[PATCH] libext2fs: Only link an inode into a directory once

[PATCH] libext2fs: Only link an inode into a directory once

[Darrick J. Wong]

The ext2fs_link helper function link_proc does not check the value of ls->done,
which means that if the function finds multiple empty spaces that will fit the
new directory entry, it will create a directory entry in each of the spaces.
Instead of doing that, check the done value and don't do anything more if we've
already added the directory entry.

Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
---
 lib/ext2fs/link.c |    3 +++
 1 files changed, 3 insertions(+), 0 deletions(-)


diff --git a/lib/ext2fs/link.c b/lib/ext2fs/link.c
index 2d03b57..2f4f54a 100644
--- a/lib/ext2fs/link.c
+++ b/lib/ext2fs/link.c
@@ -42,6 +42,9 @@ static int link_proc(struct ext2_dir_entry *dirent,
 	unsigned int rec_len, min_rec_len, curr_rec_len;
 	int ret = 0;
 
+	if (ls->done)
+		return 0;
+
 	rec_len = EXT2_DIR_REC_LEN(ls->namelen);
 
 	ls->err = ext2fs_get_rec_len(ls->fs, dirent, &curr_rec_len);

Re: [PATCH] libext2fs: Only link an inode into a directory once

[Ted Ts'o]

On Fri, Jan 06, 2012 at 08:47:43PM -0800, Darrick J. Wong wrote:
> The ext2fs_link helper function link_proc does not check the value of ls->done,
> which means that if the function finds multiple empty spaces that will fit the
> new directory entry, it will create a directory entry in each of the spaces.
> Instead of doing that, check the done value and don't do anything more if we've
> already added the directory entry.
> 
> Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>

This should't necessary, since when we insert the directory entry, we
return with the DIRENT_ABORT bit set:

        dirent->inode = ls->inode;
        dirent->name_len = ls->namelen;
        strncpy(dirent->name, ls->name, ls->namelen);
        if (ls->sb->s_feature_incompat & EXT2_FEATURE_INCOMPAT_FILETYPE)
                dirent->name_len |= (ls->flags & 0x7) << 8;

        ls->done++;
        return DIRENT_ABORT|DIRENT_CHANGED;

Did you actually observe this happening?

Thanks,

					- Ted

Re: [PATCH] libext2fs: Only link an inode into a directory once

[Darrick J. Wong]

On Wed, Feb 15, 2012 at 02:25:37PM -0500, Ted Ts'o wrote:
> On Fri, Jan 06, 2012 at 08:47:43PM -0800, Darrick J. Wong wrote:
> > The ext2fs_link helper function link_proc does not check the value of ls->done,
> > which means that if the function finds multiple empty spaces that will fit the
> > new directory entry, it will create a directory entry in each of the spaces.
> > Instead of doing that, check the done value and don't do anything more if we've
> > already added the directory entry.
> > 
> > Signed-off-by: Darrick J. Wong <djwong@us.ibm.com>
> 
> This should't necessary, since when we insert the directory entry, we
> return with the DIRENT_ABORT bit set:
> 
>         dirent->inode = ls->inode;
>         dirent->name_len = ls->namelen;
>         strncpy(dirent->name, ls->name, ls->namelen);
>         if (ls->sb->s_feature_incompat & EXT2_FEATURE_INCOMPAT_FILETYPE)
>                 dirent->name_len |= (ls->flags & 0x7) << 8;
> 
>         ls->done++;
>         return DIRENT_ABORT|DIRENT_CHANGED;
> 
> Did you actually observe this happening?

