System freeze (GPU hang) with i915 driver on Intel UHD 620

i915 driver causes system freeze (GPU hang) on kernel 5.4 series. Kernels 4.19 and 5.5 don't have this issue. Here is a part of journal after GPU hang with kernel 5.4.6-2:

...
Dec 25 14:14:04 andrey-lenovo kernel: i915 0000:00:02.0: GPU HANG: ecode 9:1:0x00000000, hang on rcs0
Dec 25 14:14:04 andrey-lenovo kernel: GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
Dec 25 14:14:04 andrey-lenovo kernel: Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel
Dec 25 14:14:04 andrey-lenovo kernel: drm/i915 developers can then reassign to the right component if it's not a kernel issue.
Dec 25 14:14:04 andrey-lenovo kernel: The GPU crash dump is required to analyze GPU hangs, so please always attach it.
Dec 25 14:14:04 andrey-lenovo kernel: GPU crash dump saved to /sys/class/drm/card0/error
Dec 25 14:14:04 andrey-lenovo kernel: i915 0000:00:02.0: Resetting rcs0 for hang on rcs0
Dec 25 14:14:04 andrey-lenovo kernel: [drm:gen8_reset_engines [i915]] *ERROR* rcs0 reset request timed out: {request: 00000001, RESET_CTL: 00000001}
Dec 25 14:14:04 andrey-lenovo kernel: i915 0000:00:02.0: Resetting chip for hang on rcs0
Dec 25 14:14:04 andrey-lenovo kernel: [drm:gen8_reset_engines [i915]] *ERROR* rcs0 reset request timed out: {request: 00000001, RESET_CTL: 00000001}
Dec 25 14:14:04 andrey-lenovo kernel: [drm:gen8_reset_engines [i915]] *ERROR* rcs0 reset request timed out: {request: 00000001, RESET_CTL: 00000001}
Dec 25 14:14:04 andrey-lenovo kernel: [drm] GuC communication enabled
Dec 25 14:14:04 andrey-lenovo kernel: i915 0000:00:02.0: GuC firmware i915/kbl_guc_33.0.0.bin version 33.0 submission:disabled
Dec 25 14:14:04 andrey-lenovo kernel: i915 0000:00:02.0: HuC firmware i915/kbl_huc_ver02_00_1810.bin version 2.0 authenticated:yes
...

System information:

~ >>> inxi -GC
CPU:       Topology: Quad Core model: Intel Core i7-8565U bits: 64 type: MT MCP L2 cache: 8192 KiB 
           Speed: 900 MHz min/max: 400/4600 MHz Core speeds (MHz): 1: 966 2: 1319 3: 1929 4: 1906 5: 3405 6: 3607 7: 1071 
           8: 2607 
Graphics:  Device-1: Intel UHD Graphics 620 driver: i915 v: kernel 
           Display: x11 server: X.Org 1.20.6 driver: modesetting resolution: 1920x1080~60Hz 
           OpenGL: renderer: Mesa DRI Intel UHD Graphics (Whiskey Lake 3x8 GT2) v: 4.6 Mesa 19.3.1 

Can be the crash error file be provided? GPU crash dump saved to /sys/class/drm/card0/error

It has just

No error state collected

https://bugs.archlinux.org/task/64860 https://gitlab.freedesktop.org/drm/intel/issues/161#note_363380

ArchLinux forum on the issue: https://bbs.archlinux.org/viewtopic.php?id=250765

https://gitlab.freedesktop.org/drm/intel/issues/673

Today I've managed to fetch /sys/class/drm/card0/error file:

