X-Git-Url: https://wimlib.net/git/?p=wimlib;a=blobdiff_plain;f=src%2Fresource.c;h=7a5d1718a631cae9b24c194b531850bacf7de8cb;hp=2dde8d55578435c39b64151746a6c1e1dbe88abd;hb=062465170ddb4520ca6fbc64a0f6e27e786de101;hpb=cfc2cfc859a047e24d002aa149f73d45d4979d24 diff --git a/src/resource.c b/src/resource.c index 2dde8d55..7a5d1718 100644 --- a/src/resource.c +++ b/src/resource.c @@ -1,83 +1,114 @@ /* * resource.c * - * Read uncompressed and compressed metadata and file resources. + * Read uncompressed and compressed metadata and file resources from a WIM file. */ /* - * Copyright (C) 2010 Carl Thijssen - * Copyright (C) 2012 Eric Biggers + * Copyright (C) 2012, 2013 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 "wimlib_internal.h" +#include "dentry.h" #include "lookup_table.h" -#include "io.h" -#include "lzx.h" -#include "xpress.h" +#include "buffer_io.h" #include "sha1.h" -#include "dentry.h" -#include "config.h" -#include + +#ifdef __WIN32__ +# include "win32.h" +#endif + #include -#include +#include +#include +#include +#include +/* Write @n bytes from @buf to the file descriptor @fd, retrying on internupt + * and on short writes. + * + * Returns short count and set errno on failure. */ +static ssize_t +full_write(int fd, const void *buf, size_t n) +{ + const void *p = buf; + ssize_t ret; + ssize_t total = 0; + + while (total != n) { + ret = write(fd, p, n); + if (ret < 0) { + if (errno == EINTR) + continue; + else + break; + } + total += ret; + p += ret; + } + return total; +} -/* - * Reads all or part of a compressed resource into an in-memory buffer. +/* Read @n bytes from the file descriptor @fd to the buffer @buf, retrying on + * internupt and on short reads. * - * @fp: The FILE* for the WIM file. - * @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. - * @len: The number of bytes of uncompressed data to read from - * the resource. - * @offset: The offset of the bytes to read within the uncompressed - * resource. - * @contents_len: An array into which the uncompressed data is written. - * It must be at least @len bytes long. + * Returns short count and set errno on failure. */ +static size_t +full_read(int fd, void *buf, size_t n) +{ + size_t bytes_remaining = n; + while (bytes_remaining) { + ssize_t bytes_read = read(fd, buf, bytes_remaining); + if (bytes_read < 0) { + if (errno == EINTR) + continue; + break; + } + bytes_remaining -= bytes_read; + buf += bytes_read; + } + return n - bytes_remaining; +} + +/* + * Reads all or part of a compressed WIM resource. * * 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, - u64 len, u64 offset, u8 contents_ret[]) +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, + consume_data_callback_t cb, + void *ctx_or_buf) { + int ret; - 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); + int (*decompress)(const void *, unsigned, void *, unsigned); /* Set the appropriate decompress function. */ - if (resource_ctype == WIM_COMPRESSION_TYPE_LZX) - decompress = lzx_decompress; + if (resource_ctype == WIMLIB_COMPRESSION_TYPE_LZX) + decompress = wimlib_lzx_decompress; else - decompress = xpress_decompress; + decompress = wimlib_xpress_decompress; /* The structure of a compressed resource consists of a table of chunk * offsets followed by the chunks themselves. Each chunk consists of @@ -88,7 +119,7 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, * The chunk offsets are measured relative to the end of the chunk * table. The first chunk is omitted from the table in the WIM file * because its offset is implicitly given by the fact that it directly - * follows the chunk table and therefore must have an offset of 0. + * follows the chunk table and therefore must have an offset of 0. */ /* Calculate how many chunks the resource conists of in its entirety. */ @@ -136,7 +167,7 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, /* According to M$'s documentation, if the uncompressed size of * the file is greater than 4 GB, the chunk entries are 8-byte * integers. Otherwise, they are 4-byte integers. */ - u64 chunk_entry_size = (resource_uncompressed_size >= (u64)1 << 32) ? + u64 chunk_entry_size = (resource_uncompressed_size >= (u64)1 << 32) ? 