X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Fresource.c;h=ffcd7530a35f5ab590ad4ac9cc226701d7c389a1;hp=fd75a6294d55095d91f082dce88fc82bfdb864ad;hb=d0e7f039e4ab206b9fd973c983e3fb841fcd2bf2;hpb=f432fb40ba0b161cae850dd4d552754e775418f2 diff --git a/src/resource.c b/src/resource.c index fd75a629..ffcd7530 100644 --- a/src/resource.c +++ b/src/resource.c @@ -5,25 +5,34 @@ */ /* - * 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 + +#ifdef WITH_NTFS_3G +#include +#include +#include +#endif + #include "wimlib_internal.h" #include "lookup_table.h" #include "io.h" @@ -31,26 +40,22 @@ #include "xpress.h" #include "sha1.h" #include "dentry.h" -#include "config.h" #include #include +#ifdef HAVE_ALLOCA_H #include - -#ifdef WITH_NTFS_3G -#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 @@ -60,9 +65,9 @@ * * 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[]) { @@ -93,7 +98,7 @@ 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. */ @@ -141,7 +146,7 @@ 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. */ @@ -155,11 +160,11 @@ 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 " @@ -186,17 +191,17 @@ 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 " @@ -232,7 +237,7 @@ 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 { @@ -240,15 +245,15 @@ 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 @@ -275,7 +280,7 @@ 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, @@ -295,7 +300,7 @@ 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 { @@ -303,7 +308,7 @@ 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; @@ -314,7 +319,7 @@ 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; @@ -345,7 +350,7 @@ err: return WIMLIB_ERR_READ; } -/* +/* * Reads uncompressed data from an open file stream. */ int read_uncompressed_resource(FILE *fp, u64 offset, u64 len, @@ -385,8 +390,21 @@ 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; } @@ -508,7 +526,7 @@ int read_wim_resource(const struct lookup_table_entry *lte, u8 buf[], } } -/* +/* * Reads all the data from the resource corresponding to a WIM lookup table * entry. * @@ -537,7 +555,7 @@ struct chunk_table { u64 offsets[0]; }; -/* +/* * Allocates and initializes a chunk table, and reserves space for it in the * output file. */ @@ -549,11 +567,10 @@ begin_wim_resource_chunk_tab(const struct lookup_table_entry *lte, { 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; + size_t alloc_size = sizeof(struct chunk_table) + num_chunks * sizeof(u64); + struct chunk_table *chunk_tab = CALLOC(1, alloc_size); + int ret; - wimlib_assert(size != 0); if (!chunk_tab) { ERROR("Failed to allocate chunk table for %"PRIu64" byte " @@ -578,12 +595,13 @@ begin_wim_resource_chunk_tab(const struct lookup_table_entry *lte, goto out; } - *chunk_tab_ret = chunk_tab; + ret = 0; out: + *chunk_tab_ret = chunk_tab; return ret; } -/* +/* * Compresses a chunk of a WIM resource. * * @chunk: Uncompressed data of the chunk. @@ -627,7 +645,7 @@ static int compress_chunk(const u8 chunk[], unsigned chunk_size, * @chunk: Uncompressed data of the chunk. * @chunk_size: Size of the chunk (<= WIM_CHUNK_SIZE) * @out_fp: FILE * to write tho chunk to. - * @out_ctype: Compression type to use when writing the chunk (ignored if no + * @out_ctype: Compression type to use when writing the chunk (ignored if no * chunk table provided) * @chunk_tab: Pointer to chunk table being created. It is updated with the * offset of the chunk we write. @@ -661,7 +679,7 @@ static int write_wim_resource_chunk(const u8 chunk[], unsigned chunk_size, *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; @@ -669,7 +687,7 @@ static int write_wim_resource_chunk(const u8 chunk[], unsigned chunk_size, return 0; } -/* +/* * Finishes a WIM chunk tale and writes it to the output file at the correct * offset. * @@ -688,11 +706,11 @@ finish_wim_resource_chunk_tab(struct chunk_table *chunk_tab, } if (chunk_tab->bytes_per_chunk_entry == 