Re: [PATCH v13 9/9] drm/xe/ras: Move xe_ras_process_errors() to xe_ras

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] drm/xe/ras: early wedging in xe_ras_init() breaks GT init and DRM registration
--

commit 9ceb08e30ef3aa269ebad9c4f296fb67c41a76af
Author: Raag Jadav <raag.jadav@intel.com>

drm/xe/ras: Move xe_ras_process_errors() to xe_ras

This commit moves the boot-time call to xe_ras_process_errors() from the end
of xe_device_probe() back into xe_ras_init(), reverting commit 20bc4883c7c0
("drm/xe/ras: Fix boot-time ras error processing") now that xe_wedge_fini()
is registered during xe_device_probe_early().

> diff --git a/drivers/gpu/drm/xe/xe_ras.c b/drivers/gpu/drm/xe/xe_ras.c
> index 0072bb4cc5dfe..db3cbc186a514 100644
> --- a/drivers/gpu/drm/xe/xe_ras.c
> +++ b/drivers/gpu/drm/xe/xe_ras.c
> @@ -1212,6 +1212,13 @@ int xe_ras_init(struct xe_device *xe)
>  	get_offline_list(xe);
>  	get_offline_queue(xe);
>  
> +	/*
> +	 * During probe, process and log any errors detected by firmware while the driver was not
> +	 * loaded. Critical errors such as Punit and CSC are reported through Pcode init failure,
> +	 * causing the driver to enter survivability mode.
> +	 */
> +	xe_ras_process_errors(xe);

[Severity: High]
Can moving xe_ras_process_errors() into xe_ras_init() break probe when a
boot-time RAS error or cold-reset wedge fault injection declares the device
wedged?

While registering xe_wedge_fini() in xe_device_probe_early() resolves the PM
reference cleanup issue, xe_ras_init() still runs in xe_device_probe() before
both xe_gt_init() and drm_dev_register():

drivers/gpu/drm/xe/xe_device.c:xe_device_probe() {
	err = xe_ras_init(xe);
	if (err)
		return err;
	...
	for_each_gt(gt, xe, id) {
		err = xe_gt_init(gt);
		if (err)
			return err;
	}
	...
	err = drm_dev_register(&xe->drm, 0);
}

Inside xe_ras_process_errors(), xe_wedge_declare() can be called via three
paths:
  - handle_soc_internal_errors() -> xe_survivability_mode_runtime_enable() on
    a CSC firmware error (csc_error->hec_fw_error)
  - handle_soc_internal_errors() -> punit_error_handler() on an IEH Punit
    error (XE_RAS_SOC_IEH_PUNIT)
  - xe_fault_wedge_cold_reset() -> punit_error_handler()

When xe_wedge_declare() runs during xe_ras_init(), it sets xe->wedged.flag = 1
and schedules xe->wedged.work (wedge_work):

drivers/gpu/drm/xe/xe_wedge.c:xe_wedge_declare() {
	if (atomic_xchg(&xe->wedged.flag, 1))
		return;
	...
	for_each_initialized_gt(gt, xe, id)
		xe_gt_wedge_declare(gt);
	...
	schedule_work(&xe->wedged.work);
}

Because xe_ras_init() returns 0, xe_device_probe() continues into
xe_gt_init() -> gt_init_with_all_forcewake() -> xe_uc_load_hw(). With
xe_wedged(xe) already set, guc_exec_queue_run_job() drops submissions (causing
xe_gt_record_default_lrcs() / emit_wa_job() to time out with -ETIME), and
__guc_ct_send_locked() fails immediately:

drivers/gpu/drm/xe/xe_guc_ct.c:__guc_ct_send_locked() {
	if (xe_wedged(ct_to_xe(ct))) {
		ret = -ENOTRECOVERABLE;
		goto out;
	}
}

That causes xe_guc_post_load_init(), xe_guc_pc_start(), and xe_guc_rc_enable()
to fail, aborting xe_gt_init() and xe_device_probe() and unwinding devres
actions (including xe_survivability_mode_fini(), which removes the runtime
survivability sysfs attributes just created by
xe_survivability_mode_runtime_enable()).

Concurrently, wedge_work() invokes drm_dev_wedged_event(), which sends a
KOBJ_CHANGE uevent on dev->primary->kdev->kobj before drm_dev_register() ->
drm_minor_register() -> device_add() has registered the primary minor device:

drivers/gpu/drm/xe/xe_wedge.c:wedge_work()
  -> drm_dev_wedged_event(&xe->drm, xe->wedged.method, NULL)
       -> kobject_uevent_env(&dev->primary->kdev->kobj, KOBJ_CHANGE, envp)

In addition,xe_ras_init() returns early when !xe->info.has_sysctrl before
reaching xe_ras_process_errors(), whereas xe_ras_process_errors() checks
xe_fault_wedge_cold_reset() before checking !xe->info.has_sysctrl:

drivers/gpu/drm/xe/xe_ras.c:xe_ras_process_errors() {
	if (xe_fault_wedge_cold_reset()) {
		xe_err(xe, "[RAS]: cold-reset wedge injected\n");
		punit_error_handler(xe);
		return XE_RAS_RECOVERY_ACTION_DISCONNECT;
	}

	if (!xe->info.has_sysctrl)
		return XE_RAS_RECOVERY_ACTION_RESET;
}

Does moving this call into xe_ras_init() also bypass cold-reset wedge fault
injection on platforms without sysctrl?

> +
>  	ret = devm_device_add_group(xe->drm.dev, &gpu_health_group);
>  	if (ret)
>  		xe_err(xe, "Failed to create GPU health sysfs, err=%d\n", ret);

[ ... ]

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261020023641.200658-1-raag.jadav@intel.com?part=9