On Sun, May 14, 2023 at 8:46 PM Zhang, Cathy <cathy.zhang@xxxxxxxxx> wrote:
>
>
>
> > -----Original Message-----
> > From: Shakeel Butt <shakeelb@xxxxxxxxxx>
> > Sent: Saturday, May 13, 2023 1:17 AM
> > To: Zhang, Cathy <cathy.zhang@xxxxxxxxx>
> > Cc: Shakeel Butt <shakeelb@xxxxxxxxxx>; Eric Dumazet
> > <edumazet@xxxxxxxxxx>; Linux MM <linux-mm@xxxxxxxxx>; Cgroups
> > <cgroups@xxxxxxxxxxxxxxx>; Paolo Abeni <pabeni@xxxxxxxxxx>;
> > davem@xxxxxxxxxxxxx; kuba@xxxxxxxxxx; Brandeburg@xxxxxxxxxx;
> > Brandeburg, Jesse <jesse.brandeburg@xxxxxxxxx>; Srinivas, Suresh
> > <suresh.srinivas@xxxxxxxxx>; Chen, Tim C <tim.c.chen@xxxxxxxxx>; You,
> > Lizhen <lizhen.you@xxxxxxxxx>; eric.dumazet@xxxxxxxxx;
> > netdev@xxxxxxxxxxxxxxx
> > Subject: Re: [PATCH net-next 1/2] net: Keep sk->sk_forward_alloc as a proper
> > size
> >
> > On Fri, May 12, 2023 at 05:51:40AM +0000, Zhang, Cathy wrote:
> > >
> > >
> > [...]
> > > >
> > > > Thanks a lot. This tells us that one or both of following scenarios
> > > > are
> > > > happening:
> > > >
> > > > 1. In the softirq recv path, the kernel is processing packets from
> > > > multiple memcgs.
> > > >
> > > > 2. The process running on the CPU belongs to memcg which is
> > > > different from the memcgs whose packets are being received on that CPU.
> > >
> > > Thanks for sharing the points, Shakeel! Is there any trace records you
> > > want to collect?
> > >
> >
> > Can you please try the following patch and see if there is any improvement?
>
> Hi Shakeel,
>
> Try the following patch, the data of 'perf top' from system wide indicates that
> the overhead of page_counter_cancel is dropped from 15.52% to 4.82%.
>
> Without patch:
> 15.52% [kernel] [k] page_counter_cancel
> 12.30% [kernel] [k] page_counter_try_charge
> 11.97% [kernel] [k] try_charge_memcg
>
> With patch:
> 10.63% [kernel] [k] page_counter_try_charge
> 9.49% [kernel] [k] try_charge_memcg
> 4.82% [kernel] [k] page_counter_cancel
>
> The patch is applied on the latest net-next/main:
> befcc1fce564 ("sfc: fix use-after-free in efx_tc_flower_record_encap_match()")
>
Thanks a lot Cathy for testing. Do you see any performance improvement
for the memcached benchmark with the patch?
