]> wimlib.net Git - wimlib/blobdiff - src/dentry.c
Fixed bug when capturing NTFS file with multiple named data streams
[wimlib] / src / dentry.c
index 471552e6acf8d1b2c2455bc29786930c30e36052..1f0fd86c7e047a4121757ed12c84318aa21e99c6 100644 (file)
  */
 
 /*
- *
- * Copyright (C) 2010 Carl Thijssen
  * Copyright (C) 2012 Eric Biggers
  *
  * This file is part of wimlib, a library for working with WIM files.
  *
  * wimlib is free software; you can redistribute it and/or modify it under the
- * terms of the GNU Lesser General Public License as published by the Free
- * Software Foundation; either version 2.1 of the License, or (at your option)
- * any later version.
+ * terms of the GNU General Public License as published by the Free Software
+ * Foundation; either version 3 of the License, or (at your option) any later
+ * version.
  *
  * wimlib is distributed in the hope that it will be useful, but WITHOUT ANY
  * WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR
- * A PARTICULAR PURPOSE. See the GNU Lesser General Public License for more
- * details.
+ * A PARTICULAR PURPOSE. See the GNU General Public License for more details.
  *
- * You should have received a copy of the GNU Lesser General Public License
- * along with wimlib; if not, see http://www.gnu.org/licenses/.
+ * You should have received a copy of the GNU General Public License along with
+ * wimlib; if not, see http://www.gnu.org/licenses/.
  */
 
 #include <errno.h>
@@ -54,14 +51,48 @@ static bool dentry_has_name(const struct dentry *dentry, const char *name,
        return memcmp(dentry->file_name_utf8, name, name_len) == 0;
 }
 
+static u64 __dentry_correct_length_unaligned(u16 file_name_len,
+                                            u16 short_name_len)
+{
+       u64 length = WIM_DENTRY_DISK_SIZE;
+       if (file_name_len)
+               length += file_name_len + 2;
+       if (short_name_len)
+               length += short_name_len + 2;
+       return length;
+}
+
+static u64 dentry_correct_length_unaligned(const struct dentry *dentry)
+{
+       return __dentry_correct_length_unaligned(dentry->file_name_len,
+                                                dentry->short_name_len);
+}
+
+/* Return the "correct" value to write in the length field of the dentry, based
+ * on the file name length and short name length */
+static u64 dentry_correct_length(const struct dentry *dentry)
+{
+       return (dentry_correct_length_unaligned(dentry) + 7) & ~7;
+}
+
+static u64 __dentry_total_length(const struct dentry *dentry, u64 length)
+{
+       for (u16 i = 0; i < dentry->num_ads; i++)
+               length += ads_entry_total_length(&dentry->ads_entries[i]);
+       return (length + 7) & ~7;
+}
+
+u64 dentry_correct_total_length(const struct dentry *dentry)
+{
+       return __dentry_total_length(dentry,
+                                    dentry_correct_length_unaligned(dentry));
+}
+
 /* Real length of a dentry, including the alternate data stream entries, which
  * are not included in the dentry->length field... */
 u64 dentry_total_length(const struct dentry *dentry)
 {
-       u64 length = (dentry->length + 7) & ~7;
-       for (u16 i = 0; i < dentry->num_ads; i++)
-               length += ads_entry_total_length(&dentry->ads_entries[i]);
-       return length;
+       return __dentry_total_length(dentry, dentry->length);
 }
 
 /* Transfers file attributes from a `stat' buffer to a struct dentry. */
@@ -76,9 +107,9 @@ void stbuf_to_dentry(const struct stat *stbuf, struct dentry *dentry)
                dentry->attributes = FILE_ATTRIBUTE_NORMAL;
        }
        if (sizeof(ino_t) >= 8)
-               dentry->hard_link = (u64)stbuf->st_ino;
+               dentry->link_group_id = (u64)stbuf->st_ino;
        else
-               dentry->hard_link = (u64)stbuf->st_ino |
+               dentry->link_group_id = (u64)stbuf->st_ino |
                                   ((u64)stbuf->st_dev << (sizeof(ino_t) * 8));
        /* Set timestamps */
        dentry->creation_time = timespec_to_wim_timestamp(&stbuf->st_mtim);
@@ -136,19 +167,51 @@ struct ads_entry *dentry_add_ads(struct dentry *dentry, const char *stream_name)
        DEBUG("Add alternate data stream %s:%s",
               dentry->file_name_utf8, stream_name);
 
