On 9.01.25 9:15, Pavel Tikhomirov wrote:
@@ -177,10 +180,9 @@ static bool ploop_try_delay_enospc(struct
ploop_rq *prq, struct pio *pio)
pr_err_once(PL_FMT("underlying disk is almost full"),
ploop_device_name(ploop));
- ploop->event_enospc = true;
- list_add_tail(&pio->list, &ploop->enospc_pios);
+ WRITE_ONCE(ploop->event_enospc, true);
Do we still need deferred_lock in ploop_get_event to protect
event_enospc after this? Probably we can also use READ/WRITE_ONCE there.
Yes, it does test and set, so it needs a lock.
I also restored the lock here , otherwise a event can be missed.
I will rework it when i rework the timer.
Events are better delivered via queue.
--
Regards,
Alexander Atanasov
_______________________________________________
Devel mailing list
Devel@openvz.org
https://lists.openvz.org/mailman/listinfo/devel