netdev - Re: [PATCH net] net: don't let netpoll invoke NAPI if in xmit context

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: <20230330202333.108dadd9@kernel.org>
Date:   Thu, 30 Mar 2023 20:23:33 -0700
From:   Jakub Kicinski <kuba@...nel.org>
To:     Eric Dumazet <edumazet@...gle.com>
Cc:     davem@...emloft.net, netdev@...r.kernel.org, pabeni@...hat.com,
        Roman Gushchin <roman.gushchin@...ux.dev>, leitao@...ian.org,
        shemminger@...ux.foundation.org
Subject: Re: [PATCH net] net: don't let netpoll invoke NAPI if in xmit
 context

On Fri, 31 Mar 2023 04:41:23 +0200 Eric Dumazet wrote:
> Note that we update WRITE_ONCE(txq->xmit_lock_owner, cpu) _after_
> spin_lock(&txq->_xmit_lock);
> 
> So there is a tiny window I think, for missing that we got the
> spinlock, but I do not see how to avoid it without excessive cost.

Ugh, true. Hopefully the chances of taking an IRQ which tries to print
something between those two instructions are fairly low. 

I was considering using dev_recursion_level() but AFAICT we don't
currently bump it when dequeuing from the qdisc..