-       if (dentry->num_ads == 0xffff)
+       if (dentry->num_ads == 0xffff) {
+               ERROR("Too many alternate data streams in one dentry!");
                return NULL;
+       }
        num_ads = dentry->num_ads + 1;
        ads_entries = REALLOC(dentry->ads_entries,
-                             num_ads * sizeof(struct ads_entry));
-       if (!ads_entries)
+                             num_ads * sizeof(dentry->ads_entries[0]));
+       if (!ads_entries) {
+               ERROR("Failed to allocate memory for new alternate data stream");
                return NULL;
+       }
+       wimlib_assert(ads_entries == dentry->ads_entries ||
+                       ads_entries < dentry->ads_entries ||
+                       ads_entries > dentry->ads_entries + dentry->num_ads);
        if (ads_entries != dentry->ads_entries) {
                /* We moved the ADS entries.  Adjust the stream lists. */
                for (u16 i = 0; i < dentry->num_ads; i++) {
                        struct list_head *cur = &ads_entries[i].lte_group_list.list;
-                       cur->prev->next = cur;
-                       cur->next->prev = cur;
+                       struct list_head *prev = cur->prev;
+                       struct list_head *next = cur->next;
+                       if ((u8*)prev >= (u8*)dentry->ads_entries
+                           && (u8*)prev < (u8*)(dentry->ads_entries + dentry->num_ads)) {
+                               /* Previous entry was located in the same ads_entries
+                                * array!  Adjust our own prev pointer. */
+                               u16 idx = (struct ads_entry*)prev -
+                                           (struct ads_entry*)&dentry->ads_entries[0].lte_group_list.list;
+                               cur->prev = &ads_entries[idx].lte_group_list.list;
+                       } else {
+                               /* Previous entry is located in a different ads_entries
+                                * array.  Adjust its next pointer. */
+                               prev->next = cur;
+                       }
+                       next = cur->next;
+                       if ((u8*)next >= (u8*)dentry->ads_entries
+                           && (u8*)next < (u8*)(dentry->ads_entries + dentry->num_ads)) {
+                               /* Next entry was located in the same ads_entries array!
+                                * Adjust our own next pointer. */
+                               u16 idx = (struct ads_entry*)next -
+                                           (struct ads_entry*)&dentry->ads_entries[0].lte_group_list.list;
+                               cur->next = &ads_entries[idx].lte_group_list.list;
+                       } else {
+                               /* Next entry is located in a different ads_entries
+                                * array.  Adjust its prev pointer. */
+                               next->prev = cur;
+                       }
                }
        }
 
@@ -313,7 +376,7 @@ void calculate_subdir_offsets(struct dentry *dentry, u64 *subdir_offset_p)
                /* Advance the subdir offset by the amount of space the children
                 * of this dentry take up. */
                do {
-                       *subdir_offset_p += dentry_total_length(child);
+                       *subdir_offset_p += dentry_correct_total_length(child);
                        child = child->next;
                } while (child != dentry->children);
 
@@ -445,6 +508,8 @@ int print_dentry(struct dentry *dentry, void *lookup_table)
 {
        const u8 *hash;
        struct lookup_table_entry *lte;
+       time_t time;
+       char *p;
 
        printf("[DENTRY]\n");
        printf("Length            = %"PRIu64"\n", dentry->length);
@@ -466,15 +531,23 @@ int print_dentry(struct dentry *dentry, void *lookup_table)
 #endif
 
        /* Translate the timestamps into something readable */
-       time_t creat_time = wim_timestamp_to_unix(dentry->creation_time);
-       time_t access_time = wim_timestamp_to_unix(dentry->last_access_time);
-       time_t mod_time = wim_timestamp_to_unix(dentry->last_write_time);
-       printf("Creation Time     = %s", asctime(gmtime(&creat_time)));
-       printf("Last Access Time  = %s", asctime(gmtime(&access_time)));
-       printf("Last Write Time   = %s", asctime(gmtime(&mod_time)));
+       time = wim_timestamp_to_unix(dentry->creation_time);
+       p = asctime(gmtime(&time));
+       *(strrchr(p, '\n')) = '\0';
+       printf("Creation Time     = %s UTC\n", p);
+
+       time = wim_timestamp_to_unix(dentry->last_access_time);
+       p = asctime(gmtime(&time));
+       *(strrchr(p, '\n')) = '\0';
+       printf("Last Access Time  = %s UTC\n", p);
+
+       time = wim_timestamp_to_unix(dentry->last_write_time);
+       p = asctime(gmtime(&time));
+       *(strrchr(p, '\n')) = '\0';
+       printf("Last Write Time   = %s UTC\n", p);
 
        printf("Reparse Tag       = 0x%"PRIx32"\n", dentry->reparse_tag);
-       printf("Hard Link Group   = 0x%"PRIx64"\n", dentry->hard_link);
+       printf("Hard Link Group   = 0x%"PRIx64"\n", dentry->link_group_id);
        printf("Number of Alternate Data Streams = %hu\n", dentry->num_ads);
        printf("Filename          = \"");
        print_string(dentry->file_name, dentry->file_name_len);
@@ -550,6 +623,7 @@ struct dentry *new_dentry(const char *name)
        dentry->prev   = dentry;
        dentry->parent = dentry;
        INIT_LIST_HEAD(&dentry->link_group_list);
+       INIT_LIST_HEAD(&dentry->lte_group_list.list);
        return dentry;
 err:
        FREE(dentry);
@@ -572,7 +646,8 @@ static void __destroy_dentry(struct dentry *dentry)
        FREE(dentry->file_name_utf8);
        FREE(dentry->short_name);
        FREE(dentry->full_path_utf8);
-       FREE(dentry->extracted_file);
+       if (dentry->extracted_file != dentry->full_path_utf8)
+               FREE(dentry->extracted_file);
 }
 
 /* Frees a WIM dentry. */
@@ -603,7 +678,6 @@ void put_dentry(struct dentry *dentry)
                }
                dentry->ads_entries_status = ADS_ENTRIES_USER;
        }
-       struct list_head *next;
        list_del(&dentry->link_group_list);
        free_dentry(dentry);
 }
@@ -722,17 +796,6 @@ void unlink_dentry(struct dentry *dentry)
        }
 }
 
-
-/* Recalculates the length of @dentry based on its file name length and short
- * name length.  */
-static inline void recalculate_dentry_size(struct dentry *dentry)
-{
-       dentry->length = WIM_DENTRY_DISK_SIZE + dentry->file_name_len + 
-                        2 + dentry->short_name_len;
-       /* Must be multiple of 8. */
-       dentry->length = (dentry->length + 7) & ~7;
-}
-
 /* Duplicates a UTF-8 name into UTF-8 and UTF-16 strings and returns the strings
  * and their lengths in the pointer arguments */
 int get_names(char **name_utf16_ret, char **name_utf8_ret,
@@ -775,8 +838,10 @@ int change_dentry_name(struct dentry *dentry, const char *new_name)
        ret = get_names(&dentry->file_name, &dentry->file_name_utf8,
                        &dentry->file_name_len, &dentry->file_name_utf8_len,
                         new_name);
+       FREE(dentry->short_name);
+       dentry->short_name_len = 0;
        if (ret == 0)
-               recalculate_dentry_size(dentry);
+               dentry->length = dentry_correct_length(dentry);
        return ret;
 }
 
@@ -859,15 +924,21 @@ void calculate_dir_tree_statistics(struct dentry *root, struct lookup_table *tab
  *
  * struct ads_entry_on_disk {
  *     u64  length;          // Length of the entry, in bytes.  This includes
- *                                 all fields (including the stream name, the
- *                                 null terminator, AND the padding!).
+ *                                 all fields (including the stream name and 
+ *                                 null terminator if present, AND the padding!).
  *     u64  reserved;        // Seems to be unused
  *     u8   hash[20];        // SHA1 message digest of the uncompressed stream
  *     u16  stream_name_len; // Length of the stream name, in bytes
  *     char stream_name[];   // Stream name in UTF-16LE, @stream_name_len bytes long,
  *                                  not including null terminator
  *     u16  zero;            // UTF-16 null terminator for the stream name, NOT
- *                                 included in @stream_name_len
+ *                                 included in @stream_name_len.  Based on what
+ *                                 I've observed from filenames in dentries,
+ *                                 this field should not exist when
+ *                                 (@stream_name_len == 0), but you can't
+ *                                 actually tell because of the padding anyway
+ *                                 (provided that the padding is zeroed, which
+ *                                 it always seems to be).
  *     char padding[];       // Padding to make the size a multiple of 8 bytes.
  * };
  *
@@ -898,7 +969,7 @@ static int read_ads_entries(const u8 *p, struct dentry *dentry,
                u64 length_no_padding;
                u64 total_length;
                size_t utf8_len;
-               const char *p_save = p;
+               const u8 *p_save = p;
 