8 : 4; /* Size of the full chunk table in the WIM file. */ @@ -150,59 +181,57 @@ 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 " - "chunk table of compressed resource", - file_offset_of_needed_chunk_entries); - return WIMLIB_ERR_READ; - } + if (fseeko(fp, file_offset_of_needed_chunk_entries, SEEK_SET)) + goto read_error; /* Number of bytes we need to read from the chunk table. */ size_t size = num_needed_chunk_entries * chunk_entry_size; - u8 chunk_tab_buf[size]; + { + u8 chunk_tab_buf[size]; - if (fread(chunk_tab_buf, 1, size, fp) != size) - goto err; + if (fread(chunk_tab_buf, 1, size, fp) != size) + goto read_error; - /* Now fill in chunk_offsets from the entries we have read in - * chunk_tab_buf. */ + /* Now fill in chunk_offsets from the entries we have read in + * chunk_tab_buf. */ - u64 *chunk_tab_p = chunk_offsets; - if (start_chunk == 0) - chunk_tab_p++; + u64 *chunk_tab_p = chunk_offsets; + if (start_chunk == 0) + chunk_tab_p++; - if (chunk_entry_size == 4) { - u32 *entries = (u32*)chunk_tab_buf; - while (num_needed_chunk_entries--) - *chunk_tab_p++ = to_le32(*entries++); - } else { - u64 *entries = (u64*)chunk_tab_buf; - while (num_needed_chunk_entries--) - *chunk_tab_p++ = to_le64(*entries++); + if (chunk_entry_size == 4) { + u32 *entries = (u32*)chunk_tab_buf; + while (num_needed_chunk_entries--) + *chunk_tab_p++ = le32_to_cpu(*entries++); + } else { + u64 *entries = (u64*)chunk_tab_buf; + while (num_needed_chunk_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 " - "first chunk of compressed resource", - file_offset_of_first_needed_chunk); - return WIMLIB_ERR_READ; - } + if (fseeko(fp, file_offset_of_first_needed_chunk, SEEK_SET)) + goto read_error; /* Pointer to current position in the output buffer for uncompressed * data. */ - u8 *out_p = (u8*)contents_ret; + u8 *out_p; + if (cb) + out_p = alloca(32768); + else + out_p = ctx_or_buf; /* Buffer for compressed data. While most compressed chunks will have a * size much less than WIM_CHUNK_SIZE, WIM_CHUNK_SIZE - 1 is the maximum @@ -210,24 +239,21 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, * happen to compress to more than the uncompressed size (i.e. a * sequence of random bytes) are always stored uncompressed. But this seems * to be the case in M$'s WIM files, even though it is undocumented. */ - u8 compressed_buf[WIM_CHUNK_SIZE - 1]; - + void *compressed_buf = alloca(WIM_CHUNK_SIZE - 1); /* Decompress all the chunks. */ for (u64 i = start_chunk; i <= end_chunk; i++) { - DEBUG2("Chunk %"PRIu64" (start %"PRIu64", end %"PRIu64").", - i, start_chunk, end_chunk); - /* Calculate the sizes of the compressed chunk and of the * uncompressed chunk. */ - uint compressed_chunk_size, uncompressed_chunk_size; + unsigned compressed_chunk_size; + unsigned uncompressed_chunk_size; if (i != num_chunks - 1) { /* All the chunks except the last one in the resource * expand to WIM_CHUNK_SIZE uncompressed, and the amount * of compressed data for the chunk is given by the * difference of offsets in the chunk offset table. */ - compressed_chunk_size = chunk_offsets[i + 1 - start_chunk] - + compressed_chunk_size = chunk_offsets[i + 1 - start_chunk] - chunk_offsets[i - start_chunk]; uncompressed_chunk_size = WIM_CHUNK_SIZE; } else { @@ -235,15 +261,15 @@ 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 @@ -252,11 +278,6 @@ 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", - compressed_chunk_size, uncompressed_chunk_size); - - /* Figure out how much of this chunk we actually need to read */ u64 start_offset; if (i == start_chunk) @@ -269,82 +290,87 @@ static int read_compressed_resource(FILE *fp, u64 resource_compressed_size, else end_offset = WIM_CHUNK_SIZE - 1; - u64 partial_chunk_size = end_offset + 1 - start_offset; - bool is_partial_chunk = (partial_chunk_size != - uncompressed_chunk_size); - - DEBUG2("start_offset = %u, end_offset = %u", start_offset, - end_offset); - DEBUG2("partial_chunk_size = %u", partial_chunk_size); + unsigned partial_chunk_size = end_offset + 1 - start_offset; + bool is_partial_chunk = (partial_chunk_size != uncompressed_chunk_size); /* This is undocumented, but chunks can be uncompressed. This * appears to always be the case when the compressed chunk size * is equal to the uncompressed chunk size. */ if (compressed_chunk_size == uncompressed_chunk_size) { - /* Probably an uncompressed chunk */ + /* Uncompressed chunk */ - if (start_offset != 0) { - if (fseeko(fp, start_offset, SEEK_CUR) != 0) { - ERROR_WITH_ERRNO("Uncompressed partial " - "chunk fseek() error"); - return WIMLIB_ERR_READ; - } - } - if (fread(out_p, 1, partial_chunk_size, fp) != - partial_chunk_size) - goto err; + if (start_offset != 0) + if (fseeko(fp, start_offset, SEEK_CUR)) + goto read_error; + if (fread(out_p, 1, partial_chunk_size, fp) != partial_chunk_size) + goto read_error; } else { /* Compressed chunk */ - 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; + goto read_error; - /* For partial chunks we must buffer the uncompressed - * data because we don't need all of it. */ - if (is_partial_chunk) { + /* For partial chunks and when writing directly to a + * buffer, we must buffer the uncompressed data because + * we don't need all of it. */ + if (is_partial_chunk && !cb) { u8 uncompressed_buf[uncompressed_chunk_size]; ret = decompress(compressed_buf, - compressed_chunk_size, - uncompressed_buf, - uncompressed_chunk_size); - if (ret != 0) - return WIMLIB_ERR_DECOMPRESSION; + compressed_chunk_size, + uncompressed_buf, + uncompressed_chunk_size); + if (ret) { + ret = WIMLIB_ERR_DECOMPRESSION; + goto out; + } memcpy(out_p, uncompressed_buf + start_offset, - partial_chunk_size); + partial_chunk_size); } else { ret = decompress(compressed_buf, - compressed_chunk_size, - out_p, - uncompressed_chunk_size); - if (ret != 0) - return WIMLIB_ERR_DECOMPRESSION; + compressed_chunk_size, + out_p, + uncompressed_chunk_size); + if (ret) { + ret = WIMLIB_ERR_DECOMPRESSION; + goto out; + } } } - - /* Advance the pointer into the uncompressed output data by the - * number of uncompressed bytes that were written. */ - out_p += partial_chunk_size; + if (cb) { + /* Feed the data to the callback function */ + ret = cb(out_p, partial_chunk_size, ctx_or_buf); + if (ret) + goto out; + } else { + /* No callback function provided; we are writing + * directly to a buffer. Advance the pointer into this + * buffer by the number of uncompressed bytes that were + * written. */ + out_p += partial_chunk_size; + } } - return 0; + ret = 0; +out: + return ret; -err: +read_error: if (feof(fp)) ERROR("Unexpected EOF in compressed file resource"); else ERROR_WITH_ERRNO("Error reading compressed file resource"); - return WIMLIB_ERR_READ; + ret = WIMLIB_ERR_READ; + goto out; } -/* +/* * 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, void *contents_ret) { if (fseeko(fp, offset, SEEK_SET) != 0) { ERROR("Failed to seek to byte %"PRIu64" of input file " @@ -365,13 +391,11 @@ int read_uncompressed_resource(FILE *fp, u64 offset, u64 len, 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. * A pointer to the byte after the memory read at @p is returned. */ -const u8 *get_resource_entry(const u8 *p, struct resource_entry *entry) +const void * +get_resource_entry(const void *p, struct resource_entry *entry) { u64 size; u8 flags; @@ -380,15 +404,29 @@ 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; } /* Copies the struct resource_entry @entry to the memory pointed to by @p in the * on-disk format. A pointer to the byte after the memory written at @p is * returned. */ -u8 *put_resource_entry(u8 *p, const struct resource_entry *entry) +void * +put_resource_entry(void *p, const struct resource_entry *entry) { p = put_u56(p, entry->size); p = put_u8(p, entry->flags); @@ -397,574 +435,383 @@ u8 *put_resource_entry(u8 *p, const struct resource_entry *entry) return p; } -/* - * 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. - * @raw: If %true, compressed data is read literally rather than being - * decompressed first. - * - * Returns zero on success, nonzero on failure. - */ -int read_wim_resource(const struct lookup_table_entry *lte, u8 buf[], - size_t size, u64 offset, bool raw) +static FILE * +wim_get_fp(WIMStruct *w) { - /* We shouldn't be allowing read over-runs in any part of the library. - * */ - if (raw) - wimlib_assert(offset + size <= lte->resource_entry.size); - else - wimlib_assert(offset + size <= lte->resource_entry.original_size); - - int ctype; - int ret; +#ifdef WITH_FUSE + pthread_mutex_lock(&w->fp_tab_mutex); FILE *fp; - 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); - wimlib_assert(lte->wim->fp); - ctype = wim_resource_compression_type(lte); - - wimlib_assert(ctype != WIM_COMPRESSION_TYPE_NONE || - (lte->resource_entry.original_size == - lte->resource_entry.size)); - - if (raw || ctype == WIM_COMPRESSION_TYPE_NONE) - return read_uncompressed_resource(lte->wim->fp, - lte->resource_entry.offset + offset, - size, buf); - else - return read_compressed_resource(lte->wim->fp, - lte->resource_entry.size, - lte->resource_entry.original_size, - lte->resource_entry.offset, - ctype, size, offset, buf); - break; - case RESOURCE_IN_STAGING_FILE: - /* The WIM FUSE implementation needs to handle multiple open - * file descriptors per lookup table entry so it does not - * currently work with this function. */ - wimlib_assert(lte->staging_file_name); - wimlib_assert(0); - break; - case RESOURCE_IN_FILE_ON_DISK: - /* The resource is in some file on the external filesystem and - * needs to be read uncompressed */ - wimlib_assert(lte->file_on_disk); - /* 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); - } + + 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_unlock; } - ret = read_uncompressed_resource(fp, offset, size, buf); - if (fp != lte->file_on_disk_fp) - fclose(fp); - return ret; - break; - case RESOURCE_IN_ATTACHED_BUFFER: - /* The resource is directly attached uncompressed in an - * in-memory buffer. */ - wimlib_assert(lte->attached_buffer); - memcpy(buf, lte->attached_buffer + offset, size); - return 0; - break; - default: - assert(0); } + DEBUG("Opening extra file descriptor to `%"TS"'", w->filename); + fp = tfopen(w->filename, T("rb")); + if (!fp) + ERROR_WITH_ERRNO("Failed to open `%"TS"'", w->filename); +out_unlock: + pthread_mutex_unlock(&w->fp_tab_mutex); +#else /* WITH_FUSE */ + fp = w->fp; +#endif /* !WITH_FUSE */ + return fp; } -/* - * 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[]) +static int +wim_release_fp(WIMStruct *w, FILE *fp) { - return read_wim_resource(lte, buf, wim_resource_size(lte), 0, false); -} + int ret = 0; +#ifdef WITH_FUSE + FILE **fp_tab; -/* 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]; -}; + 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_unlock; + } + } + + fp_tab = REALLOC(w->fp_tab, sizeof(FILE*) * (w->num_allocated_fps + 4)); + if (!fp_tab) { + ret = WIMLIB_ERR_NOMEM; + fclose(fp); + goto out_unlock; + } + 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_unlock: + pthread_mutex_unlock(&w->fp_tab_mutex); +#endif /* WITH_FUSE */ + 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) +read_partial_wim_resource(const struct wim_lookup_table_entry *lte, + u64 size, + consume_data_callback_t cb, + void *ctx_or_buf, + int flags, + u64 offset) { - 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; + FILE *wim_fp; + WIMStruct *wim; + int ret; - wimlib_assert(size != 0); + wimlib_assert(lte->resource_location == RESOURCE_IN_WIM); + wimlib_assert(offset + size <= lte->resource_entry.original_size); - if (!chunk_tab) { - ERROR("Failed to allocate chunk table for %"PRIu64" byte " - "resource", size); - ret = WIMLIB_ERR_NOMEM; - goto out; - } - 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; + wim = lte->wim; + + if (flags & WIMLIB_RESOURCE_FLAG_MULTITHREADED) { + wim_fp = wim_get_fp(wim); + if (!wim_fp) { + ret = -1; + goto out; + } + } else { + wim_fp = lte->wim->fp; } - *chunk_tab_ret = chunk_tab; + wimlib_assert(wim_fp != NULL); + + if (lte->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED && + !(flags & WIMLIB_RESOURCE_FLAG_RAW)) + { + ret = read_compressed_resource(wim_fp, + lte->resource_entry.size, + lte->resource_entry.original_size, + lte->resource_entry.offset, + wimlib_get_compression_type(wim), + size, + offset, + cb, + ctx_or_buf); + } else { + offset += lte->resource_entry.offset; + + if (fseeko(wim_fp, offset, SEEK_SET)) { + ERROR_WITH_ERRNO("Failed to seek to offset %"PRIu64 + " in WIM", offset); + ret = WIMLIB_ERR_READ; + goto out_release_fp; + } + if (cb) { + /* Send data to callback function */ + u8 buf[min(WIM_CHUNK_SIZE, size)]; + while (size) { + size_t bytes_to_read = min(WIM_CHUNK_SIZE, size); + size_t bytes_read = fread(buf, 1, bytes_to_read, wim_fp); + + if (bytes_read != bytes_to_read) + goto read_error; + ret = cb(buf, bytes_read, ctx_or_buf); + if (ret) + goto out_release_fp; + size -= bytes_read; + } + } else { + /* Send