Comment # 16 on bug 1047729 from Takashi Iwai

(In reply to P. Otato from comment #15)
> > I installed an old HP laptop with GM45 and could reproduce the issue after
> > the cycle of suspend/resume.
> > 
> > With the renewed drm-kmp, it seems working.  Please check it in your side,
> > too.
> That's good to know. I've put the laptop into suspense mode about 20 times
> and everything seems okay so far.

So far, so good.

> But I still got rendering errors in Plasma 5 and also dmesg shows some errors
> 
> I got this one right after login in:
> [   27.816824] [drm] GPU HANG: ecode 4:0:0x828fffff, in X [1306], reason:
> Hang on render ring, action: reset

Is this a regression by the updated drm-kmp?  I saw the similar hang message in
your another report, so I suppose this happened before.

> And that happens when the laptop goes into suspense mode. Not every time but
> 4 out of 20 times.
> [  131.876806] [drm:i965_irq_handler [i915]] *ERROR* CPU pipe A FIFO underrun
> [  488.647149] [drm:i965_irq_handler [i915]] *ERROR* CPU pipe A FIFO underrun
> [  662.251064] [drm:i965_irq_handler [i915]] *ERROR* CPU pipe A FIFO underrun
> [  745.751916] [drm:i965_irq_handler [i915]] *ERROR* CPU pipe A FIFO underrun

The FIFO underrun isn't usually a big problem (but ugly, yes).
The question regarding this bug is, however, whether it's a regression by the
updated drm-kmp.  If it happened with the old drm-kmp, it's fine.  At least, I
can proceed the current fix for covering the crash at PM.