                /* Read the base stream entry, excluding the stream name. */
                if (remaining_size < WIM_ADS_ENTRY_DISK_SIZE) {
@@ -909,7 +980,7 @@ static int read_ads_entries(const u8 *p, struct dentry *dentry,
                }
 
                p = get_u64(p, &length);
-               p += 8;
+               p += 8; /* Skip the reserved field */
                p = get_bytes(p, SHA1_HASH_SIZE, (u8*)cur_entry->hash);
                p = get_u16(p, &cur_entry->stream_name_len);
 
@@ -920,6 +991,8 @@ static int read_ads_entries(const u8 *p, struct dentry *dentry,
                 * */
                length_no_padding = WIM_ADS_ENTRY_DISK_SIZE +
                                    cur_entry->stream_name_len;
+
+               /* Length including the null terminator and the padding */
                total_length = ((length_no_padding + 2) + 7) & ~7;
 
                wimlib_assert(total_length == ads_entry_total_length(cur_entry));
@@ -927,7 +1000,7 @@ static int read_ads_entries(const u8 *p, struct dentry *dentry,
                if (remaining_size < length_no_padding) {
                        ERROR("Stream entries go past end of metadata resource");
                        ERROR("(remaining_size = %"PRIu64" bytes, "
-                             "length_no_padding = %"PRIu16" bytes",
+                             "length_no_padding = %"PRIu64" bytes)",
                              remaining_size, length_no_padding);
                        ret = WIMLIB_ERR_INVALID_DENTRY;
                        goto out_free_ads_entries;
@@ -949,21 +1022,23 @@ static int read_ads_entries(const u8 *p, struct dentry *dentry,
                        goto out_free_ads_entries;
                }
 
-               cur_entry->stream_name = MALLOC(cur_entry->stream_name_len);
-               if (!cur_entry->stream_name) {
-                       ret = WIMLIB_ERR_NOMEM;
-                       goto out_free_ads_entries;
-               }
-               get_bytes(p, cur_entry->stream_name_len,
-                         (u8*)cur_entry->stream_name);
-               cur_entry->stream_name_utf8 = utf16_to_utf8(cur_entry->stream_name,
-                                                           cur_entry->stream_name_len,
-                                                           &utf8_len);
-               cur_entry->stream_name_utf8_len = utf8_len;
-
-               if (!cur_entry->stream_name_utf8) {
-                       ret = WIMLIB_ERR_NOMEM;
-                       goto out_free_ads_entries;
+               if (cur_entry->stream_name_len) {
+                       cur_entry->stream_name = MALLOC(cur_entry->stream_name_len);
+                       if (!cur_entry->stream_name) {
+                               ret = WIMLIB_ERR_NOMEM;
+                               goto out_free_ads_entries;
+                       }
+                       get_bytes(p, cur_entry->stream_name_len,
+                                 (u8*)cur_entry->stream_name);
+                       cur_entry->stream_name_utf8 = utf16_to_utf8(cur_entry->stream_name,
+                                                                   cur_entry->stream_name_len,
+                                                                   &utf8_len);
+                       cur_entry->stream_name_utf8_len = utf8_len;
+
+                       if (!cur_entry->stream_name_utf8) {
+                               ret = WIMLIB_ERR_NOMEM;
+                               goto out_free_ads_entries;
+                       }
                }
                /* It's expected that the size of every ADS entry is a multiple
                 * of 8.  However, to be safe, I'm allowing the possibility of
@@ -1010,18 +1085,18 @@ int read_dentry(const u8 metadata_resource[], u64 metadata_resource_len,
 {
        const u8 *p;
        u64 calculated_size;
-       char *file_name;
-       char *file_name_utf8;
-       char *short_name;
+       char *file_name = NULL;
+       char *file_name_utf8 = NULL;
+       char *short_name = NULL;
        u16 short_name_len;
        u16 file_name_len;
-       size_t file_name_utf8_len;
+       size_t file_name_utf8_len = 0;
        int ret;
 
        dentry_common_init(dentry);
 
