Re: [Qemu-devel] [sheepdog] [PATCH v3 07/10] sheepdog: try to reconnect to sheepdog after network error

[MORITA Kazutaka]

At Thu, 25 Jul 2013 17:13:46 +0800,
Liu Yuan wrote:
> 
> > +
> > +    /* Try to reconnect the sheepdog server every one second. */
> > +    while (s->fd < 0) {
> > +        s->fd = get_sheep_fd(s);
> > +        if (s->fd < 0) {
> > +            dprintf("Wait for connection to be established\n");
> > +            co_aio_sleep_ns(1000000000ULL);
> > +        }
> > +    };
> > +
> > +    /* Move all the inflight requests to the failed queue. */
> > +    QLIST_FOREACH_SAFE(aio_req, &s->inflight_aio_head, aio_siblings, next) 
> > {
> > +        QLIST_REMOVE(aio_req, aio_siblings);
> > +        QLIST_INSERT_HEAD(&s->failed_aio_head, aio_req, aio_siblings);
> > +    }
> > +
> > +    /* Resend all the failed aio requests. */
> > +    while (!QLIST_EMPTY(&s->failed_aio_head)) {
> > +        aio_req = QLIST_FIRST(&s->failed_aio_head);
> > +        QLIST_REMOVE(aio_req, aio_siblings);
> > +        QLIST_INSERT_HEAD(&s->inflight_aio_head, aio_req, aio_siblings);
> > +        resend_aioreq(s, aio_req);
> > +    }
> > +}
> > +
> 
> Is failed queue necessary? Here you just move requests from inflight queue to
> failed queue, then interate the failed queue to send them all.
> 
> Isn't it simpler we just resend the requests in the inflight queue like
> 
> > +    QLIST_FOREACH(aio_req, &s->inflight_aio_head, aio_siblings, next) {
> > +        resend_aioreq(s, aio_req);
> > +    }

resend_aioreq() can yield and a new aio request can be added to the
inflight queue during this loop.  To avoid mixing new requests and
failed ones, I think the failed queue is necessary.

Thanks,

Kazutaka