Gamescope fails with "Failed to pin framebuffer with error -12" shortly after sleep
Brief summary of the problem:
On an Ally X unit, running on a 6.11.4 kernel variant with the following patches on top of kernel-ark, after suspend and within 20 minutes of gameplay (in this case GTA IV), gamescope displays a black screen and restarts
dmesg reports:
[ 4063.692766] amdgpu 0000:64:00.0: amdgpu: 000000008ac73718 pin failed
[ 4063.692774] [drm:amdgpu_dm_plane_helper_prepare_fb [amdgpu]] *ERROR* Failed to pin framebuffer with error -12gamescope reports:
Could not resolve keysym XF86KbdInputAssistNextgrou
Errors from xkbcomp are not fatal to the X server
Errors from xkbcomp are not fatal to the X server
(EE)(EE) failed to read Wayland events: Broken pipe
failed to read Wayland events: Broken pipe
[gamescope] [Error] drm: drmModeRmFB failed: Bad file descriptor
[gamescope] [Error] drm: drmModeRmFB failed: Bad file descriptor
[gamescope] [Error] drm: drmModeRmFB failed: Bad file descriptor
[gamescope] [Error] drm: drmModeRmFB failed: Bad file descriptorHardware description:
- CPU: Z1 Extreme
- GPU: 780M
- System Memory: 24GB (512MB is vram; AutoUMA)
- Display(s): 1080p VRR built in display, 120hz.
System information:
- Distro name and Version: Bazzite Unstable on F41
- Kernel version: 6.11.4 See here. 6.10 had a similar issue and was skipped (w/ AutoUMA as well).
How to reproduce the issue:
See above.
Log files (for system lockups / game freezes / crashes)
Full journalctl of the session: 546e64c4.txt
Misses the gamescope error, that will be fixed soon. For it, see above.
Activity
added Phoenix label
added TTM label
Due to the setup required to diagnose the issue (gamescope + game + steam) and the time it takes to reproduce the bug, it is unfortunately not possible to bisect the issue. It would take too long.
Which is very unfortunate.
Any pointers on what to try to revert/change would be appreciated.
There have been the following amdgpu commits between kernel versions 6.9 and 6.10:
❯ git log v6.9..v6.10 --format="%h: %s" drivers/gpu/drm/amd/amdgpu 48880f9686b1: drm/amdgpu: Don't show false warning for reg list bcfa48ff785b: drm/amdgpu: avoid using null object of framebuffer 74fa02c4a5ea: drm/amdgpu: Fix pci state save during mode-1 reset f6f49dda49db: drm/amdgpu/atomfirmware: fix parsing of vram_info ed5a4484f074: drm/amdgpu: init TA fw for psp v14 e356d321d024: drm/amdgpu: cleanup MES11 command submission 8bd82363e2ee: drm/amdgpu: revert "take runtime pm reference when we attach a buffer" v2 49c9ffabde55: drm/amdgpu: Indicate CU havest info to CP 84801d4f1e4f: drm/amdgpu: fix locking scope when flushing tlb 14731a640e55: Merge drm/drm-fixes into drm-misc-fixes 31849bf07e0f: drm/amdgpu: Fix the BO release clear memory warning bb61cf46b66a: Merge tag 'amd-drm-fixes-6.10-2024-05-30' of https://gitlab.freedesktop.org/agd5f/linux into drm-fixes a9bc5a19e495: drm/amdgpu: Make CPX mode auto default in NPS4 1f327dfc846a: drm/amdkfd: simplify APU VRAM handling a0cf36546cc2: drm/amdgpu: fix dereference null return value for the function amdgpu_vm_pt_parent ba46b3bda296: drm/amdgpu: Adjust logic in