Re: [Qemu-devel] [PATCH] raw-posix: Linearize direct I/O on Linux NFS

[Anthony Liguori <aliguori@us.ibm.com>]

On 04/15/2011 10:34 AM, Christoph Hellwig wrote:
> On Fri, Apr 15, 2011 at 04:26:41PM +0100, Stefan Hajnoczi wrote:
>> On Fri, Apr 15, 2011 at 4:05 PM, Christoph Hellwig<hch@lst.de>  wrote:
>>> NAK. ?Just wait for the bloody NFS client fix to get in instead of
>>> adding crap like that.
>> That's totally fine if NFS client will be fixed in the near future but
>> this doesn't seem likely:
>>
>> http://www.spinics.net/lists/linux-nfs/msg20462.html
> The code to use preadv/pwritev has been in qemu for over 2 years,
> and it took people to notice the NFS slowdown until now, so don't
> expect it to be fixed three days layer.
>
> I can't event see you in the relevent threads arguing for getting it
> fixed, so don't complain.

In general, since we are userspace, we should try to run well on 
whatever kernel we're on.

What I don't like about this patch is that likelihood of false 
positives.  We check for Linux and for NFS but that means an old 
userspace is doing unoptimal things on newer kernels.  Even if we had a 
kernel version check, if the fix gets backported to an older kernel, 
we'll still get a false positive.

Ideally, we'd be able to query the kernel to see whether we should 
bounce or not.  But AFAIK there is nothing even close to an interface to 
do this today.

Regards,

Anthony Liguori