habanalabs: skip device idle check in hpriv_release if in reset
authorTomer Tayar <ttayar@habana.ai>
Wed, 30 Nov 2022 10:07:06 +0000 (12:07 +0200)
committerOded Gabbay <ogabbay@kernel.org>
Thu, 26 Jan 2023 08:56:21 +0000 (10:56 +0200)
When user context is released and hpriv_release() is called, there is a
device idle status check, to understand if user has left the device not
idle and then a reset is required.

However, if the user process is killed because of device hard reset,
the device at this point would always be not idle, because the device
engines were already forcefully halted.

Modify hpriv_release() to skip the idle check if reset is in progress.

Signed-off-by: Tomer Tayar <ttayar@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>
drivers/misc/habanalabs/common/device.c

index afd9d4d..71f958a 100644 (file)
@@ -428,8 +428,10 @@ static void hpriv_release(struct kref *ref)
         */
        reset_device = hdev->reset_upon_device_release || hdev->reset_info.watchdog_active;
 
-       /* Unless device is reset in any case, check idle status and reset if device is not idle */
-       if (!reset_device && hdev->pdev && !hdev->pldm)
+       /* Check the device idle status and reset if not idle.
+        * Skip it if already in reset, or if device is going to be reset in any case.
+        */
+       if (!hdev->reset_info.in_reset && !reset_device && hdev->pdev && !hdev->pldm)
                device_is_idle = hdev->asic_funcs->is_device_idle(hdev, idle_mask,
                                                        HL_BUSY_ENGINES_MASK_EXT_SIZE, NULL);
        if (!device_is_idle) {