8) { - array_to_le64(chunk_tab->offsets, chunk_tab->num_chunks); + array_cpu_to_le64(chunk_tab->offsets, chunk_tab->num_chunks); } else { for (u64 i = 0; i < chunk_tab->num_chunks; i++) ((u32*)chunk_tab->offsets)[i] = - to_le32(chunk_tab->offsets[i]); + cpu_to_le32(chunk_tab->offsets[i]); } bytes_written = fwrite((u8*)chunk_tab->offsets + chunk_tab->bytes_per_chunk_entry, @@ -725,7 +743,7 @@ finish_wim_resource_chunk_tab(struct chunk_table *chunk_tab, * the same as the compression type of the WIM resource we * need to read, we simply copy the data (i.e. we do not * uncompress it, then compress it again). - * @out_res_entry: If non-NULL, a resource entry that is filled in with the + * @out_res_entry: If non-NULL, a resource entry that is filled in with the * offset, original size, compressed size, and compression flag * of the output resource. * @@ -748,6 +766,8 @@ static int write_wim_resource(struct lookup_table_entry *lte, ntfs_inode *ni = NULL; #endif + wimlib_assert(lte); + /* Original size of the resource */ original_size = wim_resource_size(lte); @@ -761,7 +781,7 @@ static int write_wim_resource(struct lookup_table_entry *lte, "stream"); return WIMLIB_ERR_WRITE; } - + /* Are the compression types the same? If so, do a raw copy (copy * without decompressing and recompressing the data). */ raw = (wim_resource_compression_type(lte) == out_ctype @@ -854,18 +874,27 @@ static int write_wim_resource(struct lookup_table_entry *lte, offset += to_read; } while (bytes_remaining); - /* If writing a compressed resource and not doing a raw copy, write the - * chunk table, and finish_wim_resource_chunk_tab() will provide the - * compressed size of the resource we wrote. Otherwise, the compressed - * size of the written resource is the same as the compressed size of - * the existing resource. */ - if (out_ctype != WIM_COMPRESSION_TYPE_NONE && !raw) { - ret = finish_wim_resource_chunk_tab(chunk_tab, out_fp, - &new_compressed_size); - if (ret != 0) - goto out_fclose; - } else { + /* Raw copy: The new compressed size is the same as the old compressed + * size + * + * Using WIM_COMPRESSION_TYPE_NONE: The new compressed size is the + * original size + * + * Using a different compression type: Call + * finish_wim_resource_chunk_tab() and it will provide the new + * compressed size. + */ + if (raw) { new_compressed_size = old_compressed_size; + } else { + if (out_ctype == WIM_COMPRESSION_TYPE_NONE) + new_compressed_size = original_size; + else { + ret = finish_wim_resource_chunk_tab(chunk_tab, out_fp, + &new_compressed_size); + if (ret != 0) + goto out_fclose; + } } /* Verify SHA1 message digest of the resource, unless we are doing a raw @@ -889,8 +918,8 @@ static int write_wim_resource(struct lookup_table_entry *lte, } } - if (new_compressed_size >= original_size && - out_ctype != WIM_COMPRESSION_TYPE_NONE && !raw) + if (!raw && new_compressed_size >= original_size && + out_ctype != WIM_COMPRESSION_TYPE_NONE) { /* Oops! We compressed the resource to larger than the original * size. Write the resource uncompressed instead. */ @@ -915,7 +944,7 @@ static int write_wim_resource(struct lookup_table_entry *lte, } goto out_fclose; } - wimlib_assert(new_compressed_size <= original_size); + wimlib_assert(new_compressed_size <= original_size || raw); if (out_res_entry) { out_res_entry->size = new_compressed_size; out_res_entry->original_size = original_size; @@ -973,10 +1002,10 @@ static int write_wim_resource_from_buffer(const u8 *buf, u64 buf_size, 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, @@ -997,7 +1026,7 @@ int extract_wim_resource_to_fd(const struct lookup_table_entry *lte, int fd, if (ret != 0) break; sha1_update(&ctx, buf, to_read); - if (full_write(fd, buf, to_read) < 0) { + if (full_write(fd, buf, to_read) < to_read) { ERROR_WITH_ERRNO("Error extracting WIM resource"); return WIMLIB_ERR_WRITE; } @@ -1015,9 +1044,9 @@ int extract_wim_resource_to_fd(const struct lookup_table_entry *lte, int fd, 