From: Eric Biggers Date: Fri, 13 Dec 2013 08:15:27 +0000 (-0600) Subject: resource.c: Cleanup and refactor X-Git-Tag: v1.6.0~161 X-Git-Url: https://wimlib.net/git/?p=wimlib;a=commitdiff_plain;h=4f9ccdbed3ee79171d0b861c4ba93b54ce8feaac resource.c: Cleanup and refactor --- diff --git a/include/wimlib/resource.h b/include/wimlib/resource.h index 91b07ce8..a16572b7 100644 --- a/include/wimlib/resource.h +++ b/include/wimlib/resource.h @@ -73,13 +73,15 @@ resource_is_compressed(const struct resource_entry *entry) return (entry->flags & WIM_RESHDR_FLAG_COMPRESSED); } -static inline void copy_resource_entry(struct resource_entry *dst, - const struct resource_entry *src) +static inline void +copy_resource_entry(struct resource_entry *dst, + const struct resource_entry *src) { memcpy(dst, src, sizeof(struct resource_entry)); } -static inline void zero_resource_entry(struct resource_entry *entry) +static inline void +zero_resource_entry(struct resource_entry *entry) { memset(entry, 0, sizeof(struct resource_entry)); } @@ -99,7 +101,6 @@ put_resource_entry(const struct resource_entry *entry, #define WIMLIB_READ_RESOURCE_FLAG_RAW_FULL 0x80000000 #define WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS 0x40000000 -#define WIMLIB_READ_RESOURCE_FLAG_SEEK_ONLY 0x20000000 #define WIMLIB_READ_RESOURCE_FLAG_RAW (WIMLIB_READ_RESOURCE_FLAG_RAW_FULL | \ WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS) #define WIMLIB_READ_RESOURCE_MASK 0xffff0000 diff --git a/include/wimlib/util.h b/include/wimlib/util.h index c61c1f97..cd0d599f 100644 --- a/include/wimlib/util.h +++ b/include/wimlib/util.h @@ -49,6 +49,10 @@ #define BUILD_BUG_ON(condition) ((void)sizeof(char[1 - 2*!!(condition)])) +/* Maximum number of array elements to allocate on the stack (used in various + * places when large temporary buffers are needed). */ +#define STACK_MAX 32768 + #ifdef ENABLE_CUSTOM_MEMORY_ALLOCATOR extern void * wimlib_malloc(size_t) _malloc_attribute; diff --git a/src/extract.c b/src/extract.c index 4d5ba7e0..4448a6ea 100644 --- a/src/extract.c +++ b/src/extract.c @@ -1378,16 +1378,22 @@ read_error: return ret; } +static int +skip_pwm_chunk_cb(const void *chunk, size_t chunk_size, void *_ctx) +{ + return 0; +} + /* Skip over an unneeded stream in a pipable WIM being read from a pipe. */ static int skip_pwm_stream(struct wim_lookup_table_entry *lte) { return read_partial_wim_resource(lte, wim_resource_size(lte), - NULL, + skip_pwm_chunk_cb, wim_resource_chunk_size(lte), NULL, - WIMLIB_READ_RESOURCE_FLAG_SEEK_ONLY, + WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS, 0); } diff --git a/src/resource.c b/src/resource.c index 466de1a5..b38c65d0 100644 --- a/src/resource.c +++ b/src/resource.c @@ -27,7 +27,6 @@ #endif #include "wimlib.h" -#include "wimlib/dentry.h" #include "wimlib/endianness.h" #include "wimlib/error.h" #include "wimlib/file_io.h" @@ -50,21 +49,20 @@ #endif #include #include -#include #include #include /* - * Compressed resources + * Compressed WIM resources * - * A compressed resource in a WIM consists of a number of consecutive LZX or - * XPRESS-compressed chunks, each of which decompresses to 32768 bytes of data, - * except possibly the last, which always decompresses to any remaining bytes. - * In addition, immediately before the chunks, a table (the "chunk table") - * provides the offset, in bytes relative to the end of the chunk table, of the - * start of each compressed chunk, except for the first chunk which is omitted - * as it always has an offset of 0. Therefore, a compressed resource with N - * chunks will have a chunk table with N - 1 entries. + * A compressed resource in a WIM consists of a number of compressed chunks, + * each of which decompresses to a fixed chunk size (given in the WIM header; + * usually 32768) except possibly the last, which always decompresses to any + * remaining bytes. In addition, immediately before the chunks, a table (the + * "chunk table") provides the offset, in bytes relative to the end of the chunk + * table, of the start of each compressed chunk, except for the first chunk + * which is omitted as it always has an offset of 0. Therefore, a compressed + * resource with N chunks will have a chunk table with N - 1 entries. * * Additional information: * @@ -87,13 +85,19 @@ * writing. To make sequential writing possible, the chunk table is placed * after the chunks rather than before the chunks, and to make sequential * reading possible, each chunk is prefixed with a 4-byte header giving its - * compressed size as a 32-bit, unsigned, little-endian integer (less than or - * equal to 32768). Otherwise the details are the same. + * compressed size as a 32-bit, unsigned, little-endian integer. Otherwise the + * details are the same. */ -static int decompress(const void *cchunk, unsigned clen, - void *uchunk, unsigned ulen, - int ctype, u32 wim_chunk_size) + +/* Decompress the specified chunk that uses the specified compression type + * @ctype, part of a WIM with default chunk size @wim_chunk_size. For LZX the + * separate @wim_chunk_size is needed because it determines the window size used + * for LZX compression. */ +static int +decompress(const void *cchunk, unsigned clen, + void *uchunk, unsigned ulen, + int ctype, u32 wim_chunk_size) { switch (ctype) { case WIMLIB_COMPRESSION_TYPE_XPRESS: @@ -113,52 +117,21 @@ static int decompress(const void *cchunk, unsigned clen, } } -/* - * read_compressed_resource()- - * - * Read data from a compressed resource being read from a seekable WIM file. - * The resource may be either pipable or non-pipable. - * - * @flags may be: - * - * 0: - * Just do a normal read, decompressing the data if necessary. - * - * WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS: - * Read the raw contents of the compressed chunks of the compressed - * resource. For pipable resources, this does *not* include the chunk - * headers. If a callback function is being used, it will be called once - * for each compressed chunk. For non-pipable resources, this mode - * excludes the chunk table. For pipable resources, this mode excludes the - * stream and chunk headers. - */ +/* Read data from a compressed WIM resource. Assumes parameters were already + * verified by read_partial_wim_resource(). */ static int -read_compressed_resource(const struct wim_lookup_table_entry * const lte, - u64 size, const consume_data_callback_t cb, - const u32 in_chunk_size, void * const ctx_or_buf, - const int flags, const u64 offset) +read_compressed_wim_resource(const