MIME-Version: 1.0
References: <20211229215646.830451-1-void@manifault.com> <1f1b9b01-cfab-8a84-f35f-c21172e5d64d@redhat.com>
 <CAPhsuW4Ua2hDs5WMtF0s_CQki-ZdYMvkU2s+Nc7Rvs=-D6WL=Q@mail.gmail.com> <4df1f9a0-3b26-903b-fe63-af5e75ed98d4@redhat.com>
In-Reply-To: <4df1f9a0-3b26-903b-fe63-af5e75ed98d4@redhat.com>
From:   Song Liu <song@kernel.org>
Date:   Mon, 10 Jan 2022 17:49:38 -0800
Message-ID: <CAPhsuW7O6br8BtAb0Tnx2aMJODfp5Rvxn-YABPUCuL1ofNPcyw@mail.gmail.com>
Subject: Re: [PATCH] livepatch: Avoid CPU hogging with cond_resched
To:     Joe Lawrence <joe.lawrence@redhat.com>
Cc:     David Vernet <void@manifault.com>, live-patching@vger.kernel.org,
        open list <linux-kernel@vger.kernel.org>, jpoimboe@redhat.com,
        pmladek@suse.com, jikos@kernel.org, mbenes@suse.cz
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Mon, Jan 10, 2022 at 8:17 AM Joe Lawrence <joe.lawrence@redhat.com> wrote:
>
> On 1/7/22 11:46 AM, Song Liu wrote:
> > On Fri, Jan 7, 2022 at 6:13 AM Joe Lawrence <joe.lawrence@redhat.com> wrote:
> >>
> >> On 12/29/21 4:56 PM, David Vernet wrote:
> >>> For example, under certain workloads, enabling a KLP patch with
> >>> many objects or functions may cause ksoftirqd to be starved, and thus for
> >>> interrupts to be backlogged and delayed.
> >>
> >> Just curious, approximately how many objects/functions does it take to
> >> hit this condition?  While the livepatching kselftests are more about
> >> API and kernel correctness, this sounds like an interesting test case
> >> for some of the other (out of tree) test suites.
> >
> > Not many patched functions. We only do small fixes at the moment. In the recent
> > example, we hit the issue with ~10 patched functions. Another version
> > with 2 to 3
> > patched function seems fine.
> >
> > Yes, I think this is an important test case.
> >
>
> Thanks, Song.  If you can share any test setup details, I'll pass those
> along to our internal QE group.  And once merged, we'll be adding this
> one to the list of backports for our distro.

We get this on shadow production traffic of our web service, which we cannot
recreate out of our data centers.
I tried to use iperf to generate traffic (among a few hosts), but it
doesn't work.
At the moment, I am not sure what's the easiest way to repro this.

Thanks,
Song