X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Fresource.c;h=1871cfcccb8617ba005124652507c555f933f480;hp=4b71cb3f6d60990db4a3e819cf09bd72804aa095;hb=051a59e4c6d114fc7abfa14ff78436adab5defb9;hpb=004c7fd634328ab9b3c4b479097a10ca12ae38d7 diff --git a/src/resource.c b/src/resource.c index 4b71cb3f..1871cfcc 100644 --- a/src/resource.c +++ b/src/resource.c @@ -5,46 +5,58 @@ */ /* - * 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" + +#ifdef WITH_NTFS_3G +#include +#include +#include +#endif + #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 +#ifdef HAVE_ALLOCA_H +#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 @@ -54,9 +66,9 @@ * * 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[]) { @@ -87,7 +99,7 @@ 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. */ @@ -135,7 +147,7 @@ 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. */ @@ -149,11 +161,11 @@ 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 " @@ -180,17 +192,17 @@ 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 " @@ -226,7 +238,7 @@ 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 { @@ -234,15 +246,15 @@ 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 @@ -269,7 +281,7 @@ 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, @@ -289,7 +301,7 @@ 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 { @@ -297,7 +309,7 @@ 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; @@ -308,7 +320,7 @@ 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; @@ -339,11 +351,11 @@ err: 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 " @@ -365,144 +377,7 @@ int read_uncompressed_resource(FILE *fp, u64 offset, u64 len, } -/* - * 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, but its actual size is " - "%"PRIu64" bytes", - resource_original_size, 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; -} - -/* - * 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. - * - * The output_resource_entry, out_refcnt, and part_number fields of @lte are - * updated. - * - * Metadata resources are not copied (they are handled elsewhere for joining and - * splitting). - */ -int copy_resource(struct lookup_table_entry *lte, void *w) -{ - if ((lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) && - !((WIMStruct*)w)->write_metadata) { - return 0; - } - - FILE *in_fp = ((WIMStruct*)w)->fp; - FILE *out_fp = ((WIMStruct*)w)->out_fp; - int ret; - u64 size = lte->resource_entry.size; - u64 offset = lte->resource_entry.offset; - off_t new_offset = ftello(out_fp); - - if (new_offset == -1) - return WIMLIB_ERR_WRITE; - - ret = copy_between_files(in_fp, offset, out_fp, size); - 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 = ((WIMStruct*)w)->hdr.part_number; - return 0; -} /* 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. @@ -516,8 +391,21 @@ 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; } @@ -533,372 +421,772 @@ 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) +static FILE *wim_get_fp(WIMStruct *w) { - if (wim_ctype != WIM_COMPRESSION_TYPE_NONE && - (reshdr_flags & WIM_RESHDR_FLAG_COMPRESSED)) - return wim_ctype; - else - return WIM_COMPRESSION_TYPE_NONE; + 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; +} /* - * Copies bytes between two file streams. + * Reads some data from the resource corresponding to a WIM lookup table entry. * - * Copies @len bytes from @in_fp to @out_fp, at the current position in @out_fp, - * and at an offset of @in_offset in @in_fp. + * @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 copy_between_files(FILE *in_fp, off_t in_offset, FILE *out_fp, size_t len) +int read_wim_resource(const struct lookup_table_entry *lte, u8 buf[], + size_t size, u64 offset, int flags) { - u8 buf[BUFFER_SIZE]; - size_t n; + int ctype; + int ret = 0; + FILE *fp; - if (fseeko(in_fp, in_offset, SEEK_SET) != 0) { - ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" of " - "input file", in_offset); - return WIMLIB_ERR_READ; - } - /* To reduce memory usage and improve speed, read and write BUFFER_SIZE - * bytes at a time. */ - while (len != 0) { - n = min(len, BUFFER_SIZE); - if (fread(buf, 1, n, in_fp) != n) { - if (feof(in_fp)) { - ERROR("Unexpected EOF when copying data " - "between files"); - } else { - ERROR_WITH_ERRNO("Error copying data between " - "files"); - } - return WIMLIB_ERR_READ; + /* 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); + + if (flags & WIMLIB_RESOURCE_FLAG_MULTITHREADED) { + fp = wim_get_fp(lte->wim); + if (!fp) + return WIMLIB_ERR_OPEN; + } else { + wimlib_assert(lte->wim->fp != NULL); + fp = lte->wim->fp; } - if (fwrite(buf, 1, n, out_fp) != n) { - ERROR_WITH_ERRNO("Error copying data between files"); - return WIMLIB_ERR_WRITE; + 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 + ret = read_compressed_resource(fp, + lte->resource_entry.size, + lte->resource_entry.original_size, + lte->resource_entry.offset, + ctype, size, offset, buf); + if (flags & WIMLIB_RESOURCE_FLAG_MULTITHREADED) { + int ret2 = wim_release_fp(lte->wim, fp); + if (ret == 0) + ret = ret2; + } + 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; } - len -= n; +#endif + default: + wimlib_assert(0); + ret = -1; + break; } - return 0; + return ret; } - -/* - * Uncompresses a WIM file resource and writes it uncompressed to a file stream. +/* + * 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. * - * @in_fp: The file stream that contains the file resource. - * @size: The size of the resource in the input file. - * @original_size: The original (uncompressed) size of the resource. - * @offset: The offset of the start of the resource in @in. - * @input_ctype: The compression type of the resource in @in. - * @out_fp: The file stream to write the file resource to. + * Returns 0 on success; nonzero on failure. */ -static int uncompress_resource(FILE *in_fp, u64 size, u64 original_size, - off_t offset, int input_ctype, FILE *out_fp) +int read_full_wim_resource(const struct lookup_table_entry *lte, u8 buf[], + int flags) { - int ret; - u8 buf[WIM_CHUNK_SIZE]; - /* Determine how many compressed chunks the file is divided into. */ - u64 num_chunks; - u64 i; - u64 uncompressed_offset; - u64 uncompressed_chunk_size; - - num_chunks = (original_size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE; - - for (i = 0; i < num_chunks; i++) { + return read_wim_resource(lte, buf, wim_resource_size(lte), 0, flags); +} - uncompressed_offset = i * WIM_CHUNK_SIZE; - uncompressed_chunk_size = min(WIM_CHUNK_SIZE, original_size - - uncompressed_offset); +/* Chunk table that's located at the beginning of each compressed resource in + * the WIM. (This is not the on-disk format; the on-disk format just has an + * array of offsets.) */ +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]; +}; - ret = read_resource(in_fp, size, original_size, offset, - input_ctype, uncompressed_chunk_size, - uncompressed_offset, buf); - if (ret != 0) - return ret; +/* + * Allocates and initializes a chunk table, and reserves space for it in the + * output file. + */ +static int +begin_wim_resource_chunk_tab(const struct lookup_table_entry *lte, + FILE *out_fp, + off_t file_offset, + struct chunk_table **chunk_tab_ret) +{ + u64 size = wim_resource_size(lte); + u64 num_chunks = (size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE; + size_t alloc_size = sizeof(struct chunk_table) + num_chunks * sizeof(u64); + struct chunk_table *chunk_tab = CALLOC(1, alloc_size); + int ret; - if (fwrite(buf, 1, uncompressed_chunk_size, out_fp) != - uncompressed_chunk_size) - { - ERROR_WITH_ERRNO("Failed to write file resource"); - return WIMLIB_ERR_WRITE; - } + if (!chunk_tab) { + ERROR("Failed to allocate chunk table for %"PRIu64" byte " + "resource", size); + ret = WIMLIB_ERR_NOMEM; + goto out; } - return 0; + chunk_tab->file_offset = file_offset; + chunk_tab->num_chunks = num_chunks; + chunk_tab->original_resource_size = size; + chunk_tab->bytes_per_chunk_entry = (size >= (1ULL << 32)) ? 