        /*Make sure the dentry really fits into the metadata resource.*/
-       if (offset + 8 > metadata_resource_len) {
+       if (offset + 8 > metadata_resource_len || offset + 8 < offset) {
                ERROR("Directory entry starting at %"PRIu64" ends past the "
                      "end of the metadata resource (size %"PRIu64")",
                      offset, metadata_resource_len);
@@ -1045,7 +1120,9 @@ int read_dentry(const u8 metadata_resource[], u64 metadata_resource_len,
         * not too short, read the rest of it (excluding the alternate data
         * streams, but including the file name and short name variable-length
         * fields) into memory. */
-       if (offset + dentry->length >= metadata_resource_len) {
+       if (offset + dentry->length >= metadata_resource_len
+           || offset + dentry->length < offset)
+       {
                ERROR("Directory entry at offset %"PRIu64" and with size "
                      "%"PRIu64" ends past the end of the metadata resource "
                      "(size %"PRIu64")",
@@ -1087,7 +1164,7 @@ int read_dentry(const u8 metadata_resource[], u64 metadata_resource_len,
                p += 4;
        } else {
                p = get_u32(p, &dentry->reparse_tag);
-               p = get_u64(p, &dentry->hard_link);
+               p = get_u64(p, &dentry->link_group_id);
        }
 
        /* By the way, the reparse_reserved field does not actually exist (at
@@ -1098,85 +1175,100 @@ int read_dentry(const u8 metadata_resource[], u64 metadata_resource_len,
        p = get_u16(p, &short_name_len);
        p = get_u16(p, &file_name_len);
 
-       /* We now know the length of the file name and short name.  These should
-        * be included in the dentry length, but make sure the numbers are
-        * consistent. */
-       calculated_size = WIM_DENTRY_DISK_SIZE + file_name_len + 2 +
-                         short_name_len;
+       /* We now know the length of the file name and short name.  Make sure
+        * the length of the dentry is large enough to actually hold them. 
+        *
+        * The calculated length here is unaligned to allow for the possibility
+        * that the dentry->length names an unaligned length, although this
+        * would be unexpected. */
+       calculated_size = __dentry_correct_length_unaligned(file_name_len,
+                                                           short_name_len);
 
        if (dentry->length < calculated_size) {
                ERROR("Unexpected end of directory entry! (Expected "
-                     "%"PRIu64" bytes, got %"PRIu64" bytes. "
+                     "at least %"PRIu64" bytes, got %"PRIu64" bytes. "
                      "short_name_len = %hu, file_name_len = %hu)", 
                      calculated_size, dentry->length,
                      short_name_len, file_name_len);
                return WIMLIB_ERR_INVALID_DENTRY;
        }
 
-       /* Read the filename. */
-       file_name = MALLOC(file_name_len);
-       if (!file_name) {
-               ERROR("Failed to allocate %hu bytes for dentry file name",
-                     file_name_len);
-               return WIMLIB_ERR_NOMEM;
-       }
-       p = get_bytes(p, file_name_len, file_name);
+       /* Read the filename if present.  Note: if the filename is empty, there
+        * is no null terminator following it. */
+       if (file_name_len) {
+               file_name = MALLOC(file_name_len);
+               if (!file_name) {
+                       ERROR("Failed to allocate %hu bytes for dentry file name",
+                             file_name_len);
+                       return WIMLIB_ERR_NOMEM;
+               }
+               p = get_bytes(p, file_name_len, file_name);
 
-       /* Convert filename to UTF-8. */
-       file_name_utf8 = utf16_to_utf8(file_name, file_name_len, 
-                                      &file_name_utf8_len);
+               /* Convert filename to UTF-8. */
+               file_name_utf8 = utf16_to_utf8(file_name, file_name_len, 
+                                              &file_name_utf8_len);
 
-       if (!file_name_utf8) {
-               ERROR("Failed to allocate memory to convert UTF-16 "
-                     "filename (%hu bytes) to UTF-8", file_name_len);
-               ret = WIMLIB_ERR_NOMEM;
-               goto out_free_file_name;
+               if (!file_name_utf8) {
+                       ERROR("Failed to allocate memory to convert UTF-16 "
+                             "filename (%hu bytes) to UTF-8", file_name_len);
+                       ret = WIMLIB_ERR_NOMEM;
+                       goto out_free_file_name;
+               }
+               if (*(u16*)p)
+                       WARNING("Expected two zero bytes following the file name "
+                               "`%s', but found non-zero bytes", file_name_utf8);
+               p += 2;
        }
 
