Jakub Sitnicki wrote:
> On Tue, May 02, 2023 at 08:51 AM -07, John Fastabend wrote:
> > We noticed some rare sk_buffs were stepping past the queue when system was
> > under memory pressure. The general theory is to skip enqueueing
> > sk_buffs when its not necessary which is the normal case with a system
> > that is properly provisioned for the task, no memory pressure and enough
> > cpu assigned.
> >
> > But, if we can't allocate memory due to an ENOMEM error when enqueueing
> > the sk_buff into the sockmap receive queue we push it onto a delayed
> > workqueue to retry later. When a new sk_buff is received we then check
> > if that queue is empty. However, there is a problem with simply checking
> > the queue length. When a sk_buff is being processed from the ingress queue
> > but not yet on the sockmap msg receive queue its possible to also recv
> > a sk_buff through normal path. It will check the ingress queue which is
> > zero and then skip ahead of the pkt being processed.
> >
> > Previously we used sock lock from both contexts which made the problem
> > harder to hit, but not impossible.
> >
> > To fix instead of popping the skb from the queue entirely we peek the
> > skb from the queue and do the copy there. This ensures checks to the
> > queue length are non-zero while skb is being processed. Then finally
> > when the entire skb has been copied to user space queue or another
> > socket we pop it off the queue. This way the queue length check allows
> > bypassing the queue only after the list has been completely processed.
> >
> > To reproduce issue we run NGINX compliance test with sockmap running and
> > observe some flakes in our testing that we attributed to this issue.
> >
> > Fixes: 04919bed948dc ("tcp: Introduce tcp_read_skb()")
> > Tested-by: William Findlay <will@xxxxxxxxxxxxx>
> > Suggested-by: Jakub Sitnicki <jakub@xxxxxxxxxxxxxx>
> > Signed-off-by: John Fastabend <john.fastabend@xxxxxxxxx>
> > ---
[...]
> > @@ -677,8 +666,7 @@ static void sk_psock_backlog(struct work_struct *work)
> > len, ingress);
> > if (ret <= 0) {
> > if (ret == -EAGAIN) {
> > - sk_psock_skb_state(psock, state, skb,
> > - len, off);
> > + sk_psock_skb_state(psock, state, len, off);
> >
> > /* Delay slightly to prioritize any
> > * other work that might be here.
>
> I've been staring at this bit and I think it doesn't matter if we update
> psock->work_state when SK_PSOCK_TX_ENABLED has been cleared.
>
> But what I think we shouldn't be doing here is scheduling
> sk_psock_backlog again if SK_PSOCK_TX_ENABLED got cleared by
> sk_psock_stop.
Yeah I agree we shouldn't be scheduling with TX_ENABLED cleared. Otherwise
while we cancle and sync the worker from the destroy path we could queue
up more work here.
Also spotted another case of this where its not wrapped in a check. I guess
we should fix it. Nice catch.
v8 it is I guess thanks.