8 : 4; + chunk_tab->table_disk_size = chunk_tab->bytes_per_chunk_entry * + (num_chunks - 1); + chunk_tab->cur_offset = 0; + chunk_tab->cur_offset_p = chunk_tab->offsets; + + 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; + } + + ret = 0; +out: + *chunk_tab_ret = chunk_tab; + return ret; } -/* - * Transfers a file resource between two files, writing it compressed. The file - * resource in the input file may be either compressed or uncompressed. - * Alternatively, the input resource may be in-memory, but it must be - * uncompressed. +/* + * Compresses a chunk of a WIM resource. * - * @in_fp: The file stream that contains the file resource. Ignored - * if uncompressed_resource != NULL. - * @uncompressed_resource: If this pointer is not NULL, it points to an - * array of @original_size bytes that are - * the uncompressed input resource. - * @size: The size of the resource in the input file. - * @original_size: The original (uncompressed) size of the resource. - * @offset: The offset of the start of the resource in @in. Ignored - * if uncompressed_resource != NULL. - * @input_ctype: The compression type of the resource in @in. Ignored if - * uncompressed_resource != NULL. - * @out_fp: The file stream to write the file resource to. - * @output_type: The compression type to use when writing the resource to - * @out. - * @new_size_ret: A location into which the new compressed size of the file - * resource in returned. + * @chunk: Uncompressed data of the chunk. + * @chunk_size: Size of the uncompressed chunk in bytes. + * @compressed_chunk: Pointer to output buffer of size at least + * (@chunk_size - 1) bytes. + * @compressed_chunk_len_ret: Pointer to an unsigned int into which the size + * of the compressed chunk will be + * returned. + * @ctype: Type of compression to use. Must be WIM_COMPRESSION_TYPE_LZX + * or WIM_COMPRESSION_TYPE_XPRESS. + * + * Returns zero if compressed succeeded, and nonzero if the chunk could not be + * compressed to any smaller than @chunk_size. This function cannot fail for + * any other reasons. */ -static int recompress_resource(FILE *in_fp, const u8 *uncompressed_resource, - u64 size, u64 original_size, - off_t offset, int input_ctype, FILE *out_fp, - int output_ctype, u64 *new_size_ret) +static int compress_chunk(const u8 chunk[], unsigned chunk_size, + u8 compressed_chunk[], + unsigned *compressed_chunk_len_ret, + int ctype) { - int ret; - int (*compress)(const void *, uint, void *, uint *); - if (output_ctype == WIM_COMPRESSION_TYPE_LZX) + int (*compress)(const void *, unsigned, void *, unsigned *); + switch (ctype) { + case WIM_COMPRESSION_TYPE_LZX: compress = lzx_compress; - else + break; + case WIM_COMPRESSION_TYPE_XPRESS: compress = xpress_compress; + break; + default: + wimlib_assert(0); + break; + } + return (*compress)(chunk, chunk_size, compressed_chunk, + compressed_chunk_len_ret); +} - u8 uncompressed_buf[WIM_CHUNK_SIZE]; - u8 compressed_buf[WIM_CHUNK_SIZE - 1]; - - /* Determine how many compressed chunks the file needs to be divided - * into. */ - u64 num_chunks = (original_size + WIM_CHUNK_SIZE - 1) / WIM_CHUNK_SIZE; - - u64 num_chunk_entries = num_chunks - 1; +/* + * Writes a chunk of a WIM resource to an output file. + * + * @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 + * chunk table provided) + * @chunk_tab: Pointer to chunk table being created. It is updated with the + * offset of the chunk we write. + * + * Returns 0 on success; nonzero on failure. + */ +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; - /* Size of the chunk entries--- 8 bytes for files over 4GB, otherwise 4 - * bytes */ - uint chunk_entry_size = (original_size >= (u64)1 << 32) ? 8 : 4; + wimlib_assert(chunk_size <= WIM_CHUNK_SIZE); - /* Array in which to construct the chunk offset table. */ - u64 chunk_offsets[num_chunk_entries]; + if (!chunk_tab) { + out_chunk = chunk; + out_chunk_size = chunk_size; + } else { + u8 *compressed_chunk = alloca(chunk_size); + int ret; - /* Offset of the start of the chunk table in the output file. */ - off_t chunk_tab_offset = ftello(out_fp); + ret = compress_chunk(chunk, chunk_size, compressed_chunk, + &out_chunk_size, out_ctype); + if (ret == 0) { + out_chunk = compressed_chunk; + } else { + out_chunk = chunk; + out_chunk_size = chunk_size; + } + *chunk_tab->cur_offset_p++ = chunk_tab->cur_offset; + chunk_tab->cur_offset += out_chunk_size; + } - if (chunk_tab_offset == -1) { - ERROR_WITH_ERRNO("Failed to get offset of output file"); + 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; +} - /* Total size of the chunk table (as written to the file) */ - u64 chunk_tab_size = chunk_entry_size * num_chunk_entries; - - /* Reserve space for the chunk table. */ - if (fwrite(chunk_offsets, 1, chunk_tab_size, out_fp) != - chunk_tab_size) - { - ERROR_WITH_ERRNO("Failed to write chunk offset table"); +/* + * Finishes a WIM chunk tale and writes it to the output file at the correct + * offset. + * + * The final size of the full compressed resource is returned in the + * @compressed_size_p. + */ +static int +finish_wim_resource_chunk_tab(struct chunk_table *chunk_tab, + FILE *out_fp, u64 *compressed_size_p) +{ + size_t bytes_written; + if (fseeko(out_fp, chunk_tab->file_offset, SEEK_SET) != 0) { + ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" of output " + "WIM file", chunk_tab->file_offset); return WIMLIB_ERR_WRITE; } - /* Read each chunk of the file, compress it, write it to the output - * file, and update th chunk offset table. */ - u64 cur_chunk_offset = 0; - for (u64 i = 0; i < num_chunks; i++) { + if (chunk_tab->bytes_per_chunk_entry == 8) { + 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] = + cpu_to_le32(chunk_tab->offsets[i]); + } + bytes_written = fwrite((u8*)chunk_tab->offsets + + chunk_tab->bytes_per_chunk_entry, + 1, chunk_tab->table_disk_size, out_fp); + if (bytes_written != 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, 0, SEEK_END) != 0) { + ERROR_WITH_ERRNO("Failed to seek to end of output WIM file"); + return WIMLIB_ERR_WRITE; + } + *compressed_size_p = chunk_tab->cur_offset + chunk_tab->table_disk_size; + return 0; +} - u64 uncompressed_offset = i * WIM_CHUNK_SIZE; - u64 uncompressed_chunk_size = min(WIM_CHUNK_SIZE, - original_size - uncompressed_offset); +/* + * Writes a WIM resource to a FILE * opened for writing. The resource may be + * written uncompressed or compressed depending on the @out_ctype parameter. + * + * If by chance the resource compresses to more than the original size (this may + * happen with random data or files than are pre-compressed), the resource is + * instead written uncompressed (and this is reflected in the @out_res_entry by + * removing the WIM_RESHDR_FLAG_COMPRESSED flag). + * + * @lte: The lookup table entry for the WIM resource. + * @out_fp: The FILE * to write the resource to. + * @out_ctype: The compression type of the resource to write. Note: if this is + * 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 + * offset, original size, compressed size, and compression flag + * of the output resource. + * + * Returns 0 on success; nonzero on failure. + */ +static int write_wim_resource(struct lookup_table_entry *lte, + FILE *out_fp, int out_ctype, + struct resource_entry *out_res_entry, + int flags) +{ + u64 bytes_remaining; + u64 original_size; + u64 old_compressed_size; + u64 new_compressed_size; + u64 offset; + int ret; + struct chunk_table *chunk_tab = NULL; + bool raw; + off_t file_offset; +#ifdef WITH_NTFS_3G + ntfs_inode *ni = NULL; +#endif + + wimlib_assert(lte); + + /* Original size of the resource */ + original_size = wim_resource_size(lte); + + /* Compressed size of the resource (as it exists now) */ + old_compressed_size = wim_resource_compressed_size(lte); + + /* Current offset in output file */ + file_offset = ftello(out_fp); + if (file_offset == -1) { + ERROR_WITH_ERRNO("Failed to get offset in output " + "stream"); + return WIMLIB_ERR_WRITE; + } - const u8 *uncompressed_p; - if (uncompressed_resource != NULL) { - uncompressed_p = uncompressed_resource + - uncompressed_offset; + /* 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 + && out_ctype != WIM_COMPRESSION_TYPE_NONE); - } else { - /* Read chunk i of the file into uncompressed_buf. */ - ret = read_resource(in_fp, size, original_size, offset, - input_ctype, - uncompressed_chunk_size, - uncompressed_offset, - uncompressed_buf); - if (ret != 0) - return ret; - uncompressed_p = uncompressed_buf; - } + if (raw) { + flags |= WIMLIB_RESOURCE_FLAG_RAW; + bytes_remaining = old_compressed_size; + } else { + flags &= ~WIMLIB_RESOURCE_FLAG_RAW; + bytes_remaining = original_size; + } - if (i != 0) - chunk_offsets[i - 1] = cur_chunk_offset; + /* Empty resource; nothing needs to be done, so just return success. */ + if (bytes_remaining == 0) + return 0; - uint compressed_len; + /* Buffer for reading chunks for the resource */ + u8 buf[min(WIM_CHUNK_SIZE, bytes_remaining)]; - ret = compress(uncompressed_p, uncompressed_chunk_size, - compressed_buf, &compressed_len); + /* If we are writing a compressed resource and not doing a raw copy, we + * need to initialize the chunk table */ + if (out_ctype != WIM_COMPRESSION_TYPE_NONE && !raw) { + ret = begin_wim_resource_chunk_tab(lte, out_fp, file_offset, + &chunk_tab); + if (ret != 0) + goto out; + } - /* if compress() returned nonzero, the compressed chunk would - * have been at least as large as the uncompressed chunk. In - * this situation, the WIM format requires that the uncompressed - * chunk be written instead. */ - const u8 *buf_to_write; - uint len_to_write; - if (ret == 0) { - buf_to_write = compressed_buf; - len_to_write = compressed_len; - } else { - buf_to_write = uncompressed_p; - len_to_write = uncompressed_chunk_size; + /* If the WIM resource is in an external file, open a FILE * to it so we + * don't have to open a temporary one in read_wim_resource() for each + * chunk. */ + if (lte->resource_location == RESOURCE_IN_FILE_ON_DISK + && !lte->file_on_disk_fp) + { + wimlib_assert(lte->file_on_disk); + lte->file_on_disk_fp = fopen(lte->file_on_disk, "rb"); + if (!lte->file_on_disk_fp) { + ERROR_WITH_ERRNO("Failed to open the file `%s' for " + "reading", lte->file_on_disk); + ret = WIMLIB_ERR_OPEN; + goto out; } - - if (fwrite(buf_to_write, 1, len_to_write, out_fp) != - len_to_write) - { - ERROR_WITH_ERRNO("Failed to write compressed " - "file resource"); - return WIMLIB_ERR_WRITE; + } +#ifdef WITH_NTFS_3G + else if (lte->resource_location == RESOURCE_IN_NTFS_VOLUME + && !lte->attr) + { + struct ntfs_location *loc = lte->ntfs_loc; + wimlib_assert(loc); + ni = ntfs_pathname_to_inode(*loc->ntfs_vol_p, NULL, loc->path_utf8); + if (!ni) { + ERROR_WITH_ERRNO("Failed to open inode `%s' in NTFS " + "volume", loc->path_utf8); + ret = WIMLIB_ERR_NTFS_3G; + goto out; + } + lte->attr = ntfs_attr_open(ni, + loc->is_reparse_point ? AT_REPARSE_POINT : AT_DATA, + (ntfschar*)loc->stream_name_utf16, + loc->stream_name_utf16_num_chars); + if (!lte->attr) { + ERROR_WITH_ERRNO("Failed to open attribute of `%s' in " + "NTFS volume", loc->path_utf8); + ret = WIMLIB_ERR_NTFS_3G; + goto out_fclose; } - cur_chunk_offset += len_to_write; } +#endif + + /* If we aren't doing a raw copy, we will compute the SHA1 message + * digest of the resource as we read it, and verify it's the same as the + * hash given in the lookup table entry once we've finished reading the + * resource. */ + SHA_CTX ctx; + if (!raw) + sha1_init(&ctx); + + /* While there are still bytes remaining in the WIM resource, read a + * chunk of the resource, update SHA1, then write that chunk using the + * desired compression type. */ + offset = 0; + do { + u64 to_read = min(bytes_remaining, WIM_CHUNK_SIZE); + ret = read_wim_resource(lte, buf, to_read, offset, flags); + if (ret != 0) + goto out_fclose; + if (!raw) + sha1_update(&ctx, buf, to_read); + ret = write_wim_resource_chunk(buf, to_read, out_fp, + out_ctype, chunk_tab); + if (ret != 0) + goto out_fclose; + bytes_remaining -= to_read; + offset += to_read; + } while (bytes_remaining); - /* The chunk offset after the last chunk, plus the size of the chunk - * table, gives the total compressed size of the resource. */ - *new_size_ret = cur_chunk_offset + chunk_tab_size; - - /* Now that all entries of the chunk table are determined, rewind the - * stream to where the chunk table was, and write it back out. */ + /* 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; + } + } - if (fseeko(out_fp, chunk_tab_offset, SEEK_SET) != 0) { - ERROR_WITH_ERRNO("Failed to seek to beginning of chunk table"); - return WIMLIB_ERR_READ; + /* Verify SHA1 message digest of the resource, unless we are doing a raw + * write (in which case we never even saw the uncompressed data). Or, + * if the hash we had before is all 0's, just re-set it to be the new + * hash. */ + if (!raw) { + u8 md[SHA1_HASH_SIZE]; + sha1_final(md, &ctx); + if (is_zero_hash(lte->hash)) { + copy_hash(lte->hash, md); + } else if (!hashes_equal(md, lte->hash)) { + ERROR("WIM resource has incorrect