Use read_stream_list() for extraction
authorEric Biggers <ebiggers3@gmail.com>
Fri, 20 Dec 2013 20:21:03 +0000 (14:21 -0600)
committerEric Biggers <ebiggers3@gmail.com>
Fri, 20 Dec 2013 20:21:03 +0000 (14:21 -0600)
include/wimlib/apply.h
include/wimlib/dentry.h
include/wimlib/file_io.h
include/wimlib/ntfs_3g.h
include/wimlib/resource.h
src/extract.c
src/lookup_table.c
src/mount_image.c
src/ntfs-3g_capture.c
src/resource.c
src/write.c

index f8023a3..11a0365 100644 (file)
@@ -1,6 +1,7 @@
 #ifndef _WIMLIB_APPLY_H
 #define _WIMLIB_APPLY_H
 
+#include "wimlib/file_io.h"
 #include "wimlib/types.h"
 #include "wimlib/list.h"
 #include "wimlib.h"
@@ -220,6 +221,8 @@ struct apply_ctx {
        unsigned long invalid_sequence;
        unsigned long partial_security_descriptors;
        unsigned long no_security_descriptors;
+       struct filedes tmpfile_fd;
+       tchar *tmpfile_name;
        u64 num_streams_remaining;
        bool root_dentry_is_special;
        uint64_t next_progress;
index 12a8f17..a2735e5 100644 (file)
@@ -230,10 +230,6 @@ struct wim_dentry {
         * alias file_name.  */
        tchar *extraction_name;
        size_t extraction_name_nchars;
-
-       /* (Extraction only) List head for building a list of dentries that
-        * contain a certain stream. */
-       struct list_head extraction_stream_list;
 };
 
 #define rbnode_dentry(node) container_of(node, struct wim_dentry, rb_node)
index 85f293f..01f9717 100644 (file)
@@ -3,6 +3,7 @@
 
 #include <stddef.h>
 #include <sys/types.h>
+#include <stdbool.h>
 
 /* Wrapper around a file descriptor that keeps track of offset (including in
  * pipes, which don't support lseek()) and a cached flag that tells whether the
index fd93fa7..b8ff372 100644 (file)
@@ -14,8 +14,7 @@ extern int
 read_ntfs_file_prefix(const struct wim_lookup_table_entry *lte,
                      u64 size,
                      consume_data_callback_t cb,
-                     u32 in_chunk_size,
-                     void *ctx_or_buf,
+                     void *cb_ctx,
                      int _ignored_flags);
 
 
index 249b0bc..1d68649 100644 (file)
@@ -151,17 +151,15 @@ put_wim_reshdr(const struct wim_reshdr *reshdr,
 #define WIMLIB_READ_RESOURCE_MASK                      0xffff0000
 
 
-/* Functions to read a resource.  */
-
-extern int
-read_partial_wim_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);
+/* Functions to read streams  */
 
 extern int
 read_partial_wim_stream_into_buf(const struct wim_lookup_table_entry *lte,
                                 size_t size, u64 offset, void *buf);
+
+extern int
+skip_wim_stream(struct wim_lookup_table_entry *lte);
+
 extern int
 read_full_stream_into_buf(const struct wim_lookup_table_entry *lte, void *buf);
 
@@ -176,35 +174,59 @@ wim_reshdr_to_data(const struct wim_reshdr *reshdr,
 extern int
 read_stream_prefix(const struct wim_lookup_table_entry *lte,
                   u64 size, consume_data_callback_t cb,
-                  u32 in_chunk_size, void *ctx_or_buf, int flags);
+                  void *cb_ctx, int flags);
+
+typedef int (*read_stream_list_begin_stream_t)(struct wim_lookup_table_entry *lte,
+                                              bool is_partial_res,
+                                              void *ctx);
+typedef int (*read_stream_list_end_stream_t)(struct wim_lookup_table_entry *lte,
+                                            int status,
+                                            void *ctx);
+
+/* Callbacks for read_stream_list().  */
+struct read_stream_list_callbacks {
 
-/* Functions to read a list of resources.  */
+       /* Called when a stream is about to be read.  */
+       read_stream_list_begin_stream_t begin_stream;
 
-typedef int (*read_stream_list_begin_stream_t)(struct wim_lookup_table_entry *lte, void *ctx);
-typedef int (*read_stream_list_end_stream_t)(struct wim_lookup_table_entry *lte, void *ctx);
+       /* Called when a chunk of data has been read.  */
+       consume_data_callback_t consume_chunk;
+
+       /* Called when a stream has been fully read.  */
+       read_stream_list_end_stream_t end_stream;
+
+       /* Parameter passed to @begin_stream.  */
+       void *begin_stream_ctx;
+
+       /* Parameter passed to @consume_chunk.  */
+       void *consume_chunk_ctx;
+
+       /* Parameter passed to @end_stream.  */
+       void *end_stream_ctx;
+};
 
 extern int
 read_stream_list(struct list_head *stream_list,
                 size_t list_head_offset,
-                read_stream_list_begin_stream_t begin_stream,
-                consume_data_callback_t consume_chunk,
-                read_stream_list_end_stream_t end_stream,
                 u32 cb_chunk_size,
-                void *cb_ctx);
+                const struct read_stream_list_callbacks *cbs);
 
-/* Functions to extract a resource.  */
+/* Functions for stream extraction.  */
 
 extern int
-extract_stream(const struct wim_lookup_table_entry *lte,
+extract_stream(struct wim_lookup_table_entry *lte,
               u64 size,
               consume_data_callback_t extract_chunk,
               void *extract_chunk_arg);
 
 extern int
-extract_stream_to_fd(const struct wim_lookup_table_entry *lte,
+extract_stream_to_fd(struct wim_lookup_table_entry *lte,
                     struct filedes *fd, u64 size);
 
-/* Miscellaneous resource functions.  */
+extern int
+extract_chunk_to_fd(const void *chunk, size_t size, void *_fd_p);
+
+/* Miscellaneous stream functions.  */
 
 extern int
 sha1_stream(struct wim_lookup_table_entry *lte);
index 1907e69..5939671 100644 (file)
@@ -165,7 +165,7 @@ ref_stream_to_extract(struct wim_lookup_table_entry *lte,
                                lte_dentries = REALLOC(prev_lte_dentries,
                                                       alloc_lte_dentries *
                                                        sizeof(lte_dentries[0]));
-                               if (!lte_dentries)
+                               if (lte_dentries == NULL)
                                        return WIMLIB_ERR_NOMEM;
                                if (prev_lte_dentries == NULL) {
                                        memcpy(lte_dentries,
@@ -1202,23 +1202,49 @@ dentry_extract(struct wim_dentry *dentry, void *_ctx)
        return extract_streams(path, ctx, dentry, NULL, NULL);
 }
 
+/* Creates a temporary file opened for writing.  The open file descriptor is
+ * returned in @fd_ret and its name is returned in @name_ret (dynamically
+ * allocated).  */
+static int
+create_temporary_file(struct filedes *fd_ret, tchar **name_ret)
+{
+       tchar *name;
+       int raw_fd;
+
+retry:
+       name = ttempnam(NULL, T("wimlib"));
+       if (name == NULL) {
+               ERROR_WITH_ERRNO("Failed to create temporary filename");
+               return WIMLIB_ERR_NOMEM;
+       }
+
+       raw_fd = topen(name, O_WRONLY | O_CREAT | O_EXCL | O_BINARY, 0600);
+
+       if (raw_fd < 0) {
+               int errno_save = errno;
+               FREE(name);
+               if (errno_save == EEXIST)
+                       goto retry;
+               ERROR_WITH_ERRNO("Failed to open temporary file \"%"TS"\"", name);
+               return WIMLIB_ERR_OPEN;
+       }
+
+       filedes_init(fd_ret, raw_fd);
+       *name_ret = name;
+       return 0;
+}
+
 /* Extract all instances of the stream @lte that are being extracted in this
- * call of extract_tree().  @can_seek specifies whether the WIM file descriptor
- * is seekable or not (e.g. is a pipe).  If not and the stream needs to be
- * extracted multiple times, it is extracted to a temporary file first.
- *
- * This is intended for use with sequential extraction of a WIM image
- * (WIMLIB_EXTRACT_FLAG_SEQUENTIAL specified).  */
+ * call of extract_tree(), but actually read the stream data from @lte_override.
+ */
 static int
 extract_stream_instances(struct wim_lookup_table_entry *lte,
-                        struct apply_ctx *ctx, bool can_seek)
+                        struct wim_lookup_table_entry *lte_override,
+                        struct apply_ctx *ctx)
 {
        struct wim_dentry **lte_dentries;
-       struct wim_lookup_table_entry *lte_tmp = NULL;
-       struct wim_lookup_table_entry *lte_override;
-       tchar *stream_tmp_filename = NULL;
        tchar path[ctx->ops->path_max];
-       unsigned i;
+       size_t i;
        int ret;
 
        if (lte->out_refcnt <= ARRAY_LEN(lte->inline_lte_dentries))
@@ -1226,49 +1252,6 @@ extract_stream_instances(struct wim_lookup_table_entry *lte,
        else
                lte_dentries = lte->lte_dentries;
 
