xfs
[Top] [All Lists]

Re: [PATCH 1/2] xfs: fix xfs_mark_inode_dirty during umount

To: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Subject: Re: [PATCH 1/2] xfs: fix xfs_mark_inode_dirty during umount
From: Dave Chinner <david@xxxxxxxxxxxxx>
Date: Tue, 30 Aug 2011 16:24:16 +1000
Cc: xfs@xxxxxxxxxxx
In-reply-to: <20110827055744.GA28351@xxxxxxxxxxxxx>
References: <20110827055731.GA24159@xxxxxxxxxxxxx> <20110827055744.GA28351@xxxxxxxxxxxxx>
User-agent: Mutt/1.5.21 (2010-09-15)
On Sat, Aug 27, 2011 at 01:57:44AM -0400, Christoph Hellwig wrote:
> During umount we do not add a dirty inode to the lru and wait for it to
> become clean first, but force writeback of data and metadata with
> I_WILL_FREE set.  Currently there is no way for XFS to detect that the
> inode has been redirtied for metadata operations, as we skip the
> mark_inode_dirty call during teardown.  Fix this by setting i_update_core
> nanually in that case, so that the inode gets flushed during inode reclaim.
> 
> Alternatively we could enable calling mark_inode_dirty for inodes in
> I_WILL_FREE state, and let the VFS dirty tracking handle this.  I decided
> against this as we will get better I/O patterns from reclaim compared to
> the synchronous writeout in write_inode_now, and always marking the inode
> dirty in some way from xfs_mark_inode_dirty is a better safetly net in
> either case.
> 
> Signed-off-by: Christoph Hellwig <hch@xxxxxx>
> 
> Index: linux-2.6/fs/xfs/xfs_iops.c
> ===================================================================
> --- linux-2.6.orig/fs/xfs/xfs_iops.c  2011-08-26 12:31:19.090631739 +0200
> +++ linux-2.6/fs/xfs/xfs_iops.c       2011-08-26 12:35:43.692531800 +0200
> @@ -70,9 +70,8 @@ xfs_synchronize_times(
>  }
>  
>  /*
> - * If the linux inode is valid, mark it dirty.
> - * Used when committing a dirty inode into a transaction so that
> - * the inode will get written back by the linux code
> + * If the linux inode is valid, mark it dirty, else mark the dirty state
> + * in the XFS inode to make sure we pick it up when reclaiming the inode.
>   */
>  void
>  xfs_mark_inode_dirty_sync(
> @@ -82,6 +81,10 @@ xfs_mark_inode_dirty_sync(
>  
>       if (!(inode->i_state & (I_WILL_FREE|I_FREEING)))
>               mark_inode_dirty_sync(inode);
> +     else {
> +             barrier();
> +             ip->i_update_core = 1;
> +     }
>  }

Why the barrier()? Isn't that just a compiler barrier? If you are
worried about catching the update vs clearing it in transaction
commit, shouldn't that use smp_mb() instead (in both places)?

Cheers,

Dave.
-- 
Dave Chinner
david@xxxxxxxxxxxxx

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