error file contents

GPU HANG: ecode 9:1:0x00000000, hang on rcs0
Kernel: 5.4.6-2-MANJARO x86_64
Driver: 20190822
Time: 1577355721 s 910731 us
Boottime: 515 s 608672 us
Uptime: 513 s 121577 us
Epoch: 4295031936 jiffies (300 HZ)
Capture: 4295031936 jiffies; 4010 ms ago, 0 ms after epoch
Reset count: 0
Suspend count: 0
Platform: COFFEELAKE
Subplatform: 0x1
PCI ID: 0x3ea0
PCI Revision: 0x00
PCI Subsystem: 17aa:39fd
IOMMU enabled?: 0
DMC loaded: yes
DMC fw version: 1.4
GT awake: yes
RPM wakelock: yes
PM suspended: no
EIR: 0x00000000
IER: 0x08080000
GTIER[0]: 0x01010101
GTIER[1]: 0x01010101
GTIER[2]: 0x80000070
GTIER[3]: 0x00000101
PGTBL_ER: 0x00000000
FORCEWAKE: 0x00010001
DERRMR: 0x2077efef
CCID: 0x00000000
  fence[0] = 00000000
  fence[1] = 00000000
  fence[2] = 00000000
  fence[3] = 00000000
  fence[4] = 00000000
  fence[5] = 00000000
  fence[6] = 00000000
  fence[7] = 00000000
  fence[8] = 00000000
  fence[9] = 00000000
  fence[10] = 00000000
  fence[11] = 00000000
  fence[12] = 00000000
  fence[13] = 00000000
  fence[14] = 00000000
  fence[15] = 00000000
  fence[16] = 00000000
  fence[17] = 00000000
  fence[18] = 00000000
  fence[19] = 00000000
  fence[20] = 00000000
  fence[21] = 00000000
  fence[22] = 00000000
  fence[23] = 00000000
  fence[24] = 00000000
  fence[25] = 00000000
  fence[26] = 00000000
  fence[27] = 00000000
  fence[28] = 00000000
  fence[29] = 00000000
  fence[30] = 00000000
  fence[31] = 00000000
ERROR: 0x00000000
DONE_REG: 0xffffffff
FAULT_TLB_DATA: 0x00000011 0x30ed7356
Num Pipes: 3
Pipe [0]:
  Power: on
  SRC: 077f0437
  STAT: 00000000
Plane [0]:
  CNTR: 00000000
  STRIDE: 0000001e
  SURF: 00000000
  TILEOFF: 00000000
Cursor [0]:
  CNTR: 00000000
  POS: 00000000
  BASE: 00000000
Pipe [1]:
  Power: on
  SRC: 077f0437
  STAT: 00000000
Plane [1]:
  CNTR: c4842400
  STRIDE: 0000000f
  SURF: 01940000
  TILEOFF: 00000000
Cursor [1]:
  CNTR: 05000023
  POS: 0261009e
  BASE: 01880000
Pipe [2]:
  Power: on
  SRC: 00000000
  STAT: 00000000
Plane [2]:
  CNTR: 00000000
  STRIDE: 00000000
  SURF: 00000000
  TILEOFF: 00000000
Cursor [2]:
  CNTR: 00000000
  POS: 00000000
  BASE: 00000000
CPU transcoder: A
  Power: on
  CONF: 00000000
  HTOTAL: 00000000
  HBLANK: 00000000
  HSYNC: 00000000
  VTOTAL: 00000000
  VBLANK: 00000000
  VSYNC: 00000000
CPU transcoder: B
  Power: on
  CONF: c0000000
  HTOTAL: 0897077f
  HBLANK: 0897077f
  HSYNC: 080307d7
  VTOTAL: 04640437
  VBLANK: 04640437
  VSYNC: 0440043b
CPU transcoder: C
  Power: on
  CONF: 00000000
  HTOTAL: 00000000
  HBLANK: 00000000
  HSYNC: 00000000
  VTOTAL: 00000000
  VBLANK: 00000000
  VSYNC: 00000000
CPU transcoder: EDP
  Power: on
  CONF: 00000000
  HTOTAL: 081f077f
  HBLANK: 081f077f
  HSYNC: 07e307b9
  VTOTAL: 04750437
  VBLANK: 04750437
  VSYNC: 043f043a
is_mobile: no
is_lp: no
require_force_probe: no
has_64bit_reloc: yes
gpu_reset_clobbers_display: no
has_reset_engine: yes
has_fpga_dbg: yes
has_global_mocs: no
has_gt_uc: yes
has_l3_dpf: no
has_llc: yes
has_logical_ring_contexts: yes
has_logical_ring_elsq: no
has_logical_ring_preemption: yes
has_pooled_eu: no
has_rc6: yes
has_rc6p: no
has_rps: yes
has_runtime_pm: yes
has_snoop: no
has_coherent_ggtt: yes
unfenced_needs_alignment: no
hws_needs_physical: no
cursor_needs_physical: no
has_csr: yes
has_ddi: yes
has_dp_mst: yes
has_fbc: yes
has_gmch: no
has_hotplug: yes
has_ipc: yes
has_modular_fia: no
has_overlay: no
has_psr: yes
overlay_needs_physical: no
supports_tv: no
Has logical contexts? yes
scheduler: 1f
slice0: 3 subslice(s) (0x7):
	subslice0: 8 EUs (0xff)
	subslice1: 8 EUs (0xff)
	subslice2: 8 EUs (0xff)
	subslice3: 0 EUs (0x0)
slice1: 0 subslice(s) (0x0):
	subslice0: 0 EUs (0x0)
	subslice1: 0 EUs (0x0)
	subslice2: 0 EUs (0x0)
	subslice3: 0 EUs (0x0)
slice2: 0 subslice(s) (0x0):
	subslice0: 0 EUs (0x0)
	subslice1: 0 EUs (0x0)
	subslice2: 0 EUs (0x0)
	subslice3: 0 EUs (0x0)
i915.vbt_firmware=(null)
i915.modeset=-1
i915.lvds_channel_mode=0
i915.panel_use_ssc=-1
i915.vbt_sdvo_panel_type=-1
i915.enable_dc=-1
i915.enable_fbc=1
i915.enable_psr=-1
i915.disable_power_well=1
i915.enable_ips=1
i915.invert_brightness=0
i915.enable_guc=2
i915.guc_log_level=-1
i915.guc_firmware_path=(null)
i915.huc_firmware_path=(null)
i915.dmc_firmware_path=(null)
i915.mmio_debug=0
i915.edp_vswing=0
i915.reset=2
i915.inject_load_failure=0
i915.fastboot=1
i915.enable_dpcd_backlight=0
i915.force_probe=*
i915.alpha_support=yes
i915.enable_hangcheck=yes
i915.prefault_disable=no
i915.load_detect_test=no
i915.force_reset_modeset_test=no
i915.error_capture=yes
i915.disable_display=no
i915.verbose_state_checks=yes
i915.nuclear_pageflip=no
i915.enable_dp_mst=yes
i915.enable_gvt=no
GuC firmware: i915/kbl_guc_33.0.0.bin
	status: RUNNING
	version: wanted 33.0, found 33.0
	uCode: 182528 bytes
	RSA: 256 bytes
HuC firmware: i915/kbl_huc_ver02_00_1810.bin
	status: RUNNING
	version: wanted 2.0, found 2.0
	uCode: 218304 bytes
	RSA: 256 bytes
global --- GuC log buffer = 0x00000000 007ef000
:fBo)Q5lqBJ4hq%+$O^_acIUaR.iA'e0=W5+n?fDZWEn6jNH5Y,_A20A2F,JhD'2/]]m!u_@80?Wo_Ua2,P4CPhB&Ni#MK%*V>pSr+27.:#QF0Jzzzzzzzzhr"J75%W4)ZsLD'VV/A&.%C-?%].M,!.OSqzzzzzzzz$h+,r.tY?.!!!"F