-       if (likely(can_seek || lte->out_refcnt < 2)) {
-               lte_override = lte;
-       } else {
-               /* Need to extract stream to temporary file.  */
-               struct filedes fd;
-               int raw_fd;
-
-               stream_tmp_filename = ttempnam(NULL, T("wimlib"));
-               if (!stream_tmp_filename) {
-                       ERROR_WITH_ERRNO("Failed to create temporary filename");
-                       ret = WIMLIB_ERR_OPEN;
-                       goto out;
-               }
-
-               lte_tmp = memdup(lte, sizeof(struct wim_lookup_table_entry));
-               if (!lte_tmp) {
-                       ret = WIMLIB_ERR_NOMEM;
-                       goto out_free_stream_tmp_filename;
-               }
-               lte_tmp->resource_location = RESOURCE_IN_FILE_ON_DISK;
-               lte_tmp->file_on_disk = stream_tmp_filename;
-               lte_override = lte_tmp;
-
-               raw_fd = topen(stream_tmp_filename,
-                              O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, 0600);
-               if (raw_fd < 0) {
-                       ERROR_WITH_ERRNO("Failed to open temporary file");
-                       ret = WIMLIB_ERR_OPEN;
-                       goto out_free_lte_tmp;
-               }
-               filedes_init(&fd, raw_fd);
-               ret = extract_stream_to_fd(lte, &fd, lte->size);
-               if (filedes_close(&fd) && !ret)
-                       ret = WIMLIB_ERR_WRITE;
-               if (ret)
-                       goto out_unlink_stream_tmp_file;
-       }
-
-       /* Extract all instances of the stream, reading either from the stream
-        * in the WIM file or from the temporary file containing the stream.
-        * dentry->tmp_flag is used to ensure that each dentry is processed only
-        * once regardless of how many times this stream appears in the streams
-        * of the corresponding inode.  */
        for (i = 0; i < lte->out_refcnt; i++) {
                struct wim_dentry *dentry = lte_dentries[i];
 
@@ -1276,8 +1259,7 @@ extract_stream_instances(struct wim_lookup_table_entry *lte,
                        continue;
                if (!build_extraction_path(path, dentry, ctx))
                        continue;
-               ret = extract_streams(path, ctx, dentry,
-                                     lte, lte_override);
+               ret = extract_streams(path, ctx, dentry, lte, lte_override);
                if (ret)
                        goto out_clear_tmp_flags;
                dentry->tmp_flag = 1;
@@ -1286,38 +1268,140 @@ extract_stream_instances(struct wim_lookup_table_entry *lte,
 out_clear_tmp_flags:
        for (i = 0; i < lte->out_refcnt; i++)
                lte_dentries[i]->tmp_flag = 0;
-out_unlink_stream_tmp_file:
-       if (stream_tmp_filename)
-               tunlink(stream_tmp_filename);
-out_free_lte_tmp:
-       FREE(lte_tmp);
-out_free_stream_tmp_filename:
-       FREE(stream_tmp_filename);
-out:
+       return ret;
+}
+
+/* Determine whether the specified stream needs to be extracted to a temporary
+ * file or not.
+ *
+ * @lte->out_refcnt specifies the number of instances of this stream that must
+ * be extracted.
+ *
+ * @is_partial_res is %true if this stream is just one of multiple in a single
+ * WIM resource being extracted.  */
+static bool
+need_tmpfile_to_extract(struct wim_lookup_table_entry *lte,
+                       bool is_partial_res)
+{
+       /* Temporary file is always required when reading a partial resource,
+        * since in that case we retrieve all the contained streams in one pass.
+        * */
+       if (is_partial_res)
+               return true;
+
+       /* Otherwise we don't need a temporary file if only a single instance of
+        * the stream is needed.  */
+       if (lte->out_refcnt == 1)
+               return false;
+
+       wimlib_assert(lte->out_refcnt >= 2);
+
+       /* We also don't need a temporary file if random access to the stream is
+        * allowed.  */
+       if (lte->resource_location != RESOURCE_IN_WIM ||
+           filedes_is_seekable(&lte->rspec->wim->in_fd))
+               return false;
+
+       return true;
+}
+
+static int
+begin_extract_stream_to_tmpfile(struct wim_lookup_table_entry *lte,
+                               bool is_partial_res, void *_ctx)
+{
+       struct apply_ctx *ctx = _ctx;
+       int ret;
+
+       if (!need_tmpfile_to_extract(lte, is_partial_res)) {
+               DEBUG("Temporary file not needed "
+                     "for stream (size=%"PRIu64")", lte->size);
+               ret = extract_stream_instances(lte, lte, ctx);
+               if (ret)
+                       return ret;
+
+               /* Negative return value here means the function was successful,
+                * but the consume_chunk and end_chunk callbacks need not be
+                * called.  */
+               return -1;
+       }
+
+       DEBUG("Temporary file needed for stream (size=%"PRIu64")", lte->size);
+       return create_temporary_file(&ctx->tmpfile_fd, &ctx->tmpfile_name);
+}
+
+static int
+end_extract_stream_to_tmpfile(struct wim_lookup_table_entry *lte,
+                             int status, void *_ctx)
+{
+       struct apply_ctx *ctx = _ctx;
+       struct wim_lookup_table_entry lte_override;
+       int ret;
+       int errno_save = errno;
+
+       ret = filedes_close(&ctx->tmpfile_fd);
+
+       if (status) {
+               ret = status;
+               errno = errno_save;
+               goto out_delete_tmpfile;
+       }
+
+       if (ret) {
+               ERROR_WITH_ERRNO("Error writing temporary file %"TS, ctx->tmpfile_name);
+               ret = WIMLIB_ERR_WRITE;
+               goto out_delete_tmpfile;
+       }
+
+       /* Now that a full stream has been extracted to a temporary file,
+        * extract all instances of it to the actual target.  */
+
+       memcpy(&lte_override, lte, sizeof(struct wim_lookup_table_entry));
+       lte_override.resource_location = RESOURCE_IN_FILE_ON_DISK;
+       lte_override.file_on_disk = ctx->tmpfile_name;
+
+       ret = extract_stream_instances(lte, &lte_override, ctx);
+
+out_delete_tmpfile:
+       errno_save = errno;
+       tunlink(ctx->tmpfile_name);
+       FREE(ctx->tmpfile_name);
+       errno = errno_save;
        return ret;
 }
 
 /* Extracts a list of streams (ctx.stream_list), assuming that the directory
  * structure and empty files were already created.  This relies on the
  * per-`struct wim_lookup_table_entry' list of dentries that reference each
- * stream that was constructed earlier.  Streams are extracted exactly in the
- * order of the stream list; however, unless the WIM's file descriptor is
- * detected to be non-seekable, streams may be read from the WIM file more than
- * one time if multiple copies need to be extracted.  */
+ * stream that was constructed earlier.  */
 static int
 extract_stream_list(struct apply_ctx *ctx)
 {
-       struct wim_lookup_table_entry *lte;
-       bool can_seek;
-       int ret;
+       if (ctx->extract_flags & WIMLIB_EXTRACT_FLAG_SEQUENTIAL) {
+               /* Sequential extraction: read the streams in the order in which
+                * they appear in the WIM file.  */
+               struct read_stream_list_callbacks cbs = {
+                       .begin_stream           = begin_extract_stream_to_tmpfile,
+                       .begin_stream_ctx       = ctx,
+                       .consume_chunk          = extract_chunk_to_fd,
+                       .consume_chunk_ctx      = &ctx->tmpfile_fd,
+                       .end_stream             = end_extract_stream_to_tmpfile,
+                       .end_stream_ctx         = ctx,
+               };
+               return read_stream_list(&ctx->stream_list,
+                                       offsetof(struct wim_lookup_table_entry, extraction_list),
+                                       0, &cbs);
+       } else {
+               /* Extract the streams in unsorted order.  */
+               struct wim_lookup_table_entry *lte;
+               int ret;
 
-       can_seek = (lseek(ctx->wim->in_fd.fd, 0, SEEK_CUR) != -1);
-       list_for_each_entry(lte, &ctx->stream_list, extraction_list) {
-               ret = extract_stream_instances(lte, ctx, can_seek);
-               if (ret)
-                       return ret;
+               list_for_each_entry(lte, &ctx->stream_list, extraction_list) {
+                       ret = extract_stream_instances(lte, lte, ctx);
+                       if (ret)
+                               return ret;
+               }
+               return 0;
        }
-       return 0;
 }
 
 #define PWM_ALLOW_WIM_HDR 0x00001
@@ -1377,25 +1461,6 @@ read_error:
 }
 
 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,
-                                        lte->size,
-                                        skip_pwm_chunk_cb,
-                                        lte_cchunk_size(lte),
-                                        NULL,
-                                        WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS,
-                                        0);
-}
-
-static int
 extract_streams_from_pipe(struct apply_ctx *ctx)
 {
        struct wim_lookup_table_entry *found_lte;
@@ -1444,19 +1509,60 @@ extract_streams_from_pipe(struct apply_ctx *ctx)
                    && (needed_lte = lookup_resource(lookup_table, found_lte->hash))
                    && (needed_lte->out_refcnt))
                {
-                       lte_unbind_wim_resource_spec(found_lte);
-                       lte_bind_wim_resource_spec(needed_lte, rspec);
+                       char *tmpfile_name = NULL;
+                       struct wim_lookup_table_entry *lte_override;
+                       struct wim_lookup_table_entry tmpfile_lte;
+
                        needed_lte->offset_in_res = found_lte->offset_in_res;
                        needed_lte->flags = found_lte->flags;
                        needed_lte->size = found_lte->size;
-                       ret = extract_stream_instances(needed_lte, ctx, false);
-                       lte_unbind_wim_resource_spec(needed_lte);
 