data directly to a buffer */ + if (fread(ctx_or_buf, 1, size, wim_fp) != size) + goto read_error; + } + ret = 0; + } + goto out_release_fp; +read_error: + ERROR_WITH_ERRNO("Error reading data from WIM"); + ret = WIMLIB_ERR_READ; +out_release_fp: + if (flags & WIMLIB_RESOURCE_FLAG_MULTITHREADED) + ret |= wim_release_fp(wim, wim_fp); out: + if (ret) { + if (errno == 0) + errno = EIO; + } return ret; } -/* - * Compresses a chunk of a WIM resource. - * - * @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 compress_chunk(const u8 chunk[], unsigned chunk_size, - u8 compressed_chunk[], - unsigned *compressed_chunk_len_ret, - int ctype) + +int +read_partial_wim_resource_into_buf(const struct wim_lookup_table_entry *lte, + size_t size, u64 offset, void *buf, + bool threadsafe) { - unsigned compressed_chunk_sz; - int (*compress)(const void *, unsigned, void *, unsigned *); - switch (ctype) { - case WIM_COMPRESSION_TYPE_LZX: - compress = lzx_compress; - 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); + return read_partial_wim_resource(lte, size, NULL, buf, + threadsafe ? WIMLIB_RESOURCE_FLAG_MULTITHREADED : 0, + offset); } -/* - * 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) +static int +read_wim_resource_prefix(const struct wim_lookup_table_entry *lte, + u64 size, + consume_data_callback_t cb, + void *ctx_or_buf, + int flags) { - const u8 *out_chunk; - unsigned out_chunk_size; - - wimlib_assert(chunk_size <= WIM_CHUNK_SIZE); + return read_partial_wim_resource(lte, size, cb, ctx_or_buf, flags, 0); +} - if (!chunk_tab) { - out_chunk = chunk; - out_chunk_size = chunk_size; - } else { - u8 *compressed_chunk = alloca(chunk_size); - int ret; - unsigned compressed_chunk_len; - 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; +static int +read_file_on_disk_prefix(const struct wim_lookup_table_entry *lte, + u64 size, + consume_data_callback_t cb, + void *ctx_or_buf, + int _ignored_flags) +{ + const tchar *filename = lte->file_on_disk; + int ret; + int fd; + size_t bytes_read; + + fd = open(filename, O_RDONLY); + if (fd < 0) { + ERROR_WITH_ERRNO("Can't open \"%"TS"\"", filename); + return WIMLIB_ERR_OPEN; + } + if (cb) { + /* Send data to callback function */ + u8 buf[min(WIM_CHUNK_SIZE, size)]; + size_t bytes_to_read; + while (size) { + bytes_to_read = min(WIM_CHUNK_SIZE, size); + bytes_read = full_read(fd, buf, bytes_to_read); + if (bytes_read != bytes_to_read) + goto read_error; + ret = cb(buf, bytes_read, ctx_or_buf); + if (ret) + goto out_close; + size -= bytes_read; } - *chunk_tab->cur_offset_p++ = chunk_tab->cur_offset; - chunk_tab->cur_offset += out_chunk_size; - } - - if (fwrite(out_chunk, 1, out_chunk_size, out_fp) != out_chunk_size) { - ERROR_WITH_ERRNO("Failed to write WIM resource chunk"); - return WIMLIB_ERR_WRITE; - } - return 0; + } else { + /* Send data directly to a buffer */ + bytes_read = full_read(fd, ctx_or_buf, size); + if (bytes_read != size) + goto read_error; + } + ret = 0; + goto out_close; +read_error: + ERROR_WITH_ERRNO("Error reading \"%"TS"\"", filename); + ret = WIMLIB_ERR_READ; +out_close: + close(fd); + return ret; } -/* - * 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) +read_buffer_prefix(const struct wim_lookup_table_entry *lte, + u64 size, consume_data_callback_t cb, + void *ctx_or_buf, int _ignored_flags) { - 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; - } + const void *inbuf = lte->attached_buffer; + int ret; - if (chunk_tab->bytes_per_chunk_entry == 8) { - array_to_le64(chunk_tab->offsets, chunk_tab->num_chunks); + if (cb) { + while (size) { + size_t chunk_size = min(WIM_CHUNK_SIZE, size); + ret = cb(inbuf, chunk_size, ctx_or_buf); + if (ret) + return ret; + size -= chunk_size; + inbuf += chunk_size; + } } else { - for (u64 i = 0; i < chunk_tab->num_chunks; i++) - ((u32*)chunk_tab->offsets)[i] = - 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; + memcpy(ctx_or_buf, inbuf, size); } - 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; } +typedef int (*read_resource_prefix_handler_t)(const struct wim_lookup_table_entry *lte, + u64 size, + consume_data_callback_t cb, + void *ctx_or_buf, + int flags); + /* - * Writes a WIM resource to a FILE * opened for writing. The resource may be - * written uncompressed or compressed depending