xfs
[Top] [All Lists]

Re: [PATCH] xfs: force buffer writeback before blocking on the ilock in

To: xfs@xxxxxxxxxxx
Subject: Re: [PATCH] xfs: force buffer writeback before blocking on the ilock in inode reclaim
From: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Date: Sun, 20 Nov 2011 02:59:18 -0500
Cc: Simon Kirby <sim@xxxxxxxxxx>
In-reply-to: <20111120072334.GA27386@xxxxxxxxxxxxx>
References: <20111120072334.GA27386@xxxxxxxxxxxxx>
User-agent: Mutt/1.5.21 (2010-09-15)
I should probably keep Simon on Cc for this one..

On Sun, Nov 20, 2011 at 02:23:34AM -0500, Christoph Hellwig wrote:
> If we are doing synchronous inode reclaim we block the VM from making
> progress in memory reclaim.  So if we encouter a flush locked inode
> make sure we force out all delayed buffers ASAP to speed up the wait
> for it to be unlocked.  Without this we can get hangs of up to 30
> seconds during workloads hitting synchronous inode reclaim.
> 
> Reported-by: Simon Kirby <sim@xxxxxxxxxx>
> Tested-by: Simon Kirby <sim@xxxxxxxxxx>
> Signed-off-by: Christoph Hellwig <hch@xxxxxx>
> 
> Index: xfs/fs/xfs/xfs_sync.c
> ===================================================================
> --- xfs.orig/fs/xfs/xfs_sync.c        2011-11-19 20:14:52.110141228 +0100
> +++ xfs/fs/xfs/xfs_sync.c     2011-11-19 20:40:17.381878121 +0100
> @@ -762,7 +762,8 @@ xfs_reclaim_inode(
>       struct xfs_perag        *pag,
>       int                     sync_mode)
>  {
> -     int     error;
> +     struct xfs_mount        *mp = ip->i_mount;
> +     int                     error;
>  
>  restart:
>       error = 0;
> @@ -770,12 +771,25 @@ restart:
>       if (!xfs_iflock_nowait(ip)) {
>               if (!(sync_mode & SYNC_WAIT))
>                       goto out;
> +
> +             /*
> +              * If we only have a single dirty inode in a cluster there is
> +              * a fair chance that the AIL push may have pushed it into
> +              * the buffer, but xfsbufd won't touch it until 30 seconds
> +              * from now, and thus we will lock up here.
> +              *
> +              * Wakeup xfsbufd now, and force it to write back even
> +              * recently dirtied buffers.
> +              */
> +             set_bit(XBT_FORCE_FLUSH, &mp->m_ddev_targp->bt_flags);
> +             wake_up_process(mp->m_ddev_targp->bt_task);
> +
>               xfs_iflock(ip);
>       }
>  
>       if (is_bad_inode(VFS_I(ip)))
>               goto reclaim;
> -     if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
> +     if (XFS_FORCED_SHUTDOWN(mp)) {
>               xfs_iunpin_wait(ip);
>               goto reclaim;
>       }
> @@ -829,8 +843,8 @@ restart:
>        * is permanent then the next sync reclaim will reclaim the inode and
>        * pass on the error.
>        */
> -     if (error && error != EAGAIN && !XFS_FORCED_SHUTDOWN(ip->i_mount)) {
> -             xfs_warn(ip->i_mount,
> +     if (error && error != EAGAIN && !XFS_FORCED_SHUTDOWN(mp)) {
> +             xfs_warn(mp,
>                       "inode 0x%llx background reclaim flush failed with %d",
>                       (long long)ip->i_ino, error);
>       }
> @@ -860,7 +874,7 @@ reclaim:
>        */
>       spin_lock(&pag->pag_ici_lock);
>       if (!radix_tree_delete(&pag->pag_ici_root,
> -                             XFS_INO_TO_AGINO(ip->i_mount, ip->i_ino)))
> +                             XFS_INO_TO_AGINO(mp, ip->i_ino)))
>               ASSERT(0);
>       __xfs_inode_clear_reclaim(pag, ip);
>       spin_unlock(&pag->pag_ici_lock);
> 
> _______________________________________________
> xfs mailing list
> xfs@xxxxxxxxxxx
> http://oss.sgi.com/mailman/listinfo/xfs
---end quoted text---

<Prev in Thread] Current Thread [Next in Thread>