+                       lte_unbind_wim_resource_spec(found_lte);
+                       lte_bind_wim_resource_spec(needed_lte, rspec);
+
+                       if (needed_lte->out_refcnt > 1) {
+
+                               struct filedes tmpfile_fd;
+
+                               /* Extract stream to temporary file.  */
+                               ret = create_temporary_file(&tmpfile_fd, &tmpfile_name);
+                               if (ret)
+                                       goto out_free_found_lte;
+
+                               ret = extract_stream_to_fd(needed_lte, &tmpfile_fd,
+                                                          needed_lte->size);
+                               if (ret) {
+                                       filedes_close(&tmpfile_fd);
+                                       goto delete_tmpfile;
+                               }
+
+                               if (filedes_close(&tmpfile_fd)) {
+                                       ERROR_WITH_ERRNO("Error writing to temporary "
+                                                        "file \"%"TS"\"", tmpfile_name);
+                                       ret = WIMLIB_ERR_WRITE;
+                                       goto delete_tmpfile;
+                               }
+                               memcpy(&tmpfile_lte, needed_lte,
+                                      sizeof(struct wim_lookup_table_entry));
+                               tmpfile_lte.resource_location = RESOURCE_IN_FILE_ON_DISK;
+                               tmpfile_lte.file_on_disk = tmpfile_name;
+                               lte_override = &tmpfile_lte;
+                       } else {
+                               lte_override = needed_lte;
+                       }
+
+                       ret = extract_stream_instances(needed_lte, lte_override, ctx);
+               delete_tmpfile:
+                       lte_unbind_wim_resource_spec(needed_lte);
+                       if (tmpfile_name) {
+                               tunlink(tmpfile_name);
+                               FREE(tmpfile_name);
+                       }
                        if (ret)
                                goto out_free_found_lte;
                        ctx->num_streams_remaining--;
                } else if (found_lte->resource_location != RESOURCE_NONEXISTENT) {
-                       ret = skip_pwm_stream(found_lte);
+                       ret = skip_wim_stream(found_lte);
                        if (ret)
                                goto out_free_found_lte;
                } else {
@@ -2268,21 +2374,6 @@ extract_tree(WIMStruct *wim, const tchar *wim_source_path, const tchar *target,
                goto out_teardown_stream_list;
        }
 
-       /* If a sequential extraction was specified, sort the streams to be
-        * extracted by their position in the WIM file so that the WIM file can
-        * be read sequentially.  */
-       if ((extract_flags & (WIMLIB_EXTRACT_FLAG_SEQUENTIAL |
-                             WIMLIB_EXTRACT_FLAG_FROM_PIPE))
-                                       == WIMLIB_EXTRACT_FLAG_SEQUENTIAL)
-       {
-               ret = sort_stream_list_by_sequential_order(
-                               &ctx.stream_list,
-                               offsetof(struct wim_lookup_table_entry,
-                                        extraction_list));
-               if (ret)
-                       goto out_teardown_stream_list;
-       }
-
        if (ctx.ops->realpath_works_on_nonexisting_files &&
            ((extract_flags & WIMLIB_EXTRACT_FLAG_RPFIX) ||
             ctx.ops->requires_realtarget_in_paths))
@@ -2920,7 +3011,7 @@ wimlib_extract_image_from_pipe(int pipe_fd, const tchar *image_num_or_name,
                } else {
                        /* Metadata resource is not for the image being
                         * extracted.  Skip over it.  */
-                       ret = skip_pwm_stream(metadata_lte);
+                       ret = skip_wim_stream(metadata_lte);
                        if (ret)
                                goto out_wimlib_free;
                }
index 4328734..fe0872c 100644 (file)
@@ -82,7 +82,6 @@ new_lookup_table_entry(void)
        }
        lte->refcnt = 1;
        BUILD_BUG_ON(RESOURCE_NONEXISTENT != 0);
-       BUILD_BUG_ON(WIMLIB_COMPRESSION_TYPE_NONE != 0);
        return lte;
 }
 
@@ -311,6 +310,7 @@ cmp_streams_by_sequential_order(const void *p1, const void *p2)
        const struct wim_lookup_table_entry *lte1, *lte2;
        int v;
        WIMStruct *wim1, *wim2;
+       u64 offset1, offset2;
 
        lte1 = *(const struct wim_lookup_table_entry**)p1;
        lte2 = *(const struct wim_lookup_table_entry**)p2;
@@ -339,9 +339,12 @@ cmp_streams_by_sequential_order(const void *p1, const void *p2)
                        return v;
 
                /* Compare by offset.  */
-               if (lte1->rspec->offset_in_wim < lte2->rspec->offset_in_wim)
+               offset1 = lte1->rspec->offset_in_wim + lte1->offset_in_res;
+               offset2 = lte2->rspec->offset_in_wim + lte2->offset_in_res;
+
+               if (offset1 < offset2)
                        return -1;
-               if (lte1->rspec->offset_in_wim > lte2->rspec->offset_in_wim)
+               if (offset1 > offset2)
                        return 1;
                return 0;
        case RESOURCE_IN_FILE_ON_DISK:
index 7543b6e..99db4ab 100644 (file)
@@ -2414,6 +2414,11 @@ wimlib_mount_image(WIMStruct *wim, int image, const char *dir,
                        return ret;
        }
 
+       if (wim->hdr.wim_version == WIM_VERSION_STREAM_CONCAT) {
+               WARNING("WIM contains streams not compressed independently; "
+                       "access may be slow.");
+       }
+
        ret = select_wim_image(wim, image);
        if (ret)
                return ret;
index 829c41c..7a07cd3 100644 (file)
@@ -78,8 +78,7 @@ int
 read_ntfs_file_prefix(const struct wim_lookup_table_entry *lte,
                      u64 size,
                      consume_data_callback_t cb,
-                     u32 in_chunk_size,
-                     void *ctx_or_buf,
+                     void *cb_ctx,
                      int _ignored_flags)
 {
        struct ntfs_location *loc = lte->ntfs_loc;
@@ -88,9 +87,8 @@ read_ntfs_file_prefix(const struct wim_lookup_table_entry *lte,
        ntfs_attr *na;
        s64 pos;
        s64 bytes_remaining;
-       void *out_buf;
-       bool out_buf_malloced;
        int ret;
+       u8 buf[BUFFER_SIZE];
 
        ni = ntfs_pathname_to_inode(vol, NULL, loc->path);
        if (!ni) {
@@ -105,44 +103,22 @@ read_ntfs_file_prefix(const struct wim_lookup_table_entry *lte,
                goto out_close_ntfs_inode;
        }
 
-       out_buf_malloced = false;
-       if (cb) {
-               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_ntfs_attr;
-                       }
-                       out_buf_malloced = true;
-               }
-       } else {
-               out_buf = ctx_or_buf;
-       }
        pos = (loc->is_reparse_point) ? 8 : 0;
        bytes_remaining = size;
        while (bytes_remaining) {
-               s64 to_read = min(bytes_remaining, in_chunk_size);
-               if (ntfs_attr_pread(na, pos, to_read, out_buf) != to_read) {
+               s64 to_read = min(bytes_remaining, sizeof(buf));
+               if (ntfs_attr_pread(na, pos, to_read, buf) != to_read) {
                        ERROR_WITH_ERRNO("Error reading \"%"TS"\"", loc->path);
                        ret = WIMLIB_ERR_NTFS_3G;
-                       goto out_free_memory;
+                       goto out_close_ntfs_attr;
                }
                pos += to_read;
                bytes_remaining -= to_read;
-               if (cb) {
-                       ret = cb(out_buf, to_read, ctx_or_buf);
-                       if (ret)
-                               goto out_free_memory;
-               } else {
-                       out_buf += to_read;
-               }
+               ret = cb(buf, to_read, cb_ctx);
+               if (ret)
+                       goto out_close_ntfs_attr;
        }
        ret = 0;
-out_free_memory:
-       if (out_buf_malloced)
-               FREE(out_buf);
 out_close_ntfs_attr:
        ntfs_attr_close(na);
 out_close_ntfs_inode:
index 1539460..90197d7 100644 (file)
@@ -135,7 +135,41 @@ struct alt_chunk_table_header_disk {
         * the chunks.  */
 } _packed_attribute;
 