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) @@ -1025,7 +1054,7 @@ 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. @@ -1046,7 +1075,7 @@ int copy_resource(struct lookup_table_entry *lte, void *wim) return 0; ret = write_wim_resource(lte, w->out_fp, - wim_resource_compression_type(lte), + wim_resource_compression_type(lte), <e->output_resource_entry); if (ret != 0) return ret; @@ -1055,14 +1084,14 @@ int copy_resource(struct lookup_table_entry *lte, void *wim) return 0; } -/* +/* * Writes a dentry's resources, including the main file resource as well as all - * alternate data streams, to the output file. + * alternate data streams, to the output file. * * @dentry: The dentry for the file. * @wim_p: A pointer to the WIMStruct containing @dentry. * - * @return zero on success, nonzero on failure. + * @return zero on success, nonzero on failure. */ int write_dentry_resources(struct dentry *dentry, void *wim_p) { @@ -1076,8 +1105,8 @@ int write_dentry_resources(struct dentry *dentry, void *wim_p) printf("Writing streams for `%s'\n", dentry->full_path_utf8); } - for (unsigned i = 0; i <= dentry->num_ads; i++) { - lte = dentry_stream_lte(dentry, i, w->lookup_table); + for (unsigned i = 0; i <= dentry->d_inode->num_ads; i++) { + lte = inode_stream_lte(dentry->d_inode, i, w->lookup_table); if (lte && ++lte->out_refcnt == 1) { ret = write_wim_resource(lte, w->out_fp, ctype, <e->output_resource_entry); @@ -1088,7 +1117,7 @@ int write_dentry_resources(struct dentry *dentry, void *wim_p) 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 @@ -1112,10 +1141,11 @@ int read_metadata_resource(WIMStruct *w, struct image_metadata *imd) u32 dentry_offset; int ret; struct dentry *dentry; - struct link_group_table *lgt; + struct inode_table inode_tab; const struct lookup_table_entry *metadata_lte; u64 metadata_len; u64 metadata_offset; + struct hlist_head inode_list; metadata_lte = imd->metadata_lte; metadata_len = wim_resource_size(metadata_lte); @@ -1178,7 +1208,7 @@ int read_metadata_resource(WIMStruct *w, struct image_metadata *imd) ret = WIMLIB_ERR_NOMEM; goto out_free_security_data; } - + ret = read_dentry(buf, metadata_len, dentry_offset, dentry); /* This is the root dentry, so set its pointers correctly. */ @@ -1187,6 +1217,7 @@ int read_metadata_resource(WIMStruct *w, struct image_metadata *imd) 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"); @@ -1202,30 +1233,29 @@ int read_metadata_resource(WIMStruct *w, struct image_metadata *imd) /* Build hash table that maps hard link group IDs to dentry sets */ DEBUG("Building link group table"); - lgt = new_link_group_table(9001); - if (!lgt) - goto out_free_dentry_tree; - ret = for_dentry_in_tree(dentry, link_group_table_insert, lgt); + ret = init_inode_table(&inode_tab, 9001); if (ret != 0) - goto out_free_lgt; + goto out_free_dentry_tree; + + for_dentry_in_tree(dentry, inode_table_insert, &inode_tab); - DEBUG("Fixing inconsistencies in the link groups"); - ret = fix_link_groups(lgt); + 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("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->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: @@ -1289,7 +1319,7 @@ int write_metadata_resource(WIMStruct *w) DEBUG("Writing dentry tree."); p = write_dentry_tree(root, p); - /* + /* * Append 20 random bytes to the metadata resource so that we don't have * identical metadata resources if we happen to append exactly the same * image twice without any changes in timestamps. If this were to @@ -1317,8 +1347,9 @@ int write_metadata_resource(WIMStruct *w) if (ret != 0) goto out; - /* It's very likely the SHA1 message digest of the metadata resource, so - * re-insert the lookup table entry into the lookup table. */ + /* It's very likely the SHA1 message digest of the metadata resource + * changed, so re-insert the lookup table entry into the lookup table. + * */ lookup_table_unlink(w->lookup_table, lte); lookup_table_insert(w->lookup_table, lte);