struct wim_lookup_table_entry * const lte, + const u64 size, const consume_data_callback_t cb, + const u32 cb_chunk_size, void * const ctx_or_buf, + const int flags, const u64 offset) { int ret; + int errno_save; const u32 orig_chunk_size = wim_resource_chunk_size(lte); const u32 orig_chunk_order = bsr32(orig_chunk_size); wimlib_assert(is_power_of_2(orig_chunk_size)); - wimlib_assert(cb == NULL || is_power_of_2(in_chunk_size)); - - /* Currently, reading raw compressed chunks is only guaranteed to work - * correctly when the full resource is requested. Furthermore, in such - * cases the requested size is specified as the compressed size, but - * here we change it to an uncompressed size to avoid confusing the rest - * of this function. */ - if (flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS) { - wimlib_assert(offset == 0); - wimlib_assert(size == lte->resource_entry.size); - wimlib_assert(wim_resource_chunk_size(lte) == in_chunk_size); - size = wim_resource_size(lte); - } - - wimlib_assert(offset + size <= wim_resource_size(lte)); /* Handle the trivial case. */ if (size == 0) @@ -172,22 +145,23 @@ read_compressed_resource(const struct wim_lookup_table_entry * const lte, bool out_buf_malloced = false; bool tmp_buf_malloced = false; bool compressed_buf_malloced = false; - const size_t stack_max = 32768; /* Get the file descriptor for the WIM. */ struct filedes * const in_fd = <e->wim->in_fd; + /* Determine if we're reading a pipable resource from a pipe or not. */ + const bool is_pipe_read = !filedes_is_seekable(in_fd); + /* Calculate the number of chunks the resource is divided into. */ const u64 num_chunks = wim_resource_chunks(lte); - /* Calculate the number of entries in the chunk table; it's one less - * than the number of chunks, since the first chunk has no entry. */ - const u64 num_chunk_entries = num_chunks - 1; - /* Calculate the 0-based index of the chunk at which the read starts. */ const u64 start_chunk = offset >> orig_chunk_order; + /* For pipe reads, we always must start from the 0th chunk. */ + const u64 actual_start_chunk = (is_pipe_read ? 0 : start_chunk); + /* Calculate the offset, within the start chunk, of the first byte of * the read. */ const u32 start_offset_in_chunk = offset & (orig_chunk_size - 1); @@ -200,114 +174,114 @@ read_compressed_resource(const struct wim_lookup_table_entry * const lte, * read. */ const u32 end_offset_in_chunk = (offset + size - 1) & (orig_chunk_size - 1); - /* Calculate the number of chunk entries are actually needed to read the - * requested part of the resource. Include an entry for the first chunk - * even though that doesn't exist in the on-disk table, but take into - * account that if the last chunk required for the read is not the last - * chunk of the resource, an extra chunk entry is needed so that the - * compressed size of the last chunk of the read can be determined. */ - const u64 num_alloc_chunk_entries = end_chunk - start_chunk + - 1 + (end_chunk != num_chunks - 1); + /* Calculate the number of entries in the chunk table; it's one less + * than the number of chunks, since the first chunk has no entry. */ + const u64 num_chunk_entries = num_chunks - 1; /* Set the size of each chunk table entry based on the resource's * uncompressed size. */ - const u64 chunk_entry_size = (wim_resource_size(lte) > ((u64)1 << 32)) ? 8 : 4; + const u64 chunk_entry_size = (wim_resource_size(lte) > (1ULL << 32)) ? 8 : 4; /* Calculate the size, in bytes, of the full chunk table. */ const u64 chunk_table_size = num_chunk_entries * chunk_entry_size; - /* Allocate a buffer to hold a subset of the chunk table. It will only - * contain offsets for the chunks that are actually needed for this - * read. For speed, allocate the buffer on the stack unless it's too - * large. */ - if (num_alloc_chunk_entries <= stack_max) { - chunk_offsets = alloca(num_alloc_chunk_entries * sizeof(u64)); - chunk_offsets_malloced = false; - } else { - chunk_offsets = MALLOC(num_alloc_chunk_entries * sizeof(u64)); - if (!chunk_offsets) { - ERROR("Failed to allocate chunk table " - "with %"PRIu64" entries", num_alloc_chunk_entries); - return WIMLIB_ERR_NOMEM; + /* Current offset to read from. */ + u64 cur_read_offset = lte->resource_entry.offset; + if (!is_pipe_read) { + /* Read the chunk table into memory. */ + + /* Calculate the number of chunk entries are actually needed to + * read the requested part of the resource. Include an entry + * for the first chunk even though that doesn't exist in the + * on-disk table, but take into account that if the last chunk + * required for the read is not the last chunk of the resource, + * an extra chunk entry is needed so that the compressed size of + * the last chunk of the read can be determined. */ + const u64 num_alloc_chunk_entries = end_chunk - start_chunk + + 1 + (end_chunk != num_chunks - 1); + + /* Allocate a buffer to hold a subset of the chunk table. It + * will only contain offsets for the chunks that are actually + * needed for this read. For speed, allocate the buffer on the + * stack unless it's too large. */ + if ((size_t)(num_alloc_chunk_entries * sizeof(u64)) != + (num_alloc_chunk_entries * sizeof(u64))) + goto oom; + + if (num_alloc_chunk_entries <= STACK_MAX / sizeof(u64)) { + chunk_offsets = alloca(num_alloc_chunk_entries * sizeof(u64)); + } else { + chunk_offsets = MALLOC(num_alloc_chunk_entries * sizeof(u64)); + if (chunk_offsets == NULL) + goto oom; + chunk_offsets_malloced = true; } - chunk_offsets_malloced = true; - } - /* Set the implicit offset of the first chunk if it's included in the - * needed chunks. */ - if (start_chunk == 0) - chunk_offsets[0] = 0; - - /* Calculate the index of the first needed entry in the chunk table. */ - const u64 start_table_idx = (start_chunk == 0) ? 0 : start_chunk - 1; - - /* Calculate the number of entries that need to be read from the chunk - * table. */ - const u64 num_needed_chunk_entries = (start_chunk == 0) ? - num_alloc_chunk_entries - 1 : num_alloc_chunk_entries; - - /* Calculate the number of bytes of data that need to be read from the - * chunk table. */ - const size_t chunk_table_needed_size = - num_needed_chunk_entries * chunk_entry_size; - if ((u64)chunk_table_needed_size != - num_needed_chunk_entries * chunk_entry_size) - { - ERROR("Compressed read request too large to fit into memory!"); - ret = WIMLIB_ERR_NOMEM; - goto out_free_memory; - } + /* Set the implicit offset of the first chunk if it's included + * in the needed chunks. */ + if (start_chunk == 0) + chunk_offsets[0] = 0; + + /* Calculate the index of the first needed entry in the chunk + * table. */ + const u64 start_table_idx = (start_chunk == 0) ? 