]> wimlib.net Git - wimlib/blobdiff - src/lookup_table.c
Cache compression type in WIMStruct and wim_lookup_table_entry
[wimlib] / src / lookup_table.c
index 998eb185f1586e8af2791a58d9bb11fd82247cc9..ba832453e6750517aba7b98014426329db594ba5 100644 (file)
  * along with wimlib; if not, see http://www.gnu.org/licenses/.
  */
 
-#include "wimlib_internal.h"
-#include "lookup_table.h"
-#include "buffer_io.h"
+#ifdef HAVE_CONFIG_H
+#  include "config.h"
+#endif
+
+#include "wimlib/endianness.h"
+#include "wimlib/error.h"
+#include "wimlib/file_io.h"
+#include "wimlib/lookup_table.h"
+#include "wimlib/metadata.h"
+#include "wimlib/paths.h"
+#include "wimlib/resource.h"
+#include "wimlib/util.h"
+
 #include <errno.h>
 #include <stdlib.h>
-
 #ifdef WITH_FUSE
-#include <unistd.h>
+#  include <unistd.h> /* for unlink() */
 #endif
 
 struct wim_lookup_table *
@@ -58,7 +67,7 @@ new_lookup_table(size_t capacity)
 }
 
 struct wim_lookup_table_entry *
-new_lookup_table_entry()
+new_lookup_table_entry(void)
 {
        struct wim_lookup_table_entry *lte;
 
@@ -79,11 +88,10 @@ clone_lookup_table_entry(const struct wim_lookup_table_entry *old)
 {
        struct wim_lookup_table_entry *new;
 
-       new = MALLOC(sizeof(*new));
+       new = memdup(old, sizeof(struct wim_lookup_table_entry));
        if (!new)
                return NULL;
 
-       memcpy(new, old, sizeof(*old));
        new->extracted_file = NULL;
        switch (new->resource_location) {
 #ifdef __WIN32__
@@ -102,32 +110,30 @@ clone_lookup_table_entry(const struct wim_lookup_table_entry *old)
                        goto out_free;
                break;
        case RESOURCE_IN_ATTACHED_BUFFER:
-               new->attached_buffer = MALLOC(wim_resource_size(old));
+               new->attached_buffer = memdup(old->attached_buffer,
+                                             wim_resource_size(old));
                if (!new->attached_buffer)
                        goto out_free;
-               memcpy(new->attached_buffer, old->attached_buffer,
-                      wim_resource_size(old));
                break;
 #ifdef WITH_NTFS_3G
        case RESOURCE_IN_NTFS_VOLUME:
                if (old->ntfs_loc) {
                        struct ntfs_location *loc;
-                       loc = MALLOC(sizeof(*loc));
+                       loc = memdup(old->ntfs_loc, sizeof(struct ntfs_location));
                        if (!loc)
                                goto out_free;
-                       memcpy(loc, old->ntfs_loc, sizeof(*loc));
                        loc->path = NULL;
                        loc->stream_name = NULL;
                        new->ntfs_loc = loc;
                        loc->path = STRDUP(old->ntfs_loc->path);
                        if (!loc->path)
                                goto out_free;
-                       loc->stream_name = MALLOC((loc->stream_name_nchars + 1) * 2);
-                       if (!loc->stream_name)
-                               goto out_free;
-                       memcpy(loc->stream_name,
-                              old->ntfs_loc->stream_name,
-                              (loc->stream_name_nchars + 1) * 2);
+                       if (loc->stream_name_nchars) {
+                               loc->stream_name = memdup(old->ntfs_loc->stream_name,
+                                                         loc->stream_name_nchars * 2);
+                               if (!loc->stream_name)
+                                       goto out_free;
+                       }
                }
                break;
 #endif
@@ -342,6 +348,25 @@ for_lookup_table_entry_pos_sorted(struct wim_lookup_table *table,
        return ret;
 }
 