amdgpu_device_partner_bandwidth() 56fb6f92854f: Merge tag 'drm-next-2024-05-25' of https://gitlab.freedesktop.org/drm/kernel ec58991054e8: drm/amdgpu: correct hbm field in boot status 2c92ca849fcc: tracing/treewide: Remove second parameter of __assign_str() e64e8f7c178e: drm/amdgpu/atomfirmware: add intergrated info v2.3 table f0bae243b2bc: Merge tag 'pci-v6.10-changes' of git://git.kernel.org/pub/scm/linux/kernel/git/pci/pci eb853413d02c: drm/amdkfd: Let VRAM allocations go to GTT domain on small APUs 2a705f3e49d2: drm/amdkfd: handle duplicate BOs in reserve_bo_and_cond_vms ff9a79307f89: Merge tag 'kbuild-v6.10' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild db5d28c0bfe5: Merge tag 'drm-next-2024-05-15' of https://gitlab.freedesktop.org/drm/kernel b1992c3772e6: kbuild: use $(src) instead of $(srctree)/$(src) for source directory 05b8b6dd225d: Revert "drm: Switch DRM_DISPLAY_HELPER to depends on" 7fe302ae198a: Revert "drm: Switch DRM_DISPLAY_DP_HELPER to depends on" 95734469533c: Revert "drm: Switch DRM_DISPLAY_HDCP_HELPER to depends on" d7c128cb775e: Revert "drm: Switch DRM_DISPLAY_HDMI_HELPER to depends on" 4a56c0ed5aa0: Merge tag 'amd-drm-next-6.10-2024-04-26' of https://gitlab.freedesktop.org/agd5f/linux into drm-next b84bc948528e: Merge v6.9-rc6 into drm-next ea686fef5489: drm/amdgpu: fix the warning about the expression (int)size - len 029c2b03892b: drm/amdgpu/mes: add mes mapping legacy queue support 9a5f15d2a29d: drm/amdgpu: fix uninitialized scalar variable warning 2a8f7464d33c: drm/amdgpu: skip ip dump if devcoredump flag is set e362b7c8f8c7: drm/amdgpu: Modify the contiguous flags behaviour bd31e5026dc3: drm/amdkfd: Enable SQ watchpoint for gfx10 acce6479e30f: drm/amdgpu: Fix buffer size in gfx_v9_4_3_init_ cp_compute_microcode() and rlc_microcode() 6f3b69139c3c: drm/amdgpu: Fix ras mode2 reset failure in ras aca mode 506c245f3f1c: drm/amdgpu: fix double free err_addr pointer warnings 7bfd16d0ec37: drm/amdgpu: initialize the last_jump_jiffies in atom_exec_context 2d10c3dbde07: drm/amdgpu: add check before free wb entry cd48b97ce778: drm/amdgpu: add return result for amdgpu_i2c_{get/put}_byte 8b2faf1a4f3b: drm/amdgpu: add error handle to avoid out-of-bounds 2e55bcf3d742: drm/amdgpu: Initialize timestamp for some legacy SOCs 48fa90718b2a: drm/amdgpu: Use new interface to reserve bad page bcc093488503: drm/amdgpu: Fix address translation defect e02387408117: drm/amdgpu: support ACA logging ecc errors 370fbff4cc6f: drm/amdgpu: add poison consumption handler bfa579b38b86: drm/amdgpu: prepare to handle pasid poison consumption 314c38cde687: drm/amdgpu: retire bad pages for umc v12_0 e74313be5a71: drm/amdgpu: add condition check for amdgpu_umc_fill_error_record 2cf8e50ec381: drm/amdgpu: Add delay work to retire bad pages f27defca6882: drm/amdgpu: umc v12_0 logs ecc errors b2aa6b108dd3: drm/amdgpu: umc v12_0 converts error address 95b4063de4f4: drm/amdgpu: add interface to update umc v12_0 ecc status a734adfbcdb0: drm/amdgpu: add poison creation handler f493dd64ee66: drm/amdgpu: prepare for logging ecc errors 98b5bc878d4b: drm/amdgpu: add message fifo to handle RAS poison events 