[Intel-gfx] [PATCH v4 2/2] drm/i915/gt: make a gt sysfs group and move power management files

[Andi Shyti]

Hi Michal,

> > /sys/.../card0
> >          ├── gt
> >          │   ├── gt0
> >          │   │   ├── id
> >          │   │   ├── rc6_enable
> >          │   │   ├── rc6_residency_ms
> >          │   │   ├── rps_act_freq_mhz
> >          │   │   ├── rps_boost_freq_mhz
> >          │   │   ├── rps_cur_freq_mhz
> >          │   │   ├── rps_max_freq_mhz
> >          │   │   ├── rps_min_freq_mhz
> >          │   │   ├── rps_RP0_freq_mhz
> >          │   │   ├── rps_RP1_freq_mhz
> >          │   │   └── rps_RPn_freq_mhz
> > 	 .   .
> > 	 .   .
> > 	 .   .
> >          │   └── gt3
> 
> gtN ?

yep!

> > diff --git a/drivers/gpu/drm/i915/Makefile b/drivers/gpu/drm/i915/Makefile
> > index aa86ac33effc..5fd203c626fc 100644
> > --- a/drivers/gpu/drm/i915/Makefile
> > +++ b/drivers/gpu/drm/i915/Makefile
> > @@ -121,7 +121,9 @@ gt-y += \
> >  	gt/intel_timeline.o \
> >  	gt/intel_workarounds.o \
> >  	gt/shmem_utils.o \
> > -	gt/sysfs_engines.o
> > +	gt/sysfs_engines.o \
> > +	gt/sysfs_gt.o \
> > +	gt/sysfs_gt_pm.o
> 
> shouldn't these be named as
> 
> > +	gt/intel_gt_sysfs.o \
> > +	gt/intel_gt_pm_sysfs.o

You are right with wanting a coherent prefix, but I kept the
trend of starting with sysfs_gt*. We already have sysfs_engine.c.

And, because I wouldn't like to have part of it sysfs_gt* and
part of it intel_gt_sysfs*, then we either rename all or we leave
it as it is.

On the other hand if we are under i915/gt/... I don't expect it
to be the sysfs of another system.

To be honest, I don't have a strong opinion on this. If you do,
then I will change everything intel_gt_sysfs*.

[...]

> > +++ b/drivers/gpu/drm/i915/gt/sysfs_gt.c
> > @@ -0,0 +1,136 @@
> > +// SPDX-License-Identifier: MIT
> > +/*
> > + * Copyright © 2020 Intel Corporation
> 
> 2022 ?

Time flies... huh? :)

> > +void intel_gt_sysfs_register(struct intel_gt *gt)
> > +{
> > +	struct kobject *dir;
> > +	char name[80];
> > +
> > +	/*
> > +	 * We need to make things right with the
> > +	 * ABI compatibility. The files were originally
> > +	 * generated under the parent directory.
> > +	 *
> > +	 * We generate the files only for gt 0
> > +	 * to avoid duplicates.
> > +	 */
> > +	if (!gt->info.id)
> 
> maybe we should have gt_is_root(gt) helper ?

yes, makes sense.

> > +		intel_gt_sysfs_pm_init(gt, gt_get_parent_obj(gt));
> > +
> > +	snprintf(name, sizeof(name), "gt%d", gt->info.id);
> > +
> > +	dir = intel_gt_create_kobj(gt, gt->i915->sysfs_gt, name);
> > +	if (!dir) {
> > +		drm_warn(&gt->i915->drm,
> > +			 "failed to initialize %s sysfs root\n", name);
> > +		return;
> > +	}
> > +
> > +	if (sysfs_create_file(dir, &dev_attr_id.attr))
> > +		drm_warn(&gt->i915->drm,
> > +			 "failed to create sysfs %s info files\n", name);
> 
> can't we use default_groups in kobj_type ?

yeah... I'll try that.

[...]

> > +static ssize_t rc6_enable_show(struct device *dev,
> > +			       struct device_attribute *attr,
> > +			       char *buff)
> > +{
> > +	struct intel_gt *gt = intel_gt_sysfs_get_drvdata(dev, attr->attr.name);
> > +	u8 mask = 0;
> > +
> > +	if (HAS_RC6(gt->i915))
> > +		mask |= BIT(0);
> > +	if (HAS_RC6p(gt->i915))
> > +		mask |= BIT(1);
> > +	if (HAS_RC6pp(gt->i915))
> > +		mask |= BIT(2);
> > +
> > +	return scnprintf(buff, PAGE_SIZE, "%x\n", mask);
> 
> sysfs_emit ?

OK

[...]

> > +	ret = __intel_gt_sysfs_create_group(kobj, rc6_attr_group);
> > +	if (ret)
> > +		drm_err(&gt->i915->drm,
> > +			"failed to create gt%u RC6 sysfs files\n", gt->info.id);
> > +
> > +	if (HAS_RC6p(gt->i915)) {
> > +		ret = __intel_gt_sysfs_create_group(kobj, rc6p_attr_group);
> > +		if (ret)
> > +			drm_err(&gt->i915->drm,
> > +				"failed to create gt%u RC6p sysfs files\n",
> > +				gt->info.id);
> > +	}
> > +
> > +	if (IS_VALLEYVIEW(gt->i915) || IS_CHERRYVIEW(gt->i915)) {
> > +		ret = __intel_gt_sysfs_create_group(kobj, media_rc6_attr_group);
> > +		if (ret)
> > +			drm_err(&gt->i915->drm,
> > +				"failed to create media %u RC6 sysfs files\n",
> > +				gt->info.id);
> > +	}
> 
> did you consider using attribute_group.is_visible instead adding groups
> manually ?

I can try this, as well.

[...]

> maybe this large but simple code movement should be done in a separate
> patch so we could then apply smaller and easier to review fixes ?

I can try to split it, even though most of it is basically a
copy/paste.

> ~Michal

Thanks a lot for this review, as well!

Andi