-/* Read data from a compressed WIM resource.  */
+/*
+ * read_compressed_wim_resource() -
+ *
+ * Read data from a compressed WIM resource.
+ *
+ * @rspec
+ *     Specification of the compressed WIM resource to read from.
+ * @ranges
+ *     Nonoverlapping, nonempty ranges of the uncompressed resource data to
+ *     read, sorted by increasing offset.
+ * @num_ranges
+ *     Number of ranges in @ranges; must be at least 1.
+ * @cb
+ *     Callback function to feed the data being read.  Each call provides the
+ *     next chunk of the requested data.  Each chunk will be of nonzero size
+ *     and will not cross range boundaries, but otherwise is of unspecified
+ *     size.
+ * @cb_ctx
+ *     Parameter to pass to @cb_ctx.
+ * @raw_chunks_mode
+ *     If %true, this function will provide the raw compressed chunks of the
+ *     resource rather than the uncompressed data.  In this mode, only a single
+ *     data range can be requested, and it must cover the entire uncompressed
+ *     resource.
+ *
+ * Possible return values:
+ *
+ *     WIMLIB_ERR_SUCCESS (0)
+ *     WIMLIB_ERR_READ                   (errno set)
+ *     WIMLIB_ERR_UNEXPECTED_END_OF_FILE (errno set to 0)
+ *     WIMLIB_ERR_NOMEM                  (errno set to ENOMEM)
+ *     WIMLIB_ERR_DECOMPRESSION          (errno set to EINVAL)
+ *
+ *     or other error code returned by the @cb function.
+ */
 static int
 read_compressed_wim_resource(const struct wim_resource_spec * const rspec,
                             const struct data_range * const ranges,
@@ -161,6 +195,9 @@ read_compressed_wim_resource(const struct wim_resource_spec * const rspec,
        wimlib_assert(cb != NULL);
        wimlib_assert(num_ranges != 0);
        for (size_t i = 0; i < num_ranges; i++) {
+               DEBUG("Range %zu/%zu: %"PRIu64"@+%"PRIu64" / %"PRIu64,
+                     i + 1, num_ranges, ranges[i].size, ranges[i].offset,
+                     rspec->uncompressed_size);
                wimlib_assert(ranges[i].size != 0);
                wimlib_assert(ranges[i].offset + ranges[i].size >= ranges[i].size);
                wimlib_assert(ranges[i].offset + ranges[i].size <= rspec->uncompressed_size);
@@ -168,6 +205,12 @@ read_compressed_wim_resource(const struct wim_resource_spec * const rspec,
        for (size_t i = 0; i < num_ranges - 1; i++)
                wimlib_assert(ranges[i].offset + ranges[i].size <= ranges[i + 1].offset);
 
+       if (raw_chunks_mode) {
+               wimlib_assert(num_ranges == 1);
+               wimlib_assert(ranges[0].offset == 0);
+               wimlib_assert(ranges[0].size == rspec->uncompressed_size);
+       }
+
        /* Get the offsets of the first and last bytes of the read.  */
        const u64 first_offset = ranges[0].offset;
        const u64 last_offset = ranges[num_ranges - 1].offset + ranges[num_ranges - 1].size - 1;
@@ -242,7 +285,8 @@ read_compressed_wim_resource(const struct wim_resource_spec * const rspec,
        /* Calculate the size of the chunk table in bytes.  */
        const u64 chunk_table_size = num_chunk_entries * chunk_entry_size;
 
-       /* Includes header  */
+       /* Calculate the size of the chunk table in bytes, including the header
+        * in the case of the alternate chunk table format.  */
        const u64 chunk_table_full_size =
                (alt_chunk_table) ? chunk_table_size + sizeof(struct alt_chunk_table_header_disk)
                                  : chunk_table_size;
@@ -332,6 +376,8 @@ read_compressed_wim_resource(const struct wim_resource_spec * const rspec,
                                        *chunk_offsets_p++ = cur_offset;
                                cur_offset += entry;
                        }
+                       if (last_needed_chunk < num_chunks - 1)
+                               *chunk_offsets_p = cur_offset;
                } else {
                        if (read_start_chunk == 0)
                                *chunk_offsets_p++ = 0;
@@ -380,12 +426,13 @@ read_compressed_wim_resource(const struct wim_resource_spec * const rspec,
                }
        }
 
-       /* Read and process each needed chunk.  */
+       /* Set current data range.  */
        const struct data_range *cur_range = ranges;
        const struct data_range * const end_range = &ranges[num_ranges];
        u64 cur_range_pos = cur_range->offset;
        u64 cur_range_end = cur_range->offset + cur_range->size;
 
+       /* Read and process each needed chunk.  */
        for (u64 i = read_start_chunk; i <= last_needed_chunk; i++) {
 
                /* Calculate uncompressed size of next chunk.  */
@@ -426,7 +473,8 @@ read_compressed_wim_resource(const struct wim_resource_spec * const rspec,
                if (rspec->is_pipable)
                        cur_read_offset += sizeof(struct pwm_chunk_hdr);
 
-               /* Uncompressed offsets  */
+               /* Offsets in the uncompressed resource at which this chunk
+                * starts and ends.  */
                const u64 chunk_start_offset = i << chunk_order;
                const u64 chunk_end_offset = chunk_start_offset + chunk_usize;
 
@@ -434,7 +482,6 @@ read_compressed_wim_resource(const struct wim_resource_spec * const rspec,
 
                        /* The next range does not require data in this chunk,
                         * so skip it.  */
-
                        cur_read_offset += chunk_csize;
                        if (is_pipe_read) {
                                u8 dummy;
@@ -457,6 +504,9 @@ read_compressed_wim_resource(const struct wim_resource_spec * const rspec,
                                goto read_error;
 
                        if (chunk_csize != chunk_usize && !raw_chunks_mode) {
+                               DEBUG("Decompressing chunk %"PRIu64" "
+                                     "(csize=%"PRIu64" usize=%"PRIu64"",
+                                     i, chunk_csize, chunk_usize);
                                ret = decompress(cbuf,
                                                 chunk_csize,
                                                 ubuf,
@@ -471,16 +521,21 @@ read_compressed_wim_resource(const struct wim_resource_spec * const rspec,
                                }
                                cb_buf = ubuf;
                        } else {
+                               /* Raw chunks requested, or data stored
+                                * uncompressed.  */
                                cb_buf = cbuf;
                        }
                        cur_read_offset += chunk_csize;
 
-                       /* At least one range requires data in this chunk.
-                        * However, the data fed to the callback function must
-                        * not overlap range boundaries.  */
+                       /* At least one range requires data in this chunk.  */
                        do {
                                size_t start, end, size;
 
+                               /* Calculate how many bytes of data should be
+                                * sent to the callback function, taking into
+                                * account that data sent to the callback
+                                * function must not overlap range boundaries.
+                                */
                                start = cur_range_pos - chunk_start_offset;
                                end = min(cur_range_end, chunk_end_offset) - chunk_start_offset;
                                size = end - start;
@@ -495,6 +550,7 @@ read_compressed_wim_resource(const struct wim_resource_spec * const rspec,
 
                                cur_range_pos += size;
                                if (cur_range_pos == cur_range_end) {
+                                       /* Advance to next range.  */
                                        if (++cur_range == end_range) {
                                                cur_range_pos = ~0ULL;
                                        } else {
@@ -506,13 +562,15 @@ read_compressed_wim_resource(const struct wim_resource_spec * const rspec,
                }
        }
 
-       if (is_pipe_read
-           && last_offset == rspec->uncompressed_size - 1
-           && chunk_table_size)
+       if (is_pipe_read &&
+           last_offset == rspec->uncompressed_size - 1 &&
+           chunk_table_size)
        {
                u8 dummy;
-               /* Skip chunk table at end of pipable resource.  */
-
+               /* If reading a pipable resource from a pipe and the full data
+                * was requested, skip the chunk table at the end so that the
+                * file descriptor is fully clear of the resource after this
+                * returns.  */
                cur_read_offset += chunk_table_size;
                ret = full_pread(in_fd, &dummy, 1, cur_read_offset - 1);
                if (ret)
@@ -532,65 +590,40 @@ out_free_memory:
 
 oom:
        ERROR("Not enough memory available to read size=%"PRIu64" bytes "
-             "from compressed resource!", last_offset - first_offset + 1);
+             "from compressed WIM resource!", last_offset - first_offset + 1);
        errno = ENOMEM;
        ret = WIMLIB_ERR_NOMEM;
        goto out_free_memory;
 
 read_error:
-       ERROR_WITH_ERRNO("Error reading compressed file resource!");
+       ERROR_WITH_ERRNO("Error reading compressed WIM resource!");
        goto out_free_memory;
 }
 
-/* Read raw data from a file descriptor at the specified offset.  */
+/* Read raw data from a file descriptor at the specified offset, feeding the
+ * data it in chunks into the specified callback function.  */
 static int
-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)
+read_raw_file_data(struct filedes *in_fd, u64 size,
+                  consume_data_callback_t cb, void *cb_ctx, u64 offset)
 {
+       u8 buf[BUFFER_SIZE];
+       size_t bytes_to_read;
        int ret;
-       u8 *tmp_buf;
-       bool tmp_buf_malloced = false;
 
-       if (cb) {
-               /* Send data to callback function in chunks.  */
-               if (cb_chunk_size <= STACK_MAX) {
-                       tmp_buf = alloca(cb_chunk_size);
-               } else {
-                       tmp_buf = MALLOC(cb_chunk_size);
-                       if (tmp_buf == NULL) {
-                               ret = WIMLIB_ERR_NOMEM;
-                               goto out;
-                       }
-                       tmp_buf_malloced = true;
-               }
-
-               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;
+       while (size) {
+               bytes_to_read = min(sizeof(buf), size);
+               ret = full_pread(in_fd, buf, bytes_to_read, offset);
+               if (ret) {
+                       ERROR_WITH_ERRNO("Read error");
+                       return ret;
                }
-       } else {
-               /* Read data directly into buffer.  */
-               ret = full_pread(in_fd, ctx_or_buf, size, offset);
+               ret = cb(buf, bytes_to_read, cb_ctx);
                if (ret)
-                       goto read_error;
+                       return ret;
+               size -= bytes_to_read;
+               offset += bytes_to_read;
        }
-       ret = 0;
-       goto out;
-
-read_error:
-       ERROR_WITH_ERRNO("Read error");
-out:
-       if (tmp_buf_malloced)
-               FREE(tmp_buf);
-       return ret;
+       return 0;
 }
 
 static int
@@ -602,71 +635,15 @@ bufferer_cb(const void *chunk, size_t size, void *_ctx)
        return 0;
 }
 
