Rearrange struct wim_inode and improve comments

diff --git a/include/wimlib/dentry.h b/include/wimlib/dentry.h
index daf92a7897977618819d16bf4049176e78e0f646..0871edc2d418894edc160743346ca3ae523091ce 100644 (file)
--- a/include/wimlib/dentry.h
+++ b/include/wimlib/dentry.h
@@ -238,104 +238,142 @@ struct wim_dentry {
 /*
  * WIM inode.
  *
 /*
  * WIM inode.
  *

- * As mentioned above, in the WIM file that is no on-disk analogue of a real
- * inode, as most of these fields are duplicated in the dentries.
+ * As mentioned in the comment above `struct wim_dentry', in the WIM file that
+ * is no on-disk analogue of a real inode, as most of these fields are
+ * duplicated in the dentries.  Instead, a `struct wim_inode' is something we
+ * create ourselves to simplify the handling of hard links.

  */
 struct wim_inode {
  */
 struct wim_inode {

-       /* Timestamps for the inode.  The timestamps are the number of
-        * 100-nanosecond intervals that have elapsed since 12:00 A.M., January
-        * 1st, 1601, UTC.  This is the same format used in NTFS inodes. */
-       u64 i_creation_time;
-       u64 i_last_access_time;
-       u64 i_last_write_time;
+       /* If i_resolved == 0:
+        *      SHA1 message digest of the contents of the unnamed-data stream
+        *      of this inode.
+        *
+        * If i_resolved == 1:
+        *      Pointer to the lookup table entry for the unnamed data stream
+        *      of this inode, or NULL.
+        *
+        * i_hash corresponds to the 'unnamed_stream_hash' field of the `struct
+        * wim_dentry_on_disk' and the additional caveats documented about that
+        * field apply here (for example, the quirks regarding all-zero hashes).
+        */
+       union {
+               u8 i_hash[SHA1_HASH_SIZE];
+               struct wim_lookup_table_entry *i_lte;
+       };

 
 

-       /* The file attributes associated with this inode.  This is a bitwise OR
-        * of the FILE_ATTRIBUTE_* flags. */
+       /* Corresponds to the 'attributes' field of `struct wim_dentry_on_disk';
+        * bitwise OR of the FILE_ATTRIBUTE_* flags that give the attributes of
+        * this inode. */

        u32 i_attributes;
 
        u32 i_attributes;
 

-       /* The index of the security descriptor in the WIM image's table of
-        * security descriptors that contains this file's security information.
-        * If -1, no security information exists for this file.  */
-       int32_t i_security_id;
+       /* Root of a red-black tree storing the child dentries of this inode, if
+        * any.  Keyed by wim_dentry->file_name, case sensitively. */
+       struct rb_root i_children;

 
 

-       /* %true iff the inode's lookup table entries has been resolved (i.e.
-        * the @lte field is valid, but the @hash field is not valid)
-        *
-        * (This is not an on-disk field.) */
-       u8 i_resolved : 1;
+#ifdef __WIN32__
+       /* Root of a red-black tree storing the children of this inode, if any.
+        * Keyed by wim_dentry->file_name, case insensitively. */
+       struct rb_root i_children_case_insensitive;
+#endif

 
 

-       u8 i_visited : 1;
+       /* List of dentries that are aliases for this inode.  There will be
+        * i_nlink dentries in this list.  */
+       struct list_head i_dentry;

 
 

-       /* Used only in NTFS-mode extraction */
-       u8 i_dos_name_extracted : 1;
+       /* Field to place this inode into a list. */
+       union {
+               /* Hash list node- used in hardlink.c when the inodes are placed
+                * into a hash table keyed by inode number and optionally device
+                * number, in order to detect dentries that are aliases for the
+                * same inode. */
+               struct hlist_node i_hlist;

 
 