-       /* Undocumented padding between file name and short name.  This probably
-        * is supposed to be a terminating null character. */
-       p += 2;
-
-       /* Read the short filename. */
-       short_name = MALLOC(short_name_len);
-       if (!short_name) {
-               ERROR("Failed to allocate %hu bytes for short filename",
-                     short_name_len);
-               ret = WIMLIB_ERR_NOMEM;
-               goto out_free_file_name_utf8;
+       /* Align the calculated size */
+       calculated_size = (calculated_size + 7) & ~7;
+
+       if (dentry->length > calculated_size) {
+               /* Weird; the dentry says it's longer than it should be.  Note
+                * that the length field does NOT include the size of the
+                * alternate stream entries. */
+
+               /* Strangely, some directory entries inexplicably have a little
+                * over 70 bytes of extra data.  The exact amount of data seems
+                * to be 72 bytes, but it is aligned on the next 8-byte
+                * boundary.  It does NOT seem to be alternate data stream
+                * entries.  Here's an example of the aligned data:
+                *
+                * 01000000 40000000 6c786bba c58ede11 b0bb0026 1870892a b6adb76f
+                * e63a3e46 8fca8653 0d2effa1 6c786bba c58ede11 b0bb0026 1870892a
+                * 00000000 00000000 00000000 00000000
+                *
+                * Here's one interpretation of how the data is laid out.
+                *
+                * struct unknown {
+                *      u32 field1; (always 0x00000001)
+                *      u32 field2; (always 0x40000000)
+                *      u8  data[48]; (???)
+                *      u64 reserved1; (always 0)
+                *      u64 reserved2; (always 0)
+                * };*/
+               DEBUG("Dentry for file or directory `%s' has %zu extra "
+                     "bytes of data",
+                     file_name_utf8, dentry->length - calculated_size);
        }
 
-       p = get_bytes(p, short_name_len, short_name);
+       /* Read the short filename if present.  Note: if there is no short
+        * filename, there is no null terminator following it. */
+       if (short_name_len) {
+               short_name = MALLOC(short_name_len);
+               if (!short_name) {
+                       ERROR("Failed to allocate %hu bytes for short filename",
+                             short_name_len);
+                       ret = WIMLIB_ERR_NOMEM;
+                       goto out_free_file_name_utf8;
+               }
 
-       /* Some directory entries inexplicably have a little over 70 bytes of
-        * extra data.  The exact amount of data seems to be 72 bytes, but it is
-        * aligned on the next 8-byte boundary.  It does NOT seem to be
-        * alternate data stream entries.  Here's an example of the aligned
-        * data:
-        *
-        * 01000000 40000000 6c786bba c58ede11 b0bb0026 1870892a b6adb76f
-        * e63a3e46 8fca8653 0d2effa1 6c786bba c58ede11 b0bb0026 1870892a
-        * 00000000 00000000 00000000 00000000
-        *
-        * Here's one interpretation of how the data is laid out.
-        *
-        * struct unknown {
-        *      u32 field1; (always 0x00000001)
-        *      u32 field2; (always 0x40000000)
-        *      u8  data[48]; (???)
-        *      u64 reserved1; (always 0)
-        *      u64 reserved2; (always 0)
-        * };*/
-#if 0
-       if (dentry->length - calculated_size >= WIM_ADS_ENTRY_DISK_SIZE) {
-               printf("%s: %lu / %lu (", file_name_utf8, 
-                               calculated_size, dentry->length);
-               print_string(p + WIM_ADS_ENTRY_DISK_SIZE, dentry->length - calculated_size - WIM_ADS_ENTRY_DISK_SIZE);
-               puts(")");
-               print_byte_field(p, dentry->length - calculated_size);
-               putchar('\n');
+               p = get_bytes(p, short_name_len, short_name);
+               if (*(u16*)p)
+                       WARNING("Expected two zero bytes following the file name "
+                               "`%s', but found non-zero bytes", file_name_utf8);
+               p += 2;
        }
-#endif
 
        /* 
         * Read the alternate data streams, if present.  dentry->num_ads tells
@@ -1188,7 +1280,6 @@ int read_dentry(const u8 metadata_resource[], u64 metadata_resource_len,
         * included in the dentry->length field for some reason.
         */
        if (dentry->num_ads != 0) {
-               calculated_size = (calculated_size + 7) & ~7;
                if (calculated_size > metadata_resource_len - offset) {
                        ERROR("Not enough space in metadata resource for "
                              "alternate stream entries");
@@ -1220,6 +1311,89 @@ out_free_file_name:
        return ret;
 }
 