-struct rechunker_context {
-       u8 *buffer;
-       u32 buffer_filled;
-       u32 cb_chunk_size;
-
-       const struct data_range *ranges;
-       size_t num_ranges;
-       size_t cur_range;
-       u64 range_bytes_remaining;
-
-       consume_data_callback_t cb;
-       void *cb_ctx;
-};
-
-static int
-rechunker_cb(const void *chunk, size_t size, void *_ctx)
-{
-       struct rechunker_context *ctx = _ctx;
-       const u8 *chunkptr = chunk;
-       size_t bytes_to_copy;
-       int ret;
-
-       wimlib_assert(ctx->cur_range != ctx->num_ranges);
-
-       while (size) {
-               bytes_to_copy = size;
-
-               if (bytes_to_copy > ctx->cb_chunk_size - ctx->buffer_filled)
-                       bytes_to_copy = ctx->cb_chunk_size - ctx->buffer_filled;
-
-               if (bytes_to_copy > ctx->range_bytes_remaining - ctx->buffer_filled)
-                       bytes_to_copy = ctx->range_bytes_remaining - ctx->buffer_filled;
-
-               memcpy(&ctx->buffer[ctx->buffer_filled], chunkptr, bytes_to_copy);
-
-               ctx->buffer_filled += bytes_to_copy;
-               chunkptr += bytes_to_copy;
-               size -= bytes_to_copy;
-               ctx->range_bytes_remaining -= bytes_to_copy;
-
-               if (ctx->buffer_filled == ctx->cb_chunk_size ||
-                   ctx->range_bytes_remaining == 0)
-               {
-                       ret = (*ctx->cb)(ctx->buffer, ctx->buffer_filled, ctx->cb_ctx);
-                       if (ret)
-                               return ret;
-                       ctx->buffer_filled = 0;
-
-                       if (ctx->range_bytes_remaining == 0 &&
-                           ++ctx->cur_range != ctx->num_ranges)
-                               ctx->range_bytes_remaining = ctx->ranges[ctx->cur_range].size;
-               }
-       }
-       return 0;
-}
-
 /*
  * read_partial_wim_resource()-
  *
  * Read a range of data from an uncompressed or compressed resource in a WIM
- * file.  Data is written into a buffer or fed into a callback function, as
- * documented in read_stream_prefix().
+ * file.  Data is fed chunk-by-chunk into the callback function @cb, passing it
+ * the argument @cb_ctx.
  *
  * By default, this function provides the uncompressed data of the resource, and
- * @size and @offset and interpreted relative to the uncompressed contents of
+ * @offset and @size and interpreted relative to the uncompressed contents of
  * the resource.  This behavior can be modified by either of the following
  * flags:
  *
@@ -689,135 +666,99 @@ rechunker_cb(const void *chunk, size_t size, void *_ctx)
  *
  *     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 cb_chunk_size, void *ctx_or_buf,
-                         int flags, u64 offset)
+static int
+read_partial_wim_resource(const struct wim_resource_spec *rspec,
+                         u64 offset, u64 size, consume_data_callback_t cb,
+                         void *cb_ctx, int flags)
 {
-       const struct wim_resource_spec *rspec;
-       struct filedes *in_fd;
-
-       /* Verify parameters.  */
-       wimlib_assert(lte->resource_location == RESOURCE_IN_WIM);
-       rspec = lte->rspec;
-       in_fd = &rspec->wim->in_fd;
-       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(!lte_is_partial(lte));
-               wimlib_assert(!(flags & WIMLIB_READ_RESOURCE_FLAG_RAW_FULL));
-               wimlib_assert(cb_chunk_size == rspec->cchunk_size);
-               wimlib_assert(size == lte->size);
+       /* Sanity checks.  */
+       if (flags & WIMLIB_READ_RESOURCE_FLAG_RAW_FULL) {
+               wimlib_assert(!(flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS));
+               wimlib_assert(offset + size >= offset);
+               wimlib_assert(offset + size <= rspec->size_in_wim);
+       } else if (flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS) {
                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(!lte_is_partial(lte));
-               wimlib_assert(offset + size >= size &&
-                             offset + size <= rspec->size_in_wim);
+               wimlib_assert(offset == rspec->uncompressed_size);
        } else {
-               /* Normal mode:  read must not overrun end of original size.  */
-               wimlib_assert(offset + size >= size &&
-                             offset + size <= lte->size);
+               wimlib_assert(offset + size >= offset);
+               wimlib_assert(offset + size <= rspec->uncompressed_size);
        }
 
-       DEBUG("Reading WIM resource: %"PRIu64" @ +%"PRIu64"[+%"PRIu64"] "
-             "from %"PRIu64"(%"PRIu64") @ +%"PRIu64" "
-             "(readflags 0x%08x, resflags 0x%02x%s)",
-             size, offset, lte->offset_in_res,
-             rspec->size_in_wim,
-             rspec->uncompressed_size,
-             rspec->offset_in_wim,
-             flags, lte->flags,
-             (rspec->is_pipable ? ", pipable" : ""));
+       DEBUG("Reading %"PRIu64" @ %"PRIu64" from WIM resource  "
+             "%"PRIu64" => %"PRIu64" @ %"PRIu64" (flags 0x%08x)",
+             size, offset, rspec->uncompressed_size,
+             rspec->size_in_wim, rspec->offset_in_wim, flags);
+
+       /* Trivial case.  */
+       if (size == 0)
+               return 0;
 
        if ((flags & WIMLIB_READ_RESOURCE_FLAG_RAW_FULL) ||
            rspec->ctype == WIMLIB_COMPRESSION_TYPE_NONE)
        {
-               return read_raw_file_data(in_fd,
+               return read_raw_file_data(&rspec->wim->in_fd,
                                          size,
                                          cb,
-                                         cb_chunk_size,
-                                         ctx_or_buf,
-                                         rspec->offset_in_wim + lte->offset_in_res + offset);
+                                         cb_ctx,
+                                         rspec->offset_in_wim + offset);
        } else {
-               bool raw_chunks;
-               struct data_range range;
-               consume_data_callback_t internal_cb;
-               void *internal_cb_ctx;
-               u8 *buf;
-               bool rechunker_buf_malloced = false;
-               struct rechunker_context *rechunker_ctx;
-               int ret;
-
-               if (size == 0)
-                       return 0;
-
-               range.offset = lte->offset_in_res + offset;
-               range.size = size;
-               raw_chunks = !!(flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS);
-
-               if (cb != NULL &&
-                   cb_chunk_size == rspec->cchunk_size &&
-                   !(rspec->flags & WIM_RESHDR_FLAG_CONCAT))
-               {
-                       internal_cb = cb;
-                       internal_cb_ctx = ctx_or_buf;
-               } else if (cb == NULL) {
-                       buf = ctx_or_buf;
-                       internal_cb = bufferer_cb;
-                       internal_cb_ctx = &buf;
-               } else {
-                       rechunker_ctx = alloca(sizeof(struct rechunker_context));
-
-                       if (cb_chunk_size <= STACK_MAX) {
-                               rechunker_ctx->buffer = alloca(cb_chunk_size);
-                       } else {
-                               rechunker_ctx->buffer = MALLOC(cb_chunk_size);
-                               if (rechunker_ctx->buffer == NULL)
-                                       return WIMLIB_ERR_NOMEM;
-                               rechunker_buf_malloced = true;
-                       }
-                       rechunker_ctx->buffer_filled = 0;
-                       rechunker_ctx->cb_chunk_size = cb_chunk_size;
-
-                       rechunker_ctx->ranges = &range;
-                       rechunker_ctx->num_ranges = 1;
-                       rechunker_ctx->cur_range = 0;
-                       rechunker_ctx->range_bytes_remaining = range.size;
+               bool raw_chunks = !!(flags & WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS);
+               struct data_range range = {
+                       .offset = offset,
+                       .size = size,
+               };
+               return read_compressed_wim_resource(rspec, &range, 1,
+                                                   cb, cb_ctx, raw_chunks);
+       }
+}
 
-                       rechunker_ctx->cb = cb;
-                       rechunker_ctx->cb_ctx = ctx_or_buf;
+int
+read_partial_wim_stream_into_buf(const struct wim_lookup_table_entry *lte,
+                                size_t size, u64 offset, void *_buf)
+{
+       u8 *buf = _buf;
 
-                       internal_cb = rechunker_cb;
-                       internal_cb_ctx = rechunker_ctx;
-               }
+       wimlib_assert(lte->resource_location == RESOURCE_IN_WIM);
 
-               ret = read_compressed_wim_resource(rspec, &range, 1,
-                                                  internal_cb, internal_cb_ctx,
-                                                  raw_chunks);
-               if (rechunker_buf_malloced)
-                       FREE(rechunker_ctx->buffer);
+       return read_partial_wim_resource(lte->rspec,
+                                        lte->offset_in_res + offset,
+                                        size,
+                                        bufferer_cb,
+                                        &buf,
+                                        0);
+}
 
-               return ret;
-       }
+static int
+skip_chunk_cb(const void *chunk, size_t size, void *_ctx)
+{
+       return 0;
 }
 
