[Intel-gfx] [PATCH 1/2] drm/i915/perf: Register sysctl path globally

[Lucas De Marchi]

On Wed, Dec 11, 2019 at 09:13:18AM -0800, Venkata Sandeep Dhanalakota wrote:
>On 19/12/11 04:39, Tvrtko Ursulin wrote:
>>
>> On 11/12/2019 16:31, Tvrtko Ursulin wrote:
>> > On 11/12/2019 16:07, Venkata Sandeep Dhanalakota wrote:
>> > > We do not require to register the sysctl paths per instance,
>> > > so making registration global.
>> > >
>> > > Cc: Sudeep Dutt <sudeep.dutt at intel.com>
>> > > Cc: Rodrigo Vivi <rodrigo.vivi at intel.com>
>> > > Cc: Daniel Vetter <daniel.vetter at ffwll.ch>
>> > > Cc: Chris Wilson <chris at chris-wilson.co.uk>
>> > > Cc: Jani Nikula <jani.nikula at intel.com>
>> > > Signed-off-by: Venkata Sandeep Dhanalakota
>> > > <venkata.s.dhanalakota at intel.com>
>> > > ---
>> > >   drivers/gpu/drm/i915/i915_perf.c       | 10 ++++++++--
>> > >   drivers/gpu/drm/i915/i915_perf_types.h |  1 -
>> > >   2 files changed, 8 insertions(+), 3 deletions(-)
>> > >
>> > > diff --git a/drivers/gpu/drm/i915/i915_perf.c
>> > > b/drivers/gpu/drm/i915/i915_perf.c
>> > > index 8d2e37949f46..426d04214a5d 100644
>> > > --- a/drivers/gpu/drm/i915/i915_perf.c
>> > > +++ b/drivers/gpu/drm/i915/i915_perf.c
>> > > @@ -387,6 +387,8 @@ struct i915_oa_config_bo {
>> > >       struct i915_vma *vma;
>> > >   };
>> > > +static struct ctl_table_header *sysctl_header;
>> > > +
>> > >   static enum hrtimer_restart oa_poll_check_timer_cb(struct hrtimer
>> > > *hrtimer);
>> > >   void i915_oa_config_release(struct kref *ref)
>> > > @@ -4345,7 +4347,8 @@ void i915_perf_init(struct drm_i915_private *i915)
>> > >           oa_sample_rate_hard_limit = 1000 *
>> > >               (RUNTIME_INFO(i915)->cs_timestamp_frequency_khz / 2);
>> > > -        perf->sysctl_header = register_sysctl_table(dev_root);
>> > > +        if (!sysctl_header)
>> > > +            sysctl_header = register_sysctl_table(dev_root);
>> > >           mutex_init(&perf->metrics_lock);
>> > >           idr_init(&perf->metrics_idr);
>> > > @@ -4395,7 +4398,10 @@ void i915_perf_fini(struct drm_i915_private *i915)
>> > >       idr_for_each(&perf->metrics_idr, destroy_config, perf);
>> > >       idr_destroy(&perf->metrics_idr);
>> > > -    unregister_sysctl_table(perf->sysctl_header);
>> > > +    if (sysctl_header) {
>> > > +        unregister_sysctl_table(sysctl_header);
>> > > +        sysctl_header = NULL;
>> > > +    }
>> >
>> > I am not sure if this could be racy with manual unbind from sysfs. Does
>> > PCI core serialize for us?
>>
>> Actually with two devices you also need to reference count it since you
>> don't want removal of the first device to remove the node but last.
>>
>Apparently this is not called during module exit, using krefs is
>way go to or have some helper which are called during module init/exit.
>
>void i915_perf_sysctl_register() {
>	sysctl_header = register_sysctl_table(dev_root);
>}
>
>void i915_perf_sysctl_unregister() {
>	unregister_sysctl_table(sysctl_header);
>}

yeah, since you are moving this to be global, I don't think it belongs
to the device, but rather to the module. So, IMO it makes more sense to
use this approach than kref/kunref on bind/unbind.

Lucas De Marchi

>
>
>> > Regards,
>> >
>> > Tvrtko
>> >
>> > >       memset(&perf->ops, 0, sizeof(perf->ops));
>> > >       perf->i915 = NULL;
>> > > diff --git a/drivers/gpu/drm/i915/i915_perf_types.h
>> > > b/drivers/gpu/drm/i915/i915_perf_types.h
>> > > index 74ddc20a0d37..45e581455f5d 100644
>> > > --- a/drivers/gpu/drm/i915/i915_perf_types.h
>> > > +++ b/drivers/gpu/drm/i915/i915_perf_types.h
>> > > @@ -380,7 +380,6 @@ struct i915_perf {
>> > >       struct drm_i915_private *i915;
>> > >       struct kobject *metrics_kobj;
>> > > -    struct ctl_table_header *sysctl_header;
>> > >       /*
>> > >        * Lock associated with adding/modifying/removing OA configs
>> > >
>_______________________________________________
>Intel-gfx mailing list
>Intel-gfx at lists.freedesktop.org
>https://lists.freedesktop.org/mailman/listinfo/intel-gfx