88a9a467c548: drm/amdgpu: Using uninitialized value *size when calling amdgpu_vce_cs_reloc eef016ba8986: drm/amdgpu/mes11: Use a separate fence per transaction 497d7cee2457: drm/amdgpu: add a spinlock to wb allocation 754c366e41d2: drm/amdgpu: update fw_share for VCN5 8e1d1905951d: drm/amdgpu: Fix VRAM memory accounting c551316e150b: drm/amdgpu: update jpeg max decode resolution af8644121e3e: drm/amdgpu: add ip dump for each ip in devcoredump e043a35dc244: drm/amdgpu: dump ip state before reset for each ip c8732c80debb: drm/amdgpu: add support for gfx v10 print 40356542c361: drm/amdgpu: add protype for print ip state c395dbb68b29: drm/amdgpu: add support of gfx10 register dump e21d253bd74b: drm/amdgpu: add prototype for ip dump af730e082035: drm/amdgpu: Add interface to reserve bad page 60c448439f3b: drm/amdgpu: Fix uninitialized variable warnings f88da7fbf665: drm/amdgpu/mes: fix use-after-free issue e0a9bbeea002: drm/amdgpu/sdma5.2: use legacy HDP flush for SDMA2/3 a16b95158644: drm/amdgpu: Update CGCG settings for GFXIP 9.4.3 ea9238a81b3a: drm/amdgpu: replace tmz flag into buffer flag 92ed1e9cd5f6: drm/amdgpu: init microcode chip name from ip versions 26de73bc0a73: drm/amdgpu: Fix the ring buffer size for queue VM flush a522ec528cc7: drm/amdgpu/umsch: don't execute umsch test when GPU is in reset/suspend 6927b0168059: drm/amdgpu: Use PCI_IRQ_INTX instead of PCI_IRQ_LEGACY e76691f45a60: drm/amdgpu: Update BO eviction priorities 6e042cee748f: drm/amdgpu/vcn: fix unitialized variable warnings 3f0664110a40: drm/amdgpu/mes11: print MES opcodes rather than numbers 2476c6bd950e: drm/amdgpu/vpe: fix vpe dpm setup failed 7b19f1f3466f: drm/amdgpu: Assign correct bits for SDMA HDP flush 939c4751819b: drm/amdgpu: Support setting reset_method at runtime c058e7a8f8af: Merge drm/drm-next into drm-misc-next a68c7eaa7a8f: drm/amdgpu: Enable clear page functionality 96950929eb23: drm/buddy: Implement tracking clear page feature 377b5b397d07: Merge tag 'amd-drm-next-6.10-2024-04-19' of https://gitlab.freedesktop.org/agd5f/linux into drm-next 81bf14519a8c: drm/amdkfd: make sure VM is ready for updating operations e53a1713de31: drm/amdgpu: Fix leak when GPU memory allocation fails 6a009ca1bf94: drm/amdgpu: remove virt_init_data_exchange from poison consumption handler 8954c3fbe764: drm/amdgpu: Change AID detection logic 93522c19488e: drm/amdgpu: enable redirection of irq's for IH V6.1 ea137071ada1: drm/amdgpu: Skip the coredump collection on reset during driver reload ca0afa2f4161: drm/amdgpu: enable redirection of irq's for IH V6.0 5adcd78fa2bc: drm:amdgpu: enable IH ring1 for IH v6.1 eefc85a2779d: drm:amdgpu: enable IH RB ring1 for IH v6.0 34633158b8eb: Merge tag 'amd-drm-next-6.10-2024-04-13' of https://gitlab.freedesktop.org/agd5f/linux into drm-next 0c1195ca0d02: drm/amd/swsmu: support smu block discovery for smu v14 577cbed31818: drm/amdgpu: rename DBG_DRV to HAD_DRV for psp v14 1347853271ed: drm/amdgpu: refactoring the runtime pm mode detection code 6c6acc5f33ab: drm/amdgpu: Load ipkeymgr drv for psp v14 12b8b4e68510: drm/amdgpu: Add missing space to DRM_WARN() message 959056982a9b: drm/amdgpu: Fix discovery initialization failure during pci rescan 394ae0603a67: drm/amdgpu: fix visible VRAM handling during faults 98856136c485: drm/amdgpu: validate the parameters of bo mapping operations more clearly f23558627f2b: drm/amdgpu: add new aca smu callback func parse_error_code() f7c161a4c250: drm/amdgpu: increase mes submission timeout 3c858cf65e9a: drm/amdgpu: add missing vbios version from devcoredump 8b9130bae048: drm/amdgpu/gfx11: properly handle regGRBM_GFX_CNTL in soft reset c8962679af35: drm/amdgpu: remove invalid resource->start check v2 a0e002cdac42: drm/amdgpu/sdma6: set sdma hang watchdog 6b0d78032f98: drm/amd/amdgpu: Update PF2VF Header 526b184e8883: drm/amdgpu: differentiate external rev id for gfx 11.5.0 d6d6561f936b: drm/amdgpu: fix incorrect number of active RBs for gfx11 d1999b4017d4: amd/amdgpu: improve VF recover time b41f742d6fa6: drm/amdgpu: Set fatal errror detected flag earlier 05e40141685f: drm/amdgpu: clear set_q_mode_offs when VM changed 4b0cb230bdb7: drm/amdgpu: retire UMC v12 mca_addr_to_pa f6ac0842364a: drm/amd/amdgpu: support MES command SET_HW_RESOURCE1 in sriov 9ecef5b2d0a0: drm/amdgpu: update check condition for XGMI ACA UE 91bc86011661: drm/amdgpu: Fix VCN allocation in CPX partition fcc0735b0087: drm/amdgpu: Add support for BAMACO mode checking 327eec542746: drm/amdgpu: Bypass asd if display hw is not available b2207dc6989f: drm/amdgpu/pm: Add support for MACO flag checking 7c1d9e10e664: drm/amd/pm: fix the high voltage issue after unload 3e2dacca5406: drm/amdgpu: use vm_update_mode=0 as default in sriov for gfx10.3 onwards b7a1a0ef12b8: drm/amd/amdgpu: add pipe1 hardware support 0453e5f2202e: drm/amdgpu: select HDP ref/mask according to gfx ring pipe 029faefb7302: drm/amdgpu: implement IRQ_STATE_ENABLE for SDMA v4.4.2 f5a3507c4abf: drm/amdgpu: add smu 14.0.1 discovery support 8966c3167402: drm/amdgpu : Increase the mes log buffer size as per new MES FW version e58acb7613aa: drm/amdgpu : Add mes_log_enable to control mes log feature 81d96e8b5a85: drm/amdgpu: refine function signature of amdgpu_aca_get_error_data() df3c7dc5c58b: drm/amdgpu: Reset dGPU if suspend got aborted 6a0e1bafd70f: drm/amdgpu: add IP's FW information to devcoredump 108ab31be9d5: drm/amdgpu/umsch: reinitialize write pointer in hw init cd409dbc6986: drm/amdgpu: Refine IB schedule error logging fee54d08bc83: Merge tag 'drm-misc-next-2024-03-28' of https://gitlab.freedesktop.org/drm/misc/kernel into drm-next 0d21364c6e8d: Merge drm/drm-next into drm-misc-next f6d2dc03fa85: drm: Switch DRM_DISPLAY_HDMI_HELPER to depends on 3166e7e6d935: drm: Switch DRM_DISPLAY_HDCP_HELPER to depends on 0323287de87d: drm: Switch DRM_DISPLAY_DP_HELPER to depends on e075e496f516: drm: Switch DRM_DISPLAY_HELPER to depends on d7f148764355: drm/amdgpu: always force full reset for SOC21 c25d09bcb79f: drm/amdgpu: fix deadlock while reading mqd from debugfs b9a8aee136b7: drm/amdgpu: enable UMSCH 4.0.6 f3e698978cfb: drm/amdgpu/umsch: update UMSCH 4.0 FW interface 0355b24bdec3: drm/amd: Flush GFXOFF requests in prepare stage eb4f6eca2632: drm/amdgpu: Fix truncations in gfx_v11_0_init_microcode() 