-       /* Set to 0 if reparse point fixups have been done.  Otherwise set to 1.
-        *
-        * Note: this actually may reflect the SYMBOLIC_LINK_RELATIVE flag.  */
-       u16 i_not_rpfixed;
+               /* Normal list node- used to connect all the inodes of a WIM image
+                * into a single linked list referenced from the
+                * `struct wim_image_metadata' for that image. */
+               struct list_head i_list;
+       };
+
+       /* Number of dentries that are aliases for this inode.  */
+       u32 i_nlink;

 
        /* Number of alternate data streams associated with this inode */
        u16 i_num_ads;
 
 
        /* Number of alternate data streams associated with this inode */
        u16 i_num_ads;
 

-       /* Unused/unknown fields that we just read into memory so we can
-        * re-write them unchanged.  */
-       u32 i_rp_unknown_1;
-       u16 i_rp_unknown_2;
+       /* Flag that indicates whether this inode's streams have been
+        * "resolved".  By default, the inode starts as "unresolved", meaning
+        * that the i_hash field, along with the hash field of any associated
+        * wim_ads_entry's, are valid and should be used as keys in the WIM
+        * lookup table to find the associated `struct wim_lookup_table_entry'.
+        * But if the inode has been resolved, then each of these fields is
+        * replaced with a pointer directly to the appropriate `struct
+        * wim_lookup_table_entry', or NULL if the stream is empty.  */
+       u8 i_resolved : 1;

 
 

-       /* If i_resolved == 0:
-        *      SHA1 message digest of the contents of the unnamed-data stream
-        *      of this inode, or all zeroes if this inode has no unnamed data
-        *      stream, or optionally all zeroes if this inode has an empty
-        *      unnamed data stream.
-        *
-        * If i_resolved == 1:
-        *      Pointer to the lookup table entry for the unnamed data stream
-        *      of this inode, or NULL if this inode has no unnamed data stream,
-        *      or optionally all zeroes if this inode has an empty unnamed data
-        *      stream.
-        */
-       union {
-               u8 i_hash[SHA1_HASH_SIZE];
-               struct wim_lookup_table_entry *i_lte;
-       };
+       /* Flag used to mark this inode as visited; this is used when visiting
+        * all the inodes in a dentry tree exactly once.  It will be 0 by
+        * default and must be cleared following the tree traversal, even in
+        * error paths.  */
+       u8 i_visited : 1;
+
+       /* For NTFS-3g extraction:  Set after the DOS name for this inode has
+        * been extracted.  */
+       u8 i_dos_name_extracted : 1;
+
+       /* Pointer to a malloc()ed array of i_num_ads alternate data stream
+        * entries for this inode.  */
+       struct wim_ads_entry *i_ads_entries;
+
+       /* Creation time, last access time, and last write time for this inode, in
+        * 100-nanosecond intervals since 12:00 a.m UTC January 1, 1601.  They
+        * should correspond to the times gotten by calling GetFileTime() on
+        * Windows. */
+       u64 i_creation_time;
+       u64 i_last_access_time;
+       u64 i_last_write_time;
+
+       /* Corresponds to 'security_id' in `struct wim_dentry_on_disk':  The
+        * index of this inode's security descriptor in the WIM image's table of
+        * security descriptors, or -1.  Note: in verify_inode(), called
+        * whenever a WIM image is loaded, out-of-bounds indices are set to -1,
+        * so the extraction code does not need to do bounds checks.  */
+       int32_t i_security_id;

 
        /* Identity of a reparse point.  See
         * http://msdn.microsoft.com/en-us/library/windows/desktop/aa365503(v=vs.85).aspx
         * for what a reparse point is. */
        u32 i_reparse_tag;
 
 
        /* Identity of a reparse point.  See
         * http://msdn.microsoft.com/en-us/library/windows/desktop/aa365503(v=vs.85).aspx
         * for what a reparse point is. */
        u32 i_reparse_tag;
 

-       /* Number of dentries that reference this inode */
-       u32 i_nlink;
+       /* Unused/unknown fields that we just read into memory so we can
+        * re-write them unchanged.  */
+       u32 i_rp_unknown_1;
+       u16 i_rp_unknown_2;

 
 