0 : start_chunk - 1; + + /* Calculate the number of entries that need to be read from the + * chunk table. */ + const u64 num_needed_chunk_entries = (start_chunk == 0) ? + num_alloc_chunk_entries - 1 : num_alloc_chunk_entries; + + /* Calculate the number of bytes of data that need to be read + * from the chunk table. */ + const size_t chunk_table_needed_size = + num_needed_chunk_entries * chunk_entry_size; + + /* Calculate the byte offset, in the WIM file, of the first + * chunk table entry to read. Take into account that if the WIM + * file is in the special "pipable" format, then the chunk table + * is at the end of the resource, not the beginning. */ + const u64 file_offset_of_needed_chunk_entries = + lte->resource_entry.offset + + (start_table_idx * chunk_entry_size) + + (lte->is_pipable ? (lte->resource_entry.size - chunk_table_size) : 0); + + /* Read the needed chunk table entries into the end of the + * chunk_offsets buffer. */ + void * const chunk_tab_data = (u8*)&chunk_offsets[num_alloc_chunk_entries] - + chunk_table_needed_size; + ret = full_pread(in_fd, chunk_tab_data, chunk_table_needed_size, + file_offset_of_needed_chunk_entries); + if (ret) + goto read_error; - /* Calculate the byte offset, in the WIM file, of the first chunk table - * entry to read. Take into account that if the WIM file is in the - * special "pipable" format, then the chunk table is at the end of the - * resource, not the beginning. */ - const u64 file_offset_of_needed_chunk_entries = - lte->resource_entry.offset - + (start_table_idx * chunk_entry_size) - + (lte->is_pipable ? (lte->resource_entry.size - chunk_table_size) : 0); - - /* Read the needed chunk table entries into the end of the chunk_offsets - * buffer. */ - void * const chunk_tab_data = (u8*)&chunk_offsets[num_alloc_chunk_entries] - - chunk_table_needed_size; - ret = full_pread(in_fd, chunk_tab_data, chunk_table_needed_size, - file_offset_of_needed_chunk_entries); - if (ret) - goto read_error; - - /* Now fill in chunk_offsets from the entries we have read in - * chunk_tab_data. Careful: chunk_offsets aliases chunk_tab_data, which - * breaks C's aliasing rules when we read 32-bit integers and store - * 64-bit integers. But since the operations are safe as long as the - * compiler doesn't mess with their order, we use the gcc may_alias - * extension to tell the compiler that loads from the 32-bit integers - * may alias stores to the 64-bit integers. */ - { - typedef le64 __attribute__((may_alias)) aliased_le64_t; - typedef le32 __attribute__((may_alias)) aliased_le32_t; - u64 * const chunk_offsets_p = chunk_offsets + (start_chunk == 0); - u64 i; - - if (chunk_entry_size == 4) { - aliased_le32_t *raw_entries = (aliased_le32_t*)chunk_tab_data; - for (i = 0; i < num_needed_chunk_entries; i++) - chunk_offsets_p[i] = le32_to_cpu(raw_entries[i]); - } else { - aliased_le64_t *raw_entries = (aliased_le64_t*)chunk_tab_data; - for (i = 0; i < num_needed_chunk_entries; i++) - chunk_offsets_p[i] = le64_to_cpu(raw_entries[i]); + /* Now fill in chunk_offsets from the entries we have read in + * chunk_tab_data. Careful: chunk_offsets aliases + * chunk_tab_data, which breaks C's aliasing rules when we read + * 32-bit integers and store 64-bit integers. But since the + * operations are safe as long as the compiler doesn't mess with + * their order, we use the gcc may_alias extension to tell the + * compiler that loads from the 32-bit integers may alias stores + * to the 64-bit integers. */ + { + typedef le64 __attribute__((may_alias)) aliased_le64_t; + typedef le32 __attribute__((may_alias)) aliased_le32_t; + u64 * const chunk_offsets_p = chunk_offsets + (start_chunk == 0); + u64 i; + + if (chunk_entry_size == 4) { + aliased_le32_t *raw_entries = (aliased_le32_t*)chunk_tab_data; + for (i = 0; i < num_needed_chunk_entries; i++) + chunk_offsets_p[i] = le32_to_cpu(raw_entries[i]); + } else { + aliased_le64_t *raw_entries = (aliased_le64_t*)chunk_tab_data; + for (i = 0; i < num_needed_chunk_entries; i++) + chunk_offsets_p[i] = le64_to_cpu(raw_entries[i]); + } } - } - /* Calculate file offset of the first chunk that needs to be read. - * Note: if the resource is pipable, the entries in the chunk table do - * *not* include the chunk headers. */ - u64 cur_read_offset = lte->resource_entry.offset + chunk_offsets[0]; - if (!lte->is_pipable) - cur_read_offset += chunk_table_size; - else - cur_read_offset += start_chunk * sizeof(struct pwm_chunk_hdr); + /* Set offset to beginning of first chunk to read. */ + cur_read_offset += chunk_table_size + chunk_offsets[0]; + } /* If using a callback function, allocate a temporary buffer that will * be used to pass data to it. If writing directly to a buffer instead, @@ -315,15 +289,13 @@ read_compressed_resource(const struct wim_lookup_table_entry * const lte, size_t out_buf_size; u8 *out_buf_end, *out_p; if (cb) { - out_buf_size = max(in_chunk_size, orig_chunk_size); - if (out_buf_size <= stack_max) { + out_buf_size = max(cb_chunk_size, orig_chunk_size); + if (out_buf_size <= STACK_MAX) { out_buf = alloca(out_buf_size); } else { out_buf = MALLOC(out_buf_size); - if (out_buf == NULL) { - ret = WIMLIB_ERR_NOMEM; - goto out_free_memory; - } + if (out_buf == NULL) + goto oom; out_buf_malloced = true; } } else { @@ -335,36 +307,32 @@ read_compressed_resource(const struct wim_lookup_table_entry * const lte, /* Unless the raw compressed data was requested, allocate a temporary * buffer for reading compressed chunks, each of which can be at most - * orig_chunk_size - 1 bytes. This excludes compressed chunks that are - * a full orig_chunk_size bytes, which are actually stored uncompressed. - */ + * @orig_chunk_size - 1 bytes. This excludes compressed chunks that are + * a full @orig_chunk_size bytes, which are actually stored + * uncompressed. */ if (!(flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS)) { - if (orig_chunk_size - 1 <= stack_max) { + if (orig_chunk_size - 1 <= STACK_MAX) { compressed_buf = alloca(orig_chunk_size - 1); } else { compressed_buf = MALLOC(orig_chunk_size - 1); - if (compressed_buf == NULL) { - ret = WIMLIB_ERR_NOMEM; - goto out_free_memory; - } + if (compressed_buf == NULL) + goto oom; compressed_buf_malloced = true; } } - /* Allocate yet another temporary buffer, this one for reading partial - * chunks. */ + /* Allocate yet another temporary buffer, this one for decompressing + * chunks for which only part of the data is needed. */ if (start_offset_in_chunk != 0 || (end_offset_in_chunk != orig_chunk_size - 1 && offset + size != wim_resource_size(lte))) { - if (orig_chunk_size <= stack_max) { + if (orig_chunk_size <= STACK_MAX) { tmp_buf = alloca(orig_chunk_size); } else { tmp_buf = MALLOC(orig_chunk_size); - if (tmp_buf == NULL) { - ret = WIMLIB_ERR_NOMEM; - goto out_free_memory; - } + if (tmp_buf == NULL) + goto oom; tmp_buf_malloced = true; } } @@ -372,136 +340,159 @@ read_compressed_resource(const struct wim_lookup_table_entry * const lte, /* Read, and possibly decompress, each needed chunk, either writing the * data directly into the @ctx_or_buf buffer or passing it to the @cb * callback function. */ - for (u64 i = start_chunk; i <= end_chunk; i++) { + for (u64 i = actual_start_chunk; i <= end_chunk; i++) { - /* If the resource is pipable, skip the chunk header. */ - if (lte->is_pipable) - cur_read_offset += sizeof(struct pwm_chunk_hdr); - - /* Calculate the sizes of the compressed chunk and of the - * uncompressed chunk. */ - u32 compressed_chunk_size; - u32 uncompressed_chunk_size; - if (i != num_chunks - 1) { - /* Not the last chunk. Compressed size is given by - * difference of chunk table entries; uncompressed size - * is always the WIM chunk size. */ - compressed_chunk_size = chunk_offsets[i + 1 - start_chunk] - - chunk_offsets[i - start_chunk]; - uncompressed_chunk_size = orig_chunk_size; - } else { - /* Last chunk. Compressed size is the remaining size in - * the compressed resource; uncompressed size is the - * remaining size in the uncompressed resource. */ - compressed_chunk_size = lte->resource_entry.size - - chunk_table_size - - chunk_offsets[i - start_chunk]; - if (lte->is_pipable) - compressed_chunk_size -= num_chunks * - sizeof(struct pwm_chunk_hdr); - - if ((wim_resource_size(lte) & (orig_chunk_size - 1)) == 0) - uncompressed_chunk_size = orig_chunk_size; - else - uncompressed_chunk_size = wim_resource_size(lte) & - (orig_chunk_size - 1); - } - - /* Calculate how much of this chunk needs to be read. */ + /* Calculate uncompressed size of next chunk. */ + u32 chunk_usize; + if ((i == num_chunks - 1) && (wim_resource_size(lte) & (orig_chunk_size - 1))) + chunk_usize = (wim_resource_size(lte) & (orig_chunk_size - 1)); + else + chunk_usize = orig_chunk_size; - u32 partial_chunk_size; - u32 start_offset = 0; - u32 end_offset = orig_chunk_size - 1; + /* Calculate compressed size of next chunk. */ + u32 chunk_csize; + if (is_pipe_read) { + struct pwm_chunk_hdr chunk_hdr; - if (flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS) { - partial_chunk_size = compressed_chunk_size; + ret = full_pread(in_fd, &chunk_hdr, + sizeof(chunk_hdr), cur_read_offset); + if (ret) + goto read_error; + chunk_csize = le32_to_cpu(chunk_hdr.compressed_size); } else { - if (i == start_chunk) - start_offset = start_offset_in_chunk; + if (i == num_chunks - 1) { + chunk_csize = lte->resource_entry.size - + chunk_table_size - + chunk_offsets[i - start_chunk]; + if (lte->is_pipable) + chunk_csize -= num_chunks * sizeof(struct pwm_chunk_hdr); + } else { + chunk_csize = chunk_offsets[i + 1 - start_chunk] - + chunk_offsets[i - start_chunk]; + } + } + if (chunk_csize == 0 || chunk_csize > orig_chunk_size) { + ERROR("Invalid chunk size in compressed resource!"); + errno = EINVAL; + ret = WIMLIB_ERR_INVALID_CHUNK_SIZE; + goto out_free_memory; + } + if (lte->is_pipable) + cur_read_offset += sizeof(struct pwm_chunk_hdr); - if (i == end_chunk) - end_offset = end_offset_in_chunk; + if (i >= start_chunk) { + /* Calculate how much of this chunk needs to be read. */ + u32 chunk_needed_size; + u32 start_offset = 0; + u32 end_offset = orig_chunk_size - 1; - partial_chunk_size = end_offset + 1 - start_offset; - } + if (flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS) { + chunk_needed_size = chunk_csize; + } else { + if (i == start_chunk) + start_offset = start_offset_in_chunk; - if (compressed_chunk_size == uncompressed_chunk_size || - (flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS)) - { - /* Chunk stored uncompressed, or reading raw chunk data. */ - ret = full_pread(in_fd, - out_p, - partial_chunk_size, - cur_read_offset + start_offset); - if (ret) - goto read_error; - } else { - /* Compressed chunk and not doing raw read. */ - u8 *target; - - /* Read the compressed data into compressed_buf. */ - ret = full_pread(in_fd, - compressed_buf, - compressed_chunk_size, - cur_read_offset); - if (ret) - goto read_error; + if (i == end_chunk) + end_offset = end_offset_in_chunk; - /* For partial chunks we must buffer the uncompressed - * data because we don't need all of it. */ - if (partial_chunk_size == uncompressed_chunk_size) - target = out_p; - else - target = tmp_buf; - - /* Decompress the chunk. */ - ret = decompress(compressed_buf, - compressed_chunk_size, - target, - uncompressed_chunk_size, - wim_resource_compression_type(lte), - orig_chunk_size); - if (ret) { - ERROR("Failed to decompress data."); - ret = WIMLIB_ERR_DECOMPRESSION; - errno = EINVAL; - goto out_free_memory; + chunk_needed_size = end_offset + 1 - start_offset; } - if (partial_chunk_size != uncompressed_chunk_size) - memcpy(out_p, tmp_buf + start_offset, - partial_chunk_size); - } - out_p += partial_chunk_size; + if (chunk_csize == chunk_usize || + (flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS)) + { + /* Read the raw chunk data. */ - if (cb) { - /* Feed the data to the callback function. */ - wimlib_assert(offset == 0); + ret = full_pread(in_fd, + out_p, + chunk_needed_size, + cur_read_offset + start_offset); + if (ret) + goto read_error; + } else { + /* Read and decompress the chunk. */ - if (flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS) { - ret = cb(out_buf, out_p - out_buf, ctx_or_buf); + u8 *target; + + ret = full_pread(in_fd, + compressed_buf, + chunk_csize, + cur_read_offset); if (ret) + goto read_error; + + if (chunk_needed_size == chunk_usize) + target = out_p; + else + target = tmp_buf; + + ret = decompress(compressed_buf, + chunk_csize, + target, + chunk_usize, + wim_resource_compression_type(lte), + orig_chunk_size); + if (ret) { + ERROR("Failed to decompress data!"); + ret = WIMLIB_ERR_DECOMPRESSION; + errno = EINVAL; goto out_free_memory; - out_p = out_buf; + } + if (chunk_needed_size != chunk_usize) + memcpy(out_p, tmp_buf + start_offset, + chunk_needed_size); + } + + out_p += chunk_needed_size; - } else if (i == end_chunk || out_p == out_buf_end) { - size_t bytes_sent; - const u8 *p; + if (cb) { + /* Feed the data to the callback function. */ - for (p = out_buf; p != out_p; p += bytes_sent) { - bytes_sent = min(in_chunk_size, out_p - p); - ret = cb(p, bytes_sent, ctx_or_buf); + if (flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS) { + ret = cb(out_buf, out_p - out_buf, ctx_or_buf); if (ret) goto out_free_memory; + out_p = out_buf; + } else if (i == end_chunk || out_p == out_buf_end) { + size_t bytes_sent; + const u8 *p; + + for (p = out_buf; p != out_p; p += bytes_sent) { + bytes_sent = min(cb_chunk_size, out_p - p); + ret = cb(p, bytes_sent, ctx_or_buf); + if (ret) + goto out_free_memory; + } + out_p = out_buf; } - out_p = out_buf; } + cur_read_offset += chunk_csize; + } else { + u8 dummy; + + /* Skip data only. */ + cur_read_offset += chunk_csize; + ret = full_pread(in_fd, &dummy, 1, cur_read_offset - 1); + if (ret) + goto read_error; } - cur_read_offset += compressed_chunk_size; } + if (is_pipe_read + && size == lte->resource_entry.original_size + && chunk_table_size) + { + u8 dummy; + /* Skip chunk table at end of pipable resource. */ + + cur_read_offset += chunk_table_size; + ret = full_pread(in_fd, &dummy, 1, cur_read_offset - 1); + if (ret) + goto read_error; + } ret = 0; out_free_memory: + errno_save = errno; if (chunk_offsets_malloced) FREE(chunk_offsets); if (out_buf_malloced) @@ -510,149 +501,73 @@ out_free_memory: FREE(compressed_buf); if (tmp_buf_malloced) FREE(tmp_buf); + errno = errno_save; return ret; -read_error: - ERROR_WITH_ERRNO("Error reading compressed file resource"); +oom: + ERROR("Not enough memory available to read size=%"PRIu64" bytes " + "from compressed resource!", size); + errno = ENOMEM; + ret = WIMLIB_ERR_NOMEM; goto out_free_memory; -} -/* Skip over the chunk table at the end of pipable, compressed resource being - * read from a pipe. */ -static int -skip_chunk_table(const struct wim_lookup_table_entry *lte, - struct filedes *in_fd) -{ - u64 num_chunk_entries = wim_resource_chunks(lte) - 1; - u64 chunk_entry_size = (wim_resource_size(lte) > ((u64)1 << 32)) ? 8 : 4; - u64 chunk_table_size = num_chunk_entries * chunk_entry_size; - int ret; - - if (num_chunk_entries != 0) { - u8 dummy; - ret = full_pread(in_fd, &dummy, 1, - in_fd->offset + chunk_table_size - 1); - if (ret) - return ret; - } - return 0; +read_error: + ERROR_WITH_ERRNO("Error reading compressed file resource!"); + goto out_free_memory; } -/* Read and decompress data from a compressed, pipable resource being read from - * a pipe. */ +/* Read raw data from a file descriptor at the specified offset. */ static int -read_pipable_resource(const struct wim_lookup_table_entry *lte, - u64 size, consume_data_callback_t cb, - u32 in_chunk_size, void *ctx_or_buf, - int flags, u64 offset) +read_raw_file_data(struct filedes *in_fd, + u64 size, + consume_data_callback_t cb, + u32 cb_chunk_size, + void *ctx_or_buf, + u64 offset) { - struct filedes *in_fd; int ret; - const u32 orig_chunk_size = wim_resource_chunk_size(lte); - u8 cchunk[orig_chunk_size - 1]; + u8 *tmp_buf; + bool tmp_buf_malloced = false; - size_t out_buf_size; - u8 *out_buf, *out_buf_end, *out_p; if (cb) { - out_buf_size = max(in_chunk_size, orig_chunk_size); - out_buf = alloca(out_buf_size); - } else { - out_buf_size = size; - out_buf = ctx_or_buf; - } - out_buf_end = out_buf + out_buf_size; - out_p = out_buf; - - /* Get pointers to appropriate decompression function and the input file - * descriptor. */ - in_fd = <e->wim->in_fd; - - /* This function currently assumes the entire resource is being read at - * once and that the raw compressed data isn't being requested. This is - * based on the fact that this function currently only gets called - * during the operation of wimlib_extract_image_from_pipe(). */ - wimlib_assert(!(flags & WIMLIB_READ_RESOURCE_FLAG_RAW)); - wimlib_assert(offset == 0); - wimlib_assert(size == wim_resource_size(lte)); - wimlib_assert(in_fd->offset == lte->resource_entry.offset); - - u32 chunk_usize; - for (offset = 0; offset < size; offset += chunk_usize) { - struct pwm_chunk_hdr chunk_hdr; - u32 chunk_csize; - - /* Calculate uncompressed size of next chunk. */ - chunk_usize = min(orig_chunk_size, size - offset); - - /* Read the compressed size of the next chunk from the chunk - * header. */ - ret = full_read(in_fd, &chunk_hdr, sizeof(chunk_hdr)); - if (ret) - goto read_error; - - chunk_csize = le32_to_cpu(chunk_hdr.compressed_size); - - if (chunk_csize > orig_chunk_size) { - errno = EINVAL; - ret = WIMLIB_ERR_INVALID_PIPABLE_WIM; - goto invalid; - } - - /* Read chunk data. */ - ret = full_read(in_fd, cchunk, chunk_csize); - if (ret) - goto read_error; - - if (flags & WIMLIB_READ_RESOURCE_FLAG_SEEK_ONLY) - continue; - - /* Decompress chunk if needed. Uncompressed size same - * as compressed size means the chunk is uncompressed. - */ - if (chunk_csize == chunk_usize) { - memcpy(out_p, cchunk, chunk_usize); + /* Send data to callback function in chunks. */ + if (cb_chunk_size <= STACK_MAX) { + tmp_buf = alloca(cb_chunk_size); } else { - ret = (*decompress)(cchunk, chunk_csize, - out_p, chunk_usize, - wim_resource_compression_type(lte), - orig_chunk_size); - if (ret) { - errno = EINVAL; - ret = WIMLIB_ERR_DECOMPRESSION; - goto invalid; + tmp_buf = MALLOC(cb_chunk_size); + if (tmp_buf == NULL) { + ret = WIMLIB_ERR_NOMEM; + goto out; } + tmp_buf_malloced = true; } - out_p += chunk_usize; - - /* Feed the uncompressed data into the callback function or copy - * it into the provided buffer. */ - if (cb && (out_p == out_buf_end || - offset + chunk_usize == size)) - { - size_t bytes_sent; - const u8 *p; - for (p = out_buf; p != out_p; p += bytes_sent) { - bytes_sent = min(in_chunk_size, out_p - p); - ret = cb(p, bytes_sent, ctx_or_buf); - if (ret) - return ret; - } - out_p = out_buf; + while (size) { + size_t bytes_to_read = min(cb_chunk_size, size); + ret = full_pread(in_fd, tmp_buf, bytes_to_read, + offset); + if (ret) + goto read_error; + ret = cb(tmp_buf, bytes_to_read, ctx_or_buf); + if (ret) + goto out; + size -= bytes_to_read; + offset += bytes_to_read; } + } else { + /* Read data directly into buffer. */ + ret = full_pread(in_fd, ctx_or_buf, size, offset); + if (ret) + goto read_error; } - - ret = skip_chunk_table(lte, in_fd); - if (ret) - goto read_error; - return 0; + ret = 0; + goto out; read_error: - ERROR_WITH_ERRNO("Error reading compressed file