hash!"); + if (lte->resource_location == RESOURCE_IN_FILE_ON_DISK) { + ERROR("We were reading it from `%s'; maybe it changed " + "while we were reading it.", + lte->file_on_disk); + } + ret = WIMLIB_ERR_INVALID_RESOURCE_HASH; + goto out_fclose; + } } - if (chunk_entry_size == 8) { - array_to_le64(chunk_offsets, num_chunk_entries); + 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. */ + if (fseeko(out_fp, file_offset, SEEK_SET) != 0) { + ERROR_WITH_ERRNO("Failed to seek to byte %"PRIu64" " + "of output WIM file", file_offset); + ret = WIMLIB_ERR_WRITE; + goto out_fclose; + } + ret = write_wim_resource(lte, out_fp, WIM_COMPRESSION_TYPE_NONE, + out_res_entry, flags); + if (ret != 0) + goto out_fclose; + if (fflush(out_fp) != 0) { + ERROR_WITH_ERRNO("Failed to flush output WIM file"); + ret = WIMLIB_ERR_WRITE; + goto out_fclose; + } + if (ftruncate(fileno(out_fp), file_offset + out_res_entry->size) != 0) { + ERROR_WITH_ERRNO("Failed to truncate output WIM file"); + ret = WIMLIB_ERR_WRITE; + goto out_fclose; + } } else { - for (u64 i = 0; i < num_chunk_entries; i++) - ((u32*)chunk_offsets)[i] = to_le32(chunk_offsets[i]); + if (out_res_entry) { + out_res_entry->size = new_compressed_size; + out_res_entry->original_size = original_size; + out_res_entry->offset = file_offset; + out_res_entry->flags = lte->resource_entry.flags + & ~WIM_RESHDR_FLAG_COMPRESSED; + if (out_ctype != WIM_COMPRESSION_TYPE_NONE) + out_res_entry->flags |= WIM_RESHDR_FLAG_COMPRESSED; + } } - if (fwrite(chunk_offsets, 1, chunk_tab_size, out_fp) != chunk_tab_size) - { - ERROR_WITH_ERRNO("Failed to write chunk table"); - return WIMLIB_ERR_WRITE; + ret = 0; +out_fclose: + if (lte->resource_location == RESOURCE_IN_FILE_ON_DISK + && lte->file_on_disk_fp) { + fclose(lte->file_on_disk_fp); + lte->file_on_disk_fp = NULL; } - - if (fseeko(out_fp, 0, SEEK_END) != 0) { - ERROR_WITH_ERRNO("Failed to seek to end of output file"); - return WIMLIB_ERR_WRITE; +#ifdef WITH_NTFS_3G + else if (lte->resource_location == RESOURCE_IN_NTFS_VOLUME) { + if (lte->attr) { + ntfs_attr_close(lte->attr); + lte->attr = NULL; + } + if (ni) + ntfs_inode_close(ni); } +#endif +out: + FREE(chunk_tab); + return ret; +} +/* 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; } -int write_resource_from_memory(const u8 resource[], int out_ctype, - u64 resource_original_size, FILE *out_fp, - u64 *resource_size_ret) +/* + * 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) { - if (out_ctype == WIM_COMPRESSION_TYPE_NONE) { - if (fwrite(resource, 1, resource_original_size, out_fp) != - resource_original_size) - { - ERROR_WITH_ERRNO("Failed to write resource of length " - "%"PRIu64, resource_original_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; } - *resource_size_ret = resource_original_size; - return 0; - } else { - return recompress_resource(NULL, resource, - resource_original_size, - resource_original_size, 0, 0, out_fp, - out_ctype, resource_size_ret); + 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)); +} -/* - * Transfers a file resource from a FILE* opened for reading to a FILE* opened - * for writing, possibly changing the compression type. +/* + * 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. * - * @in_fp: The FILE* that contains the file resource. - * @size: The (compressed) size of the file resource. - * @original_size: The uncompressed size of the file resource. - * @offset: The offset of the file resource in the input file. - * @input_ctype: The compression type of the file resource in the input - * file. - * @out_fp: The FILE* for the output file. The file resource is - * written at the current position of @out. - * @output_ctype: The compression type to which the file resource will be - * converted. - * @output_res_entry: A pointer to a resource entry that, upon successful - * return of this function, will have the size, - * original size, offset, and flags fields filled - * in for the file resource written to the output - * file. + * The output_resource_entry, out_refcnt, and part_number fields of @lte are + * updated. + * + * Metadata resources are not copied (they are handled elsewhere for joining and + * splitting). */ -static int