netdev - Re: [net] bonding: annotate data-races around slave->last

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]

Message-ID: <c7e1b0e6-c405-4dde-aeec-c1e4a95532be@meta.com>
Date: Thu, 22 Jan 2026 12:42:35 -0500
From: Chris Mason <clm@...a.com>
To: Jakub Kicinski <kuba@...nel.org>, Eric Dumazet <edumazet@...gle.com>
Cc: netdev@...r.kernel.org, syzkaller@...glegroups.com, andy@...yhouse.net,
        eric.dumazet@...il.com, pabeni@...hat.com, j.vosburgh@...il.com,
        davem@...emloft.net
Subject: Re: [net] bonding: annotate data-races around slave->last_rx



On 1/22/26 11:58 AM, Jakub Kicinski wrote:
> On Thu, 22 Jan 2026 16:56:33 +0100 Eric Dumazet wrote:
>> On Thu, Jan 22, 2026 at 4:50 PM Eric Dumazet <edumazet@...gle.com> wrote:
>>> On Thu, Jan 22, 2026 at 4:29 PM Jakub Kicinski <kuba@...nel.org> wrote:  
>>>> I guess if nothing else we should do it for consistency?  
>>>
>>> To be clear, I will cook a V2 with this part.
>>>  
>>
>> _bond_options_ns_ip6_target_set() probably needs a fix as well.
>>
>> Is AI review stopping at the first error ?
> 
> I don't think it explicitly stops after the first occurrence but it
> seems to have limited ability to dig around. Simplifying, it only digs
> around for 5-10min, and it spends a lot of time proving that the
> issue is real. So once it's spend time proving one issue it's often out
> of time, and may not search for another one :(
> 
> Ccing Chris, hopefully he'll chime in later if I'm wrong.

It's supposed to keep hunting for bugs until it has reviewed the entire
thing, and then run all the bugs it found through the false positive pass.

But, sometimes it gets really excited and jumps out early, or as Jakub
mentions sometimes the context window fills and it starts skipping
steps.  I've tried different ways to force things, but unfortunately
none of them have worked 100% of the time.

-chris