+/* Skip over the data of the specified stream, which must correspond to a full
+ * WIM resource.  */
 int
-read_partial_wim_stream_into_buf(const struct wim_lookup_table_entry *lte,
-                                size_t size, u64 offset, void *buf)
+skip_wim_stream(struct wim_lookup_table_entry *lte)
 {
-       return read_partial_wim_resource(lte, size, NULL, 0, buf, 0, offset);
+       wimlib_assert(lte->resource_location == RESOURCE_IN_WIM);
+       wimlib_assert(!lte_is_partial(lte));
+       return read_partial_wim_resource(lte->rspec,
+                                        0,
+                                        lte->rspec->uncompressed_size,
+                                        skip_chunk_cb,
+                                        NULL,
+                                        WIMLIB_READ_RESOURCE_FLAG_RAW_CHUNKS);
 }
 
 static int
 read_wim_stream_prefix(const struct wim_lookup_table_entry *lte, u64 size,
-                      consume_data_callback_t cb, u32 cb_chunk_size,
-                      void *ctx_or_buf, int flags)
+                      consume_data_callback_t cb, void *cb_ctx, int flags)
 {
-       return read_partial_wim_resource(lte, size, cb, cb_chunk_size,
-                                        ctx_or_buf, flags, 0);
+       return read_partial_wim_resource(lte->rspec,
+                                        lte->offset_in_res,
+                                        size,
+                                        cb,
+                                        cb_ctx,
+                                        flags);
 }
 
 #ifndef __WIN32__
@@ -828,18 +769,18 @@ read_wim_stream_prefix(const struct wim_lookup_table_entry *lte, u64 size,
  * 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.
- */
+ * 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 cb_chunk_size,
-                        void *ctx_or_buf, int _ignored_flags)
+                        consume_data_callback_t cb, void *cb_ctx,
+                        int _ignored_flags)
 {
        int ret;
        int raw_fd;
        struct filedes fd;
 
        wimlib_assert(size <= lte->size);
+
        DEBUG("Reading %"PRIu64" bytes from \"%"TS"\"", size, lte->file_on_disk);
 
        raw_fd = open(lte->file_on_disk, O_BINARY | O_RDONLY);
@@ -848,7 +789,7 @@ read_file_on_disk_prefix(const struct wim_lookup_table_entry *lte, u64 size,
                return WIMLIB_ERR_OPEN;
        }
        filedes_init(&fd, raw_fd);
-       ret = read_raw_file_data(&fd, size, cb, cb_chunk_size, ctx_or_buf, 0);
+       ret = read_raw_file_data(&fd, size, cb, cb_ctx, 0);
        filedes_close(&fd);
        return ret;
 }
@@ -859,34 +800,16 @@ read_file_on_disk_prefix(const struct wim_lookup_table_entry *lte, u64 size,
 static int
 read_buffer_prefix(const struct wim_lookup_table_entry *lte,
                   u64 size, consume_data_callback_t cb,
-                  u32 cb_chunk_size, void *ctx_or_buf, int _ignored_flags)
+                  void *cb_ctx, int _ignored_flags)
 {
        wimlib_assert(size <= lte->size);
-
-       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(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;
+       return (*cb)(lte->attached_buffer, size, cb_ctx);
 }
 
 typedef int (*read_stream_prefix_handler_t)(const struct wim_lookup_table_entry *lte,
-                                           u64 size, consume_data_callback_t cb,
-                                           u32 cb_chunk_size, void *ctx_or_buf,
-                                           int flags);
+                                           u64 size,
+                                           consume_data_callback_t cb,
+                                           void *cb_ctx, int flags);
 
 /*
  * read_stream_prefix()-
@@ -895,33 +818,20 @@ typedef int (*read_stream_prefix_handler_t)(const struct wim_lookup_table_entry
  * 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.
  *
- * 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 (@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
- * stream 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.
+ * This function feeds the data to a callback function @cb.
  *
  * If the stream 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 stream data cannot be successfully read (for a number of different
- * reasons, depending on the stream location), or if a callback function was
- * specified and it returned nonzero.
+ * reasons, depending on the stream location), or if @cb returned nonzero in
+ * which case that error code will be returned.
  */
 int
 read_stream_prefix(const struct wim_lookup_table_entry *lte, u64 size,
-                  consume_data_callback_t cb, u32 cb_chunk_size,
-                  void *ctx_or_buf, int flags)
+                  consume_data_callback_t cb, void *cb_ctx, int flags)
 {
-       /* This function merely verifies several preconditions, then passes
-        * control to an appropriate function for understanding each possible
-        * stream location.  */
        static const read_stream_prefix_handler_t handlers[] = {
                [RESOURCE_IN_WIM]             = read_wim_stream_prefix,
        #ifdef __WIN32__
@@ -942,17 +852,16 @@ read_stream_prefix(const struct wim_lookup_table_entry *lte, u64 size,
        };
        wimlib_assert(lte->resource_location < ARRAY_LEN(handlers)
                      && handlers[lte->resource_location] != NULL);
-       wimlib_assert(cb == NULL || cb_chunk_size > 0);
-       return handlers[lte->resource_location](lte, size, cb, cb_chunk_size,
-                                               ctx_or_buf, flags);
+       return handlers[lte->resource_location](lte, size, cb, cb_ctx, flags);
 }
 
 /* Read the full uncompressed data of the specified stream into the specified
  * buffer, which must have space for at least lte->size bytes.  */
 int
-read_full_stream_into_buf(const struct wim_lookup_table_entry *lte, void *buf)
+read_full_stream_into_buf(const struct wim_lookup_table_entry *lte, void *_buf)
 {
-       return read_stream_prefix(lte, lte->size, NULL, 0, buf, 0);
+       u8 *buf = _buf;
+       return read_stream_prefix(lte, lte->size, bufferer_cb, &buf, 0);
 }
 
 /* Read the full uncompressed data of the specified stream.  A buffer sufficient
@@ -1021,56 +930,226 @@ wim_reshdr_to_data(const struct wim_reshdr *reshdr, WIMStruct *wim, void **buf_r
        return wim_resource_spec_to_data(&rspec, buf_ret);
 }
 
-struct read_stream_list_ctx {
-       read_stream_list_begin_stream_t begin_stream;
-       consume_data_callback_t consume_chunk;
-       read_stream_list_end_stream_t end_stream;
-       void *begin_stream_ctx;
-       void *consume_chunk_ctx;
-       void *end_stream_ctx;
+struct streamifier_context {
+       struct read_stream_list_callbacks cbs;
        struct wim_lookup_table_entry *cur_stream;
        u64 cur_stream_offset;
        struct wim_lookup_table_entry *final_stream;
        size_t list_head_offset;
 };
 
+/* Callback for translating raw resource data into streams.  */
 static int
-read_stream_list_wrapper_cb(const void *chunk, size_t size, void *_ctx)
+streamifier_cb(const void *chunk, size_t size, void *_ctx)
 {
-       struct read_stream_list_ctx *ctx = _ctx;
+       struct streamifier_context *ctx = _ctx;
        int ret;
 
+       DEBUG("%zu bytes passed to streamifier", size);
+
+       wimlib_assert(ctx->cur_stream != NULL);
+
        if (ctx->cur_stream_offset == 0) {
                /* Starting a new stream.  */
-               ret = (*ctx->begin_stream)(ctx->cur_stream, ctx->begin_stream_ctx);
+               DEBUG("Begin new stream (size=%"PRIu64").", ctx->cur_stream->size);
+               ret = (*ctx->cbs.begin_stream)(ctx->cur_stream, true, ctx->cbs.begin_stream_ctx);
                if (ret)
                        return ret;
        }
 
-       ret = (*ctx->consume_chunk)(chunk, size, ctx->consume_chunk_ctx);
+       /* Consume the chunk.  */
+       ret = (*ctx->cbs.consume_chunk)(chunk, size, ctx->cbs.consume_chunk_ctx);
        if (ret)
                return ret;
-
        ctx->cur_stream_offset += size;
 
        if (ctx->cur_stream_offset == ctx->cur_stream->size) {
-               /* Finished reading all the data for a stream; advance
-                * to the next one.  */
-               ret = (*ctx->end_stream)(ctx->cur_stream, ctx->end_stream_ctx);
+               /* Finished reading all the data for a stream; advance to the
+                * next one.  */
+               DEBUG("End stream (size=%"PRIu64").", ctx->cur_stream->size);
+               ret = (*ctx->cbs.end_stream)(ctx->cur_stream, 0, ctx->cbs.end_stream_ctx);
                if (ret)
                        return ret;
 
-               if (ctx->cur_stream == ctx->final_stream)
-                       return 0;
+               if (ctx->cur_stream != ctx->final_stream) {
+                       struct list_head *cur = (struct list_head *)
+                                       ((u8*)ctx->cur_stream + ctx->list_head_offset);
+                       struct list_head *next = cur->next;
 
-               struct list_head *cur = (struct list_head *)
-                               ((u8*)ctx->cur_stream + ctx->list_head_offset);
-               struct list_head *next = cur->next;
+                       ctx->cur_stream = (struct wim_lookup_table_entry *)
+                                       ((u8*)next - ctx->list_head_offset);
 
-               ctx->cur_stream = (struct wim_lookup_table_entry *)
-                               ((u8*)next - ctx->list_head_offset);
+                       ctx->cur_stream_offset = 0;
+               } else {
+                       ctx->cur_stream = NULL;
+               }
+       }
+       return 0;
+}
 
