[RFC] Patch to fix reiserfs bad path release panic on 2.6.19-rc1

[Suzuki K P]

hi,

One of our test team hit a reiserfs_panic while running fsstress tests 
on 2.6.19-rc1. The message looks like :

<0>REISERFS: panic(device Null superblock): reiserfs[5676]: assertion 
!(p->path_length != 1 ) failed at 
fs/reiserfs/stree.c:397:reiserfs_check_path: path not properly relsed.

The backtrace looked :

kernel BUG in reiserfs_panic at fs/reiserfs/prints.c:361!
3:mon> t
[c0000000688aafb0] c00000000014f964 .reiserfs_check_path+0x58/0x74
[c0000000688ab030] c000000000135474 .reiserfs_get_block+0x1444/0x1508
[c0000000688ab2a0] c0000000000cdc8c .__block_prepare_write+0x1c8/0x558
[c0000000688ab3d0] c0000000000ce050 .block_prepare_write+0x34/0x64
[c0000000688ab450] c000000000132784 .reiserfs_prepare_write+0x118/0x1d0
[c0000000688ab500] c00000000009df14 .generic_file_buffered_write+0x314/0x82c
[c0000000688ab6b0] c00000000009e77c 
.__generic_file_aio_write_nolock+0x350/0x3e0
[c0000000688ab7e0] c00000000009e988 .__generic_file_write_nolock+0x78/0xb0
[c0000000688ab950] c00000000009eb18 .generic_file_write+0x60/0xf0
[c0000000688aba10] c000000000138d2c .reiserfs_file_write+0x198/0x2038
[c0000000688abcf0] c0000000000ca1a8 .vfs_write+0xd0/0x1b4
[c0000000688abd90] c0000000000ca8c8 .sys_write+0x4c/0x8c
[c0000000688abe30] c00000000000871c syscall_exit+0x0/0x4



Upon debugging I found that the restart_transaction was not releasing 
the path if the th->refcount was > 1.

/*static*/
int restart_transaction(struct reiserfs_transaction_handle *th, 
                          			struct inode *inode, struct path *path)
{
	[...]

        /* we cannot restart while nested */
        if (th->t_refcount > 1) { <<- Path is not released in this case!
                return 0;
        }

        pathrelse(path); <<- Path released here.
	[...]



This could happen in such a situation :

In reiserfs/inode.c: reiserfs_get_block() ::

     if (repeat == NO_DISK_SPACE || repeat == QUOTA_EXCEEDED) {
         /* restart the transaction to give the journal a chance to free
          ** some blocks.  releases the path, so we have to go back to
          ** research if we succeed on the second try
          */
         SB_JOURNAL(inode->i_sb)->j_next_async_flush = 1;

       -->>  retval = restart_transaction(th, inode, &path); <<--

 We are supposed to release the path, no matter we succeed or fail. But 
if the th->refcount is > 1, the path is still valid. And,

         if (retval)
                  goto failure;
         repeat =
             _allocate_block(th, block, inode,
                            &allocated_block_nr, NULL, create);

If the above allocate_block fails with NO_DISK_SPACE or QUOTA_EXCEEDED, 
we would have path which is not released.


        if (repeat != NO_DISK_SPACE && repeat != QUOTA_EXCEEDED) {
                  goto research;
        }
        if (repeat == QUOTA_EXCEEDED)
                  retval = -EDQUOT;
        else
                  retval = -ENOSPC;
        goto failure;
	[...]

      failure:
	[...]
        reiserfs_check_path(&path); << Panics here !


Attached here is a patch which could fix the issue.

Comments ?

Thanks,

Suzuki K P
Linux Technology Center,
IBM Systems & Technology Labs.
* fix reiserfs/inode.c : restart_transaction() to release the path in all cases.

The restart_transaction() doesn't release the path when the the journal handle has a refcount > 1. This would trigger a reiserfs_panic() if we encounter an -ENOSPC / -EDQUOT in reiserfs_get_block(). 


Signed-off-by: Suzuki K P <suzuki@xxxxxxxxxx>

Index: linux-2.6.19-rc1/fs/reiserfs/inode.c
===================================================================
--- linux-2.6.19-rc1.orig/fs/reiserfs/inode.c	2006-10-12 02:13:12.000000000 -0700
+++ linux-2.6.19-rc1/fs/reiserfs/inode.c	2006-10-13 16:38:32.000000000 -0700
@@ -216,11 +216,12 @@
 	BUG_ON(!th->t_trans_id);
 	BUG_ON(!th->t_refcount);
 
+	pathrelse(path);
+
 	/* we cannot restart while nested */
 	if (th->t_refcount > 1) {
 		return 0;
 	}
-	pathrelse(path);
 	reiserfs_update_sd(th, inode);
 	err = journal_end(th, s, len);
 	if (!err) {