xfs
[Top] [All Lists]

Re: [PATCH 1/5] xfs: remove dead ENODEV handling in xfs_destroy_ioend

To: Dave Chinner <david@xxxxxxxxxxxxx>
Subject: Re: [PATCH 1/5] xfs: remove dead ENODEV handling in xfs_destroy_ioend
From: Christoph Hellwig <hch@xxxxxxxxxxxxx>
Date: Mon, 22 Aug 2011 10:27:46 -0400
Cc: xfs@xxxxxxxxxxx
In-reply-to: <20110822063245.GK3162@dastard>
References: <20110814222412.359079843@xxxxxxxxxxxxxxxxxxxxxx> <20110814222517.607954695@xxxxxxxxxxxxxxxxxxxxxx> <20110822063245.GK3162@dastard>
User-agent: Mutt/1.5.21 (2010-09-15)
On Mon, Aug 22, 2011 at 04:32:45PM +1000, Dave Chinner wrote:
> On Sun, Aug 14, 2011 at 06:24:13PM -0400, Christoph Hellwig wrote:
> > No driver returns ENODEV from it bio completion handler, not has this
> > ever been documented.  Remove the dead code dealing with it.
> > 
> > Signed-off-by: Christoph Hellwig <hch@xxxxxx>
> 
> IIRC, this was added years ago for XVM to cause filesystems to shut
> down when the storage went away (e.g. someone tripped over a cable
> bundle or fenced the machine).  It would be nice if there was some
> standard way of handling this rather than having the filesytem
> continue to allow applications to dirty memory when there is no
> chance of ever cleaning it....
> 
> As it is, I see no point in keeping something that won't ever be
> used, so

Initially I just made sure to defer contexts with ENODEV returns
to the workqueues as well, as we can't really shutdown the fs from
irq context.  But given that this code is dead, and we are mixing
up positive and negative errors in io_error I decided it's better
to remove it for now.

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