Re: [vm] writing to UDF DVD+RW (/dev/sr0) while under memory pressure: box ==> doorstop

[Andrew Morton <akpm@linux-foundation.org>]

On Tue, 08 Jan 2008 12:05:16 +0100 Mike Galbraith <efault@gmx.de> wrote:

> 
> On Sun, 2008-01-06 at 12:29 -0800, Andrew Morton wrote:
> > On Sun, 06 Jan 2008 19:01:10 +0100 Mike Galbraith <efault@gmx.de> wrote:
> > 
> > > 
> > > On Sun, 2008-01-06 at 10:42 +0100, Mike Galbraith wrote:
> > > > FWIW, here's box having same seizure writing to /dev/pktcdvd/sr0.
> > > 
> > > Ugh, horrid device IO ==> horrid VM AFIKT.  I'll continue rummaging, but
> > > I wonder, am I doing something _so_ stupid as to warrant zero response?
> > > 
> > 
> > Copying linux-mm might have helped.  Bad time of year, too.
> > 
> > >From the traces it looks like all IO has stopped and everyone is waiting
> > (directly or indirectly) for certain IO completions.  Is that really true,
> > or is IO still dribbling along?
> > 
> > >From the kswapd0 trace it would appear that UDF is doing synchronous writes
> > (udf_update_inode->sync_dirty_buffer) which perhaps is making things
> > mightily slow.
> 
> This is a pretty uninteresting thread except for the VM aspect, but
> rather than leave it dangling, I'll laugh at my naivety and close.
> 
> Seeks on DVD drives are apparently far FAR more horrible that I ever
> imagined, so writing to a filesystem on a DVD+RW in effect constitutes
> the stupid thing I was worried about above.  Make an image, populate it,
> write it instead... sure to be a truckload faster, and due to the VM
> connection, much safer too.  Oh well.
> 

Well.  From your earlier trace it appeared that something was causing
the filesystem to perform synchronous inode writes - sync_dirty_buffer() was
called.

This will cause many more seeks than would occur if we were doing full
delayed writing, with obvious throughput implications.