[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAL+tcoDR7O327uDMLgZ_74X_Vkvnx1ACpcs0BMNjTZRD0pY9kw@mail.gmail.com>
Date: Tue, 22 Jul 2025 19:50:57 +0800
From: Jason Xing <kerneljasonxing@...il.com>
To: Paolo Abeni <pabeni@...hat.com>
Cc: Jeroen de Borst <jeroendb@...gle.com>, netdev@...r.kernel.org, hramamurthy@...gle.com,
davem@...emloft.net, edumazet@...gle.com, kuba@...nel.org, willemb@...gle.com,
Joshua Washington <joshwash@...gle.com>, Praveen Kaligineedi <pkaligineedi@...gle.com>
Subject: Re: [PATCH net-next v2 4/5] gve: implement DQO TX datapath for AF_XDP zero-copy
On Tue, Jul 22, 2025 at 5:32 PM Paolo Abeni <pabeni@...hat.com> wrote:
>
> On 7/22/25 1:37 AM, Jason Xing wrote:
> > On Thu, Jul 17, 2025 at 11:29 PM Jeroen de Borst <jeroendb@...gle.com> wrote:
> >> +
> >> + pkt = gve_alloc_pending_packet(tx);
> >
> > I checked gve_alloc_pending_packet() and noticed there is one slight
> > chance to return NULL? If so, it will trigger a NULL dereference
> > issue.
>
> IIRC, a similar thing was already mentioned on an older patch. Still
> IIRC, the trick is that there is a previous, successful call to
> gve_has_avail_slots_tx_dqo() that ensures there are free TX packets
> avail and operations racing in between on other CPUs could only add more
> free pkts.
Oh, thanks for clarifying this. It does make sense :)
Thanks,
Jason
Powered by blists - more mailing lists