Skip to contentSkip to conversation

xfs: allow unlinked symlinks and dirs with zero size

PATCH In review Darrick J. Wong 2 participants 5 messages Jun 7, 2024 – Jun 11, 2024 XFS filesystem

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 xfs: allow unlinked symlinks and dirs with zero size

From: Darrick J. Wong <[email protected]>

For a very very long time, inode inactivation has set the inode size to zero before unmapping the extents associated with the data fork. Unfortunately, newer commit 3c6f46eacd876 changed the inode verifier to prohibit zero-length symlinks and directories. If an inode happens to get logged in this state and the system crashes before freeing the inode, log recovery will also fail on the broken inode.

Therefore, allow zero-size symlinks and directories as long as the link count is zero; nobody will be able to open these files by handle so there isn't any risk of data exposure.

Fixes 3c6f46eacd876 ("xfs: sanity check directory inode di_size")
---
 fs/xfs/libxfs/xfs_inode_buf.c |   20 +++++++++++++++++---
 1 file changed, 17 insertions(+), 3 deletions(-)
fs/xfs/libxfs/xfs_inode_buf.c +17 -3
index 9caf9aa2221d3..afe06cfd6f0cc 100644
@@ -612,9 +612,23 @@ xfs_dinode_verify(
 	if (mode && xfs_mode_to_ftype(mode) == XFS_DIR3_FT_UNKNOWN)
 		return __this_address;
 
-	/* No zero-length symlinks/dirs. */
-	if ((S_ISLNK(mode) || S_ISDIR(mode)) && di_size == 0)
-		return __this_address;
+	/*
+	 * No zero-length symlinks/dirs unless they're unlinked and hence being
+	 * inactivated.
+	 */
+	if ((S_ISLNK(mode) || S_ISDIR(mode)) && di_size == 0) {
+		if (dip->di_version > 1) {
+			if (dip->di_nlink)
+				return __this_address;
+			else
+				ASSERT(0);
+		} else {
+			if (dip->di_onlink)
+				return __this_address;
+			else
+				ASSERT(0);
+		}
+	}
 
 	fa = xfs_dinode_verify_nrext64(mp, dip);
 	if (fa)
Reviewed-by on v1 1/1
#1 lore raw
Quoted 6 lines · Darrick J. Wong

From: Darrick J. Wong <[email protected]>

For a very very long time, inode inactivation has set the inode size to zero before unmapping the extents associated with the data fork. Unfortunately, newer commit 3c6f46eacd876 changed the inode verifier to prohibit zero-length symlinks and directories. If an inode happens to

", newer commit" above reads really odd. Maybe just drop the "newer "?

fs/xfs/libxfs/xfs_inode_buf.c xfs_dinode_verify( · line 630

Quoted 12 lines
fs/xfs/libxfs/xfs_inode_buf.c +12 -0
+	if ((S_ISLNK(mode) || S_ISDIR(mode)) && di_size == 0) {
+		if (dip->di_version > 1) {
+			if (dip->di_nlink)
+				return __this_address;
+			else
+				ASSERT(0);
+		} else {
+			if (dip->di_onlink)
+				return __this_address;
+			else
+				ASSERT(0);
+		}

No need for else after a return.

With that fixed:

#2 lore raw
Quoted 25 lines · Christoph Hellwig
Quoted 6 lines · Darrick J. Wong

From: Darrick J. Wong <[email protected]>

For a very very long time, inode inactivation has set the inode size to zero before unmapping the extents associated with the data fork. Unfortunately, newer commit 3c6f46eacd876 changed the inode verifier to prohibit zero-length symlinks and directories. If an inode happens to

", newer commit" above reads really odd. Maybe just drop the "newer "?

fs/xfs/libxfs/xfs_inode_buf.c xfs_dinode_verify( · line 630

Quoted 12 lines
fs/xfs/libxfs/xfs_inode_buf.c +12 -0
+	if ((S_ISLNK(mode) || S_ISDIR(mode)) && di_size == 0) {
+		if (dip->di_version > 1) {
+			if (dip->di_nlink)
+				return __this_address;
+			else
+				ASSERT(0);
+		} else {
+			if (dip->di_onlink)
+				return __this_address;
+			else
+				ASSERT(0);
+		}

No need for else after a return.

Quoted 3 lines

With that fixed:

It turns out that even this is still buggy because directories that are being inactivated (e.g. after repair has replaced the contents) can have zero isize. Sooo I'll have a new patch in a day or two.

--D

#3 lore raw
Quoted 3 lines · Darrick J. Wong

It turns out that even this is still buggy because directories that are being inactivated (e.g. after repair has replaced the contents) can have zero isize. Sooo I'll have a new patch in a day or two.

Isn't that what this patch checks for? Or do you mean inactivated with non-zero nlink?

Btw, it might make sense to add a helper or local variable to check for an unlinked dinode instead of open coding the v1 check twice in this function.

#4 lore raw
Quoted 7 lines · Christoph Hellwig
Quoted 3 lines · Darrick J. Wong

It turns out that even this is still buggy because directories that are being inactivated (e.g. after repair has replaced the contents) can have zero isize. Sooo I'll have a new patch in a day or two.

Isn't that what this patch checks for? Or do you mean inactivated with non-zero nlink?

Nah, there's another verifier check elsewhere that also trips when di-size == 0.

Quoted 3 lines

Btw, it might make sense to add a helper or local variable to check for an unlinked dinode instead of open coding the v1 check twice in this function.

Maybe? But the logic is slightly different, isn't it? I'll have a look in the morning.

--D