[Intel-gfx] [PATCH v7 0/9] dyndbg: drm.debug adaptation

[jim.cromie at gmail.com]

On Thu, Oct 27, 2022 at 9:59 AM Ville Syrjälä
<ville.syrjala at linux.intel.com> wrote:
>
> On Thu, Oct 27, 2022 at 09:37:52AM -0600, jim.cromie at gmail.com wrote:
> > On Thu, Oct 27, 2022 at 9:08 AM Jason Baron <jbaron at akamai.com> wrote:
> > >
> > >
> > >
> > > On 10/21/22 05:18, Jani Nikula wrote:
> > > > On Thu, 20 Oct 2022, Ville Syrjälä <ville.syrjala at linux.intel.com> wrote:
> > > >> On Sat, Sep 24, 2022 at 03:02:34PM +0200, Greg KH wrote:
> > > >>> On Sun, Sep 11, 2022 at 11:28:43PM -0600, Jim Cromie wrote:
> > > >>>> hi Greg, Dan, Jason, DRM-folk,
> > > >>>>
> > > >>>> heres follow-up to V6:
> > > >>>>   rebased on driver-core/driver-core-next for -v6 applied bits (thanks)
> > > >>>>   rework drm_debug_enabled{_raw,_instrumented,} per Dan.
> > > >>>>
> > > >>>> It excludes:
> > > >>>>   nouveau parts (immature)
> > > >>>>   tracefs parts (I missed --to=Steve on v6)
> > > >>>>   split _ddebug_site and de-duplicate experiment (way unready)
> > > >>>>
> > > >>>> IOW, its the remaining commits of V6 on which Dan gave his Reviewed-by.
> > > >>>>
> > > >>>> If these are good to apply, I'll rebase and repost the rest separately.
> > > >>>
> > > >>> All now queued up, thanks.
> > > >>
> > > >> This stuff broke i915 debugs. When I first load i915 no debug prints are
> > > >> produced. If I then go fiddle around in /sys/module/drm/parameters/debug
> > > >> the debug prints start to suddenly work.
> > > >
> > > > Wait what? I always assumed the default behaviour would stay the same,
> > > > which is usually how we roll. It's a regression in my books. We've got a
> > > > CI farm that's not very helpful in terms of dmesg logging right now
> > > > because of this.
> > > >
> > > > BR,
> > > > Jani.
> > > >
> > > >
> > >
> > > That doesn't sound good - so you are saying that prior to this change some
> > > of the drm debugs were default enabled. But now you have to manually enable
> > > them?
> > >
> > > Thanks,
> > >
> > > -Jason
> >
> >
> > Im just seeing this now.
> > Any new details ?
>
> No. We just disabled it as BROKEN for now. I was just today thinking
> about sending that patch out if no solutin is forthcoming soon since
> we need this working before 6.1 is released.
>
> Pretty sure you should see the problem immediately with any driver
> (at least if it's built as a module, didn't try builtin). Or at least
> can't think what would make i915 any more special.
>

So, I should note -
99% of my time & energy on this dyndbg + drm patchset
has been done using virtme,
so my world-view (and dev-hack-test env) has been smaller, simpler
maybe its been fatally simplistic.

ive just rebuilt v6.0  (before the trouble)
and run it thru my virtual home box,
I didnt see any unfamiliar drm-debug output
that I might have inadvertently altered somehow

I have some real HW I can put a reference kernel on,0
to look for the missing output, but its all gonna take some time,
esp to tighten up my dev-test-env

in the meantime, there is:

config DRM_USE_DYNAMIC_DEBUG
bool "use dynamic debug to implement drm.debug"
default y
depends on DRM
depends on DYNAMIC_DEBUG || DYNAMIC_DEBUG_CORE
depends on JUMP_LABEL
help
  Use dynamic-debug to avoid drm_debug_enabled() runtime overheads.
  Due to callsite counts in DRM drivers (~4k in amdgpu) and 56
  bytes per callsite, the .data costs can be substantial, and
  are therefore configurable.

Does changing the default fix things for i915 dmesg ?
or is the problem deeper ?

theres also this Makefile addition, which I might have oversimplified

CFLAGS-$(CONFIG_DRM_USE_DYNAMIC_DEBUG) += -DDYNAMIC_DEBUG_MODULE