Commit 4cd454a2 authored by Oded Gabbay's avatar Oded Gabbay
Browse files

habanalabs/gaudi: recover from CPU WD event 



There are rare cases where the device CPU's watchdog has expired and as
a result, the watchdog reset has happened and the CPU will now move to
running its preboot f/w.

When that happens, the driver will only know that a heartbeat failure
occurred. As a result, the driver will send a message to the CPU's main
f/w asking it to reset the device, but because the CPU is now running
preboot, it won't respond and the re-initialization process will later
fail when trying to load the f/w.

The solution is to send the request to the preboot as well, only if the
reset was caused because of HB failure.

Signed-off-by: default avatarOded Gabbay <ogabbay@kernel.org>
parent c9d1383c

drivers/misc/habanalabs/gaudi/gaudi.c

+19 −1
Original line number Diff line number Diff line 
// SPDX-License-Identifier: GPL-2.0



/*

 * Copyright 2016-2020 HabanaLabs, Ltd.

 * Copyright 2016-2021 HabanaLabs, Ltd.

 * All Rights Reserved.

 */
@@ -4296,6 +4296,24 @@ static void gaudi_hw_fini(struct hl_device *hdev, bool hard_reset, bool fw_reset 


		WREG32(irq_handler_offset,

			gaudi_irq_map_table[GAUDI_EVENT_HALT_MACHINE].cpu_id);



		/* This is a hail-mary attempt to revive the card in the small chance that the

		 * f/w has experienced a watchdog event, which caused it to return back to preboot.

		 * In that case, triggering reset through GIC won't help. We need to trigger the

		 * reset as if Linux wasn't loaded.

		 *

		 * We do it only if the reset cause was HB, because that would be the indication

		 * of such an event.

		 *

		 * In case watchdog hasn't expired but we still got HB, then this won't do any

		 * damage.

		 */

		if (hdev->curr_reset_cause == HL_RESET_CAUSE_HEARTBEAT) {

			if (hdev->asic_prop.hard_reset_done_by_fw)

				hl_fw_ask_hard_reset_without_linux(hdev);

			else

				hl_fw_ask_halt_machine_without_linux(hdev);

		}

	} else {

		if (hdev->asic_prop.hard_reset_done_by_fw)

			hl_fw_ask_hard_reset_without_linux(hdev);