Re: [PATCH] cancel the setfilesize transation when io error happen

[Dave Chinner <david@xxxxxxxxxxxxx>]

On Tue, Sep 15, 2015 at 07:42:45AM -0400, Brian Foster wrote:
> On Thu, Sep 10, 2015 at 09:40:10AM +0800, Zhaohongjiang wrote:
> > When I ran xfstest/073 case, the remount process was blocked to wait
> > transactions to be zero. I found there was a io error happened, and the
> > setfilesize transaction was not released properly. We should add the
> > changes to cancel the io error in this case.
> > 
> 
> Thanks for the patch. Was the io error by chance or part of the test? I
> ask because it would be good to have test coverage for this if we don't
> currently.
> 
> > Reproduction steps:
> > 1. dd if=/dev/zero of=xfs1.img bs=1M count=2048
> > 2. mkfs.xfs xfs1.img
> > 3. losetup -f ./xfs1.img /dev/loop0
> > 4. mount -t xfs /dev/loop0 /home/test_dir/
> > 5. mkdir /home/test_dir/test
> > 6. mkfs.xfs -dfile,name=image,size=2g
> > 7. mount -t xfs -o loop image /home/test_dir/test
> > 8. cp a file bigger than 2g to /home/test_dir/test
> > 9. mount -t xfs -o remount,ro /home/test_dir/test
> > 
> > Signed-off-by: Zhao Hongjiang <zhaohongjiang@xxxxxxxxxx>
> > ---
> >  fs/xfs/xfs_aops.c | 15 ++++++++++++++-
> >  1 file changed, 14 insertions(+), 1 deletion(-)
> > 
> > diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c
> > index 50ab287..fab55da 100644
> > --- a/fs/xfs/xfs_aops.c
> > +++ b/fs/xfs/xfs_aops.c
> > @@ -212,8 +212,21 @@ xfs_end_io(
> >             ioend->io_error = -EIO;
> >             goto done;
> >     }
> > -   if (ioend->io_error)
> > +   if (ioend->io_error) {
> > +           /*
> > +            * We should cancel the setfilesize transation when io error
> > +            * happen.
> > +            */
> > +           if (ioend->io_append_trans) {
> > +                   
> > current_set_flags_nested(&ioend->io_append_trans->t_pflags,
> > +                                                           PF_FSTRANS);
> > +                   rwsem_acquire_read(&VFS_I(XFS_I(ioend->io_inode))->i_sb
> > +                           ->s_writers.lock_map[SB_FREEZE_FS-1], 0, 1, 
> > _THIS_IP_);
> > +                   xfs_trans_cancel(ioend->io_append_trans);               
> >         
> 
> Trailing whitespace on the line above.
> 
> That aside, I wonder if it would be cleaner to put this stuff in a
> helper similar to xfs_setfilesize_ioend(). In fact, we could probably
> just update xfs_setfilesize_ioend() to check ioend->io_error after
> fixing up the task flags and whatnot and cancel the transaction instead
> of calling xfs_setfilesize(). Then, call it here also and update the new
> comment to point out that it cancels the transaction on error. Thoughts?

Agreed, I made this modification on commit. See below.

Cheers,

Dave.
-- 
Dave Chinner
david@xxxxxxxxxxxxx

cancel the setfilesize transation when io error happen

From: Zhaohongjiang <zhaohongjiang@xxxxxxxxxx>

When I ran xfstest/073 case, the remount process was blocked to wait
transactions to be zero. I found there was a io error happened, and the
setfilesize transaction was not released properly. We should add the
changes to cancel the io error in this case.

Reproduction steps:
1. dd if=/dev/zero of=xfs1.img bs=1M count=2048
2. mkfs.xfs xfs1.img
3. losetup -f ./xfs1.img /dev/loop0
4. mount -t xfs /dev/loop0 /home/test_dir/
5. mkdir /home/test_dir/test
6. mkfs.xfs -dfile,name=image,size=2g
7. mount -t xfs -o loop image /home/test_dir/test
8. cp a file bigger than 2g to /home/test_dir/test
9. mount -t xfs -o remount,ro /home/test_dir/test

[ dchinner: moved io error detection to xfs_setfilesize_ioend() after
  transaction context restoration. ]

Signed-off-by: Zhao Hongjiang <zhaohongjiang@xxxxxxxxxx>
Signed-off-by: Dave Chinner <david@xxxxxxxxxxxxx>
---
 fs/xfs/xfs_aops.c | 13 +++++++++++--
 1 file changed, 11 insertions(+), 2 deletions(-)

diff --git a/fs/xfs/xfs_aops.c b/fs/xfs/xfs_aops.c
index aa9b441..c1c87f9 100644
--- a/fs/xfs/xfs_aops.c
+++ b/fs/xfs/xfs_aops.c
@@ -172,6 +172,12 @@ xfs_setfilesize_ioend(
        current_set_flags_nested(&tp->t_pflags, PF_FSTRANS);
        __sb_writers_acquired(VFS_I(ip)->i_sb, SB_FREEZE_FS);
 
+       /* we abort the update if there was an IO error */
+       if (ioend->io_error) {
+               xfs_trans_cancel(tp);
+               return ioend->io_error;
+       }
+
        return xfs_setfilesize(ip, tp, ioend->io_offset, ioend->io_size);
 }
 
@@ -212,14 +218,17 @@ xfs_end_io(
                ioend->io_error = -EIO;
                goto done;
        }
-       if (ioend->io_error)
-               goto done;
 
        /*
         * For unwritten extents we need to issue transactions to convert a
         * range to normal written extens after the data I/O has finished.
+        * Detecting and handling completion IO errors is done individually
+        * for each case as different cleanup operations need to be performed
+        * on error.
         */
        if (ioend->io_type == XFS_IO_UNWRITTEN) {
+               if (ioend->io_error)
+                       goto done;
                error = xfs_iomap_write_unwritten(ip, ioend->io_offset,
                                                  ioend->io_size);
        } else if (ioend->io_append_trans) {