firewire: use sleepable workqueue to handle 1394 OHCI IT/IR context events: test 2

[Edmund Raile <edmund.raile@xxxxxxxxxxxxxx>]

Hello Sakamoto-San, I came around to testing your patch [1], after RFT.

I've had to make the following changes to patch 1/5 again for it to apply to
mainline (d1f2d51b711a3b7f1ae1b46701c769c1d580fa7f), due to missing b171e20
from 2009 and a7ecbe9 from 2022.

@@ -584,9 +601,13 @@ int fw_card_add(struct fw_card *card,
 
 	generate_config_rom(card, tmp_config_rom);
 	ret = card->driver->enable(card, tmp_config_rom, config_rom_length);
 	if (ret == 0)
 		list_add_tail(&card->link, &card_list);
+	else
+		destroy_workqueue(isoc_wq);
+
+	card->isoc_wq = isoc_wq;

 	mutex_unlock(&card_mutex);

 	return ret;
@@ -709,7 +729,9 @@ void fw_core_remove_card(struct fw_card *card)
 {
 	struct fw_card_driver dummy_driver = dummy_driver_template;
 	unsigned long flags;
 
+	might_sleep();
+
 	card->driver->update_phy_reg(card, 4,
 				     PHY_LINK_ACTIVE | PHY_CONTENDER, 0);
 	fw_schedule_bus_reset(card, false, true);
@@ -719,6 +741,7 @@ void fw_core_remove_card(struct fw_card *card)
 	dummy_driver.free_iso_context	= card->driver->free_iso_context;
 	dummy_driver.stop_iso		= card->driver->stop_iso;
 	card->driver = &dummy_driver;
+	drain_workqueue(card->isoc_wq);
 
 	spin_lock_irqsave(&card->lock, flags);
 	fw_destroy_nodes(card);

Then everything applied fine.

This resulted in 6.11.0-rc6-1-mainline-00326-gd1f2d51b711a-dirty.

Testing it with TI XIO2213B and RME Fireface 800 so far:

Initially I had a buffer freeze after 3 hours of continuous ALSA playback
from mpv:
  mpv --audio-device=alsa/sysdefault:CARD=Fireface800 Spor-Ignition.flac
accompanied by stresstest (mprime).

It didn't freeze/crash the kernel, just the audio buffer kept repeating.
Gone after power-cycling the interface and restarting playback.

Can't say with certainty whether it's related, have been unable to replicate
the issue for the past 3 days (good sign I hope).
That's why I was holding this message back a bit.

Kind regards,
Edmund Raile.

Signed-off-by: Edmund Raile <edmund.raile@xxxxxxxxxxxxxx>
Tested-by: Edmund Raile <edmund.raile@xxxxxxxxxxxxxx>