Re: [PATCH 2/2] vhost: lockless enqueuing

From: Pankaj Gupta <pagupta@redhat.com>
To: Jason Wang <jasowang@redhat.com>
Cc: mst@redhat.com, kvm@vger.kernel.org,
	virtualization@lists.linux-foundation.org,
	netdev@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH 2/2] vhost: lockless enqueuing
Date: Tue, 26 Apr 2016 03:57:10 -0400 (EDT)	[thread overview]
Message-ID: <19697601.46250110.1461657430875.JavaMail.zimbra@redhat.com> (raw)
In-Reply-To: <571F1330.7030504@redhat.com>

> 
> 
> 
> On 04/26/2016 02:24 PM, Pankaj Gupta wrote:
> > Hi Jason,
> >
> > Overall patches look good. Just one doubt I have is below:
> >> We use spinlock to synchronize the work list now which may cause
> >> unnecessary contentions. So this patch switch to use llist to remove
> >> this contention. Pktgen tests shows about 5% improvement:
> >>
> >> Before:
> >> ~1300000 pps
> >> After:
> >> ~1370000 pps
> >>
> >> Signed-off-by: Jason Wang <jasowang@redhat.com>
> >> ---
> >>  drivers/vhost/vhost.c | 52
> >>  +++++++++++++++++++++++++--------------------------
> >>  drivers/vhost/vhost.h |  7 ++++---
> >>  2 files changed, 29 insertions(+), 30 deletions(-)
> 
> [...]
> 
> >> -		if (work) {
> >> +		node = llist_del_all(&dev->work_list);
> >> +		if (!node)
> >> +			schedule();
> >> +
> >> +		node = llist_reverse_order(node);
> > Can we avoid llist reverse here?
> >
> 
> Probably not, this is because:
> 
> - we should process the work exactly the same order as they were queued,
> otherwise flush won't work
> - llist can only add a node to the head of list.

Got it.

Thanks,
> 
> Thanks
> 