Skip to contentSkip to conversation

xfs: fix memory leak in xfs_dir2_node_removename

PATCH In review Mark Tinguely 4 participants 8 messages Sep 27, 2013 XFS filesystem

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 xfs: fix memory leak in xfs_dir2_node_removename

Free the memory pointed to by state before returning on error from xfs_dir2_node_removename.c

---
Found by Coverity (134681) in userspace, same patch applies there also.

 fs/xfs/xfs_dir2_node.c |   10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)
fs/xfs/xfs_dir2_node.c +5 -5
Index: b/fs/xfs/xfs_dir2_node.c
===================================================================
@@ -2131,10 +2131,9 @@ xfs_dir2_node_removename(
 	/*
 	 * Didn't find it, upper layer screwed up.
 	 */
-	if (rval != EEXIST) {
-		xfs_da_state_free(state);
-		return rval;
-	}
+	if (rval != EEXIST)
+		goto done;
+
 	blk = &state->path.blk[state->path.active - 1];
 	ASSERT(blk->magic == XFS_DIR2_LEAFN_MAGIC);
 	ASSERT(state->extravalid);
@@ -2145,7 +2144,7 @@ xfs_dir2_node_removename(
 	error = xfs_dir2_leafn_remove(args, blk->bp, blk->index,
 		&state->extrablk, &rval);
 	if (error)
-		return error;
+		goto done;
 	/*
 	 * Fix the hash values up the btree.
 	 */
@@ -2160,6 +2159,7 @@ xfs_dir2_node_removename(
 	 */
 	if (!error)
 		error = xfs_dir2_node_to_leaf(state);
+done:
 	xfs_da_state_free(state);
 	return error;
 }

_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs

Reviewed-by on v1 1/1
#1 lore raw

Looks good,

Quoted 52 lines · Mark Tinguely

Free the memory pointed to by state before returning on error from xfs_dir2_node_removename.c

---
Found by Coverity (134681) in userspace, same patch applies there also.

 fs/xfs/xfs_dir2_node.c |   10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

Index: b/fs/xfs/xfs_dir2_node.c
===================================================================

fs/xfs/xfs_dir2_node.c +5 -5
@@ -2131,10 +2131,9 @@ xfs_dir2_node_removename(
 	/*
 	 * Didn't find it, upper layer screwed up.
 	 */
-	if (rval != EEXIST) {
-		xfs_da_state_free(state);
-		return rval;
-	}
+	if (rval != EEXIST)
+		goto done;
+
 	blk = &state->path.blk[state->path.active - 1];
 	ASSERT(blk->magic == XFS_DIR2_LEAFN_MAGIC);
 	ASSERT(state->extravalid);
@@ -2145,7 +2144,7 @@ xfs_dir2_node_removename(
 	error = xfs_dir2_leafn_remove(args, blk->bp, blk->index,
 		&state->extrablk, &rval);
 	if (error)
-		return error;
+		goto done;
 	/*
 	 * Fix the hash values up the btree.
 	 */
@@ -2160,6 +2159,7 @@ xfs_dir2_node_removename(
 	 */
 	if (!error)
 		error = xfs_dir2_node_to_leaf(state);
+done:
 	xfs_da_state_free(state);
 	return error;
 }

_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs

Signature · 6 lines
Carlos

_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs
Reviewed-by on v1 1/1
#2 lore raw
In reply to Mark Tinguely
Quoted 7 lines · Mark Tinguely

Free the memory pointed to by state before returning on error from xfs_dir2_node_removename.c

---
Found by Coverity (134681) in userspace, same patch applies there also.

Heh, looks like that one has been around since the dawn of time, thanks.

how do we handle the matching userspace fixes, separate patch to be explicit? Wait for the next syncup?

Thanks,
-Eric

Quoted 40 lines
 fs/xfs/xfs_dir2_node.c |   10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

Index: b/fs/xfs/xfs_dir2_node.c
===================================================================

fs/xfs/xfs_dir2_node.c +5 -5
@@ -2131,10 +2131,9 @@ xfs_dir2_node_removename(
 	/*
 	 * Didn't find it, upper layer screwed up.
 	 */
-	if (rval != EEXIST) {
-		xfs_da_state_free(state);
-		return rval;
-	}
+	if (rval != EEXIST)
+		goto done;
+
 	blk = &state->path.blk[state->path.active - 1];
 	ASSERT(blk->magic == XFS_DIR2_LEAFN_MAGIC);
 	ASSERT(state->extravalid);
@@ -2145,7 +2144,7 @@ xfs_dir2_node_removename(
 	error = xfs_dir2_leafn_remove(args, blk->bp, blk->index,
 		&state->extrablk, &rval);
 	if (error)
-		return error;
+		goto done;
 	/*
 	 * Fix the hash values up the btree.
 	 */
@@ -2160,6 +2159,7 @@ xfs_dir2_node_removename(
 	 */
 	if (!error)
 		error = xfs_dir2_node_to_leaf(state);
+done:
 	xfs_da_state_free(state);
 	return error;
 }
Quoted 5 lines

_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs

_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs

#3 lore raw
Quoted 11 lines · Eric Sandeen
Quoted 7 lines · Mark Tinguely

Free the memory pointed to by state before returning on error from xfs_dir2_node_removename.c

---
Found by Coverity (134681) in userspace, same patch applies there also.

