[Intel-gfx] [PATCH] drm/i915: Limit the number of loops for reading a split 64bit register

Chris Wilson chris at chris-wilson.co.uk
Tue Sep 8 06:10:35 PDT 2015


On Tue, Sep 08, 2015 at 03:36:32PM +0300, Jani Nikula wrote:
> On Tue, 08 Sep 2015, Chris Wilson <chris at chris-wilson.co.uk> wrote:
> > In I915_READ64_2x32 we attempt to read a 64bit register using 2 32bit
> > reads. Due to the nature of the registers we try to read in this manner,
> > they may increment between the two instruction (e.g. a timestamp
> > counter). To keep the result accurate, we repeat the read if we detect
> > an overflow (i.e. the upper value varies). However, some harware is just
> > plain flaky and may endless loop as the the upper 32bits are not stable.
> > Just give up after a couple of tries and report whatever we read last.
> >
> > Reported-by: russianneuromancer at ya.ru
> > Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=91906
> > Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> > Cc: MichaƂ Winiarski <michal.winiarski at intel.com>
> > Cc: Daniel Vetter <daniel.vetter at ffwll.ch>
> > Cc: stable at vger.kernel.org
> > ---
> >  drivers/gpu/drm/i915/i915_drv.h | 4 ++--
> >  1 file changed, 2 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
> > index 12870073d58f..8943dcb724a8 100644
> > --- a/drivers/gpu/drm/i915/i915_drv.h
> > +++ b/drivers/gpu/drm/i915/i915_drv.h
> > @@ -3402,13 +3402,13 @@ int intel_freq_opcode(struct drm_i915_private *dev_priv, int val);
> >  #define I915_READ64(reg)	dev_priv->uncore.funcs.mmio_readq(dev_priv, (reg), true)
> >  
> >  #define I915_READ64_2x32(lower_reg, upper_reg) ({			\
> > -	u32 upper, lower, tmp;						\
> > +	u32 upper, lower, tmp, loop = 0;				\
> >  	tmp = I915_READ(upper_reg);					\
> >  	do {								\
> >  		upper = tmp;						\
> >  		lower = I915_READ(lower_reg);				\
> >  		tmp = I915_READ(upper_reg);				\
> > -	} while (upper != tmp);						\
> > +	} while (upper != tmp && loop++ != 2);				\
> 
> Do you think it matters that you'll take the previous, not the last,
> value when you give up?

Not an awful lot, but it's a good reason to respin.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre


More information about the Intel-gfx mailing list