[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20120626194634.GA14054@redhat.com>
Date: Tue, 26 Jun 2012 22:46:34 +0300
From: "Michael S. Tsirkin" <mst@...hat.com>
To: Jason Wang <jasowang@...hat.com>
Cc: Eric Dumazet <eric.dumazet@...il.com>, akong@...hat.com,
habanero@...ux.vnet.ibm.com, tahm@...ux.vnet.ibm.com,
haixiao@...iper.net, jwhan@...ewood.snu.ac.kr,
ernesto.martin@...sat.com, mashirle@...ibm.com,
davem@...emloft.net, netdev@...r.kernel.org,
linux-kernel@...r.kernel.org, krkumar2@...ibm.com,
shemminger@...tta.com, edumazet@...gle.com
Subject: Re: [PATCH 5/6] tuntap: per queue 64 bit stats\
On Tue, Jun 26, 2012 at 02:00:53PM +0800, Jason Wang wrote:
> On 06/25/2012 08:52 PM, Eric Dumazet wrote:
> >On Mon, 2012-06-25 at 19:59 +0800, Jason Wang wrote:
> >>As we've added multiqueue support for tun/tap, this patch convert the statistics
> >>to use per-queue 64 bit statistics.
> >LLTX means you can have several cpus calling TX path in parallel.
> >
> >So tx stats are wrong (even before this patch), and racy after this
> >patch (if several cpu access same queue, it seems to be possible)
> >
> > u64_stats_update_begin(&tfile->stats.tx_syncp);
> > tfile->stats.tx_packets++;
> > tfile->stats.tx_bytes += total;
> > u64_stats_update_end(&tfile->stats.tx_syncp);
> >
> >This can break horribly if several cpus run this code using same 'tfile'
> >pointer.
>
> Yes, looks like it's hard to use NETIF_F_LLTX without breaking the
> u64 statistics, may worth to use tx lock and alloc_netdev_mq().
Or make them per cpu as most everyone did.
> >I suggest this patch comes before 'tuntap: multiqueue support' in the
> >serie.
>
> Sure, thanks.
> >
> >
> >
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Powered by blists - more mailing lists