>
> There's a couple of different ways I can see to fix the problem -
> the first is to not reference the buffer in xlog_iodone() after
> running the callbacks that may trigger it being freed. I'd prfer to
> see if this fixes the problem before having to do more invasive
> surgery. Can you try the patch below to see if it fixes the
> problem?
>
> fs/xfs/xfs_log.c | 11 ++++-------
> 1 file changed, 4 insertions(+), 7 deletions(-)
>
> Index: 2.6.x-xfs-new/fs/xfs/xfs_log.c
> ===================================================================
> --- 2.6.x-xfs-new.orig/fs/xfs/xfs_log.c 2007-04-03 09:09:36.000000000
> +1000
> +++ 2.6.x-xfs-new/fs/xfs/xfs_log.c 2007-04-16 11:40:21.655306665 +1000
> @@ -988,14 +988,11 @@ xlog_iodone(xfs_buf_t *bp)
> } else if (iclog->ic_state & XLOG_STATE_IOERROR) {
> aborted = XFS_LI_ABORTED;
> }
> + /* log I/O is always issued ASYNC, so we should see that here */
I guess this is a left over because at a prior time
xlog_sync() took an extra flags param (which could have XFS_LOG_SYNC set)
which could do a SYNC write of the iclog.
IIRC, we took this extra param out because nobody was ever calling with
it set for xlog_sync().
> + WARN_ON(!(XFS_BUF_ISASYNC(bp)));
> xlog_state_done_syncing(iclog, aborted);
> - if (!(XFS_BUF_ISASYNC(bp))) {
> - /*
> - * Corresponding psema() will be done in bwrite(). If we don't
> - * vsema() here, panic.
> - */
> - XFS_BUF_V_IODONESEMA(bp);
> - }
> + /* do not reference bp here - it may have been freed during unmount */
> +
> } /* xlog_iodone */
>
> /*
>
>
--Tim
|