8e4617c25d53: drm/amdgpu: simplify convert_error_address interface for UMC v12 539ff12ee5e4: drm/amdgpu: Fix truncation issues in gfx_v9_0.c 927a8a800ebb: drm/amdgpu: Fix truncation in gfx_v10_0_init_microcode 20fd14460f45: drm/amdgpu: Fix 'fw_name' buffer size to prevent truncations in amdgpu_mes_init_microcode 7c2bc34ab926: drm/amdgpu: Fix format character cut-off issues in amdgpu_vcn_early_init() 8b3495eafb4d: drm/amdgpu: add socket id parameter for psp query address cmd f88a7dd06ab4: drm/amdgpu: Add a NULL check for freeing root PT 9022f01b9709: drm/amdgpu: refactor code to split devcoredump code 9ddafd1d1404: drm/amdgpu/vpe: power on vpe when hw_init 31fd330b97ba: drm/amdgpu: add ras event id support for ACA bd15bf742f6d: drm/amdgpu: avoid update aca bank multi times during ras isr f7bcfb7a56b2: drm/amdgpu: retrieve umc odecc error count for aca umc v12.0 b6c4f90b3819: drm/amdgpu: sync page table freeing with tlb flush a61e2ce9d425: drm/amdgpu: Enable smuio v14_0_2 callbacks d80e44a34e25: drm/amdgpu: Add smuio callback to get gpu clk counter 2d93151de890: drm/amdgpu: Add smuio v14_0_2 ip block support b93d759f540a: drm/amdgpu: add umc v12.0.0 deferred error support 865d3397630b: drm/amdgpu: add aca deferred error type support 2fc46e0b2fe8: drm/amdgpu: make reset method configurable for RAS poison e3d4de8d8b24: drm/amdgpu: retire unused aca_bank_report data structure f26c4e3fc999: drm/amdgpu: Update setting EEPROM table version 69bf42fbb227: drm/amdgpu: refine aca error cache for umc v12.0 87428b405437: drm/amdgpu: refine aca error cache for sdma v4.4.2 62d2aaa7d466: drm/amdgpu: refine aca error cache for xgmi v6.4.0 d8070c424108: drm/amdgpu: support utcl2 RAS poison query for mmhub 176c3e89567f: drm/amdgpu: add utcl2 RAS poison query for mmhub 5275114a7043: drm/amdgpu: refine aca error cache for mmhub v1.8 d8a3f0a0348d: drm/amdgpu: implement TLB flush fence e6136150cd26: drm/amdgpu: refine aca error cache for gfx v9.4.3 949899cbacf5: drm/amdgpu: add new api to save error count into aca cache abc3b5d21d34: drm/amdgpu: add new aca_smu_type support 6fe4dab331a7: drm/amdgpu: remove the adev check for NULL 3cfaadbe0fcb: drm/amdgpu: add support for atom fw version v3_5 765bea0d73b1: drm/amdgpu: Apply retry to IP discovery v2 and v4 9dc57c2adf2c: drm/amdgpu: add ras event id support ab66c832847f: drm/amdgpu: trigger flr_work if reading pf2vf data failed d72e2bdac4ad: drm/amdgpu: add the hw_ip version of all IP's 6bb89d134042: drm/amdgpu: Skip virt_exchange_init on SDMA poison consumption dfe9c3cde229: drm/amdgpu: Do a basic health check before reset 97d5aa60306d: drm/amdgpu: cleanup unused variable 0c501d3c11bb: drm/amdgpu: skip GFX FED error in page fault handling 71a8d61ebc38: drm/amdgpu: retire gfx ras query_utcl2_poison_status 3eb899c40a61: drm/amdgpu: add ring buffer information in devcoredump 583681d4a417: drm/amdgpu: add vm fault information to devcoredump fb0f5f541475: drm/amdgpu: add utcl2 poison query for gfxhub dc406d92a097: drm/amdgpu: add recent pagefault info in vm_manager d6eb77731c45: Merge drm/drm-next into drm-misc-next 216c1282dde3: drm/amdgpu: use GTT only as fallback for VRAM|GTTFrom those, the following look suspicious:
31849bf07e0f: drm/amdgpu: Fix the BO release clear memory warning 1f327dfc846a: drm/amdkfd: simplify APU VRAM handling eb853413d02c: drm/amdkfd: Let VRAM allocations go to GTT domain on small APUs 2a705f3e49d2: drm/amdkfd: handle duplicate BOs in reserve_bo_and_cond_vms 216c1282dde3: drm/amdgpu: use GTT only as fallback for VRAM|GTT 394ae0603a67: drm/amdgpu: fix visible VRAM handling during faults e76691f45a60: drm/amdgpu: Update BO eviction priorities 8e1d1905951d: drm/amdgpu: Fix VRAM memory accounting 8b2faf1a4f3b: drm/amdgpu: add error handle to avoid out-of-boundsAnd from those, the following have a description that seems relevant
216c1282dde3: drm/amdgpu: use GTT only as fallback for VRAM|GTT eb853413d02c: drm/amdkfd: Let VRAM allocations go to GTT domain on small APUs 394ae0603a67: drm/amdgpu: fix visible VRAM handling during faults e76691f45a60: drm/amdgpu: Update BO eviction priorities 8e1d1905951d: drm/amdgpu: Fix VRAM memory accountingAnd from those, the following have a content that is very relevant: 216c1282dde3: drm/amdgpu: use GTT only as fallback for VRAM|GTT eb853413d02c: drm/amdkfd: Let VRAM allocations go to GTT domain on small APUs 394ae0603a67: drm/amdgpu: fix visible VRAM handling during faults 8e1d1905951d: drm/amdgpu: Fix VRAM memory accounting
Edited by Antheas Kapenekakiseb853413d02c: drm/amdkfd: Let VRAM allocations go to GTT domain on small APUsonly affects ROCm applications so probably not that one. @ckoenig I think there may have been an issue on APUs caused by the fix for handling buffer migration back to VRAM after suspend on dGPUs for vulkan. That should not be necessary on APUs. Do you remember whether the details on that?
Ok messing with a revert of 216c1282dde3: drm/amdgpu: use GTT only as fallback for VRAM|GTT
Cannot get it to crash with AutoUMA for the last hour. Seems like that was it, perhaps combined with something else. Full patch series below. We will continue testing it ofc.
https://github.com/hhd-dev/patchwork/releases/tag/bazzite-6.11.4-305
Is the game running before the suspend?
So the process to reproduce the bug would be:
- start the game, play for a while
- suspend, resume
- play for a while -> hit the issue
Is that right?
edit: if you can still reproduce, would you be able to attach the memory usage statistics before/after suspend? (= save the content of
/sys/class/drm/renderD128/device/mem_info_gtt_used,/sys/class/drm/renderD128/device/mem_info_vram_usedand/sys/class/drm/renderD128/device/mem_info_vis_vram_used)Edited by Pierre-Eric Pelloux-PrayerHi Pierre, Yes. I think I hit it once without even suspending or just being in SteamUI.
The pattern was that randomly within 2 hours after booting there would be a crash. I think suspending made it happen a lot faster (within 1-2 minutes).
I cannot reproduce the bug anymore with the new kernel. I played GTA IV for 2 days with dozens of suspends without rebooting and getting that crash. Then got a hard lock with no logs which is very likely for a completely different reason.
Here is what you asked for with the new kernel (after the revert).
Before suspend:
bazzite@antheas-ally-x:~$ for f in /sys/class/drm/renderD128/device/mem_*; do echo "$f:"; sudo cat $f; done /sys/class/drm/renderD128/device/mem_info_gtt_total: 12784238592 /sys/class/drm/renderD128/device/mem_info_gtt_used: 1611788288 /sys/class/drm/renderD128/device/mem_info_preempt_used: 0 /sys/class/drm/renderD128/device/mem_info_vis_vram_total: 536870912 /sys/class/drm/renderD128/device/mem_info_vis_vram_used: 504123392 /sys/class/drm/renderD128/device/mem_info_vram_total: 536870912 /sys/class/drm/renderD128/device/mem_info_vram_used: 500355072After Suspend:
bazzite@antheas-ally-x:~$ for f in /sys/class/drm/renderD128/device/mem_*; do echo "$f:"; sudo cat $f; done /sys/class/drm/renderD128/device/mem_info_gtt_total: 12784238592 /sys/class/drm/renderD128/device/mem_info_gtt_used: 1611382784 /sys/class/drm/renderD128/device/mem_info_preempt_used: 0 /sys/class/drm/renderD128/device/mem_info_vis_vram_total: 536870912 /sys/class/drm/renderD128/device/mem_info_vis_vram_used: 479014912 /sys/class/drm/renderD128/device/mem_info_vram_total: 536870912 /sys/class/drm/renderD128/device/mem_info_vram_used: 473935872If you think it is worth it I could test with the broken kernel, given it would take me an hour or so.
Edited by Antheas KapenekakisYes, ideally reporting these numbers with a broken kernel would be useful. Capturing:
/sys/kernel/debug/dri/0/amdgpu_vram_mmand/sys/kernel/debug/dri/0/amdgpu_gtt_mmcould also be helpful to identify a fragmentation issue.(even better, if possible, would be to log these files regularly, so we can see their evolution until the "Failed to pin" error)
Here you go. Performance up to the crash was significantly worse on the broken kernel.
Two folders: kernel-working and kernel-broken + the capture script
Inside each folder there is a log containing summary information about the capture and in the folders are the files you asked. You can monitor the wakeups to see when the suspend happens. Before the suspend, wakeup count increases by one, and after by 2-3.
Kernels:
Edited by Antheas KapenekakisThanks!
Performance up to the crash was significantly worse on the broken kernel.
Did the crash occured at the end of the capture or between these 2 lines:
169: VRAM: 470MB/537MB, GTT: 1102MB/12784MB, Wakeups: 12, Dir: kernel-broken/00169
171: VRAM: 80MB/537MB, GTT: 14MB/12784MB, Wakeups: 12, Dir: kernel-broken/00171
Well, I have to be honest and say that since the recording is every 2 seconds I cannot really tell exactly. But after 163 there was a change as well.
163: VRAM: 518MB/537MB, GTT: 1689MB/12784MB, Wakeups: 11, Dir: kernel-broken/00163 165: VRAM: 470MB/537MB, GTT: 1102MB/12784MB, Wakeups: 12, Dir: kernel-broken/00165I can tell you that definitely it was crashed after 171.
I stopped the capture manually around 10 seconds after it stopped responding.
Edited by Antheas Kapenekakis
Had the other crash happen again and remembered there was an issue mentioned about it: https://www.reddit.com/r/ROGAllyX/comments/1eoch1d/rog_ally_x_black_screen_crashes_has_it_happened/
That looks like what I am experiencing now (after the revert).
The only thing I can think of about it is that we are missing a firmware notification for AC/DC and perhaps it is running a bit hotter?
At least it is probably not 6.11 related.
Reverting works 100%. Was asked about the reverted patch as the revert affects dGPUs.
Related to: #2893 (closed)
Seems like this patch is 90% of the gain.
So we will test this the following few days: https://github.com/hhd-dev/patchwork/commit/f683e78eec449d9106f8c01a7ddfc34c6e6a2a18
Edited by Antheas KapenekakisThanks, saved me from some testing. Tweaked the patch.
Here's the patch: https://patchwork.kernel.org/project/dri-devel/patch/20240604160503.43359-4-christian.koenig@amd.com/ I guess it never landed?
Looks like this whole series never landed: https://patchwork.freedesktop.org/series/134458/
mentioned in commit agd5f/linux@e8fc090d
mentioned in commit agd5f/linux@5a67c316
mentioned in commit nouveau@7eb1bd60
