Re: [PATCH 3/3] drm/xe: Force wedged state and block GT reset upon any GPU hang

[Rodrigo Vivi <rodrigo.vivi@intel.com>]

On Wed, Mar 13, 2024 at 04:54:38PM -0500, Lucas De Marchi wrote:
> On Wed, Mar 13, 2024 at 05:44:00PM -0400, Rodrigo Vivi wrote:
> > On Wed, Mar 13, 2024 at 03:49:56PM -0500, Lucas De Marchi wrote:
> > > On Wed, Mar 13, 2024 at 03:54:59PM -0400, Rodrigo Vivi wrote:
> > > > In many validation situations when debugging GPU Hangs,
> > > > it is useful to preserve the GT situation from the moment
> > > > that the timeout occurred.
> > > >
> > > > This patch introduces a module parameter that could be used
> > > > on situations like this.
> > > >
> > > > If xe.wedged module parameter is set to 2, Xe will be declared
> > > > wedged on every single execution timeout (a.k.a. GPU hang) right
> > > > after devcoredump snapshot capture and without attempting any
> > > > kind of GT reset and blocking entirely any kind of execution.
> > > >
> > > > Cc: Alan Previn <alan.previn.teres.alexis@intel.com>
> > > > Cc: Himanshu Somaiya <himanshu.somaiya@intel.com>
> > > > Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
> > > > ---
> > > > drivers/gpu/drm/xe/xe_device.c     | 22 ++++++++++++++++++++++
> > > > drivers/gpu/drm/xe/xe_device.h     |  6 +-----
> > > > drivers/gpu/drm/xe/xe_guc_submit.c |  4 ++++
> > > > drivers/gpu/drm/xe/xe_module.c     |  5 +++++
> > > > drivers/gpu/drm/xe/xe_module.h     |  1 +
> > > > 5 files changed, 33 insertions(+), 5 deletions(-)
> > > >
> > > > diff --git a/drivers/gpu/drm/xe/xe_device.c b/drivers/gpu/drm/xe/xe_device.c
> > > > index 5f0a2bdb7c24..296bc75a55f7 100644
> > > > --- a/drivers/gpu/drm/xe/xe_device.c
> > > > +++ b/drivers/gpu/drm/xe/xe_device.c
> > > > @@ -774,3 +774,25 @@ u64 xe_device_uncanonicalize_addr(struct xe_device *xe, u64 address)
> > > > {
> > > > 	return address & GENMASK_ULL(xe->info.va_bits - 1, 0);
> > > > }
> > > > +
> > > > +/**
> > > > + * xe_device_declare_wedged - Declare device wedged
> > > > + * @xe: xe device instance
> > > > + *
> > > > + * This is a final state that can only be cleared with a module reload.
> > > 
> > > same thing I mentioned about module reload elsewhere.
> > > 
> > > > + * In this state every IOCTL will be blocked so the GT cannot be used.
> > > > + * In general it will be called upon any critical error such as gt reset
> > > > + * failure or guc loading failure.
> > > > + * If xe.wedged module parameter is set to 2, this function will be called
> > > > + * on every single execution timeout (a.k.a. GPU hang) right after devcoredump
> > > > + * snapshot capture. In this mode, GT reset won't be attempted so the state of
> > > > + * the issue is preserved for further debugging.
> > > 
> > > Maybe make it a little bit more drastic so people don't use it when they
> > > shouldn't. I really hate the amount of reset=0 I see on machines being
> > > used with i915 with the additional patches to achieve a similar goal.
> > > IMO it should be a very targeted tool in a toolbox, not the hammer it
> > > became.
> > > 
> > > "In this mode, GT reset won't be attempted so there will be no recovery.
> > > Any GPU hang will completely kill the GPU so an autopsy may be
> > > attempted.  Use with care, if you know what you're doing."
> > > 
> > > > + */
> > > > +void xe_device_declare_wedged(struct xe_device *xe)
> > > > +{
> > > > +	if (xe_modparam.wedged == 0)
> > > > +		return;
> > > > +
> > > > +	atomic_set(&xe->wedged, 1);
> > > > +	drm_err(&xe->drm, "CRITICAL: Xe has been declared wedged. A module reload is required.\n");
> > > > +}
> > > > diff --git a/drivers/gpu/drm/xe/xe_device.h b/drivers/gpu/drm/xe/xe_device.h
> > > > index d10664d32f7f..de3149baf82f 100644
> > > > --- a/drivers/gpu/drm/xe/xe_device.h
> > > > +++ b/drivers/gpu/drm/xe/xe_device.h