on the @out_ctype parameter. + * Read the first @size bytes from a generic "resource", which may be located in + * the WIM (compressed or uncompressed), in an external file, or directly in an + * in-memory buffer. * - * 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). + * Feed the data either to a callback function (cb != NULL, passing it + * ctx_or_buf), or write it directly into a buffer (cb == NULL, ctx_or_buf + * specifies the buffer, which must have room for @size bytes). * - * @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. + * When using a callback function, it is called with chunks up to 32768 bytes in + * size until the resource is exhausted. * - * Returns 0 on success; nonzero on failure. + * If the resource is located in a WIM file, @flags can be + * WIMLIB_RESOURCE_FLAG_MULTITHREADED if it must be safe to access the resource + * concurrently by multiple threads, or WIMLIB_RESOURCE_FLAG_RAW if the raw + * compressed data is to be supplied instead of the uncompressed data. */ -static int write_wim_resource(struct lookup_table_entry *lte, - FILE *out_fp, int out_ctype, - struct resource_entry *out_res_entry) +int +read_resource_prefix(const struct wim_lookup_table_entry *lte, + u64 size, consume_data_callback_t cb, void *ctx_or_buf, + int flags) { - u64 bytes_remaining; - u64 original_size; - u64 old_compressed_size; - u64 new_compressed_size; - u64 offset = 0; - int ret = 0; - struct chunk_table *chunk_tab = NULL; - bool raw; - off_t file_offset; - - /* 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; - } - - /* 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); - if (raw) - bytes_remaining = old_compressed_size; - else - bytes_remaining = original_size; - - /* Empty resource; nothing needs to be done, so just return success. */ - if (bytes_remaining == 0) - return 0; + static const read_resource_prefix_handler_t handlers[] = { + [RESOURCE_IN_WIM] = read_wim_resource_prefix, + [RESOURCE_IN_FILE_ON_DISK] = read_file_on_disk_prefix, + [RESOURCE_IN_ATTACHED_BUFFER] = read_buffer_prefix, + #ifdef WITH_FUSE + [RESOURCE_IN_STAGING_FILE] = read_file_on_disk_prefix, + #endif + #ifdef WITH_NTFS_3G + [RESOURCE_IN_NTFS_VOLUME] = read_ntfs_file_prefix, + #endif + #ifdef __WIN32__ + [RESOURCE_WIN32] = read_win32_file_prefix, + [RESOURCE_WIN32_ENCRYPTED] = read_win32_encrypted_file_prefix, + #endif + }; + wimlib_assert(lte->resource_location < ARRAY_LEN(handlers) + && handlers[lte->resource_location] != NULL); + return handlers[lte->resource_location](lte, size, cb, ctx_or_buf, flags); +} - /* Buffer for reading chunks for the resource */ - char buf[min(WIM_CHUNK_SIZE, bytes_remaining)]; +int +read_full_resource_into_buf(const struct wim_lookup_table_entry *lte, + void *buf, bool thread_safe) +{ + return read_resource_prefix(lte, + wim_resource_size(lte), + NULL, buf, + thread_safe ? WIMLIB_RESOURCE_FLAG_MULTITHREADED : 0); +} - /* 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; - } +struct extract_ctx { + SHA_CTX sha_ctx; + consume_data_callback_t extract_chunk; + void *extract_chunk_arg; +}; - /* 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; - } - } +static int +extract_chunk_sha1_wrapper(const void *chunk, size_t chunk_size, + void *_ctx) +{ + struct extract_ctx *ctx = _ctx; - /* 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. */ - do { - u64 to_read = min(bytes_remaining, WIM_CHUNK_SIZE); - ret = read_wim_resource(lte, buf, to_read, offset, raw); - 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); - - /* 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 { - new_compressed_size = old_compressed_size; - } + sha1_update(&ctx->sha_ctx, chunk, chunk_size); + return ctx->extract_chunk(chunk, chunk_size, ctx->extract_chunk_arg); +} - /* 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); +/* Extracts the first @size bytes of a WIM resource to somewhere. In the + * process, the SHA1 message digest of the resource is checked if the full + * resource is being extracted. + * + * @extract_chunk is a function that is called to extract each chunk of the + * resource. */ +int +extract_wim_resource(const struct wim_lookup_table_entry *lte, + u64 size, + consume_data_callback_t extract_chunk, + void *extract_chunk_arg) +{ + int ret; + if (size == wim_resource_size(lte)) { + /* Do SHA1 */ + struct extract_ctx ctx; + ctx.extract_chunk = extract_chunk; + ctx.extract_chunk_arg = extract_chunk_arg; + sha1_init(&ctx.sha_ctx); + ret = read_resource_prefix(lte, size, + extract_chunk_sha1_wrapper, + &ctx, 0); + if (ret == 0) { + u8 hash[SHA1_HASH_SIZE]; + sha1_final(hash, &ctx.sha_ctx); + if (!hashes_equal(hash, lte->hash)) { + #ifdef ENABLE_ERROR_MESSAGES + ERROR_WITH_ERRNO("Invalid SHA1 message digest " + "on the following WIM resource:"); + print_lookup_table_entry(lte, stderr); + if (lte->resource_location == RESOURCE_IN_WIM) + ERROR("The WIM file appears to be corrupt!"); + ret = WIMLIB_ERR_INVALID_RESOURCE_HASH; + #endif } - ret = WIMLIB_ERR_INVALID_RESOURCE_HASH; - goto out_fclose; - } - } - - if (new_compressed_size > original_size) { - /* 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); - 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; - } - wimlib_assert(new_compressed_size <= original_size); - 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; - } -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; + } else { + /* Don't do SHA1 */ + ret = read_resource_prefix(lte, size, extract_chunk, + extract_chunk_arg, 0); } -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 that we 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_hash(lte.hash); - ret = write_wim_resource(<e, out_fp, out_ctype, out_res_entry); - 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) +static int +extract_wim_chunk_to_fd(const void *buf, size_t len, void *_fd_p) { - u64 bytes_remaining = size; - char 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, false); - if (ret != 0) - break; - sha1_update(&ctx, buf, to_read); - if (full_write(fd, buf, to_read) < 0) { - 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; + int fd = *(int*)_fd_p; + ssize_t ret = full_write(fd, buf, len); + if (ret < len) { + ERROR_WITH_ERRNO("Error writing to file descriptor"); + return WIMLIB_ERR_WRITE; + } else { + return 0; } - 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) +int +extract_wim_resource_to_fd(const struct wim_lookup_table_entry *lte, + int fd, u64 size) { - return extract_wim_resource_to_fd(lte, fd, wim_resource_size(lte)); + return extract_wim_resource(lte, size, extract_wim_chunk_to_fd, &fd); } -/* +/* * 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. @@ -972,253 +819,20 @@ int extract_full_wim_resource_to_fd(const struct lookup_table_entry *lte, int fd * 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). + * (This function is confusing and should be refactored somehow.) */ -int copy_resource(struct lookup_table_entry *lte, void *wim) +int +copy_resource(struct wim_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); - if (ret != 0) - return ret; - lte->out_refcnt = lte->refcnt; - lte->part_number = w->hdr.part_number; - return 0; -} - -/* - * Writes a dentry's resources, including the main file resource as well as all - * alternate data streams, to the output file. - * - * @dentry: The dentry for the file. - * @wim_p: A pointer to the WIMStruct 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); - - if (w->write_flags & WIMLIB_WRITE_FLAG_VERBOSE) { - wimlib_assert(dentry->full_path_utf8); - 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); - if (lte && ++lte->out_refcnt == 1) { - ret = write_wim_resource(lte, w->out_fp, ctype, - <e->output_resource_entry); - if (ret != 0) - break; - } - } - return ret; -} - -/* - * Reads the metadata metadata resource from the WIM file. The metadata - * resource consists of the security data, followed by the directory entry for - * the root directory, followed by all the other directory entries in the - * filesystem. The subdir_offset field of each directory entry gives the start - * of its child entries from the beginning of the metadata resource. An - * end-of-directory is signaled by a directory entry of length '0', really of - * length 8, because that's how long the 'length' field is. - * - * @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 `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) -{ - u8 *buf; - int ctype; - u32 dentry_offset; - int ret; - struct dentry *dentry; - struct wim_security_data *sd; - struct link_group_table *lgt; - const struct lookup_table_entry *metadata_lte; - const struct resource_entry *res_entry; - - metadata_lte = imd->metadata_lte; - res_entry = &metadata_lte->resource_entry; - - DEBUG("Reading metadata resource: length = %"PRIu64", " - "offset = %"PRIu64"", - res_entry->original_size, res_entry->offset); - - if (res_entry->original_size < 8) { - ERROR("Expected at least 8 bytes for the metadata resource"); - return WIMLIB_ERR_INVALID_RESOURCE_SIZE; - } - - /* Allocate memory for the uncompressed metadata resource. */ - buf = MALLOC(res_entry->original_size); - - if (!buf) { - ERROR("Failed to allocate %"PRIu64" bytes for uncompressed " - "metadata resource", res_entry->original_size); - return WIMLIB_ERR_NOMEM; - } - - /* Determine the compression type of the metadata resource. */ - - /* Read the metadata resource into memory. (It may be compressed.) */ - ret = read_full_wim_resource(metadata_lte, buf); - 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 security data starts with a 4-byte integer giving its total - * length. */ - - /* Read the security data into a wim_security_data structure. */ - ret = read_security_data(buf, res_entry->original_size, &sd); - if (ret != 0) - goto out_free_buf; - - dentry = MALLOC(sizeof(struct dentry)); - if (!dentry) { - ERROR("Failed to allocate %zu bytes for root dentry", - sizeof(struct dentry)); - ret = WIMLIB_ERR_NOMEM; - goto out_free_security_data; + wim_resource_compression_type(lte), + <e->output_resource_entry, 0); + if (ret == 0) { + lte->out_refcnt = lte->refcnt; + lte->part_number = w->hdr.part_number; } - - 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. */ - dentry->parent = dentry; - dentry->next = dentry; - dentry->prev = dentry; - if (ret != 0) - goto out_free_dentry_tree; - - DEBUG("Reading dentry tree"); - /* Now read the entire directory entry tree. */ - ret = read_dentry_tree(buf, res_entry->original_size, dentry); - if (ret != 0) - goto out_free_dentry_tree; - - DEBUG("Calculating dentry full paths"); - /* Calculate the full paths in the dentry tree. */ - ret = for_dentry_in_tree(dentry, calculate_dentry_full_path, NULL); - if (ret != 0) - goto out_free_dentry_tree; - - DEBUG("Building link group table"); - /* Build hash table that maps hard link group IDs to dentry sets */ - lgt = new_link_group_table(9001); - if (!lgt) - goto out_free_dentry_tree; - ret = for_dentry_in_tree(dentry, link_group_table_insert, lgt); - if (ret != 0) - goto out_free_lgt; - - DEBUG("Freeing duplicate ADS entries in link group table"); - ret = link_groups_free_duplicate_data(lgt); - if (ret != 0) - goto out_free_lgt; - DEBUG("Done reading image metadata"); - - imd->lgt = lgt; - imd->security_data = sd; - imd->root_dentry = dentry; - goto out_free_buf; -out_free_lgt: - free_link_group_table(lgt); -out_free_dentry_tree: - free_dentry_tree(dentry, NULL); -out_free_security_data: - free_security_data(sd); -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, *duplicate_lte; - u64 metadata_original_size; - - /* - * We 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 - * happen, it would cause confusion about the number and order of images - * in the WIM. - */ - const unsigned random_tail_len = 20; - - DEBUG("Writing metadata resource for image %d", w->current_image); - - root = wim_root_dentry(w); - - 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_subdir_offsets(root, &subdir_offset); - metadata_original_size = subdir_offset + random_tail_len; - buf = MALLOC(metadata_original_size); - if (!buf) { - ERROR("Failed to allocate %"PRIu64" bytes for " - "metadata resource", metadata_original_size); - return WIMLIB_ERR_NOMEM; - } - - p = write_security_data(sd, buf); - - DEBUG("Writing dentry tree."); - p = write_dentry_tree(root, p); - randomize_byte_array(p, random_tail_len); - wimlib_assert(p - buf + random_tail_len == metadata_original_size); - - lte = wim_metadata_lookup_table_entry(w); - - ret = write_wim_resource_from_buffer(buf, metadata_original_size, - w->out_fp, - wimlib_get_compression_type(w), - <e->output_resource_entry, - lte->hash); - lookup_table_unlink(w->lookup_table, lte); - lookup_table_insert(w->lookup_table, lte); - wimlib_assert(lte->out_refcnt == 0); - lte->out_refcnt++; - lte->output_resource_entry.flags |= WIM_RESHDR_FLAG_METADATA; -out: - FREE(buf); return ret; }