Revert "mailbox: avoid timer start from callback"

raspberrypi inclusion
category: feature
bugzilla: 50432

--------------------------------

This reverts commit c7dacf5b.

The Pi 400 shutdown/poweroff mechanism relies on being able to set
a GPIO on the expander in the pm_power_off handler, something that
requires two mailbox calls - GET_GPIO_STATE and SET_GPIO_STATE. A
recent kernel change introduces a reasonable possibility that the
GET call doesn't completes, and bisecting led to a commit from
October that changes the timer usage of the mailbox.

My theory is that there is a race condition in the new code that breaks
the poll timer, but that it normally goes unnoticed because subsequent
mailbox activity wakes it up again. The power-off mailbox calls happen
at a time when other subsystems have been shut down, so if one of them
fails then there is nothing to allow it to recover.

See: https://github.com/raspberrypi/linux/issues/3941Signed-off-by: NPhil Elwell <phil@raspberrypi.com>
Signed-off-by: NFang Yafen <yafen@iscas.ac.cn>
Signed-off-by: NZheng Zengkai <zhengzengkai@huawei.com>

drivers/mailbox/mailbox.c

@@ -82,12 +82,9 @@ static void msg_submit(struct mbox_chan *chan)
 exit:
 	spin_unlock_irqrestore(&chan->lock, flags);
 
+	if (!err && (chan->txdone_method & TXDONE_BY_POLL))
 		/* kick start the timer immediately to avoid delays */
-	if (!err && (chan->txdone_method & TXDONE_BY_POLL)) {
-		/* but only if not already active */
-		if (!hrtimer_active(&chan->mbox->poll_hrt))
 		hrtimer_start(&chan->mbox->poll_hrt, 0, HRTIMER_MODE_REL);
-	}
 }
 
 static void tx_tick(struct mbox_chan *chan, int r)
@@ -125,10 +122,11 @@ static enum hrtimer_restart txdone_hrtimer(struct hrtimer *hrtimer)
 		struct mbox_chan *chan = &mbox->chans[i];
 
 		if (chan->active_req && chan->cl) {
-			resched = true;
 			txdone = chan->mbox->ops->last_tx_done(chan);
 			if (txdone)
 				tx_tick(chan, 0);
+			else
+				resched = true;
 		}
 	}