-               ctx->cur_stream_offset = 0;
+struct hasher_context {
+       SHA_CTX sha_ctx;
+       struct read_stream_list_callbacks cbs;
+};
+
+/* Callback for starting to read a stream while calculating its SHA1 message
+ * digest.  */
+static int
+hasher_begin_stream(struct wim_lookup_table_entry *lte, bool is_partial_res,
+                   void *_ctx)
+{
+       struct hasher_context *ctx = _ctx;
+
+       sha1_init(&ctx->sha_ctx);
+
+       if (ctx->cbs.begin_stream == NULL)
+               return 0;
+       else
+               return (*ctx->cbs.begin_stream)(lte, is_partial_res,
+                                               ctx->cbs.begin_stream_ctx);
+}
+
+/* Callback for continuing to read a stream while calculating its SHA1 message
+ * digest.  */
+static int
+hasher_consume_chunk(const void *chunk, size_t size, void *_ctx)
+{
+       struct hasher_context *ctx = _ctx;
+
+       sha1_update(&ctx->sha_ctx, chunk, size);
+       if (ctx->cbs.consume_chunk == NULL)
+               return 0;
+       else
+               return (*ctx->cbs.consume_chunk)(chunk, size, ctx->cbs.consume_chunk_ctx);
+}
+
+/* Callback for finishing reading a stream while calculating its SHA1 message
+ * digest.  */
+static int
+hasher_end_stream(struct wim_lookup_table_entry *lte, int status, void *_ctx)
+{
+       struct hasher_context *ctx = _ctx;
+       u8 hash[SHA1_HASH_SIZE];
+       int ret;
+
+       if (status) {
+               ret = status;
+               goto out_next_cb;
+       }
+
+       sha1_final(hash, &ctx->sha_ctx);
+
+       if (lte->unhashed) {
+               /* No SHA1 message digest was present before; fill it in with
+                * the calculated value.  */
+               DEBUG("Set SHA1 message digest for stream (size=%"PRIu64").", lte->size);
+               copy_hash(lte->hash, hash);
+       } else {
+               /* A SHA1 message digest was present before.  Verify that it is
+                * the same as the calculated value.  */
+               if (!hashes_equal(hash, lte->hash)) {
+                       if (wimlib_print_errors) {
+                               ERROR("Invalid SHA1 message digest "
+                                     "on the following WIM stream:");
+                               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;
+                       errno = EINVAL;
+                       goto out_next_cb;
+               }
+               DEBUG("SHA1 message digest okay for stream (size=%"PRIu64").", lte->size);
+       }
+       ret = 0;
+out_next_cb:
+       if (ctx->cbs.end_stream == NULL)
+               return ret;
+       else
+               return (*ctx->cbs.end_stream)(lte, ret, ctx->cbs.end_stream_ctx);
+}
+
+/* Read the full data of the stream @lte, passing the data into the specified
+ * callbacks (all of which are optional) and either checking or computing the
+ * SHA1 message digest of the stream.  */
+static int
+read_full_stream_with_sha1(struct wim_lookup_table_entry *lte,
+                          const struct read_stream_list_callbacks *cbs)
+{
+       int ret;
+
+       struct hasher_context hasher_ctx = {
+               .cbs = *cbs,
+       };
+
+       ret = hasher_begin_stream(lte, false, &hasher_ctx);
+       if (ret)
+               return ret;
+
+       ret = read_stream_prefix(lte, lte->size, hasher_consume_chunk,
+                                &hasher_ctx, 0);
+
+       return hasher_end_stream(lte, ret, &hasher_ctx);
+}
+
+struct rechunkifier_context {
+       u8 *buffer;
+       u32 buffer_filled;
+       u32 cb_chunk_size;
+
+       const struct data_range *ranges;
+       size_t num_ranges;
+       size_t cur_range;
+       u64 range_bytes_remaining;
+
+       consume_data_callback_t cb;
+       void *cb_ctx;
+};
+
+/* Wrapper callback for adjusting the data chunk size.  */
+static int
+rechunkifier_cb(const void *chunk, size_t size, void *_ctx)
+{
+       struct rechunkifier_context *ctx = _ctx;
+       const u8 *chunkptr = chunk;
+       size_t bytes_to_copy;
+       int ret;
+
+       wimlib_assert(ctx->cur_range != ctx->num_ranges);
+
+       while (size) {
+
+               /* Append more data to the buffer.  */
+               bytes_to_copy = size;
+
+               if (bytes_to_copy > ctx->cb_chunk_size - ctx->buffer_filled)
+                       bytes_to_copy = ctx->cb_chunk_size - ctx->buffer_filled;
+
+               if (bytes_to_copy > ctx->range_bytes_remaining - ctx->buffer_filled)
+                       bytes_to_copy = ctx->range_bytes_remaining - ctx->buffer_filled;
+
+               memcpy(&ctx->buffer[ctx->buffer_filled], chunkptr, bytes_to_copy);
+
+               ctx->buffer_filled += bytes_to_copy;
+               chunkptr += bytes_to_copy;
+               size -= bytes_to_copy;
+               ctx->range_bytes_remaining -= bytes_to_copy;
+
+               if (ctx->buffer_filled == ctx->cb_chunk_size ||
+                   ctx->range_bytes_remaining == 0)
+               {
+                       /* Maximum chunk size reached, or current range ended.
+                        * Call the next consume_data_callback_t and empty the
+                        * buffer  */
+                       ret = (*ctx->cb)(ctx->buffer, ctx->buffer_filled, ctx->cb_ctx);
+                       if (ret)
+                               return ret;
+                       ctx->buffer_filled = 0;
+
+                       if (ctx->range_bytes_remaining == 0 &&
+                           ++ctx->cur_range != ctx->num_ranges)
+                               ctx->range_bytes_remaining = ctx->ranges[ctx->cur_range].size;
+               }
        }
        return 0;
 }
@@ -1088,18 +1167,11 @@ read_stream_list_wrapper_cb(const void *chunk, size_t size, void *_ctx)
  * @list_head_offset
  *     Offset of the `struct list_head' within each `struct
  *     wim_lookup_table_entry' that makes up the @stream_list.
- * @begin_stream
- *     Callback for starting to process a stream.
- * @consume_chunk
- *     Callback for receiving a chunk of stream data.
- * @end_stream
- *     Callback for finishing the processing of a stream.
  * @cb_chunk_size
  *     Size of chunks to provide to @consume_chunk.  For a given stream, all
  *     the chunks will be this size, except possibly the last which will be the
- *     remainder.
- * @cb_ctx
- *     Parameter to pass to the callback functions.
+ *     remainder.  If @
+ * @cbs TODO
  *
  * Returns 0 on success; a nonzero error code on failure.  Failure can occur due
  * to an error reading the data or due to an error status being returned by any
@@ -1108,11 +1180,8 @@ read_stream_list_wrapper_cb(const void *chunk, size_t size, void *_ctx)
 int
 read_stream_list(struct list_head *stream_list,
                 size_t list_head_offset,
-                read_stream_list_begin_stream_t begin_stream,
-                consume_data_callback_t consume_chunk,
-                read_stream_list_end_stream_t end_stream,
                 u32 cb_chunk_size,
-                void *cb_ctx)
+                const struct read_stream_list_callbacks *cbs)
 {
        int ret;
        struct list_head *cur, *next;
@@ -1162,6 +1231,9 @@ read_stream_list(struct list_head *stream_list,
                                 * read and @lte_last specifies the last stream
                                 * in the resource that needs to be read.  */
 
+                               DEBUG("Reading %zu streams combined in same "
+                                     "WIM resource", stream_count);
+
                                next = next2;
 
                                struct data_range ranges[stream_count];
@@ -1181,28 +1253,73 @@ read_stream_list(struct list_head *stream_list,
                                        }
                                }
 
-                               struct rechunker_context rechunker_ctx = {
-                                       .buffer = MALLOC(cb_chunk_size),
-                                       .buffer_filled = 0,
-                                       .cb_chunk_size = cb_chunk_size,
-                                       .ranges = ranges,
-                                       .num_ranges = stream_count,
-                                       .cur_range = 0,
-                                       .range_bytes_remaining = ranges[0].size,
-                                       .cb = consume_chunk,
-                                       .cb_ctx = cb_ctx,
-                               };
+                               /* Set up a chain of callbacks.
+                                *
+                                * The first level is the
+                                * streamifier_cb,
+                                * which takes in chunks of data and divides
+                                * them into the constituent streams.
+                                *
+                                * The second level are the SHA1 message digest
+                                * callbacks, which checksum each stream.
+                                *
+                                * rechunkifier_cb handles dividing the read
+                                * data into chunks of maximum size
+                                * @cb_chunk_size.  If @cb_chunk_size is 0, then
+                                * this callback is not needed.
+                                *
+                                * Finally, the last level of callbacks are
+                                * @cbs, passed as arguments to this function.
+                                */
+
+                               struct rechunkifier_context *rechunkifier_ctx = NULL;
+                               consume_data_callback_t last_cb;
+                               void *last_cb_ctx;
+
+                               if (cb_chunk_size != 0) {
+                                       rechunkifier_ctx = alloca(sizeof(*rechunkifier_ctx));
+                                       *rechunkifier_ctx = (struct rechunkifier_context) {
+                                               .buffer                 = MALLOC(cb_chunk_size),
+                                               .buffer_filled          = 0,
+                                               .cb_chunk_size          = cb_chunk_size,
+                                               .ranges                 = ranges,
+                                               .num_ranges             = stream_count,
+                                               .cur_range              = 0,
+                                               .range_bytes_remaining  = ranges[0].size,
+                                               .cb                     = cbs->consume_chunk,
+                                               .cb_ctx                 = cbs->consume_chunk_ctx,
+                                       };
+
+                                       if (rechunkifier_ctx->buffer == NULL)
+                                               return WIMLIB_ERR_NOMEM;
+                                       last_cb = rechunkifier_cb;
+                                       last_cb_ctx = rechunkifier_ctx;
+                               } else {
+                                       rechunkifier_ctx = NULL;
+                                       last_cb = cbs->consume_chunk;
+                                       last_cb_ctx = cbs->consume_chunk_ctx;
+                               }
 