transfer_file_resource(FILE *in_fp, u64 size, u64 original_size, - off_t offset, int input_ctype, FILE *out_fp, - int output_ctype, - struct resource_entry *output_res_entry) +int copy_resource(struct lookup_table_entry *lte, void *wim) { + WIMStruct *w = wim; int ret; - /* Handle zero-length files */ - if (original_size == 0) { - memset(output_res_entry, 0, sizeof(*output_res_entry)); + if ((lte->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) && + !w->write_metadata) return 0; - } - /* Get current offset in the output file. */ - off_t out_offset = ftello(out_fp); - if (out_offset == -1) { - ERROR_WITH_ERRNO("Failed to get output position"); - return WIMLIB_ERR_WRITE; - } - output_res_entry->offset = (u64)out_offset; + 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; +} - if (output_ctype == input_ctype) { - /* The same compression types; simply copy the resource. */ +/* + * 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 containing @dentry. + * + * @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); - ret = copy_between_files(in_fp, offset, out_fp, size); - if (ret != 0) - return ret; - output_res_entry->size = size; - } else { - /* Different compression types. */ + if (w->write_flags & WIMLIB_WRITE_FLAG_VERBOSE) { + wimlib_assert(dentry->full_path_utf8); + printf("Writing streams for `%s'\n", dentry->full_path_utf8); + } - if (output_ctype == WIM_COMPRESSION_TYPE_NONE) { - /* Uncompress a compressed file resource */ - ret = uncompress_resource(in_fp, size, - original_size, offset, - input_ctype, out_fp); + 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, 0); if (ret != 0) - return ret; - output_res_entry->size = original_size; - } else { - u64 new_size; - /* Compress an uncompressed file resource, or compress a - * compressed file resource using a different - * compression type */ - ret = recompress_resource(in_fp, NULL, size, - original_size, - offset, input_ctype, out_fp, - output_ctype, &new_size); - if (ret != 0) - return ret; - output_res_entry->size = new_size; + break; } - } - - output_res_entry->original_size = original_size; - if (output_ctype == WIM_COMPRESSION_TYPE_NONE) - output_res_entry->flags = 0; - else - output_res_entry->flags = WIM_RESHDR_FLAG_COMPRESSED; - return 0; + 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 @@ -909,67 +1197,89 @@ static int transfer_file_resource(FILE *in_fp, u64 size, u64 original_size, * * @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; - const struct resource_entry *res_entry; 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; + u64 metadata_len; + u64 metadata_offset; + struct hlist_head inode_list; - res_entry = &imd->metadata_lte->resource_entry; + metadata_lte = imd->metadata_lte; + 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. */ - 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 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", @@ -978,88 +1288,101 @@ 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. */ - ret = read_dentry_tree(buf, res_entry->original_size, dentry); + /* 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 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 out_free_dentry_tree; - lgt = new_link_group_table(9001); - if (!lgt) + /* 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; - 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; - 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, false); + 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; - struct resource_entry *res_entry; - off_t metadata_offset; u64 metadata_original_size; - u64 metadata_compressed_size; - int metadata_ctype; - u8 hash[WIM_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; + 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; - 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; + /* 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 " @@ -1067,169 +1390,55 @@ 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); - - ret = write_resource_from_memory(buf, - metadata_ctype, - metadata_original_size, - out, - &metadata_compressed_size); - FREE(buf); - if (ret != 0) - return ret; + /* 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); - /* Update the lookup table entry, including