+/* On-disk format of a WIM lookup table entry (stream entry). */
+struct wim_lookup_table_entry_disk {
+       /* Location, offset, compression status, and metadata status of the
+        * stream. */
+       struct resource_entry_disk resource_entry;
+
+       /* Which part of the split WIM this stream is in; indexed from 1. */
+       le16 part_number;
+
+       /* Reference count of this stream over all WIM images. */
+       le32 refcnt;
+
+       /* SHA1 message digest of the uncompressed data of this stream, or
+        * optionally all zeroes if this stream is of zero length. */
+       u8 hash[SHA1_HASH_SIZE];
+} _packed_attribute;
+
+#define WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE 50
+
 /*
  * Reads the lookup table from a WIM file.
  *
@@ -356,11 +381,16 @@ read_lookup_table(WIMStruct *w)
        size_t num_entries;
        struct wim_lookup_table *table;
        struct wim_lookup_table_entry *cur_entry, *duplicate_entry;
-       u8 table_buf[(BUFFER_SIZE / WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE) *
-                       WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE];
-       const u8 *p;
+       struct wim_lookup_table_entry_disk
+                       table_buf[BUFFER_SIZE / sizeof(struct wim_lookup_table_entry_disk)]
+                               _aligned_attribute(8);
+
+       BUILD_BUG_ON(sizeof(struct wim_lookup_table_entry_disk) !=
+                    WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE);
+
        off_t offset;
        size_t buf_entries_remaining;
+       const struct wim_lookup_table_entry_disk *disk_entry;
 
        DEBUG("Reading lookup table: offset %"PRIu64", size %"PRIu64"",
              w->hdr.lookup_table_res_entry.offset,
@@ -373,23 +403,25 @@ read_lookup_table(WIMStruct *w)
        }
 
        num_entries = w->hdr.lookup_table_res_entry.size /
-                     WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE;
+                     sizeof(struct wim_lookup_table_entry_disk);
        table = new_lookup_table(num_entries * 2 + 1);
-       if (!table)
+       if (!table) {
+               ERROR("Failed to allocate stream hash table of size %zu",
+                     num_entries * 2 + 1);
                return WIMLIB_ERR_NOMEM;
+       }
 
        w->current_image = 0;
        offset = w->hdr.lookup_table_res_entry.offset;
        buf_entries_remaining = 0;
-       for (; num_entries != 0; num_entries--, buf_entries_remaining--) {
+       for (; num_entries != 0;
+            num_entries--, buf_entries_remaining--, disk_entry++)
+       {
                if (buf_entries_remaining == 0) {
                        size_t entries_to_read, bytes_to_read;
 
-                       entries_to_read = min(sizeof(table_buf) /
-                                               WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE,
-                                             num_entries);
-                       bytes_to_read = entries_to_read *
-                                               WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE;
+                       entries_to_read = min(ARRAY_LEN(table_buf), num_entries);
+                       bytes_to_read = entries_to_read * sizeof(struct wim_lookup_table_entry_disk);
                        if (full_pread(w->in_fd, table_buf,
                                       bytes_to_read, offset) != bytes_to_read)
                        {
@@ -399,7 +431,7 @@ read_lookup_table(WIMStruct *w)
                                goto out_free_lookup_table;
                        }
                        offset += bytes_to_read;
-                       p = table_buf;
+                       disk_entry = table_buf;
                        buf_entries_remaining = entries_to_read;
                }
                cur_entry = new_lookup_table_entry();
@@ -410,63 +442,77 @@ read_lookup_table(WIMStruct *w)
 
                cur_entry->wim = w;
                cur_entry->resource_location = RESOURCE_IN_WIM;
-               p = get_resource_entry(p, &cur_entry->resource_entry);
-               p = get_u16(p, &cur_entry->part_number);
-               p = get_u32(p, &cur_entry->refcnt);
-               p = get_bytes(p, SHA1_HASH_SIZE, cur_entry->hash);
+               get_resource_entry(&disk_entry->resource_entry, &cur_entry->resource_entry);
+               cur_entry->part_number = le16_to_cpu(disk_entry->part_number);
+               cur_entry->refcnt = le32_to_cpu(disk_entry->refcnt);
+               copy_hash(cur_entry->hash, disk_entry->hash);
+
+               if (cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED)
+                       cur_entry->compression_type = w->compression_type;
+               else
+                       BUILD_BUG_ON(WIMLIB_COMPRESSION_TYPE_NONE != 0);
 