> > > > @@ -181,10 +181,6 @@ static inline bool xe_device_wedged(struct xe_device *xe)
> > > > 	return atomic_read(&xe->wedged);
> > > > }
> > > >
> > > > -static inline void xe_device_declare_wedged(struct xe_device *xe)
> > > > -{
> > > > -	atomic_set(&xe->wedged, 1);
> > > > -	drm_err(&xe->drm, "CRITICAL: Xe has been declared wedged. A module reload is required.\n");
> > > > -}
> > > > +void xe_device_declare_wedged(struct xe_device *xe);
> > > >
> > > > #endif
> > > > diff --git a/drivers/gpu/drm/xe/xe_guc_submit.c b/drivers/gpu/drm/xe/xe_guc_submit.c
> > > > index 19efdb2f881f..987a57205fc4 100644
> > > > --- a/drivers/gpu/drm/xe/xe_guc_submit.c
> > > > +++ b/drivers/gpu/drm/xe/xe_guc_submit.c
> > > > @@ -34,6 +34,7 @@
> > > > #include "xe_macros.h"
> > > > #include "xe_map.h"
> > > > #include "xe_mocs.h"
> > > > +#include "xe_module.h"
> > > > #include "xe_ring_ops_types.h"
> > > > #include "xe_sched_job.h"
> > > > #include "xe_trace.h"
> > > > @@ -949,6 +950,9 @@ guc_exec_queue_timedout_job(struct drm_sched_job *drm_job)
> > > > 	simple_error_capture(q);
> > > > 	xe_devcoredump(job);
> > > >
> > > > +	if (xe_modparam.wedged == 2)
> > > > +		xe_device_declare_wedged(xe);
> > > > +
> > > > 	trace_xe_sched_job_timedout(job);
> > > >
> > > > 	/* Kill the run_job entry point */
> > > > diff --git a/drivers/gpu/drm/xe/xe_module.c b/drivers/gpu/drm/xe/xe_module.c
> > > > index 110b69864656..61272553f40f 100644
> > > > --- a/drivers/gpu/drm/xe/xe_module.c
> > > > +++ b/drivers/gpu/drm/xe/xe_module.c
> > > > @@ -17,6 +17,7 @@ struct xe_modparam xe_modparam = {
> > > > 	.enable_display = true,
> > > > 	.guc_log_level = 5,
> > > > 	.force_probe = CONFIG_DRM_XE_FORCE_PROBE,
> > > > +	.wedged = 1,
> > > > 	/* the rest are 0 by default */
> > > > };
> > > >
> > > > @@ -48,6 +49,10 @@ module_param_named_unsafe(force_probe, xe_modparam.force_probe, charp, 0400);
> > > > MODULE_PARM_DESC(force_probe,
> > > > 		 "Force probe options for specified devices. See CONFIG_DRM_XE_FORCE_PROBE for details.");
> > > >
> > > > +module_param_named_unsafe(wedged, xe_modparam.wedged, int, 0600);
> > > > +MODULE_PARM_DESC(wedged,
> > > > +		 "Wedged mode - 0=never, 1=upon-critical-errors[default], 2=upon-any-hang");
> > > 
> > > as we chatted earlier today, I think there's one thing missing. If an
> > > engine hangs, we won't notice as GuC will reset the engine by itself.
> > > We still need to pass GLOBAL_POLICY_DISABLE_ENGINE_RESET as a policy to
> > > GuC in the ADS structure. This is what I was attempting earlier today.
> > > 
> > > I was wondering if we could even update that in runtime with
> > > INTEL_GUC_ACTION_GLOBAL_SCHED_POLICY_CHANGE so we can target the debug
> > > only to when we are ready to reproduce the issues and only on the
> > > specific device.
> > 
> > I like that... Perhaps instead of the module parameter we have a debugfs
> > entry where we switch the state?
> > 
> > So, on a debugfs write to switch to this mode that gets 'busted'/'wedged'/'dead'
> > on every timeout/hang, then at that time we send this sched policy change?
> 
> the shortcome that we always have is that it doesn't cover the probing
> part. One common problem on early device support (either because of hw
> issues or driver/guc bugs) is on default context submission. If that
> receives a timeout, shouldn't we declare it busted?

hmm... indeed.

so, perhaps continue with the mod param xe.busted=2 but then right before any
submission we check for xe.modparam.busted == 2 to send the
INTEL_GUC_ACTION_GLOBAL_SCHED_POLICY_CHANGE before sending the command.

so on the next upcoming execution it would avoid the gt reset entirely.

then we could use it either at boot with xe.busted=2
or at runtime by
echo 2 | sudo tee /sys/module/xe/parameters/busted

> 
> I like "busted" :)
> 
> Lucas De Marchi