MIME-Version: 1.0
References: <20230221110344.82818-1-kerneljasonxing@gmail.com>
 <48429c16fdaee59867df5ef487e73d4b1bf099af.camel@redhat.com> <CANn89iJjCXfwUQ4XxtCrNFChdCHciBMuWcK=Az4X1acBeqVDiQ@mail.gmail.com>
In-Reply-To: <CANn89iJjCXfwUQ4XxtCrNFChdCHciBMuWcK=Az4X1acBeqVDiQ@mail.gmail.com>
From:   Jason Xing <kerneljasonxing@gmail.com>
Date:   Tue, 21 Feb 2023 21:44:25 +0800
Message-ID: <CAL+tcoAdYO_NnkWLYbxxRgw0=muhM0TJo3FBEeCBYtBmsnfWUw@mail.gmail.com>
Subject: Re: [PATCH net] udp: fix memory schedule error
To:     Eric Dumazet <edumazet@google.com>
Cc:     Paolo Abeni <pabeni@redhat.com>, willemdebruijn.kernel@gmail.com,
        davem@davemloft.net, dsahern@kernel.org, kuba@kernel.org,
        netdev@vger.kernel.org, linux-kernel@vger.kernel.org,
        bpf@vger.kernel.org, Jason Xing <kernelxing@tencent.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Tue, Feb 21, 2023 at 8:35 PM Eric Dumazet <edumazet@google.com> wrote:
>
> On Tue, Feb 21, 2023 at 1:27 PM Paolo Abeni <pabeni@redhat.com> wrote:
> >
> > On Tue, 2023-02-21 at 19:03 +0800, Jason Xing wrote:
> > > From: Jason Xing <kernelxing@tencent.com>
> > >
> > > Quoting from the commit 7c80b038d23e ("net: fix sk_wmem_schedule()
> > > and sk_rmem_schedule() errors"):
> > >
> > > "If sk->sk_forward_alloc is 150000, and we need to schedule 150001 bytes,
> > > we want to allocate 1 byte more (rounded up to one page),
> > > instead of 150001"
> >
> > I'm wondering if this would cause measurable (even small) performance
> > regression? Specifically under high packet rate, with BH and user-space
> > processing happening on different CPUs.
> >
> > Could you please provide the relevant performance figures?
> >
> > Thanks!
> >
> > Paolo
> >
>
> Probably not a big deal.
>

> TCP skb truesize can easily reach 180 KB, but for UDP it's 99% below
> or close to a 4K page.

Yes.

>
> I doubt this change makes any difference for UDP.

Based on my understanding of this part, it could not neither cause
much regression nor improve much performance. I think what you've done
to the TCP stack is the right way to go so the UDP can probably follow
this.
Calculating extra memory is a little bit odd because we actually don't
need that much when receiving skb everytime.

Thanks,
Jason