On 10/13/2015 8:09 PM, Bart Van Assche wrote:
On 10/13/2015 09:41 AM, Sagi Grimberg wrote:
void transport_free_session(struct se_session *se_sess)
{
+ spin_lock_irq(&se_sess->sess_cmd_lock);
+ while (wait_event_interruptible_lock_irq(se_sess->cmd_list_wq,
+ list_empty(&se_sess->sess_cmd_list),
+ se_sess->sess_cmd_lock) < 0)
+ ;
+ spin_unlock_irq(&se_sess->sess_cmd_lock);
+
Can you please explain to me why not to use
wait_for_completion_interruptible() here?
Hello Sagi,
The above loop should only finish after sess_cmd_list became empty and
after target_release_cmd_kref() has unlocked the sess_cmd_lock spinlock.
target_release_cmd_kref() is not allowed to dereference the session
pointer after the above loop has finished. I had not found a way to
implement that with a single completion object without introducing race
conditions. Hence the above construct.
I see,
Another question, I notice that you wake_up cmd_list_wq even if the
session is not tearing down (where transport_free_session is not
running). Won't that introduce extra work in the hot path of command
processing?
--
To unsubscribe from this list: send the line "unsubscribe target-devel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
