netdev - Re: [PATCH net 3/3] bnxt_en: Update MRU and RSS table of RSS contexts on queue reset

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: <20250522082650.3c4a5bb2@kernel.org>
Date: Thu, 22 May 2025 08:26:50 -0700
From: Jakub Kicinski <kuba@...nel.org>
To: David Wei <dw@...idwei.uk>
Cc: Michael Chan <michael.chan@...adcom.com>, davem@...emloft.net,
 netdev@...r.kernel.org, edumazet@...gle.com, pabeni@...hat.com,
 andrew+netdev@...n.ch, pavan.chebbi@...adcom.com,
 andrew.gospodarek@...adcom.com
Subject: Re: [PATCH net 3/3] bnxt_en: Update MRU and RSS table of RSS
 contexts on queue reset

On Thu, 22 May 2025 12:01:34 +0100 David Wei wrote:
> On 5/20/25 19:29, Michael Chan wrote:
> > On Tue, May 20, 2025 at 7:17 PM Jakub Kicinski <kuba@...nel.org> wrote:  
> >> "reliable" is a bit of a big word that some people would reserve
> >> for code which is production tested or at the very least very
> >> heavily validated.  
> > 
> > FWIW, queue_mgmt_ops was heavily tested by Somnath under heavy traffic
> > conditions.  Obviously RSS contexts were not included during testing
> > and this problem was missed.  
> 
> IIRC from the initial testing w/ Somnath even though the VNICs are reset
> the traffic on unrelated queues are unaffected.

How did you check that? IIUC the device does not currently report
packet loss due to MRU clamp (!?!)

> If we ensure that is the cse with this patchset, would that resolve
> your concerns Jakub?

For ZC we expect the queues to be taken out of the main context.
IIUC it'd be a significant improvement over the status quo if
we could check which contexts the queue is in (incl. context 0)
and only clamp MRU on those.