int read_resource(FILE *fp, u64 resource_size, u64 resource_original_size, - u64 resource_offset, int resource_ctype, u64 len, - u64 offset, void *contents_ret) -{ - if (resource_ctype == WIM_COMPRESSION_TYPE_NONE) { - if (resource_size != resource_original_size) { - ERROR("Resource with original size %"PRIu64" " - "bytes is marked as uncompressed, \n", - resource_original_size); - ERROR(" but its actual size is %"PRIu64" " - "bytes!\n", - resource_size); - return WIMLIB_ERR_INVALID_RESOURCE_SIZE; - } - return read_uncompressed_resource(fp, - resource_offset + offset, - len, contents_ret); - } else { - return read_compressed_resource(fp, resource_size, - resource_original_size, resource_offset, - resource_ctype, len, offset, contents_ret); - } -} - - -/* - * Extracts the first @size bytes file resource specified by @entry to the open - * file @fd. Returns nonzero on error. - * - * XXX - * This function is somewhat redundant with uncompress_resource(). The - * main difference is that this function writes to a file descriptor using - * low-level calls to write() rather than to a FILE* with fwrite(); also this - * function allows only up to @size bytes to be extracted. - */ -int extract_resource_to_fd(WIMStruct *w, const struct resource_entry *entry, - int fd, u64 size) -{ - u64 num_chunks; - u64 n; - u8 buf[min(size, WIM_CHUNK_SIZE)]; - int res_ctype; - u64 offset; - u64 i; - int ret; - - errno = 0; - - num_chunks = (size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE; - n = WIM_CHUNK_SIZE; - res_ctype = wim_resource_compression_type(w, entry); - offset = 0; - for (i = 0; i < num_chunks; i++) { - if (i == num_chunks - 1) { - n = size % WIM_CHUNK_SIZE; - if (n == 0) { - n = WIM_CHUNK_SIZE; - } - } - ret = read_resource(w->fp, entry->size, entry->original_size, - entry->offset, res_ctype, n, offset, buf); - if (ret != 0) - return ret; - if (full_write(fd, buf, n) != n) - return WIMLIB_ERR_WRITE; - offset += n; - } - return ret; -} /* Reads the contents of a struct resource_entry, as represented in the on-disk * format, from the memory pointed to by @p, and fills in the fields of @entry. @@ -474,8 +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; } @@ -491,21 +419,302 @@ u8 *put_resource_entry(u8 *p, const struct resource_entry *entry) return p; } -/* Given the compression type for the WIM file as a whole as the flags field of - * a resource entry, returns the compression type for that resource entry. */ -int resource_compression_type(int wim_ctype, int reshdr_flags) +#ifdef WITH_FUSE +static FILE *wim_get_fp(WIMStruct *w) { - if (wim_ctype == WIM_COMPRESSION_TYPE_NONE) { - return WIM_COMPRESSION_TYPE_NONE; - } else { - if (reshdr_flags & WIM_RESHDR_FLAG_COMPRESSED) - return wim_ctype; + 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) +{ + int ctype; + 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: + /* 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; + } + + ctype = wim_resource_compression_type(lte); + + wimlib_assert(ctype != WIMLIB_COMPRESSION_TYPE_NONE || + (lte->resource_entry.original_size == + lte->resource_entry.size)); + + if ((flags & WIMLIB_RESOURCE_FLAG_RAW) + || ctype == WIMLIB_COMPRESSION_TYPE_NONE) + ret = read_uncompressed_resource(fp, + lte->resource_entry.offset + offset, + size, buf); else - return WIM_COMPRESSION_TYPE_NONE; + 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: + 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 { + fp = fopen(lte->file_on_disk, "rb"); + if (!fp) { + ERROR_WITH_ERRNO("Failed to open the file " + "`%s'", lte->file_on_disk); + ret = WIMLIB_ERR_OPEN; + break; + } + } + ret = read_uncompressed_resource(fp, offset, size, buf); + if (fp != lte->file_on_disk_fp) + fclose(fp); + break; + case RESOURCE_IN_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); + 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: + wimlib_assert(0); + ret = -1; + break; } + return ret; } -/* +/* + * 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) +{ + 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, + 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; + 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; + + 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 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; + 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, 0); + if (ret != 0) + break; + 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, wim_resource_size(lte)); +} + +/* + * Copies the file resource specified by the lookup table entry @lte from the + * 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. + */ +int copy_resource(struct lookup_table_entry *lte, void *wim) +{ + WIMStruct *w = wim; + int ret; + + if ((lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) && + !w->write_metadata) + return 0; + + ret = write_wim_resource(lte, w->out_fp, + wim_resource_compression_type(lte), + <e->output_resource_entry, 0); + if (ret != 0) + return ret; + lte->out_refcnt = lte->refcnt; + lte->part_number = w->hdr.part_number; + return 0; +} + +/* * 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 @@ -515,103 +724,250 @@ int resource_compression_type(int wim_ctype, int reshdr_flags) * length 8, because that's how long the 'length' field is. * * @fp: The FILE* for the input WIM file. - * @res_entry: The resource entry for the metadata resource (a.k.a the metadata - * for the metadata) * @wim_ctype: The compression type of the WIM file. - * @root_dentry_p: A pointer to a pointer to a struct dentry structure into which the - * root dentry is allocated and returned. + * @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: True on success, false on failure. + * @return: Zero on success, nonzero on failure. */ -int read_metadata_resource(FILE *fp, const struct resource_entry *res_entry, - int wim_ctype, struct dentry **root_dentry_p) +int read_metadata_resource(WIMStruct *w, struct image_metadata *imd) { u8 *buf; - int ctype; u32 dentry_offset; int ret; struct dentry *dentry; + struct inode_table inode_tab; + const struct lookup_table_entry *metadata_lte; + u64 metadata_len; + struct hlist_head inode_list; + + metadata_lte = imd->metadata_lte; + metadata_len = wim_resource_size(metadata_lte); + + DEBUG("Reading metadata resource: length = %"PRIu64", " + "offset = %"PRIu64"", metadata_len, + metadata_lte->resource_entry.