[Intel-gfx] 5.15-rc1 i915 blank screen booting on ThinkPads

[Hugh Dickins]

On Fri, 17 Sep 2021, Matthew Brost wrote:
> On Sat, Sep 18, 2021 at 01:52:48AM +0300, Jani Nikula wrote:
> > On Fri, 17 Sep 2021, Matthew Brost <matthew.brost at intel.com> wrote:
> > > On Fri, Sep 17, 2021 at 02:26:48PM -0700, Hugh Dickins wrote:
> > >> On Thu, 16 Sep 2021, Jani Nikula wrote:
> > >> > On Thu, 16 Sep 2021, Tvrtko Ursulin <tvrtko.ursulin at linux.intel.com> wrote:
> > >> > > On 16/09/2021 05:37, Hugh Dickins wrote:
> > >> > >> Two Lenovo ThinkPads, old T420s (2011), newer X1 Carbon 5th gen (2017):
> > >> > >> i915 working fine on both up to 5.14, but blank screens booting 5.15-rc1,
> > >> > >> kernel crashed in some way.
> > >> ...
> > >> > > Kernel logs with drm.debug=0xe, with the broken black screen state, 
> > >> > > would probably answer a lot of questions if you could gather it from 
> > >> > > both machines?
> > >> > 
> > >> > And for that, I think it's best to file separate bugs at [1] and attach
> > >> > the logs there. It helps keep the info in one place. Thanks.
> > >> > 
> > >> > BR,
> > >> > Jani.
> > >> > 
> > >> > [1] https://gitlab.freedesktop.org/drm/intel/issues/new
> > >> 
> > >> Thanks for the quick replies: but of course, getting kernel logs was
> > >> the difficult part, this being bootup, with just a blank screen, and
> > >> no logging to disk at this stage.  I've never needed it before, but
> > >> netconsole to the rescue.
> > >> 
> > >> Problem then obvious, both machines now working,
> > >> please let me skip the bug reports, here's a patch:
> > >> 
> > >
> > > Thanks for finding / fixing this Hugh. I will post this patch in a way
> > > our CI system can understand.
> > 
> > Thanks indeed!
> > 
> > Matt, please get rid of the BUG_ON while at it, and make it a
> > WARN. Oopsing doesn't do anyone any good.
> > 
> 
> Sure. Will do. Long term we should just look to rip out crap this (i.e.
> stealing bits from aligned addresses for flags).

It just crossed my mind, that I never did due diligence on _other_
callers of i915_sw_fence_init().  In fact they're okay, but that's
because their fence functions are all declared with the
#define __i915_sw_fence_call __aligned(4)
from i915_sw_fence.h, which I had not seen when I sent the patch.

I'm not going to resend, but if I were you, I'd quietly edit that
patch to use __i915_sw_fence_call in place of my __aligned(4).

Thanks,
Hugh

> 
> Matt
> 
> > BR,
> > Jani.
> > 
> > >
> > > Matt 
> > >
> > >> [PATCH] drm/i915: fix blank screen booting crashes
> > >> 
> > >> 5.15-rc1 crashes with blank screen when booting up on two ThinkPads
> > >> using i915.  Bisections converge convincingly, but arrive at different
> > >> and surprising "culprits", none of them the actual culprit.
> > >> 
> > >> netconsole (with init_netconsole() hacked to call i915_init() when
> > >> logging has started, instead of by module_init()) tells the story:
> > >> 
> > >> kernel BUG at drivers/gpu/drm/i915/i915_sw_fence.c:245!
> > >> with RSI: ffffffff814d408b pointing to sw_fence_dummy_notify().
> > >> I've been building with CONFIG_CC_OPTIMIZE_FOR_SIZE=y, and that
> > >> function needs to be 4-byte aligned.
> > >> 
> > >> Fixes: 62eaf0ae217d ("drm/i915/guc: Support request cancellation")
> > >> Signed-off-by: Hugh Dickins <hughd at google.com>
> > >> ---
> > >> 
> > >>  drivers/gpu/drm/i915/gt/intel_context.c |    1 +
> > >>  1 file changed, 1 insertion(+)
> > >> 
> > >> --- a/drivers/gpu/drm/i915/gt/intel_context.c
> > >> +++ b/drivers/gpu/drm/i915/gt/intel_context.c
> > >> @@ -362,6 +362,7 @@ static int __intel_context_active(struct
> > >>  	return 0;
> > >>  }
> > >>  
> > >> +__aligned(4)	/* Respect the I915_SW_FENCE_MASK */
> > >>  static int sw_fence_dummy_notify(struct i915_sw_fence *sf,
> > >>  				 enum i915_sw_fence_notify state)
> > >>  {
> > 
> > -- 
> > Jani Nikula, Intel Open Source Graphics Center