-                               if (rechunker_ctx.buffer == NULL)
-                                       return WIMLIB_ERR_NOMEM;
+                               struct hasher_context hasher_ctx = {
+                                       .cbs = {
+                                               .begin_stream           = cbs->begin_stream,
+                                               .begin_stream_ctx       = cbs->begin_stream_ctx,
+                                               .consume_chunk          = last_cb,
+                                               .consume_chunk_ctx      = last_cb_ctx,
+                                               .end_stream             = cbs->end_stream,
+                                               .end_stream_ctx         = cbs->end_stream_ctx,
+                                       },
+                               };
 
-                               struct read_stream_list_ctx ctx = {
-                                       .begin_stream           = begin_stream,
-                                       .begin_stream_ctx       = cb_ctx,
-                                       .consume_chunk          = rechunker_cb,
-                                       .consume_chunk_ctx      = &rechunker_ctx,
-                                       .end_stream             = end_stream,
-                                       .end_stream_ctx         = cb_ctx,
+                               struct streamifier_context streamifier_ctx = {
+                                       .cbs = {
+                                               .begin_stream           = hasher_begin_stream,
+                                               .begin_stream_ctx       = &hasher_ctx,
+                                               .consume_chunk          = hasher_consume_chunk,
+                                               .consume_chunk_ctx      = &hasher_ctx,
+                                               .end_stream             = hasher_end_stream,
+                                               .end_stream_ctx         = &hasher_ctx,
+                                       },
                                        .cur_stream             = lte,
                                        .cur_stream_offset      = 0,
                                        .final_stream           = lte_last,
@@ -1212,133 +1329,88 @@ read_stream_list(struct list_head *stream_list,
                                ret = read_compressed_wim_resource(lte->rspec,
                                                                   ranges,
                                                                   stream_count,
-                                                                  read_stream_list_wrapper_cb,
-                                                                  &ctx,
+                                                                  streamifier_cb,
+                                                                  &streamifier_ctx,
                                                                   false);
-                               FREE(rechunker_ctx.buffer);
-                               if (ret)
+                               if (rechunkifier_ctx != NULL)
+                                       FREE(rechunkifier_ctx->buffer);
+
+                               if (ret) {
+                                       if (streamifier_ctx.cur_stream_offset != 0) {
+                                               ret = (*streamifier_ctx.cbs.end_stream)
+                                                       (streamifier_ctx.cur_stream,
+                                                        ret,
+                                                        streamifier_ctx.cbs.end_stream_ctx);
+                                       }
                                        return ret;
+                               }
                                continue;
                        }
                }
-               ret = (*begin_stream)(lte, cb_ctx);
-               if (ret)
-                       return ret;
-
-               ret = read_stream_prefix(lte, lte->size, consume_chunk,
-                                        cb_chunk_size, cb_ctx, 0);
-               if (ret)
-                       return ret;
 
-               ret = (*end_stream)(lte, cb_ctx);
-               if (ret)
+               ret = read_full_stream_with_sha1(lte, cbs);
+               if (ret > 0)
                        return ret;
        }
        return 0;
 }
 
-struct extract_ctx {
-       SHA_CTX sha_ctx;
-       consume_data_callback_t extract_chunk;
-       void *extract_chunk_arg;
-};
-
-static int
-extract_chunk_sha1_wrapper(const void *chunk, size_t chunk_size, void *_ctx)
-{
-       struct extract_ctx *ctx = _ctx;
-
-       sha1_update(&ctx->sha_ctx, chunk, chunk_size);
-       return ctx->extract_chunk(chunk, chunk_size, ctx->extract_chunk_arg);
-}
-
 /* Extracts the first @size bytes of a stream to somewhere.  In the process, the
  * SHA1 message digest of the uncompressed stream is checked if the full stream
  * is being extracted.
  *
- * @extract_chunk is a function that will be called to extract each chunk of the
- * stream.  */
+ * @extract_chunk is the callback to extract each chunk of the stream.  */
 int
-extract_stream(const struct wim_lookup_table_entry *lte, u64 size,
+extract_stream(struct wim_lookup_table_entry *lte, u64 size,
               consume_data_callback_t extract_chunk, void *extract_chunk_arg)
 {
-       int ret;
        if (size == lte->size) {
-               /* 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_stream_prefix(lte, size,
-                                        extract_chunk_sha1_wrapper,
-                                        lte_cchunk_size(lte),
-                                        &ctx, 0);
-               if (ret == 0) {
-                       u8 hash[SHA1_HASH_SIZE];
-                       sha1_final(hash, &ctx.sha_ctx);
-                       if (!hashes_equal(hash, lte->hash)) {
-                               if (wimlib_print_errors) {
-                                       ERROR("Invalid SHA1 message digest "
-                                             "on the following WIM stream:");
-                                       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;
-                       }
-               }
+               /* Do SHA1.  */
+               struct read_stream_list_callbacks cbs = {
+                       .consume_chunk          = extract_chunk,
+                       .consume_chunk_ctx      = extract_chunk_arg,
+               };
+               return read_full_stream_with_sha1(lte, &cbs);
        } else {
-               /* Don't do SHA1 */
-               ret = read_stream_prefix(lte, size, extract_chunk,
-                                        lte_cchunk_size(lte),
-                                        extract_chunk_arg, 0);
+               /* Don't do SHA1.  */
+               return read_stream_prefix(lte, size, extract_chunk,
+                                         extract_chunk_arg, 0);
        }
-       return ret;
 }
 
-static int
-extract_wim_chunk_to_fd(const void *buf, size_t len, void *_fd_p)
+/* Write a chunk of data to a file descriptor.  This function can be passed as a
+ * consume_data_callback_t.  */
+int
+extract_chunk_to_fd(const void *chunk, size_t size, void *_fd_p)
 {
        struct filedes *fd = _fd_p;
-       int ret = full_write(fd, buf, len);
-       if (ret)
+
+       int ret = full_write(fd, chunk, size);
+       if (ret) {
                ERROR_WITH_ERRNO("Error writing to file descriptor");
-       return ret;
+               return ret;
+       }
+
+       return 0;
 }
 
 /* Extract the first @size bytes of the specified stream to the specified file
  * descriptor.  If @size is the full size of the stream, its SHA1 message digest
  * is also checked.  */
 int
-extract_stream_to_fd(const struct wim_lookup_table_entry *lte,
+extract_stream_to_fd(struct wim_lookup_table_entry *lte,
                     struct filedes *fd, u64 size)
 {
-       return extract_stream(lte, size, extract_wim_chunk_to_fd, fd);
-}
-
-
-static int
-sha1_chunk(const void *buf, size_t len, void *ctx)
-{
-       sha1_update(ctx, buf, len);
-       return 0;
+       return extract_stream(lte, size, extract_chunk_to_fd, fd);
 }
 
 /* Calculate the SHA1 message digest of a stream, storing it in @lte->hash.  */
 int
 sha1_stream(struct wim_lookup_table_entry *lte)
 {
-       int ret;
-       SHA_CTX sha_ctx;
-
-       sha1_init(&sha_ctx);
-       ret = read_stream_prefix(lte, lte->size,
-                                sha1_chunk, lte_cchunk_size(lte),
-                                &sha_ctx, 0);
-       if (ret == 0)
-               sha1_final(lte->hash, &sha_ctx);
-
-       return ret;
+       struct read_stream_list_callbacks cbs = {
+       };
+       return read_full_stream_with_sha1(lte, &cbs);
 }
 
 /* Convert a WIM resource header to a stand-alone resource specification.  */
@@ -1390,7 +1462,7 @@ get_wim_reshdr(const struct wim_reshdr_disk *disk_reshdr,
        reshdr->uncompressed_size = le64_to_cpu(disk_reshdr->uncompressed_size);
        reshdr->flags = disk_reshdr->flags;
 
-       /* Truncate numbers to 62 bits to avoid possible overflows.  */
+       /* Avoid possible overflows.  */
        if (reshdr->offset_in_wim & 0xc000000000000000ULL)
                return WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
 
index 63c91b2..a3f4213 100644 (file)
@@ -532,7 +532,7 @@ try_write_again:
        else
                in_chunk_size = out_chunk_size;
        ret = read_stream_prefix(lte, read_size, write_resource_cb,
-                                in_chunk_size, &write_ctx, resource_flags);
+                                &write_ctx, resource_flags);
        if (ret)
                goto out_free_chunk_tab;
 
@@ -1480,8 +1480,7 @@ submit_stream_for_compression(struct wim_lookup_table_entry *lte,
        ctx->next_lte = lte;
        INIT_LIST_HEAD(&lte->msg_list);
        list_add_tail(&lte->being_compressed_list, &ctx->outstanding_streams);
-       ret = read_stream_prefix(lte, lte->size, main_writer_thread_cb,
-                                ctx->out_chunk_size, ctx, 0);
+       ret = read_stream_prefix(lte, lte->size, main_writer_thread_cb, ctx, 0);
        if (ret)
                return ret;
        wimlib_assert(ctx->next_chunk == ctx->next_num_chunks);