                if (cur_entry->part_number != w->hdr.part_number) {
-                       ERROR("A lookup table entry in part %hu of the WIM "
-                             "points to part %hu",
-                             w->hdr.part_number, cur_entry->part_number);
-                       ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
-                       goto out_free_cur_entry;
+                       WARNING("A lookup table entry in part %hu of the WIM "
+                               "points to part %hu (ignoring it)",
+                               w->hdr.part_number, cur_entry->part_number);
+                       free_lookup_table_entry(cur_entry);
+                       continue;
                }
 
                if (is_zero_hash(cur_entry->hash)) {
-                       ERROR("The WIM lookup table contains an entry with a "
-                             "SHA1 message digest of all 0's");
-                       ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
-                       goto out_free_cur_entry;
+                       WARNING("The WIM lookup table contains an entry with a "
+                               "SHA1 message digest of all 0's (ignoring it)");
+                       free_lookup_table_entry(cur_entry);
+                       continue;
                }
 
                if (!(cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_COMPRESSED)
                    && (cur_entry->resource_entry.size !=
                        cur_entry->resource_entry.original_size))
                {
-               #ifdef ENABLE_ERROR_MESSAGES
-                       ERROR("Found uncompressed resource with original size "
-                             "not the same as compressed size");
-                       ERROR("The lookup table entry for the resource is as follows:");
-                       print_lookup_table_entry(cur_entry, stderr);
-               #endif
-                       ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
-                       goto out_free_cur_entry;
+                       if (wimlib_print_errors) {
+                               WARNING("Found uncompressed resource with "
+                                       "original size (%"PRIu64") not the same "
+                                       "as compressed size (%"PRIu64")",
+                                       cur_entry->resource_entry.original_size,
+                                       cur_entry->resource_entry.size);
+                               if (cur_entry->resource_entry.original_size) {
+                                       WARNING("Overriding compressed size with original size.");
+                                       cur_entry->resource_entry.size =
+                                               cur_entry->resource_entry.original_size;
+                               } else {
+                                       WARNING("Overriding original size with compressed size");
+                                       cur_entry->resource_entry.original_size =
+                                               cur_entry->resource_entry.size;
+                               }
+                       }
                }
 
                if (cur_entry->resource_entry.flags & WIM_RESHDR_FLAG_METADATA) {
                        /* Lookup table entry for a metadata resource */
                        if (cur_entry->refcnt != 1) {
-                       #ifdef ENABLE_ERROR_MESSAGES
-                               ERROR("Found metadata resource with refcnt != 1:");
-                               print_lookup_table_entry(cur_entry, stderr);
-                       #endif
+                               if (wimlib_print_errors) {
+                                       ERROR("Found metadata resource with refcnt != 1:");
+                                       print_lookup_table_entry(cur_entry, stderr);
+                               }
                                ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
                                goto out_free_cur_entry;
                        }
 
                        if (w->hdr.part_number != 1) {
-                               ERROR("Found a metadata resource in a "
-                                     "non-first part of the split WIM!");
-                               ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
-                               goto out_free_cur_entry;
+                               WARNING("Ignoring metadata resource found in a "
+                                       "non-first part of the split WIM");
+                               free_lookup_table_entry(cur_entry);
+                               continue;
                        }
                        if (w->current_image == w->hdr.image_count) {
-                               ERROR("The WIM header says there are %u images "
-                                     "in the WIM, but we found more metadata "
-                                     "resources than this", w->hdr.image_count);
-                               ret = WIMLIB_ERR_IMAGE_COUNT;
-                               goto out_free_cur_entry;
+                               WARNING("The WIM header says there are %u images "
+                                       "in the WIM, but we found more metadata "
+                                       "resources than this (ignoring the extra)",
+                                       w->hdr.image_count);
+                               free_lookup_table_entry(cur_entry);
+                               continue;
                        }
 
