On Tue, Feb 11, 2020 at 10:15:29PM +0100, Miguel A. Vallejo wrote:
> Because everytime my system hangs / freezes I found something like
> this in syslog:
(On what GPU?)
> [ 135.116721] i915 0000:00:02.0: GPU HANG: ecode 9:1:0x00000000, hang on rcs0
> [ 135.116724] GPU hangs can indicate a bug anywhere in the entire gfx
> stack, including userspace.
> [ 135.116725] Please file a _new_ bug report on bugs.freedesktop.org
> against DRI -> DRM/Intel
> [ 135.116726] drm/i915 developers can then reassign to the right
> component if it's not a kernel issue.
> [ 135.116727] The GPU crash dump is required to analyze GPU hangs, so
> please always attach it.
> [ 135.116729] GPU crash dump saved to /sys/class/drm/card0/error
> [ 135.117739] i915 0000:00:02.0: Resetting rcs0 for hang on rcs0
> [ 135.118508] [drm:gen8_reset_engines [i915]] *ERROR* rcs0 reset
> request timed out: {request: 00000001, RESET_CTL: 00000001}
[...]
> More people with my very same experience here:
>
> https://gitlab.freedesktop.org/drm/intel/issues/673
Hmm. "Paul" in that thread reports having
"Gpu: HD Graphics 530 (Skylake GT2 / 0x1912 / mesa: 19.2.7)"
My workstation here, the one I'm typing on, has:
00:02.0 VGA compatible controller [0300]: Intel Corporation HD Graphics 530
[8086:1912] (rev 06)
I'm guessing the 0x1912 in Paul's report matches the 8086:1912 in my
PCI ID.
For whatever it's worth, I do *not* see any problems like this. Not
even close. Intel integrated graphics of this generation have been
rock solid for me on Debian 10, with non-free firmware + microcode.
The web page you linked also talks about "transition to idle", which
I'm guessing is related to CPU power management...? I'm not using
a laptop, and I'm not doing anything related to power management,
apart from whatever defaults Linux and Debian are doing.
I suspect there's some triggering event/bug that only some users
are encountering due to different installed packages or usage patterns.
I'm probably a fairly nontypical user by today's standards (desktop PC
with fvwm, no desktop environment, limited or no 3D graphics stuff),
but I might match pretty closely against the usage patterns of
the *developers* of some of this stuff... maybe they just don't use
the things that trigger the bugs, so they never see the bugs, so they
never fix the bugs.