resource"); - return ret; - -invalid: - ERROR("Compressed file resource is invalid"); + ERROR_WITH_ERRNO("Read error"); +out: + if (tmp_buf_malloced) + FREE(tmp_buf); return ret; } @@ -663,46 +578,19 @@ invalid: * file. Data is written into a buffer or fed into a callback function, as * documented in read_resource_prefix(). * - * @flags can be: - * - * 0: - * Just do a normal read, decompressing the data if necessary. @size and - * @offset are interpreted relative to the uncompressed contents of the - * stream. + * By default, this function provides the uncompressed data of the resource, and + * @size and @offset and interpreted relative to the uncompressed contents of + * the resource. The behavior can be modified by any of the following flags: * * WIMLIB_READ_RESOURCE_FLAG_RAW_FULL: - * Only valid when the resource is compressed: Read the raw contents of - * the compressed resource. If the resource is non-pipable, this includes - * the chunk table as well as the compressed chunks. If the resource is - * pipable, this includes the compressed chunks--- including the chunk - * headers--- and the chunk table. The stream header is still *not* - * included. - * - * In this mode, @offset is relative to the beginning of the raw contents - * of the compressed resource--- that is, the chunk table if the resource - * is non-pipable, or the header for the first compressed chunk if the - * resource is pipable. @size is the number of raw bytes to read, which - * must not overrun the end of the resource. For example, if @offset is 0, - * then @size can be at most the raw size of the compressed resource - * (@lte->resource_entry.size). + * Read @size bytes at @offset of the raw contents of the compressed + * resource. In the case of pipable resources, this excludes the stream + * header. Exclusive with WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS. * * WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS: - * Only valid when the resource is compressed and is not being read from a - * pipe: Read the raw contents of the compressed chunks of the compressed - * resource. For pipable resources, this does *not* include the chunk - * headers. If a callback function is being used, it will be called once - * for each compressed chunk. The chunk table is excluded. Also, for - * pipable resources, the stream and chunk headers are excluded. In this - * mode, @size must be exactly the raw size of the compressed resource - * (@lte->resource_entry.size) and @offset must be 0. - * - * WIMLIB_READ_RESOURCE_FLAG_SEEK_ONLY: - * Only valid when the resource is being read from a pipe: Skip over the - * requested data rather than feed it to the callback function or write it - * into the buffer. No decompression is done. - * WIMLIB_READ_RESOURCE_FLAG_RAW_* may not be combined with this flag. - * @offset must be 0 and @size must be the uncompressed size of the - * resource. + * Read the raw compressed chunks of the compressed resource. @size must + * be the full uncompressed size, @offset must be 0, and @cb_chunk_size + * must be the resource chunk size. * * Return values: * WIMLIB_ERR_SUCCESS (0) @@ -710,109 +598,65 @@ invalid: * WIMLIB_ERR_UNEXPECTED_END_OF_FILE (errno set to 0) * WIMLIB_ERR_NOMEM (errno set to ENOMEM) * WIMLIB_ERR_DECOMPRESSION (errno set to EINVAL) - * WIMLIB_ERR_INVALID_PIPABLE_WIM (errno set to EINVAL) + * WIMLIB_ERR_INVALID_CHUNK_SIZE (errno set to EINVAL) * * or other error code returned by the @cb function. */ int read_partial_wim_resource(const struct wim_lookup_table_entry *lte, u64 size, consume_data_callback_t cb, - u32 in_chunk_size, + u32 cb_chunk_size, void *ctx_or_buf, int flags, u64 offset) { struct filedes *in_fd; int ret; - /* Make sure the resource is actually located in a WIM file and is not - * somewhere else. */ + /* Verify parameters. */ wimlib_assert(lte->resource_location == RESOURCE_IN_WIM); - - /* If a callback was specified, in_chunk_size must be a power of 2 (and - * not 0). */ - wimlib_assert(cb == NULL || is_power_of_2(in_chunk_size)); - - /* If a callback was specified, offset must be zero. */ - wimlib_assert(cb == NULL || offset == 0); - - /* Retrieve input file descriptor for the WIM file. */ in_fd = <e->wim->in_fd; - - /* Don't allow raw reads (either full or chunks) of uncompressed - * resources. */ - wimlib_assert(!(flags & WIMLIB_READ_RESOURCE_FLAG_RAW) || - resource_is_compressed(<e->resource_entry)); - - /* Don't allow seek-only reads unless reading from a pipe; also don't - * allow combining SEEK_ONLY with either RAW flag. */ - wimlib_assert(!(flags & WIMLIB_READ_RESOURCE_FLAG_SEEK_ONLY) || - (!filedes_is_seekable(in_fd) && - !(flags & WIMLIB_READ_RESOURCE_FLAG_RAW))); + if (cb) + wimlib_assert(is_power_of_2(cb_chunk_size)); + if (flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS) { + /* Raw chunks mode is subject to the restrictions noted. */ + wimlib_assert(!(flags & WIMLIB_READ_RESOURCE_FLAG_RAW_FULL)); + wimlib_assert(cb_chunk_size == wim_resource_chunk_size(lte)); + wimlib_assert(size == lte->resource_entry.original_size); + wimlib_assert(offset == 0); + } else if (flags & WIMLIB_READ_RESOURCE_FLAG_RAW_FULL) { + /* Raw full mode: read must not overrun end of store size. */ + wimlib_assert(offset + size >= size && + offset + size <= lte->resource_entry.size); + } else { + /* Normal mode: read must not overrun end of original size. */ + wimlib_assert(offset + size >= size && + offset + size <= lte->resource_entry.original_size); + } DEBUG("Reading WIM resource: %"PRIu64" @ +%"PRIu64" " - "from %"PRIu64" @ +%"PRIu64" (readflags 0x%08x, resflags 0x%02x%s)", + "from %"PRIu64"(%"PRIu64") @ +%"PRIu64" " + "(readflags 0x%08x, resflags 0x%02x%s)", size, offset, - lte->resource_entry.original_size, lte->resource_entry.offset, + lte->resource_entry.size, + lte->resource_entry.original_size, + lte->resource_entry.offset, flags, lte->resource_entry.flags, (lte->is_pipable ? ", pipable" : "")); if ((flags & WIMLIB_READ_RESOURCE_FLAG_RAW_FULL) || - !resource_is_compressed(<e->resource_entry)) - { - /* Reading raw resource contents or reading uncompressed - * resource. */ - wimlib_assert(offset + size <= lte->resource_entry.size); - offset += lte->resource_entry.offset; - if (flags & WIMLIB_READ_RESOURCE_FLAG_SEEK_ONLY) { - if (lte->resource_entry.size != 0) { - u8 dummy; - ret = full_pread(in_fd, &dummy, 1, - offset + lte->resource_entry.size - 1); - if (ret) - goto read_error; - } - } else if (cb) { - /* Send data to callback function */ - u8 