Attached error file: error

Attached are two more error files. err_3 is much longer.

err_2

err_3

The same issue with kernel 5.4.7-1

Jan 03 17:35:33 andrey-lenovo kernel: i915 0000:00:02.0: GPU HANG: ecode 9:1:0x00000000, hang on rcs0
Jan 03 17:35:33 andrey-lenovo kernel: GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
Jan 03 17:35:33 andrey-lenovo kernel: Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel
Jan 03 17:35:33 andrey-lenovo kernel: drm/i915 developers can then reassign to the right component if it's not a kernel issue.
Jan 03 17:35:33 andrey-lenovo kernel: The GPU crash dump is required to analyze GPU hangs, so please always attach it.
Jan 03 17:35:33 andrey-lenovo kernel: GPU crash dump saved to /sys/class/drm/card0/error
Jan 03 17:35:33 andrey-lenovo kernel: i915 0000:00:02.0: Resetting rcs0 for hang on rcs0
Jan 03 17:35:33 andrey-lenovo kernel: [drm:gen8_reset_engines [i915]] *ERROR* rcs0 reset request timed out: {request: 00000001, RESET_CTL: 00000001}
Jan 03 17:35:33 andrey-lenovo kernel: i915 0000:00:02.0: Resetting chip for hang on rcs0
Jan 03 17:35:33 andrey-lenovo kernel: [drm:gen8_reset_engines [i915]] *ERROR* rcs0 reset request timed out: {request: 00000001, RESET_CTL: 00000001}
Jan 03 17:35:33 andrey-lenovo kernel: [drm:gen8_reset_engines [i915]] *ERROR* rcs0 reset request timed out: {request: 00000001, RESET_CTL: 00000001}
Jan 03 17:35:33 andrey-lenovo kernel: [drm] GuC communication enabled
Jan 03 17:35:33 andrey-lenovo kernel: i915 0000:00:02.0: GuC firmware i915/kbl_guc_33.0.0.bin version 33.0 submission:disabled
Jan 03 17:35:33 andrey-lenovo kernel: i915 0000:00:02.0: HuC firmware i915/kbl_huc_ver02_00_1810.bin version 2.0 authenticated:yes

GPU crash dump: err_4

I've tested kernels 5.4.10-1 and 5.4.11-1. GPU hangs do occur sometimes , but the system hangs only for a few seconds.