-       /* Alternate data stream entries. */
-       struct wim_ads_entry *i_ads_entries;
+       /* Corresponds to not_rpfixed in `struct wim_dentry_on_disk':  Set to 0
+        * if reparse point fixups have been done.  Otherwise set to 1.  Note:
+        * this actually may reflect the SYMBOLIC_LINK_RELATIVE flag.
+        */
+       u16 i_not_rpfixed;

 
 

-       /* Inode number */
+       /* Inode number; corresponds to hard_link_group_id in the `struct
+        * wim_dentry_on_disk'.  */

        u64 i_ino;
 
        u64 i_ino;
 

-       /* Device number, used only during image capture */
-       u64 i_devno;
-
-       /* List of dentries that reference this inode (there should be i_nlink
-        * of them) */
-       struct list_head i_dentry;
-

        union {
        union {

-               struct hlist_node i_hlist;
-               struct list_head i_list;
+               /* Device number, used only during image capture, so we can
+                * identify hard linked files by the combination of inode number
+                * and device number (rather than just inode number, which could
+                * be ambigious if the captured tree spans a mountpoint).  Set
+                * to 0 otherwise.  */
+               u64 i_devno;
+
+               /* Used only during image extraction: pointer to the first path
+                * (malloc()ed buffer) at which this inode has been extracted.
+                * Freed and set to NULL after the extraction is done (either
+                * success or failure).  */
+               tchar *i_extracted_file;

        };
 
        };
 

-       tchar *i_extracted_file;
-
-       /* Root of a red-black tree storing the children of this inode (if
-        * non-empty, implies the inode is a directory, although that is also
-        * noted in the @attributes field.) */
-       struct rb_root i_children;
-
-#ifdef __WIN32__
-       struct rb_root i_children_case_insensitive;
-#endif
-

        /* Next alternate data stream ID to be assigned */
        u32 i_next_stream_id;
 
        /* Next alternate data stream ID to be assigned */
        u32 i_next_stream_id;
 

diff --git a/src/dentry.c b/src/dentry.c
index 86cbd3486db08577a51c48b4ad5dcc2e1e73726d..d6eb8a5c737f3942e80d4b753ea47a314ff697ce 100644 (file)
--- a/src/dentry.c
+++ b/src/dentry.c
@@ -110,16 +110,13 @@ struct wim_dentry_on_disk {
        le64 unused_1;
        le64 unused_2;
 
        le64 unused_1;
        le64 unused_2;
 

-       /* The following three time fields should correspond to those gotten by
-        * calling GetFileTime() on Windows. */

 
 

-       /* Creation time, in 100-nanosecond intervals since January 1, 1601. */
+       /* Creation time, last access time, and last write time, in
+        * 100-nanosecond intervals since 12:00 a.m UTC January 1, 1601.  They
+        * should correspond to the times gotten by calling GetFileTime() on
+        * Windows. */

        le64 creation_time;
        le64 creation_time;

-
-       /* Last access time, in 100-nanosecond intervals since January 1, 1601. */

        le64 last_access_time;
        le64 last_access_time;

-
-       /* Last write time, in 100-nanosecond intervals since January 1, 1601. */

        le64 last_write_time;
 
        /* Vaguely, the SHA-1 message digest ("hash") of the file's contents.
        le64 last_write_time;
 
        /* Vaguely, the SHA-1 message digest ("hash") of the file's contents.


@@ -1140,7 +1137,6 @@ free_inode(struct wim_inode *inode)
                 * hlist_del() behaves the same as list_del(). */
                if (!hlist_unhashed(&inode->i_hlist))
                        hlist_del(&inode->i_hlist);
                 * hlist_del() behaves the same as list_del(). */
                if (!hlist_unhashed(&inode->i_hlist))
                        hlist_del(&inode->i_hlist);

-               FREE(inode->i_extracted_file);

                FREE(inode);
        }
 }
                FREE(inode);
        }
 }

diff --git a/src/extract.c b/src/extract.c
index f9c4ee9894d48445e8392a63251fca9f64b844b5..3e0e6af3f971407b78cd5d1cffdffe3d600bee0d 100644 (file)
--- a/src/extract.c
+++ b/src/extract.c
@@ -634,13 +634,14 @@ skip_dentry:
 static int
 dentry_reset_needs_extraction(struct wim_dentry *dentry, void *_ignore)
 {
 static int
 dentry_reset_needs_extraction(struct wim_dentry *dentry, void *_ignore)
 {

+       struct wim_inode *inode = dentry->d_inode;
+

        dentry->needs_extraction = 0;
        dentry->not_extracted = 0;
        dentry->needs_extraction = 0;
        dentry->not_extracted = 0;

-       dentry->is_win32_name = 0;
-       dentry->d_inode->i_visited = 0;
-       dentry->d_inode->i_dos_name_extracted = 0;
-       FREE(dentry->d_inode->i_extracted_file);
-       dentry->d_inode->i_extracted_file = NULL;
+       inode->i_visited = 0;
+       inode->i_dos_name_extracted = 0;
+       FREE(inode->i_extracted_file);
+       inode->i_extracted_file = NULL;

        if ((void*)dentry->extraction_name != (void*)dentry->file_name)
                FREE(dentry->extraction_name);
        dentry->extraction_name = NULL;
        if ((void*)dentry->extraction_name != (void*)dentry->file_name)
                FREE(dentry->extraction_name);
        dentry->extraction_name = NULL;

diff --git a/src/hardlink.c b/src/hardlink.c
index 461c9523371309b3594833d82e241a4234dd6bfd..475534400ce2d5a3369085441467f579031d88ce 100644 (file)
--- a/src/hardlink.c
+++ b/src/hardlink.c
@@ -581,31 +581,31 @@ void
 inode_table_prepare_inode_list(struct wim_inode_table *table,
                               struct list_head *head)
 {
 inode_table_prepare_inode_list(struct wim_inode_table *table,
                               struct list_head *head)
 {

-       struct wim_inode *inode;
+       struct wim_inode *inode, *tmp_inode;

        struct hlist_node *cur, *tmp;
        u64 cur_ino = 1;
 
        /* Re-assign inode numbers in the existing list to avoid duplicates. */
        struct hlist_node *cur, *tmp;
        u64 cur_ino = 1;
 
        /* Re-assign inode numbers in the existing list to avoid duplicates. */

-       list_for_each_entry(inode, head, i_list) {
-               if (inode->i_nlink > 1)
-                       inode->i_ino = cur_ino++;
-               else
-                       inode->i_ino = 0;
-       }
+       list_for_each_entry(inode, head, i_list)
+               inode->i_ino = cur_ino++;

 
        /* Assign inode numbers to the new inodes and move them to the image's
         * inode list. */
        for (size_t i = 0; i < table->capacity; i++) {
                hlist_for_each_entry_safe(inode, cur, tmp, &table->array[i], i_hlist)
                {
 
        /* Assign inode numbers to the new inodes and move them to the image's
         * inode list. */
        for (size_t i = 0; i < table->capacity; i++) {
                hlist_for_each_entry_safe(inode, cur, tmp, &table->array[i], i_hlist)
                {

-                       if (inode->i_nlink > 1)
-                               inode->i_ino = cur_ino++;
-                       else
-                               inode->i_ino = 0;
+                       inode->i_ino = cur_ino++;
+                       inode->i_devno = 0;

                        list_add_tail(&inode->i_list, head);
                }
                INIT_HLIST_HEAD(&table->array[i]);
        }
                        list_add_tail(&inode->i_list, head);
                }
                INIT_HLIST_HEAD(&table->array[i]);
        }

-       list_splice_tail(&table->extra_inodes, head);
+       list_for_each_entry_safe(inode, tmp_inode, &table->extra_inodes, i_list)
+       {
+               inode->i_ino = cur_ino++;
+               inode->i_devno = 0;
+               list_add_tail(&inode->i_list, head);
+       }
+       INIT_LIST_HEAD(&table->extra_inodes);

        table->num_entries = 0;
 }
        table->num_entries = 0;
 }