netdev - Re: regression with napi/softirq ?

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date:   Fri, 19 Jul 2019 16:53:28 +0100
From:   Sudip Mukherjee <sudipm.mukherjee@...il.com>
To:     Eric Dumazet <eric.dumazet@...il.com>
Cc:     Thomas Gleixner <tglx@...utronix.de>,
        "Peter Zijlstra (Intel)" <peterz@...radead.org>,
        "David S. Miller" <davem@...emloft.net>,
        netdev <netdev@...r.kernel.org>,
        linux-kernel <linux-kernel@...r.kernel.org>
Subject: Re: regression with napi/softirq ?

On Thu, Jul 18, 2019 at 4:08 PM Eric Dumazet <eric.dumazet@...il.com> wrote:
>
>
>
> On 7/18/19 2:55 PM, Sudip Mukherjee wrote:
>
> > Thanks Eric. But there is no improvement in delay between
> > softirq_raise and softirq_entry with this change.
> > But moving to a later kernel (linus master branch? ) like Thomas has
> > said in the other mail might be difficult atm. I can definitely
> > move to v4.14.133 if that helps. Thomas ?
>
> If you are tracking max latency then I guess you have to tweak SOFTIRQ_NOW_MASK
> to include NET_RX_SOFTIRQ
>
> The patch I gave earlier would only lower the probability of events, not completely get rid of them.
>
>
>
> diff --git a/kernel/softirq.c b/kernel/softirq.c
> index 0427a86743a46b7e1891f7b6c1ff585a8a1695f5..302046dd8d7e6740e466c422954f22565fe19e69 100644
> --- a/kernel/softirq.c
> +++ b/kernel/softirq.c
> @@ -81,7 +81,7 @@ static void wakeup_softirqd(void)
>   * right now. Let ksoftirqd handle this at its own rate, to get fairness,
>   * unless we're doing some of the synchronous softirqs.
>   */
> -#define SOFTIRQ_NOW_MASK ((1 << HI_SOFTIRQ) | (1 << TASKLET_SOFTIRQ))
> +#define SOFTIRQ_NOW_MASK ((1 << HI_SOFTIRQ) | (1 << TASKLET_SOFTIRQ) | (1 << NET_RX_SOFTIRQ))
>  static bool ksoftirqd_running(unsigned long pending)
>  {
>         struct task_struct *tsk = __this_cpu_read(ksoftirqd);

Thanks Eric, this looks better than the hack that tglx gave. :)
Though the hack was good for testing.

But my original problem was a drop is network packets and till now I
was thinking
that the delay in processing the softirq is causing that. But with the
hack tglx has given
the latency has decreased but my problem is still there.
So, I am looking into it again now.

Thanks again for the patch.


-- 
Regards
Sudip