Yes.  I took a look at the block iteration code again, and I'm wondering, do we
need one of these:

if (ret & BLOCK_ABORT)
	break;

...around lib/ext2fs/block.c, line 450?  This is the code blob:

if (!(extent.e_flags & EXT2_EXTENT_FLAGS_LEAF)) {
        if (ctx.flags & BLOCK_FLAG_DATA_ONLY)
                continue;
        if ((!(extent.e_flags &
               EXT2_EXTENT_FLAGS_SECOND_VISIT) &&
             !(ctx.flags & BLOCK_FLAG_DEPTH_TRAVERSE)) ||
            ((extent.e_flags &
              EXT2_EXTENT_FLAGS_SECOND_VISIT) &&
             (ctx.flags & BLOCK_FLAG_DEPTH_TRAVERSE))) {
                ret |= (*ctx.func)(fs, &blk,
                                   -1, 0, 0, priv_data);
                if (ret & BLOCK_CHANGED) {
                        extent.e_pblk = blk;
                        ctx.errcode =
        ext2fs_extent_replace(handle, 0, &extent);
                        if (ctx.errcode)
                                break;
                }
		/* INSERT HERE? */
        }
        continue;
}

It looks to me that when ctx.func returns BLOCK_ABORT, the continue causes
the code to loop around and get the next extent, which isn't what we want.

--D

Re: [PATCH] libext2fs: Only link an inode into a directory once

[Ted Ts'o]

On Wed, Feb 15, 2012 at 12:40:15PM -0800, Darrick J. Wong wrote:
> 
> Yes.  I took a look at the block iteration code again, and I'm
> wondering, do we need one of these:
> 
> if (ret & BLOCK_ABORT)
> 	break;
> 
> ...around lib/ext2fs/block.c, line 450?

Good spotting; it turns out there was another problem later on, where the

	if (ret & BLOCK_ABORT)
		break;

does the wrong thing, since we need to break out of both the for loop
*and* the while loop.

I'll send a patch which I believe will fix BLOCK_ABORT in the block
iterator for the extent case.

						- Ted

[PATCH] libext2fs: fix BLOCK_ABORT handling in the block iterator for extents

[Theodore Ts'o]

When processing files that contain extents, the block iterator
functions were not properly handling the BLOCK_ABORT bit.  This could
cause problems such as ext2fs_link() adding a directory entry multiple
times.

Thanks to Darrick Wong <djwong@us.ibm.com> for reporting this.

Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
---
 lib/ext2fs/block.c |   14 ++++++++------
 1 files changed, 8 insertions(+), 6 deletions(-)

diff --git a/lib/ext2fs/block.c b/lib/ext2fs/block.c
index 6a5c10d..85a1803 100644
--- a/lib/ext2fs/block.c
+++ b/lib/ext2fs/block.c
@@ -414,7 +414,7 @@ errcode_t ext2fs_block_iterate3(ext2_filsys fs,
 						0, 0, priv_data);
 				ret |= r;
 				check_for_ro_violation_goto(&ctx, ret,
-							    extent_errout);
+							    extent_done);
 				if (r & BLOCK_CHANGED) {
 					ctx.errcode =
 						ext2fs_extent_set_bmap(handle,
@@ -448,6 +448,8 @@ errcode_t ext2fs_block_iterate3(ext2_filsys fs,
 						if (ctx.errcode)
 							break;
 					}
+					if (ret & BLOCK_ABORT)
+						break;
 				}
 				continue;
 			}
@@ -473,23 +475,23 @@ errcode_t ext2fs_block_iterate3(ext2_filsys fs,
 						0, 0, priv_data);
 				ret |= r;
 				check_for_ro_violation_goto(&ctx, ret,
-							    extent_errout);
+							    extent_done);
 				if (r & BLOCK_CHANGED) {
 					ctx.errcode =
 						ext2fs_extent_set_bmap(handle,
 						       (blk64_t) blockcnt,
 						       new_blk, uninit);
 					if (ctx.errcode)
-						goto extent_errout;
+						goto extent_done;
 				}
 				if (ret & BLOCK_ABORT)
-					break;
+					goto extent_done;
 			}
 		}
 
-	extent_errout:
+	extent_done:
 		ext2fs_extent_free(handle);
-		ret |= BLOCK_ERROR | BLOCK_ABORT;
+		ret |= BLOCK_ERROR; /* ctx.errcode is always valid here */
 		goto errout;
 	}
 
-- 
1.7.9.107.g97f9a