Heh, looks like that one has been around since the dawn of time, thanks.

Quoted 7 lines

how do we handle the matching userspace fixes, separate patch to be explicit? Wait for the next syncup?

Thanks,
-Eric

<patch delete>

Good question.

The user space should be kept up to date with the kernel.

Since the patches will be identical except the directory name, I was hoping to submit one copy. But I am not trying to invent a policy, just being lazy.

--Mark.

_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs

#4 lore raw
Quoted 22 lines · Mark Tinguely
Quoted 11 lines · Eric Sandeen
Quoted 7 lines · Mark Tinguely

Free the memory pointed to by state before returning on error from xfs_dir2_node_removename.c

---
Found by Coverity (134681) in userspace, same patch applies there also.

Heh, looks like that one has been around since the dawn of time, thanks.

Quoted 7 lines

how do we handle the matching userspace fixes, separate patch to be explicit? Wait for the next syncup?

Thanks,
-Eric

<patch delete>

Quoted 8 lines

Good question.

The user space should be kept up to date with the kernel.

Since the patches will be identical except the directory name, I was hoping to submit one copy. But I am not trying to invent a policy, just being lazy.

--Mark.

Was just an offhanded question; it'd just be good to know what we all expect.

I suppose that it could depend on the severity of the flaw; a minor leak before exit() isn't a big deal and could wait for a global sync-up; a data corruption fix might need to be quickly merged to both trees.

-Eric

_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs

#5 lore raw
In reply to Mark Tinguely
Quoted 10 lines · Mark Tinguely

plain text document attachment (xfs-fix-leak-in-xfs_dir2_node_removename.patch)
Free the memory pointed to by state before returning on error from xfs_dir2_node_removename.c

---
Found by Coverity (134681) in userspace, same patch applies there also.

Is the first hunk right ?

xfs_da_node_lookup_int called as

        error = xfs_da_node_lookup_int(state, &rval);

and returns with

        *result = retval;
        return(0);

so, on return, error == 0 and rval == an error code. The next lines:

        if (error)
                rval = error;

won't change that. But previously if rval != EEXIST you returned rval. With the change below, you return error, which is zero.

--
Roger

Quoted 45 lines
 fs/xfs/xfs_dir2_node.c |   10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

Index: b/fs/xfs/xfs_dir2_node.c
===================================================================

fs/xfs/xfs_dir2_node.c +5 -5
@@ -2131,10 +2131,9 @@ xfs_dir2_node_removename(
 	/*
 	 * Didn't find it, upper layer screwed up.
 	 */
-	if (rval != EEXIST) {
-		xfs_da_state_free(state);
-		return rval;
-	}
+	if (rval != EEXIST)
+		goto done;
+
 	blk = &state->path.blk[state->path.active - 1];
 	ASSERT(blk->magic == XFS_DIR2_LEAFN_MAGIC);
 	ASSERT(state->extravalid);
@@ -2145,7 +2144,7 @@ xfs_dir2_node_removename(
 	error = xfs_dir2_leafn_remove(args, blk->bp, blk->index,
 		&state->extrablk, &rval);
 	if (error)
-		return error;
+		goto done;
 	/*
 	 * Fix the hash values up the btree.
 	 */
@@ -2160,6 +2159,7 @@ xfs_dir2_node_removename(
 	 */
 	if (!error)
 		error = xfs_dir2_node_to_leaf(state);
+done:
 	xfs_da_state_free(state);
 	return error;
 }

_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs

Signature · 6 lines
Roger Willcocks <[email protected]>

_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs
#6 lore raw
Quoted 26 lines · Roger Willcocks
Quoted 10 lines · Mark Tinguely

plain text document attachment (xfs-fix-leak-in-xfs_dir2_node_removename.patch)
Free the memory pointed to by state before returning on error from xfs_dir2_node_removename.c

---
Found by Coverity (134681) in userspace, same patch applies there also.

Is the first hunk right ?

xfs_da_node_lookup_int called as

         error = xfs_da_node_lookup_int(state,&rval);

and returns with

         *result = retval;
         return(0);

so, on return, error == 0 and rval == an error code. The next lines:

Quoted 8 lines
         if (error)
                 rval = error;

won't change that. But previously if rval != EEXIST you returned rval. With the change below, you return error, which is zero.

--
Roger

Thanks, guilty as charged. need to add the error = EEXIST.

--Mark.

_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs

#7 lore raw
In reply to Roger Willcocks
Quoted 24 lines · Roger Willcocks
Quoted 10 lines · Mark Tinguely

plain text document attachment (xfs-fix-leak-in-xfs_dir2_node_removename.patch)
Free the memory pointed to by state before returning on error from xfs_dir2_node_removename.c

---
Found by Coverity (134681) in userspace, same patch applies there also.

Is the first hunk right ?

xfs_da_node_lookup_int called as

        error = xfs_da_node_lookup_int(state, &rval);

and returns with

        *result = retval;
        return(0);
Quoted 7 lines

so, on return, error == 0 and rval == an error code. The next lines:

        if (error)
                rval = error;

won't change that. But previously if rval != EEXIST you returned rval. With the change below, you return error, which is zero.

Oh geez. o_O . . . Thanks Roger.

3rd review's the charm? ;)

-Eric

_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs