On Fri, Nov 01, 2024 at 09:32:23AM -0400, Thomas Frohwein wrote: > >Synopsis: amdgpu Radeon 780M graphics hang after page fault > >Category: system > >Environment: > System : OpenBSD 7.6 > Details : OpenBSD 7.6-current (GENERIC.MP) #393: Sat Oct 26 > 21:59:25 MDT 2024 > > dera...@amd64.openbsd.org:/usr/src/sys/arch/amd64/compile/GENERIC.MP > > Architecture: OpenBSD.amd64 > Machine : amd64 > >Description: > After a variable amount of time of running a Godot game (Brotato), > generally > 1-5 minutes, the system hangs and xconsole/messages show the following > just > before the hang: > > gmc_v11_0_process_interrupt *ERROR* [gfxhub] page fault (src_id:0 ring:24 > vmid:2 pasid:32779, for process godot pid 7378 thread godot pid 387416) > gmc_v11_0_process_interrupt *ERROR* in page starting at address > 0x0000000000000000 from client 10 > gfxhub_v3_0_print_l2_protection_fault_status *ERROR* > GCVM_L2_PROTECTION_FAULT_STATUS:0x00201430 > gfxhub_v3_0_print_l2_protection_fault_status *ERROR* Faulty UTCL2 client > ID: SQC (data) (0xa) > gfxhub_v3_0_print_l2_protection_fault_status *ERROR* MORE_FAULTS: 0x0 > gfxhub_v3_0_print_l2_protection_fault_status *ERROR* WALKER_ERROR: 0x0 > gfxhub_v3_0_print_l2_protection_fault_status *ERROR* PERMISSION_FAULTS: 0x3 > gfxhub_v3_0_print_l2_protection_fault_status *ERROR* MAPPING_ERROR: 0x0 > gfxhub_v3_0_print_l2_protection_fault_status *ERROR* RW: 0x0 > > >How-To-Repeat: > So far only found with the Godot 3 game Brotato (but repeatedly, after > variable time). I've tried hard to reproduce it with 0ad or other Godot > games, but so far has only happened with Brotato. > I have reproduced it several times with it. > > >Fix: > Unknown. I have tried setting AMD_DEBUG=dcc and then AMD_DEBUG=nodcc > based on [1,2]. > > [1] https://gitlab.freedesktop.org/drm/amd/-/issues/2496 > [2] https://gitlab.freedesktop.org/drm/amd/-/issues/2690
AMD people usually flag these as problems with Mesa (without specifying specific fixes). > amdgpu0: IP DISCOVERY GC 11.0.1 12 CU rev 0x09 There is newer GC 11.0.1 firmware, does it help? The linux-firmware commit doesn't specify what changed. Index: sysutils/firmware/amdgpu/Makefile =================================================================== RCS file: /cvs/ports/sysutils/firmware/amdgpu/Makefile,v diff -u -p -r1.30 Makefile --- sysutils/firmware/amdgpu/Makefile 11 Sep 2024 06:32:58 -0000 1.30 +++ sysutils/firmware/amdgpu/Makefile 2 Nov 2024 11:34:05 -0000 @@ -1,5 +1,5 @@ FW_DRIVER= amdgpu -FW_VER= 20240909 +FW_VER= 20241017 DISTNAME= linux-firmware-${FW_VER} EXTRACT_SUFX= .tar.xz EXTRACT_FILES= ${DISTNAME}/{LICENSE.\*,\*.bin} Index: sysutils/firmware/amdgpu/distinfo =================================================================== RCS file: /cvs/ports/sysutils/firmware/amdgpu/distinfo,v diff -u -p -r1.27 distinfo --- sysutils/firmware/amdgpu/distinfo 11 Sep 2024 06:32:58 -0000 1.27 +++ sysutils/firmware/amdgpu/distinfo 2 Nov 2024 11:34:42 -0000 @@ -1,2 +1,2 @@ -SHA256 (firmware/linux-firmware-20240909.tar.xz) = lD+9GYg8+OrfieCyJCJUnbBWVXsezTClZABhWXE2lnE= -SIZE (firmware/linux-firmware-20240909.tar.xz) = 383099276 +SHA256 (firmware/linux-firmware-20241017.tar.xz) = omw471qDJy8rmM6L+MoYZahSo97qSc5ajdgEuRQ1EnM= +SIZE (firmware/linux-firmware-20241017.tar.xz) = 397400292 Index: sysutils/firmware/amdgpu/pkg/PLIST =================================================================== RCS file: /cvs/ports/sysutils/firmware/amdgpu/pkg/PLIST,v diff -u -p -r1.19 PLIST --- sysutils/firmware/amdgpu/pkg/PLIST 1 Jul 2024 06:46:15 -0000 1.19 +++ sysutils/firmware/amdgpu/pkg/PLIST 2 Nov 2024 11:35:10 -0000 @@ -160,6 +160,13 @@ firmware/amdgpu/gc_11_5_1_mes1.bin firmware/amdgpu/gc_11_5_1_mes_2.bin firmware/amdgpu/gc_11_5_1_pfp.bin firmware/amdgpu/gc_11_5_1_rlc.bin +firmware/amdgpu/gc_11_5_2_imu.bin +firmware/amdgpu/gc_11_5_2_me.bin +firmware/amdgpu/gc_11_5_2_mec.bin +firmware/amdgpu/gc_11_5_2_mes1.bin +firmware/amdgpu/gc_11_5_2_mes_2.bin +firmware/amdgpu/gc_11_5_2_pfp.bin +firmware/amdgpu/gc_11_5_2_rlc.bin firmware/amdgpu/gc_9_4_3_mec.bin firmware/amdgpu/gc_9_4_3_rlc.bin firmware/amdgpu/green_sardine_asd.bin @@ -393,6 +400,8 @@ firmware/amdgpu/psp_14_0_0_ta.bin firmware/amdgpu/psp_14_0_0_toc.bin firmware/amdgpu/psp_14_0_1_ta.bin firmware/amdgpu/psp_14_0_1_toc.bin +firmware/amdgpu/psp_14_0_4_ta.bin +firmware/amdgpu/psp_14_0_4_toc.bin firmware/amdgpu/raven2_asd.bin firmware/amdgpu/raven2_ce.bin firmware/amdgpu/raven2_gpu_info.bin @@ -438,6 +447,7 @@ firmware/amdgpu/sdma_6_0_2.bin firmware/amdgpu/sdma_6_0_3.bin firmware/amdgpu/sdma_6_1_0.bin firmware/amdgpu/sdma_6_1_1.bin +firmware/amdgpu/sdma_6_1_2.bin firmware/amdgpu/si58_mc.bin firmware/amdgpu/sienna_cichlid_ce.bin firmware/amdgpu/sienna_cichlid_dmcub.bin @@ -579,6 +589,7 @@ firmware/amdgpu/verde_smc.bin firmware/amdgpu/verde_uvd.bin firmware/amdgpu/vpe_6_1_0.bin firmware/amdgpu/vpe_6_1_1.bin +firmware/amdgpu/vpe_6_1_3.bin firmware/amdgpu/yellow_carp_asd.bin firmware/amdgpu/yellow_carp_ce.bin firmware/amdgpu/yellow_carp_dmcub.bin