buf[min(in_chunk_size, size)]; - while (size) { - size_t bytes_to_read = min(in_chunk_size, size); - ret = full_pread(in_fd, buf, bytes_to_read, - offset); - if (ret) - goto read_error; - ret = cb(buf, bytes_to_read, ctx_or_buf); - if (ret) - goto out; - size -= bytes_to_read; - offset += bytes_to_read; - } - } else { - /* Send data directly to a buffer */ - ret = full_pread(in_fd, ctx_or_buf, size, offset); - if (ret) - goto read_error; - } - ret = 0; - } else if (lte->is_pipable && !filedes_is_seekable(in_fd)) { - /* Reading compressed, pipable resource from pipe. */ - ret = read_pipable_resource(lte, size, cb, - in_chunk_size, - ctx_or_buf, flags, offset); + !resource_is_compressed(<e->resource_entry)) { + return read_raw_file_data(in_fd, + size, + cb, + cb_chunk_size, + ctx_or_buf, + offset + lte->resource_entry.offset); } else { - /* Reading compressed, possibly pipable resource from seekable - * file. */ - ret = read_compressed_resource(lte, size, cb, - in_chunk_size, - ctx_or_buf, flags, offset); + return read_compressed_wim_resource(lte, size, cb, + cb_chunk_size, + ctx_or_buf, flags, offset); } - goto out; - -read_error: - ERROR_WITH_ERRNO("Error reading data from WIM"); -out: - return ret; } - int read_partial_wim_resource_into_buf(const struct wim_lookup_table_entry *lte, size_t size, u64 offset, void *buf) @@ -824,104 +668,76 @@ static int read_wim_resource_prefix(const struct wim_lookup_table_entry *lte, u64 size, consume_data_callback_t cb, - u32 in_chunk_size, + u32 cb_chunk_size, void *ctx_or_buf, int flags) { - return read_partial_wim_resource(lte, size, cb, in_chunk_size, + return read_partial_wim_resource(lte, size, cb, cb_chunk_size, ctx_or_buf, flags, 0); } - #ifndef __WIN32__ +/* This function handles reading resource data that is located in an external + * file, such as a file that has been added to the WIM image through execution + * of a wimlib_add_command. + * + * This assumes the file can be accessed using the standard POSIX open(), + * read(), and close(). On Windows this will not necessarily be the case (since + * the file may need FILE_FLAG_BACKUP_SEMANTICS to be opened, or the file may be + * encrypted), so Windows uses its own code for its equivalent case. + */ static int read_file_on_disk_prefix(const struct wim_lookup_table_entry *lte, u64 size, consume_data_callback_t cb, - u32 in_chunk_size, + u32 cb_chunk_size, void *ctx_or_buf, int _ignored_flags) { - const tchar *filename = lte->file_on_disk; int ret; - struct filedes fd; int raw_fd; - u8 *out_buf; - bool out_buf_malloced; - const size_t stack_max = 32768; + struct filedes fd; - DEBUG("Reading %"PRIu64" bytes from \"%"TS"\"", - size, lte->file_on_disk); + wimlib_assert(size <= wim_resource_size(lte)); + DEBUG("Reading %"PRIu64" bytes from \"%"TS"\"", size, lte->file_on_disk); - raw_fd = open(filename, O_RDONLY); + raw_fd = open(lte->file_on_disk, O_BINARY | O_RDONLY); if (raw_fd < 0) { - ERROR_WITH_ERRNO("Can't open \"%"TS"\"", filename); + ERROR_WITH_ERRNO("Can't open \"%"TS"\"", lte->file_on_disk); return WIMLIB_ERR_OPEN; } filedes_init(&fd, raw_fd); - out_buf_malloced = false; - if (cb) { - /* Send data to callback function */ - if (in_chunk_size <= stack_max) { - out_buf = alloca(in_chunk_size); - } else { - out_buf = MALLOC(in_chunk_size); - if (out_buf == NULL) { - ret = WIMLIB_ERR_NOMEM; - goto out_close; - } - out_buf_malloced = true; - } - - size_t bytes_to_read; - while (size) { - bytes_to_read = min(in_chunk_size, size); - ret = full_read(&fd, out_buf, bytes_to_read); - if (ret) - goto read_error; - ret = cb(out_buf, bytes_to_read, ctx_or_buf); - if (ret) - goto out_close; - size -= bytes_to_read; - } - } else { - /* Send data directly to a buffer */ - ret = full_read(&fd, ctx_or_buf, size); - if (ret) - goto read_error; - } - ret = 0; - goto out_close; - -read_error: - ERROR_WITH_ERRNO("Error reading \"%"TS"\"", filename); -out_close: + ret = read_raw_file_data(&fd, size, cb, cb_chunk_size, ctx_or_buf, 0); filedes_close(&fd); - if (out_buf_malloced) - FREE(out_buf); return ret; } #endif /* !__WIN32__ */ +/* This function handles the trivial case of reading resource data that is, in + * fact, already located in an in-memory buffer. */ static int read_buffer_prefix(const struct wim_lookup_table_entry *lte, u64 size, consume_data_callback_t cb, - u32 in_chunk_size, + u32 cb_chunk_size, void *ctx_or_buf, int _ignored_flags) { + wimlib_assert(size <= wim_resource_size(lte)); if (cb) { + /* Feed the data into the callback function in + * appropriately-sized chunks. */ int ret; u32 chunk_size; for (u64 offset = 0; offset < size; offset += chunk_size) { - chunk_size = min(in_chunk_size, size - offset); + chunk_size = min(cb_chunk_size, size - offset); ret = cb((const u8*)lte->attached_buffer + offset, chunk_size, ctx_or_buf); if (ret) return ret; } } else { + /* Copy the data directly into the specified buffer. */ memcpy(ctx_or_buf, lte->attached_buffer, size); } return 0; @@ -930,32 +746,44 @@ read_buffer_prefix(const struct wim_lookup_table_entry *lte, typedef int (*read_resource_prefix_handler_t)(const struct wim_lookup_table_entry *lte, u64 size, consume_data_callback_t cb, - u32 in_chunk_size, + u32 cb_chunk_size, void *ctx_or_buf, int flags); /* * read_resource_prefix()- * - * 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. + * Reads the first @size bytes from a generic "resource", which may be located + * in any one of several locations, such as in a WIM file (compressed or + * uncompressed), in an external file, or directly in an in-memory buffer. * - * 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). + * This function feeds 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 at least @size + * bytes). * - * When using a callback function, it is called with chunks up to 32768 bytes in - * size until the resource is exhausted. + * When (@cb != NULL), @cb_chunk_size specifies the maximum size of data chunks + * to feed the callback function. @cb_chunk_size must be positive, and if the + * resource is in a WIM file, must be a power of 2. All chunks, except possibly + * the last one, will be this size. If (@cb == NULL), @cb_chunk_size is + * ignored. * * If the resource is located in a WIM file, @flags can be set as documented in * read_partial_wim_resource(). Otherwise @flags are ignored. + * + * Returns 0 on success; nonzero on error. A nonzero value will be returned if + * the resource data cannot be successfully read (for a number of different + * reasons, depending