[PATCH] drm/xe/debugfs: Make sysfs gt force reset synchronous

[Matthew Brost]

On Wed, Dec 27, 2023 at 08:53:03PM +0530, Nilawar, Badal wrote:
> 
> 
> On 27-12-2023 16:06, Gupta, Anshuman wrote:
> > 
> > 
> > > -----Original Message-----
> > > From: Poosa, Karthik <karthik.poosa at intel.com>
> > > Sent: Wednesday, December 27, 2023 2:02 PM
> > > To: intel-xe at lists.freedesktop.org
> > > Cc: Gupta, Anshuman <anshuman.gupta at intel.com>; Nilawar, Badal
> > > <badal.nilawar at intel.com>; Brost, Matthew <matthew.brost at intel.com>;
> > > Vivi, Rodrigo <rodrigo.vivi at intel.com>; Poosa, Karthik
> > > <karthik.poosa at intel.com>
> > > Subject: [PATCH] drm/xe/debugfs: Make sysfs gt force reset synchronous
> > > 
> > > Wait for gt reset to complete before returning from force_reset sysfs call.
> > > Without this igt test freq_reset_multiple fails sporadically in case xe_guc_pc is
> > > not started.
> > > 
> > > v2:
> > > - Changed wait for completion to interruptible (Anshuman).
> > > - Moved timeout to xe_gt.h (Anshuman).
> > > - Created a debugfs for updating timeout (Rodrigo).
> > > 
> > > Testcase: igt at xe_guc_pc@freq_reset_multiple
> > > Signed-off-by: Karthik Poosa <karthik.poosa at intel.com>
> > > ---
> > >   drivers/gpu/drm/xe/xe_gt.c         |  2 ++
> > >   drivers/gpu/drm/xe/xe_gt_debugfs.c | 12 ++++++++++++
> > >   drivers/gpu/drm/xe/xe_gt_types.h   |  6 ++++++
> > >   3 files changed, 20 insertions(+)
> > > 
> > > diff --git a/drivers/gpu/drm/xe/xe_gt.c b/drivers/gpu/drm/xe/xe_gt.c index
> > > 3af2adec1295..47abb9336c58 100644
> > > --- a/drivers/gpu/drm/xe/xe_gt.c
> > > +++ b/drivers/gpu/drm/xe/xe_gt.c
> > > @@ -65,6 +65,7 @@ struct xe_gt *xe_gt_alloc(struct xe_tile *tile)
> > > 
> > >   	gt->tile = tile;
> > >   	gt->ordered_wq = alloc_ordered_workqueue("gt-ordered-wq", 0);
> > > +	init_completion(&gt->reset_done);
> > > 
> > >   	return gt;
> > >   }
> > > @@ -633,6 +634,7 @@ static int gt_reset(struct xe_gt *gt)
> > >   	xe_device_mem_access_put(gt_to_xe(gt));
> > >   	XE_WARN_ON(err);
> > > 
> > > +	complete(&gt->reset_done);
> > >   	xe_gt_info(gt, "reset done\n");
> > > 
> > >   	return 0;
> > > diff --git a/drivers/gpu/drm/xe/xe_gt_debugfs.c
> > > b/drivers/gpu/drm/xe/xe_gt_debugfs.c
> > > index c4b67cf09f8f..fbda886c8a95 100644
> > > --- a/drivers/gpu/drm/xe/xe_gt_debugfs.c
> > > +++ b/drivers/gpu/drm/xe/xe_gt_debugfs.c
> > > @@ -58,8 +58,16 @@ static int hw_engines(struct seq_file *m, void *data)
> > > static int force_reset(struct seq_file *m, void *data)  {
> > >   	struct xe_gt *gt = node_to_gt(m->private);
> > > +	struct xe_device *xe = gt_to_xe(gt);
> > > +	long ret;
> > > 
> > >   	xe_gt_reset_async(gt);
> > > +	ret = wait_for_completion_interruptible_timeout(&gt->reset_done,
> > > +
> > This would defeat the purpose of xe_gt_reset_async(), as this will make force_reset
> > synchronous , I think we need wait_for_completion_interruptible_timeout in  xe_gt_freq sysfs
> > before reading the guc pc frequency.
> > Something like below.
> > guc_pc_freq_ready()
> > {
> > 	wait_for_completion_interruptible_timeout()
> > }
> Then how force_reset debugfs should be handled then? I am seeing only 2
> possible ways, if reset is in progress 1. Do nothing or 2. Wait for
> completion and then initiate reset.
> 

What about changing force_reset to:

write == 0 -> reset async
write == 1 -> reset sync

Matt

> For freq sysfs entries I agree we should wait for guc pc is ready.
> 
> Regards,
> Badal
> > 
> > Thanks,
> > Anshuman Gupta.
> > 						msecs_to_jiffies(gt-
> > > > reset_timeout_ms));
> > > +	if (ret <= 0) {
> > > +		drm_err(&xe->drm, "gt reset timed out/interrputed, ret
> > > %ld\n", ret);
> > > +		return -ETIMEDOUT;
> > > +	}
> > > 
> > >   	return 0;
> > >   }
> > > @@ -225,6 +233,10 @@ void xe_gt_debugfs_register(struct xe_gt *gt)
> > >   		return;
> > >   	}
> > > 
> > > +	/* set a default timeout */
> > > +	gt->reset_timeout_ms = 1000;
> > > +	debugfs_create_u32("gt_reset_timeout_ms", 0600, root,
> > > +					&gt->reset_timeout_ms);
> > >   	/*
> > >   	 * Allocate local copy as we need to pass in the GT to the debugfs
> > >   	 * entry and drm_debugfs_create_files just references the
> > > drm_info_list diff --git a/drivers/gpu/drm/xe/xe_gt_types.h
> > > b/drivers/gpu/drm/xe/xe_gt_types.h
> > > index f74684660475..824cefde20d2 100644
> > > --- a/drivers/gpu/drm/xe/xe_gt_types.h
> > > +++ b/drivers/gpu/drm/xe/xe_gt_types.h
> > > @@ -358,6 +358,12 @@ struct xe_gt {
> > >   		/** @oob: bitmap with active OOB workaroudns */
> > >   		unsigned long *oob;
> > >   	} wa_active;
> > > +
> > > +	/** @reset_done : completion for GT reset */
> > > +	struct completion reset_done;
> > > +
> > > +	/** @gt_reset_timeout_ms : gt reset timeout in ms */
> > > +	u32 reset_timeout_ms;
> > >   };
> > > 
> > >   #endif
> > > --
> > > 2.25.1
> >