[PATCH] nvmet-rdma: recheck queue state is LIVE in state lock in recv done

[Ruozhu Li]

On 2025/02/13 6:20 UTC, Christoph Hellwig wrote:
> > +++ b/drivers/nvme/target/rdma.c
> > @@ -1042,12 +1042,16 @@ static void nvmet_rdma_recv_done(struct ib_cq *cq, struct ib_wc *wc)
> >               unsigned long flags;
> >
> >               spin_lock_irqsave(&queue->state_lock, flags);
> > -             if (queue->state == NVMET_RDMA_Q_CONNECTING)
> > -                     list_add_tail(&rsp->wait_list, &queue->rsp_wait_list);
> > -             else
> > -                     nvmet_rdma_put_rsp(rsp);
> > +             if (queue->state != NVMET_RDMA_Q_LIVE) {
>
> You'll want a comment based on the commit message that the first live check is racy.
Okay, I will add a comment in next version.
>
> > +                     if (queue->state == NVMET_RDMA_Q_CONNECTING)
> > +                             list_add_tail(&rsp->wait_list, &queue->rsp_wait_list);
> > +                     else
> > +                             nvmet_rdma_put_rsp(rsp);
> > +
> > +                     spin_unlock_irqrestore(&queue->state_lock, flags);
> > +                     return;
>
> This gets a bit messy now.  What about a struture like this instead?
Looks good to me, I will retest it and send another version soon.

Thanks.