Jan 15 10:53:08 andrey-lenovo kernel: i915 0000:00:02.0: GPU HANG: ecode 9:1:0x85ff9bc3, in Xorg [1062], hang on rcs0
Jan 15 10:53:08 andrey-lenovo kernel: GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
Jan 15 10:53:08 andrey-lenovo kernel: Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel
Jan 15 10:53:08 andrey-lenovo kernel: drm/i915 developers can then reassign to the right component if it's not a kernel issue.
Jan 15 10:53:08 andrey-lenovo kernel: The GPU crash dump is required to analyze GPU hangs, so please always attach it.
Jan 15 10:53:08 andrey-lenovo kernel: GPU crash dump saved to /sys/class/drm/card0/error
Jan 15 10:53:08 andrey-lenovo kernel: i915 0000:00:02.0: Resetting rcs0 for hang on rcs0

err_8

err_9

EDIT. GPU HANG happened once more. This time hard reset was needed.

err_10

Similar problem here? With i915 & UHD 600

Odroid-H2 -> Intel Celeron J4105 / UHD Graphics 600 (Gen 9.5, GT1, up to 700MHz)

$ journalctl --no-pager -p 3 --no-hostname

Jan 13 19:03:21 kernel: ucsi_acpi USBC000:00: failed to reset PPM!
Jan 13 19:03:21 kernel: ucsi_acpi USBC000:00: PPM init failed (-110)
Jan 13 19:03:30 bluetoothd[581]: connect error: Host is down (112)
Jan 13 22:26:24 kernel: [drm:gen8_reset_engines [i915]] *ERROR* rcs0 reset request timed out: {request: 00000001, RESET_CTL: 00000001}
Jan 13 22:26:24 kernel: [drm:gen8_reset_engines [i915]] *ERROR* rcs0 reset request timed out: {request: 00000001, RESET_CTL: 00000001}
Jan 13 22:26:24 kernel: [drm:gen8_reset_engines [i915]] *ERROR* rcs0 reset request timed out: {request: 00000001, RESET_CTL: 00000001}

Jan 14 04:11:26 kernel: [drm:gen8_reset_engines [i915]] *ERROR* rcs0 reset request timed out: {request: 00000001, RESET_CTL: 00000001}
Jan 14 04:11:26 kernel: [drm:gen8_reset_engines [i915]] *ERROR* rcs0 reset request timed out: {request: 00000001, RESET_CTL: 00000001}
Jan 14 04:11:26 kernel: [drm:gen8_reset_engines [i915]] *ERROR* rcs0 reset request timed out: {request: 00000001, RESET_CTL: 00000001}
Jan 14 18:31:01 kernel: BUG: kernel NULL pointer dereference, address: 0000000000000022
Jan 14 18:31:01 kernel: #PF: supervisor read access in kernel mode
Jan 14 18:31:01 kernel: #PF: error_code(0x0000) - not-present page
-- Reboot --

Jan 15 16:42:07 kernel: ucsi_acpi USBC000:00: failed to reset PPM!
Jan 15 16:42:07 kernel: ucsi_acpi USBC000:00: PPM init failed (-110)
-- Reboot --
Jan 15 16:42:57 kernel: ucsi_acpi USBC000:00: failed to reset PPM!
Jan 15 16:42:57 kernel: ucsi_acpi USBC000:00: PPM init failed (-110)