+/* Run some miscellaneous verifications on a WIM dentry */
+int verify_dentry(struct dentry *dentry, void *wim)
+{
+       const WIMStruct *w = wim;
+       const struct lookup_table *table = w->lookup_table;
+       const struct wim_security_data *sd = wim_const_security_data(w);
+       int ret = WIMLIB_ERR_INVALID_DENTRY;
+
+       /* Check the security ID */
+       if (dentry->security_id < -1) {
+               ERROR("Dentry `%s' has an invalid security ID (%d)",
+                       dentry->full_path_utf8, dentry->security_id);
+               goto out;
+       }
+       if (dentry->security_id >= sd->num_entries) {
+               ERROR("Dentry `%s' has an invalid security ID (%d) "
+                     "(there are only %u entries in the security table)",
+                       dentry->full_path_utf8, dentry->security_id,
+                       sd->num_entries);
+               goto out;
+       }
+
+       /* Check that lookup table entries for all the resources exist, except
+        * if the SHA1 message digest is all 0's, which indicates there is
+        * intentionally no resource there.  */
+       if (w->hdr.total_parts == 1) {
+               for (unsigned i = 0; i <= dentry->num_ads; i++) {
+                       struct lookup_table_entry *lte;
+                       const u8 *hash;
+                       hash = dentry_stream_hash_unresolved(dentry, i);
+                       lte = __lookup_resource(table, hash);
+                       if (!lte && !is_zero_hash(hash)) {
+                               ERROR("Could not find lookup table entry for stream "
+                                     "%u of dentry `%s'", i, dentry->full_path_utf8);
+                               goto out;
+                       }
+               }
+       }
+
+       /* Make sure there is only one un-named stream. */
+       unsigned num_unnamed_streams = 0;
+       for (unsigned i = 0; i <= dentry->num_ads; i++) {
+               const u8 *hash;
+               hash = dentry_stream_hash_unresolved(dentry, i);
+               if (!dentry_stream_name_len(dentry, i) && !is_zero_hash(hash))
+                       num_unnamed_streams++;
+       }
+       if (num_unnamed_streams > 1) {
+               ERROR("Dentry `%s' has multiple (%u) un-named streams", 
+                     dentry->full_path_utf8, num_unnamed_streams);
+               goto out;
+       }
+
+       /* Cannot have a short name but no long name */
+       if (dentry->short_name_len && !dentry->file_name_len) {
+               ERROR("Dentry `%s' has a short name but no long name",
+                     dentry->full_path_utf8);
+               goto out;
+       }
+
+       /* Make sure root dentry is unnamed */
+       if (dentry_is_root(dentry)) {
+               if (dentry->file_name_len) {
+                       ERROR("The root dentry is named `%s', but it must "
+                             "be unnamed", dentry->file_name_utf8);
+                       goto out;
+               }
+       }
+
+#if 0
+       /* Check timestamps */
+       if (dentry->last_access_time < dentry->creation_time ||
+           dentry->last_write_time < dentry->creation_time) {
+               WARNING("Dentry `%s' was created after it was last accessed or "
+                     "written to", dentry->full_path_utf8);
+       }
+#endif
+
+       ret = 0;
+out:
+       return ret;
+}
+
 /* 
  * Writes a WIM dentry to an output buffer.
  *
@@ -1231,10 +1405,15 @@ out_free_file_name:
 static u8 *write_dentry(const struct dentry *dentry, u8 *p)
 {
        u8 *orig_p = p;
-       unsigned padding;
        const u8 *hash;
 