on the resource location), or if a callback function was + * specified and it returned nonzero. */ int read_resource_prefix(const struct wim_lookup_table_entry *lte, - u64 size, consume_data_callback_t cb, u32 in_chunk_size, + u64 size, consume_data_callback_t cb, u32 cb_chunk_size, void *ctx_or_buf, int flags) { + /* This function merely verifies several preconditions, then passes + * control to an appropriate function for understanding each possible + * resource location. */ static const read_resource_prefix_handler_t handlers[] = { [RESOURCE_IN_WIM] = read_wim_resource_prefix, #ifdef __WIN32__ @@ -976,17 +804,24 @@ read_resource_prefix(const struct wim_lookup_table_entry *lte, }; wimlib_assert(lte->resource_location < ARRAY_LEN(handlers) && handlers[lte->resource_location] != NULL); - wimlib_assert(cb == NULL || in_chunk_size > 0); - return handlers[lte->resource_location](lte, size, cb, in_chunk_size, ctx_or_buf, flags); + wimlib_assert(cb == NULL || cb_chunk_size > 0); + return handlers[lte->resource_location](lte, size, cb, cb_chunk_size, + ctx_or_buf, flags); } +/* Read the full uncompressed data of the specified resource into the specified + * buffer, which must have space for at least lte->resource_entry.original_size + * bytes. */ int read_full_resource_into_buf(const struct wim_lookup_table_entry *lte, void *buf) { - return read_resource_prefix(lte, wim_resource_size(lte), NULL, 0, buf, 0); + return read_resource_prefix(lte, wim_resource_size(lte), + NULL, 0, buf, 0); } +/* Read the full uncompressed data of the specified resource. A buffer + * sufficient to hold the data is allocated and returned in @buf_ret. */ int read_full_resource_into_alloc_buf(const struct wim_lookup_table_entry *lte, void **buf_ret) @@ -1003,7 +838,7 @@ read_full_resource_into_alloc_buf(const struct wim_lookup_table_entry *lte, } buf = MALLOC(lte->resource_entry.original_size); - if (!buf) + if (buf == NULL) return WIMLIB_ERR_NOMEM; ret = read_full_resource_into_buf(lte, buf); @@ -1016,6 +851,8 @@ read_full_resource_into_alloc_buf(const struct wim_lookup_table_entry *lte, return 0; } +/* Retrieve the full uncompressed data of the specified WIM resource, provided + * as a raw `struct resource_entry'. */ int res_entry_to_data(const struct resource_entry *res_entry, WIMStruct *wim, void **buf_ret) @@ -1024,7 +861,7 @@ res_entry_to_data(const struct resource_entry *res_entry, struct wim_lookup_table_entry *lte; lte = new_lookup_table_entry(); - if (!lte) + if (lte == NULL) return WIMLIB_ERR_NOMEM; copy_resource_entry(<e->resource_entry, res_entry); @@ -1053,12 +890,12 @@ extract_chunk_sha1_wrapper(const void *chunk, size_t chunk_size, return ctx->extract_chunk(chunk, chunk_size, ctx->extract_chunk_arg); } -/* 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 +/* Extracts the first @size bytes of a resource to somewhere. In the process, + * the SHA1 message digest of the uncompressed 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. */ + * @extract_chunk is a function that will be called to extract each chunk of the + * resource. */ int extract_wim_resource(const struct wim_lookup_table_entry *lte, u64 size, @@ -1109,6 +946,9 @@ extract_wim_chunk_to_fd(const void *buf, size_t len, void *_fd_p) return ret; } +/* Extract the first @size bytes of the specified resource to the specified file + * descriptor. If @size is the full size of the resource, its SHA1 message + * digest is also checked. */ int extract_wim_resource_to_fd(const struct wim_lookup_table_entry *lte, struct filedes *fd, u64 size) @@ -1124,7 +964,7 @@ sha1_chunk(const void *buf, size_t len, void *ctx) return 0; } -/* Calculate the SHA1 message digest of a stream. */ +/* Calculate the SHA1 message digest of a resource, storing it in @lte->hash. */ int sha1_resource(struct wim_lookup_table_entry *lte) { @@ -1137,18 +977,19 @@ sha1_resource(struct wim_lookup_table_entry *lte) &sha_ctx, 0); if (ret == 0) sha1_final(lte->hash, &sha_ctx); + return ret; } -/* Translates a WIM resource entry from the on-disk format to an in-memory - * format. */ +/* Translates a WIM resource entry from the on-disk format into an in-memory + * format. */ void get_resource_entry(const struct resource_entry_disk *disk_entry, struct resource_entry *entry) { /* Note: disk_entry may not be 8 byte aligned--- in that case, the - * offset and original_size members will be unaligned. (This should be - * okay since `struct resource_entry_disk' is declared as packed.) */ + * offset and original_size members will be unaligned. (This is okay + * since `struct resource_entry_disk' is declared as packed.) */ /* Read the size and flags into a bitfield portably... */ entry->size = (((u64)disk_entry->size[0] << 0) | @@ -1183,8 +1024,8 @@ put_resource_entry(const struct resource_entry *entry, struct resource_entry_disk *disk_entry) { /* Note: disk_entry may not be 8 byte aligned--- in that case, the - * offset and original_size members will be unaligned. (This should be - * okay since `struct resource_entry_disk' is declared as packed.) */ + * offset and original_size members will be unaligned. (This is okay + * since `struct resource_entry_disk' is declared as packed.) */ u64 size = entry->size; disk_entry->size[0] = size >> 0; diff --git a/src/win32_replacements.c b/src/win32_replacements.c index d3757a3f..41706382 100644 --- a/src/win32_replacements.c +++ b/src/win32_replacements.c @@ -88,7 +88,7 @@ realpath(const wchar_t *path, wchar_t *resolved_path) ret = GetFullPathNameW(path, ret, resolved_path, NULL); if (!ret) { err = GetLastError(); - free(resolved_path); + FREE(resolved_path); resolved_path = NULL; goto fail_win32; } @@ -455,8 +455,8 @@ globfree(glob_t *pglob) { size_t i; for (i = 0; i < pglob->gl_pathc; i++) - free(pglob->gl_pathv[i]); - free(pglob->gl_pathv); + FREE(pglob->gl_pathv[i]); + FREE(pglob->gl_pathv); } #endif /* __WIN32__ */ diff --git a/src/write.c b/src/write.c index 084edc62..de59f951 100644 --- a/src/write.c +++ b/src/write.c @@ -452,10 +452,13 @@ write_wim_resource(struct wim_lookup_table_entry *lte, */ if (lte->is_pipable == !!(resource_flags & WIMLIB_WRITE_RESOURCE_FLAG_PIPABLE)) + { resource_flags |= WIMLIB_READ_RESOURCE_FLAG_RAW_FULL; - else + read_size = lte->resource_entry.size; + } else { resource_flags |= WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS; - read_size = lte->resource_entry.size; + read_size = lte->resource_entry.original_size; + } write_ctx.doing_sha = false; } else { write_ctx.doing_sha = true;