                        /* Notice very carefully:  We are assigning the metadata
@@ -487,18 +533,19 @@ read_lookup_table(WIMStruct *w)
                         * metadata resource */
                        duplicate_entry = __lookup_resource(table, cur_entry->hash);
                        if (duplicate_entry) {
-                       #ifdef ENABLE_ERROR_MESSAGES
-                               ERROR("The WIM lookup table contains two entries with the "
-                                     "same SHA1 message digest!");
-                               ERROR("The first entry is:");
-                               print_lookup_table_entry(duplicate_entry, stderr);
-                               ERROR("The second entry is:");
-                               print_lookup_table_entry(cur_entry, stderr);
-                       #endif
-                               ret = WIMLIB_ERR_INVALID_LOOKUP_TABLE_ENTRY;
-                               goto out_free_cur_entry;
+                               if (wimlib_print_errors) {
+                                       WARNING("The WIM lookup table contains two entries with the "
+                                             "same SHA1 message digest!");
+                                       WARNING("The first entry is:");
+                                       print_lookup_table_entry(duplicate_entry, stderr);
+                                       WARNING("The second entry is:");
+                                       print_lookup_table_entry(cur_entry, stderr);
+                               }
+                               free_lookup_table_entry(cur_entry);
+                               continue;
+                       } else {
+                               lookup_table_insert(table, cur_entry);
                        }
-                       lookup_table_insert(table, cur_entry);
                }
        }
 
@@ -524,14 +571,14 @@ out:
 }
 
 
-static u8 *
-write_lookup_table_entry(const struct wim_lookup_table_entry *lte, u8 *buf_p)
+static void
+write_lookup_table_entry(const struct wim_lookup_table_entry *lte,
+                        struct wim_lookup_table_entry_disk *disk_entry)
 {
-       buf_p = put_resource_entry(buf_p, &lte->output_resource_entry);
-       buf_p = put_u16(buf_p, lte->part_number);
-       buf_p = put_u32(buf_p, lte->out_refcnt);
-       buf_p = put_bytes(buf_p, SHA1_HASH_SIZE, lte->hash);
-       return buf_p;
+       put_resource_entry(&lte->output_resource_entry, &disk_entry->resource_entry);
+       disk_entry->part_number = cpu_to_le16(lte->part_number);
+       disk_entry->refcnt = cpu_to_le32(lte->out_refcnt);
+       copy_hash(disk_entry->hash, lte->hash);
 }
 
 int
@@ -541,32 +588,34 @@ write_lookup_table_from_stream_list(struct list_head *stream_list,
 {
        int ret;
        off_t start_offset;
-       u8 table_buf[(BUFFER_SIZE / WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE) *
-                       WIM_LOOKUP_TABLE_ENTRY_DISK_SIZE];
-       u8 *buf_p;
+       struct wim_lookup_table_entry_disk
+                       table_buf[BUFFER_SIZE / sizeof(struct wim_lookup_table_entry_disk)]
+                               _aligned_attribute(8);
        size_t table_size;
        size_t bytes_to_write;
        struct wim_lookup_table_entry *lte;
+       size_t cur_idx;
 
        start_offset = filedes_offset(out_fd);
        if (start_offset == -1)
                goto write_error;
 