the hash and output resource - * entry fields, for this image's metadata resource. */ + /* Get the lookup table entry for the metadata resource so we can update + * it. */ lte = wim_metadata_lookup_table_entry(w); - res_entry = <e->output_resource_entry; - lte->out_refcnt++; - if (memcmp(hash, lte->hash, WIM_HASH_SIZE) != 0) { - lookup_table_unlink(w->lookup_table, lte); - memcpy(lte->hash, hash, WIM_HASH_SIZE); - lookup_table_insert(w->lookup_table, lte); - } - res_entry->original_size = metadata_original_size; - res_entry->offset = metadata_offset; - res_entry->size = metadata_compressed_size; - res_entry->flags = WIM_RESHDR_FLAG_METADATA; - if (metadata_ctype != WIM_COMPRESSION_TYPE_NONE) - res_entry->flags |= WIM_RESHDR_FLAG_COMPRESSED; - return 0; -} -static int write_file_resource(WIMStruct *w, const u8 hash[]) -{ - /* Get the lookup entry for the file resource. */ - struct lookup_table_entry *lte; - - lte = __lookup_resource(w->lookup_table, hash); - if (!lte) - return 0; - - /* No need to write file resources twice. */ - if (++lte->out_refcnt != 1) - return 0; - - /* do not write empty resources */ - if (lte->resource_entry.original_size == 0) - return 0; - - int out_wim_ctype = wimlib_get_compression_type(w); - struct resource_entry *output_res_entry = <e->output_resource_entry; - u64 len; - FILE *in_fp; - FILE *out_fp = w->out_fp; - int ret = 0; - - /* Figure out if we can read the resource from the WIM file, or - * if we have to read it from the filesystem outside, or if it's a - * symbolic link with the data already in memory pointed to by a field - * of the lookup table entry. */ - if (lte->is_symlink) { - off_t offset = ftello(w->out_fp); - u64 new_size; - - if (offset == -1) { - ERROR_WITH_ERRNO("Could not get position in output " - "file"); - return WIMLIB_ERR_WRITE; - } - - wimlib_assert(lte->symlink_buf); + wimlib_assert(lte != NULL); - len = lte->resource_entry.original_size; - - ret = recompress_resource(NULL, lte->symlink_buf, len, len, 0, - 0, out_fp, out_wim_ctype, &new_size); - output_res_entry->size = new_size; - output_res_entry->original_size = len; - output_res_entry->offset = offset; - output_res_entry->flags = (out_wim_ctype == WIM_COMPRESSION_TYPE_NONE) - ? 0 : WIM_RESHDR_FLAG_COMPRESSED; - } else if (lte->file_on_disk) { - - /* Read from disk (uncompressed) */ - - len = lte->resource_entry.original_size; - - in_fp = fopen(lte->file_on_disk, "rb"); - if (!in_fp) { - ERROR_WITH_ERRNO("Failed to open the file `%s'", - lte->file_on_disk); - return WIMLIB_ERR_OPEN; - } - - ret = transfer_file_resource(in_fp, len, len, 0, - WIM_COMPRESSION_TYPE_NONE, out_fp, - out_wim_ctype, output_res_entry); - fclose(in_fp); - } else { - int in_wim_ctype; - - /* Read from input WIM (possibly compressed) */ + /* 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 may be a different WIM file, in the case of - * exporting images from one WIM file to another */ - if (lte->other_wim_fp) { - /* Different WIM file. */ - in_fp = lte->other_wim_fp; - in_wim_ctype = lte->other_wim_ctype; - } else { - /* Same WIM file. */ - in_fp = w->fp; - in_wim_ctype = out_wim_ctype; - } - int input_res_ctype = resource_compression_type( - in_wim_ctype, - lte->resource_entry.flags); - - ret = transfer_file_resource(in_fp, - lte->resource_entry.size, - lte->resource_entry.original_size, - lte->resource_entry.offset, - input_res_ctype, - out_fp, - out_wim_ctype, - output_res_entry); - } + /* 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; } - -/* - * Writes a dentry's resources to the output file. - * - * @dentry: The dentry for the file resource. - * @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, and the alternate data streams. - * - * @return zero on success, nonzero on failure. - */ -int write_dentry_resources(struct dentry *dentry, void *wim_p) -{ - WIMStruct *w = wim_p; - int ret; - - /* Directories don't need file resources. */ - if (dentry_is_directory(dentry)) - return 0; - - ret = write_file_resource(w, dentry->hash); - if (ret != 0) - return ret; - for (u16 i = 0; i < dentry->num_ads; i++) { - ret = write_file_resource(w, dentry->ads_entries[i].hash); - if (ret != 0) - return ret; - } - return 0; -} -