-       p = put_u64(p, dentry->length);
+       /* We calculate the correct length of the dentry ourselves because the
+        * dentry->length field may been set to an unexpected value from when we
+        * read the dentry in (for example, there may have been unknown data
+        * appended to the end of the dentry...) */
+       u64 length = dentry_correct_length(dentry);
+
+       p = put_u64(p, length);
        p = put_u32(p, dentry->attributes);
        p = put_u32(p, dentry->security_id);
        p = put_u64(p, dentry->subdir_offset);
@@ -1243,37 +1422,37 @@ static u8 *write_dentry(const struct dentry *dentry, u8 *p)
        p = put_u64(p, dentry->creation_time);
        p = put_u64(p, dentry->last_access_time);
        p = put_u64(p, dentry->last_write_time);
-       if (dentry->resolved && dentry->lte)
-               hash = dentry->lte->hash;
-       else
-               hash = dentry->hash;
+       hash = dentry_stream_hash(dentry, 0);
        p = put_bytes(p, SHA1_HASH_SIZE, hash);
        if (dentry->attributes & FILE_ATTRIBUTE_REPARSE_POINT) {
                p = put_zeroes(p, 4);
                p = put_u32(p, dentry->reparse_tag);
                p = put_zeroes(p, 4);
        } else {
-               u64 hard_link;
+               u64 link_group_id;
                p = put_u32(p, 0);
                if (dentry->link_group_list.next == &dentry->link_group_list)
-                       hard_link = 0;
+                       link_group_id = 0;
                else
-                       hard_link = dentry->hard_link;
-               p = put_u64(p, hard_link);
+                       link_group_id = dentry->link_group_id;
+               p = put_u64(p, link_group_id);
        }
        p = put_u16(p, dentry->num_ads);
        p = put_u16(p, dentry->short_name_len);
        p = put_u16(p, dentry->file_name_len);
-       p = put_bytes(p, dentry->file_name_len, (u8*)dentry->file_name);
-       p = put_u16(p, 0); /* filename padding, 2 bytes. */
-       p = put_bytes(p, dentry->short_name_len, (u8*)dentry->short_name);
-
-       wimlib_assert(p - orig_p <= dentry->length);
-       if (p - orig_p < dentry->length)
-               p = put_zeroes(p, dentry->length - (p - orig_p));
+       if (dentry->file_name_len) {
+               p = put_bytes(p, dentry->file_name_len, (u8*)dentry->file_name);
+               p = put_u16(p, 0); /* filename padding, 2 bytes. */
+       }
+       if (dentry->short_name) {
+               p = put_bytes(p, dentry->short_name_len, (u8*)dentry->short_name);
+               p = put_u16(p, 0); /* short name padding, 2 bytes */
+       }
 
        /* Align to 8-byte boundary */
-       p = put_zeroes(p, (8 - dentry->length % 8) % 8);
+       wimlib_assert(length >= (p - orig_p)
+                       && length - (p - orig_p) <= 7);
+       p = put_zeroes(p, length - (p - orig_p));
 
        /* Write the alternate data streams, if there are any.  Please see
         * read_ads_entries() for comments about the format of the on-disk
@@ -1281,17 +1460,19 @@ static u8 *write_dentry(const struct dentry *dentry, u8 *p)
        for (u16 i = 0; i < dentry->num_ads; i++) {
                p = put_u64(p, ads_entry_total_length(&dentry->ads_entries[i]));
                p = put_u64(p, 0); /* Unused */
-               if (dentry->resolved && dentry->ads_entries[i].lte)
-                       hash = dentry->ads_entries[i].lte->hash;
-               else
-                       hash = dentry->ads_entries[i].hash;
+               hash = dentry_stream_hash(dentry, i + 1);
                p = put_bytes(p, SHA1_HASH_SIZE, hash);
                p = put_u16(p, dentry->ads_entries[i].stream_name_len);
-               p = put_bytes(p, dentry->ads_entries[i].stream_name_len,
-                                (u8*)dentry->ads_entries[i].stream_name);
-               p = put_u16(p, 0);
+               if (dentry->ads_entries[i].stream_name_len) {
+                       p = put_bytes(p, dentry->ads_entries[i].stream_name_len,
+                                        (u8*)dentry->ads_entries[i].stream_name);
+                       p = put_u16(p, 0);
+               }
                p = put_zeroes(p, (8 - (p - orig_p) % 8) % 8);
        }
+#ifdef ENABLE_ASSERTIONS
+       wimlib_assert(p - orig_p == __dentry_total_length(dentry, length));
+#endif
        return p;
 }
 
@@ -1433,7 +1614,11 @@ int read_dentry_tree(const u8 metadata_resource[], u64 metadata_resource_len,
                                break;
                }
 
-               /* Advance to the offset of the next child. */
+               /* Advance to the offset of the next child.  Note: We need to
+                * advance by the TOTAL length of the dentry, not by the length
+                * child->length, which although it does take into account the
+                * padding, it DOES NOT take into account alternate stream
+                * entries. */
                cur_offset += dentry_total_length(child);
        }