-       buf_p = table_buf;
        table_size = 0;
+       cur_idx = 0;
        list_for_each_entry(lte, stream_list, lookup_table_list) {
-               if (buf_p == table_buf + sizeof(table_buf)) {
+               if (cur_idx == ARRAY_LEN(table_buf)) {
                        bytes_to_write = sizeof(table_buf);
                        if (full_write(out_fd, table_buf,
                                       bytes_to_write) != bytes_to_write)
                                goto write_error;
                        table_size += bytes_to_write;
-                       buf_p = table_buf;
+                       cur_idx = 0;
                }
-               buf_p = write_lookup_table_entry(lte, buf_p);
+               write_lookup_table_entry(lte, &table_buf[cur_idx]);
+               cur_idx++;
        }
-       bytes_to_write = buf_p - table_buf;
-       if (bytes_to_write != 0) {
+       if (cur_idx != 0) {
+               bytes_to_write = cur_idx * sizeof(struct wim_lookup_table_entry_disk);
                if (full_write(out_fd, table_buf,
                               bytes_to_write) != bytes_to_write)
                        goto write_error;
@@ -727,9 +776,11 @@ do_print_lookup_table_entry(struct wim_lookup_table_entry *lte, void *fp)
  * Prints the lookup table of a WIM file.
  */
 WIMLIBAPI void
-wimlib_print_lookup_table(WIMStruct *w)
+wimlib_print_lookup_table(WIMStruct *wim)
 {
-       for_lookup_table_entry(w->lookup_table,
+       for (int i = 0; i < wim->hdr.image_count; i++)
+               print_lookup_table_entry(wim->image_metadata[i]->metadata_lte, stdout);
+       for_lookup_table_entry(wim->lookup_table,
                               do_print_lookup_table_entry,
                               stdout);
 }
@@ -791,7 +842,9 @@ lookup_resource(WIMStruct *w,
 
        inode = dentry->d_inode;
 
-       wimlib_assert(inode->i_resolved);
+       if (!inode->i_resolved)
+               if (inode_resolve_ltes(inode, w->lookup_table))
+                       return -EIO;
 
        if (!(lookup_flags & LOOKUP_FLAG_DIRECTORY_OK)
              && inode_is_directory(inode))
@@ -833,24 +886,52 @@ out:
  * This function always succeeds; unresolved lookup table entries are given a
  * NULL pointer.
  */
-void
+int
 inode_resolve_ltes(struct wim_inode *inode, struct wim_lookup_table *table)
 {
+       const u8 *hash;
 
        if (!inode->i_resolved) {
-               struct wim_lookup_table_entry *lte;
+               struct wim_lookup_table_entry *lte, *ads_lte;
+
                /* Resolve the default file stream */
-               lte = __lookup_resource(table, inode->i_hash);
-               inode->i_lte = lte;
-               inode->i_resolved = 1;
+               lte = NULL;
+               hash = inode->i_hash;
+               if (!is_zero_hash(hash)) {
+                       lte = __lookup_resource(table, hash);
+                       if (unlikely(!lte))
+                               goto resource_not_found;
+               }
 
                /* Resolve the alternate data streams */
+               struct wim_lookup_table_entry *ads_ltes[inode->i_num_ads];
                for (u16 i = 0; i < inode->i_num_ads; i++) {
-                       struct wim_ads_entry *cur_entry = &inode->i_ads_entries[i];
-                       lte = __lookup_resource(table, cur_entry->hash);
-                       cur_entry->lte = lte;
+                       struct wim_ads_entry *cur_entry;
+
+                       ads_lte = NULL;
+                       cur_entry = &inode->i_ads_entries[i];
+                       hash = cur_entry->hash;
+                       if (!is_zero_hash(hash)) {
+                               ads_lte = __lookup_resource(table, hash);
+                               if (unlikely(!ads_lte))
+                                       goto resource_not_found;
+                       }
+                       ads_ltes[i] = ads_lte;
                }
+               inode->i_lte = lte;
+               for (u16 i = 0; i < inode->i_num_ads; i++)
+                       inode->i_ads_entries[i].lte = ads_ltes[i];
+               inode->i_resolved = 1;
+       }
+       return 0;
+resource_not_found:
+       if (wimlib_print_errors) {
+               ERROR("\"%"TS"\": resource not found", inode_first_full_path(inode));
+               tfprintf(stderr, T("        SHA-1 message digest of missing resource:\n        "));
+               print_hash(hash, stderr);
+               tputc(T('\n'), stderr);
        }
+       return WIMLIB_ERR_RESOURCE_NOT_FOUND;
 }
 
 void