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; + } - DEBUG("Reading metadata resource: length = %lu, offset = %lu\n", - res_entry->original_size, res_entry->offset); - - if (res_entry->original_size < 8) { - ERROR("Expected at least 8 bytes for the metadata " - "resource!\n"); + 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!\n", - res_entry->original_size); + "metadata resource", metadata_len); return WIMLIB_ERR_NOMEM; } - /* Determine the compression type of the metadata resource. */ - ctype = resource_compression_type(wim_ctype, res_entry->flags); - /* Read the metadata resource into memory. (It may be compressed.) */ - ret = read_full_resource(fp, res_entry->size, - res_entry->original_size, res_entry->offset, - ctype, buf); + ret = read_full_wim_resource(metadata_lte, buf, 0); if (ret != 0) - goto err1; + goto out_free_buf; - DEBUG("Finished reading metadata resource into memory.\n"); + DEBUG("Finished reading metadata resource into memory."); -#if 0 - /* Read the security data into a WIMSecurityData structure. */ - if (!read_security_data(buf, res_entry->original_size, sd)) - goto err1; -#endif + /* 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, 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. */ + + 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; + } + DEBUG("Reading root dentry"); + + /* 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!\n", - sizeof(struct dentry)); + ERROR("Failed to allocate %zu bytes for root dentry", + sizeof(struct dentry)); ret = WIMLIB_ERR_NOMEM; - goto err1; + goto out_free_security_data; } - /* Read the root directory entry starts after security data, on an - * 8-byte aligned address. - * - * The security data starts with a 4-byte integer giving its total - * length. */ - get_u32(buf, &dentry_offset); - dentry_offset += (8 - dentry_offset % 8) % 8; - - ret = read_dentry(buf, res_entry->original_size, dentry_offset, dentry); - if (ret != 0) - goto err1; + ret = read_dentry(buf, metadata_len, dentry_offset, dentry); - /* This is the root dentry, so set its pointers correctly. */ + /* This is the root dentry, so set its parent to itself. */ dentry->parent = dentry; - dentry->next = dentry; - dentry->prev = dentry; - /* Now read the entire directory entry tree. */ - ret = read_dentry_tree(buf, res_entry->original_size, dentry); + if (ret == 0 && dentry->length == 0) { + ERROR("Metadata resource cannot begin with end-of-directory entry!"); + ret = WIMLIB_ERR_INVALID_DENTRY; + } + + if (ret != 0) { + FREE(dentry); + goto out_free_security_data; + } + + inode_add_dentry(dentry, dentry->d_inode); + + /* Now read the entire directory entry tree into memory. */ + DEBUG("Reading dentry tree"); + ret = read_dentry_tree(buf, metadata_len, dentry); if (ret != 0) - goto err2; + goto out_free_dentry_tree; /* 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 err2; + goto out_free_dentry_tree; - *root_dentry_p = dentry; - FREE(buf); - return ret; -err2: - free_dentry_tree(dentry, NULL, false); -err1: + /* Build hash table that maps hard link group IDs to dentry sets */ + DEBUG("Building link group table"); + ret = init_inode_table(&inode_tab, 9001); + if (ret != 0) + goto out_free_dentry_tree; + + 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_dentry_tree; + + 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_dentry_tree; + + DEBUG("Done reading image metadata"); + + imd->root_dentry = dentry; + imd->inode_list = inode_list; + goto out_free_buf; +out_free_dentry_tree: + free_dentry_tree(dentry, NULL); +out_free_security_data: + free_security_data(imd->security_data); + imd->security_data = NULL; +out_free_buf: FREE(buf); return ret; } +/* Write the metadata resource for the current WIM image. */ +int write_metadata_resource(WIMStruct *w) +{ + u8 *buf; + u8 *p; + int ret; + u64 subdir_offset; + struct dentry *root; + struct lookup_table_entry *lte; + u64 metadata_original_size; + const struct wim_security_data *sd; + + DEBUG("Writing metadata resource for image %d (offset = %"PRIu64")", + w->current_image, ftello(w->out_fp)); + + root = wim_root_dentry(w); + sd = wim_security_data(w); + + /* 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 = (((u64)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 " + "metadata resource", metadata_original_size); + return WIMLIB_ERR_NOMEM; + } + /* Write the security data into the resource buffer */ + p = write_security_data(sd, buf); + + /* Write the dentry tree into the resource buffer */ + p = write_dentry_tree(root, p); + + /* 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); + + /* 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, + w->out_fp, + wimlib_get_compression_type(w), + <e->output_resource_entry, + lte->hash); + if (ret != 0) + goto out; + + /* 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); + lte->out_refcnt = 1; + + /* Make sure that the lookup table entry for this metadata resource is + * 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; +}