netdev - Re: [Patch net v2 11/11] selftests/tc-testing: Add a test case for FQ

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

Message-ID: <e909b2a0-244e-4141-9fa9-1b7d96ab7d71@mojatatu.com>
Date: Mon, 7 Apr 2025 17:20:21 -0300
From: Victor Nogueira <victor@...atatu.com>
To: Jakub Kicinski <kuba@...nel.org>, Pedro Tammela <pctammela@...atatu.com>
Cc: Cong Wang <xiyou.wangcong@...il.com>, netdev@...r.kernel.org,
 jhs@...atatu.com, jiri@...nulli.us
Subject: Re: [Patch net v2 11/11] selftests/tc-testing: Add a test case for
 FQ_CODEL with ETS parent

On 4/7/25 17:09, Jakub Kicinski wrote:
> On Fri, 4 Apr 2025 16:03:26 -0300 Victor Nogueira wrote:
>>> Any ideas what is causing the IFE failure? Looks like it started
>>> happening when this series landed in the testing tree but I don't
>>> see how it could be related ?
>>
>> Yes, I saw that, but since it succeeded on retry and, as you said,
>> it doesn't seem to be related to this series, it looks more like
>> those IFE tests are a bit unstable. I talked to Pedro and we are
>> taking a look at it.
> 
> I dropped this set from the queue temporarily, and the failure
> went away (net-next-2025-04-07--18-00).
> Now I'm less inclined to think the IFE failure is not related to
> the series. But since the retry passes I'm not sure if Cong will
> be able to debug this.
> 
> Could someone on Mojatatu side take a closer look please?

We reached a different conclusion here.

Went through it during the weekend and today with Pedro.
IFE relies on some "sub-modules" for it to work (like
act_meta_skbprio and act_meta_skbtcindex). The issue is that
when running tdc in parallel (with more than 16 cores), the
act_meta_skbtcindex module is not loaded in time, which causes,
for example, the failure in test 219f.
Likely there might be a recent change in rwlock? not sure.
When we fix tdc.sh to pre load these modules, that tests
succeeded again. Will send a patch

cheers,
Victor

PS: It may have worked for you (in net-next-2025-04-07--18-00)
because we applied the patch we are going to send earlier.