I am just reporting this as a separate bug, but it may be a duplicate of #98747, #98794, #98860, #98891, #98288 and probably more. But I am not sure whether it is, and kernel log asks me to file a _new_ bug, so reporting as new bug.
I got this with 4.9.0-rc4-tp520-drm-intel-fixes-btrfstrim which is a rc4 with drm-intel-fixes branch from 8th of November, I didn´t retest with 4.9-rc7 yet as I am back at 4.8 again after having faced a GPU hang on resume after hibernation, thats #98288.
merkaba:~#1> zgrep "GPU HANG" /var/log/kern.log*
/var/log/kern.log.3.gz:Nov 8 21:19:08 merkaba kernel: [ 8401.004898] [drm] GPU HANG: ecode 6:0:0x85fffffc, in psclient.bin , reason: Hang on render ring, action: reset
Nov 8 21:19:08 merkaba kernel: [ 8401.004898] [drm] GPU HANG: ecode 6:0:0x85fffffc, in psclient.bin , reason: Hang on render ring, action: reset
Nov 8 21:19:08 merkaba kernel: [ 8401.004902] [drm] GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
Nov 8 21:19:08 merkaba kernel: [ 8401.004903] [drm] Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel
Nov 8 21:19:08 merkaba kernel: [ 8401.004904] [drm] drm/i915 developers can then reassign to the right component if it's not a kernel issue.
Nov 8 21:19:08 merkaba kernel: [ 8401.004905] [drm] The gpu crash dump is required to analyze gpu hangs, so please always attach it.
Nov 8 21:19:08 merkaba kernel: [ 8401.004905] [drm] GPU crash dump saved to /sys/class/drm/card0/error
Nov 8 21:19:08 merkaba kernel: [ 8401.004947] drm/i915: Resetting chip after gpu hang
Unfortunately I do not have a crash dump. As the hang was so long (or repeated) that I just switched off the machine with the power button.
I may retest with latest 4.9 kernel on the weekend, but not during the week anymore, but even then I cannot guarentee whether I can get a crash dump as when the machine is hanging, hmmm, maybe via ssh from another laptop.
Martin, please verify that you can reproduce this using the modesetting DDX. There are known issues with xf86-video-intel:
Identifier "Intel Graphics"
Option "AccelMethod" "glamor"
Option "DRI" "3"
Debian Sid switched to modesetting DDX by default a while ago, but I also used it before that time already.
[ 9.986] (EE) Failed to load module "vesa" (module does not exist, 0)
[ 9.986] (II) modesetting: Driver for Modesetting Kernel Drivers: kms
[ 10.014] (II) modeset(0): using drv /dev/dri/card0
[ 10.014] (II) modeset(0): Creating default Display subsection in Screen section
"Default Screen Section" for depth/fbbpp 24/32
[ 10.014] (==) modeset(0): Depth 24, (==) framebuffer bpp 32
[ 10.014] (==) modeset(0): RGB weight 888
[ 10.014] (==) modeset(0): Default visual is TrueColor
[ 10.014] (II) Loading sub module "glamoregl"
[ 10.014] (II) LoadModule: "glamoregl"
[ 10.014] (II) Loading /usr/lib/xorg/modules/libglamoregl.so
[ 10.588] (II) modeset(0): Using spanning desktop for initial modes
[ 10.588] (II) modeset(0): Output LVDS-1 using initial mode 1920x1080 +0+0
[ 10.588] (II) modeset(0): Output DP-3 using initial mode 1920x1080 +1920+0
Closing, as I do not see this with recent 4.9 and 4.10 kernels, anymore. At least not with 4.9.24 and 4.10.16, but probably also some earlier versions.