$ inxi

System:    Kernel: 5.4.6-2-MANJARO x86_64 bits: 64 compiler: gcc v: 9.2.0 
           parameters: BOOT_IMAGE=/boot/vmlinuz-5.4-x86_64 root=UUID=<filter> rw quiet apparmor=1 
           security=apparmor udev.log_priority=3 
           Desktop: Xfce info: xfce4-panel dm: LightDM 1.30.0 Distro: Manjaro Linux 
Machine:   Type: Desktop System: HARDKERNEL product: ODROID-H2 v: 1.0 serial: <filter> 
           Mobo: HARDKERNEL model: ODROID-H2 v: 1.0 serial: <filter> UEFI: American Megatrends v: 5.13 date: 08/22/2019 
CPU:       Topology: Quad Core model: Intel Celeron J4105 bits: 64 type: MCP arch: Goldmont Plus family: 6 model-id: 7A (122) 
           stepping: 1 microcode: 32 L2 cache: 4096 KiB 
           flags: lm nx pae sse sse2 sse3 sse4_1 sse4_2 ssse3 vmx bogomips: 11984 
           Speed: 1818 MHz min/max: 800/2500 MHz Core speeds (MHz): 1: 2446 2: 2396 3: 2385 4: 2395 
           Vulnerabilities: Type: itlb_multihit status: Not affected 
           Type: l1tf status: Not affected 
           Type: mds status: Not affected 
           Type: meltdown mitigation: PTI 
           Type: spec_store_bypass mitigation: Speculative Store Bypass disabled via prctl and seccomp 
           Type: spectre_v1 mitigation: usercopy/swapgs barriers and __user pointer sanitization 
           Type: spectre_v2 mitigation: Enhanced IBRS, IBPB: conditional, RSB filling 
           Type: tsx_async_abort status: Not affected 
Graphics:  Device-1: Intel UHD Graphics 605 driver: i915 v: kernel bus ID: 00:02.0 chip ID: 8086:3185 
           Display: x11 server: X.Org 1.20.6 driver: intel unloaded: modesetting alternate: fbdev,vesa tty: N/A 
           OpenGL: renderer: Mesa DRI Intel UHD Graphics 600 (Geminilake 2x6) v: 4.6 Mesa 19.3.1 compat-v: 3.0 
           direct render: Yes 

Please check if 5.4.13-2 makes it any better ...

The GPU hang still happens with 5.4.13-2.

err_12

I have also been experiencing this issue with CPU/GPU Intel i5-8250U/Intel UHD 620. Typically the hang requires a full reset, though as of 5.4.12 the system has been recoverable on occasion. Haven't had much uptime with 5.4.13 but will report back if the issue persists here and with later versions.

Issue is fixed in linux55, however we didn't found the needed patch to backport yet.

Using Git to bisect Linux sources from 219d54332a09 (Linux 5.4, i915 broken) to e42617b825f8 (Linux 5.5-rc1, i915 fixed) yielded 5f71c84038d3 as the first commit in which the error does not occur.

Applying the diff from 219d54332a09 to 5f71c84038d3 to linux54 kernel sources (see i915.patch; git diff 219d54332a09 5f71c84038d3 -- drivers/gpu/drm/i915) fixes the error for me. Patch 0001-revert-drm-i915-cmdparser-use-explicit-goto-for-error-paths.patch needs to be removed for the diff to work (see linux54.patch).

Caveat: The patch is quite large and changes 52 files in drivers/gpu/drm/i915.

So actually you didn't git-bisect to the commit which first break it? Your patch is kinda half-backed as it doesn't pin-point the issue. Some even report that i915 is now broken for them ...

here is such an example. I might revert that patch now.

@tschiatf thx for bi-secting it and providing a patch-set. Will try it out in revision 2 of 5.4.14 kernel.

I've been running kernel 5.4.14-2 for about 5 hours. No GPU hangs so far. Probably the issue is solved.

Today I had GPU hang again. Kernel 5.4.14-2. This time no hard reset was needed, GPU returned to normal after about 15 seconds.

Jan 27 11:59:48 kernel: i915 0000:00:02.0: GPU HANG: ecode 9:1:0x00000000, hang on rcs0
Jan 27 11:59:48 kernel: GPU hangs can indicate a bug anywhere in the entire gfx stack, including userspace.
Jan 27 11:59:48 kernel: Please file a _new_ bug report on bugs.freedesktop.org against DRI -> DRM/Intel
Jan 27 11:59:48 kernel: drm/i915 developers can then reassign to the right component if it's not a kernel issue.
Jan 27 11:59:48 kernel: The GPU crash dump is required to analyze GPU hangs, so please always attach it.
Jan 27 11:59:48 kernel: GPU crash dump saved to /sys/class/drm/card0/error
Jan 27 11:59:48 kernel: i915 0000:00:02.0: Resetting rcs0 for hang on rcs0
Jan 27 11:59:51 kernel: iwlwifi 0000:73:00.0: Queue 11 is inactive on fifo 2 and stuck for 10000 ms. SW [232, 240] HW [162, 162] FH TRB=0x0a5a5a5a2
Jan 27 12:00:00 kernel: i915 0000:00:02.0: Resetting rcs0 for hang on rcs0

err_13

What about 5.4.15-1?

I'll try to add i915 drm of 5.5 now to 5.4 series and only apply the differences: git diff 5.4.15 5.5 -- drivers/gpu/drm/i915 using the stable tree ...

The issue persists with the kernel 5.4.15-2.

err_14