Re: [PATCH v2 00/14] drm/msm/a8xx: GPU state capture support

[Rob Clark]

On Mon, Oct 26, 2026 at 4:06 PM Alexander Koskovich <akoskovich@xxxxx> wrote:
>
> On Wednesday, October 21st, 2026 at 4:54 PM, Rob Clark <robin.clark@xxxxxxxxxxxxxxxx> wrote:
> >
> > I'd appreciate tested-by's on Hawi, Maili, Milos, and Pakala, since I
> > don't have these devices at-hand to test myself.  In particular, do
> > things look fine after the switch to generated nonctx regs.  To test
> > devcoredump, you can use the msm_recovery test in igt.. or just:
> >
> >   cat /sys/kernel/debug/dri/*/gpu > /dev/null
> >
>
> Tested this series on Milos, series as is works fine for initial boot, navigating
> around GNOME & apps etc, but when I run 'cat /sys/kernel/debug/dri/*/gpu', run
> into an oops.
>
> That is resolved after removing 0x20400-0x20487 from milos_gmucx_registers,
> and removing 0x0f000-0x0f002 from milos_cp_cp_pipe_none_registers. Not sure
> if access to these is restricted by TZ or something isn't configured right?
> These ranges don't seem to be described downstream in adreno_gen8_3_0_snapshot.h.

Thanks, yeah, there seems to be some things blocked by
tz/hyp/something, which are not blocked on glymur/el2.  I'll have a v3
soon, but still chasing down some of these details.

>
> After that 'cat /sys/kernel/debug/dri/*/gpu' works and I get a dump, did observe these errors in dmesg after though:
>
> [   42.891106] platform 3d6c000.gmu: GMU fence error: 0x0
> [   42.891142] platform 3d6c000.gmu: GMU fence error: 0x0
> [   42.891157] platform 3d6c000.gmu: GMU fence error: 0x0
> [   42.891171] platform 3d6c000.gmu: GMU fence error: 0x0
> [   42.891184] platform 3d6c000.gmu: GMU fence error: 0x0
> [   42.891198] platform 3d6c000.gmu: GMU fence error: 0x0
> [   42.891211] platform 3d6c000.gmu: GMU fence error: 0x0
> [   42.891224] platform 3d6c000.gmu: GMU fence error: 0x0
> [   42.891237] platform 3d6c000.gmu: GMU fence error: 0x0
> [   42.891250] platform 3d6c000.gmu: GMU fence error: 0x0
> [   43.640122] ------------[ cut here ]------------
> [   43.640137] msm_dpu ae01000.display-controller: [drm] crashdumper timed out: status=5c00bd00 (-110)
> [   43.640222] WARNING: drivers/gpu/drm/msm/adreno/a8xx_gpu_state.c:115 at CD_RUN+0x20c/0x594 [msm], CPU#3: cat/1929
> [   43.640419] Modules linked in: rfcomm q6apm_dai q6prm_clocks q6apm_lpass_dais snd_q6dsp_common q6prm algif_hash algif_skcipher bnep snd_q6apm panel_raydium_rm69220 fastrpc qcom_pd_mapper ath11k_ahb ath11k mac80211 hci_uart btqca bluetooth qcom_iris cfg80211 ecdh_generic msm ledtrig_pattern videobuf2_dma_contig ecc v4l2_mem2mem drm_dp_aux_bus videobuf2_memops qcom_q6v5_pas ocmem qcom_pil_info videobuf2_v4l2 pwrseq_qcom_wcn gpu_sched ipa pwrseq_core qmi_tmd focaltech_ft3683g qcom_q6v5 drm_gpuvm videobuf2_common drm_exec rfkill drm_display_helper ubwc_config cec qcom_common mdt_loader libarc4 nft_reject_inet nf_reject_ipv6 nf_reject_ipv4 nft_reject nft_ct nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 nf_tables uhid fuse usb_f_ncm u_ether spi_geni_qcom evdev libcomposite
> [   43.640750] CPU: 3 UID: 0 PID: 1929 Comm: cat Tainted: G S                   7.3.0-next-20261026 #3 PREEMPT(full)
> [   43.640763] Tainted: [S]=CPU_OUT_OF_SPEC
> [   43.640770] Hardware name: Nothing Phone (3a) (DT)
> [   43.640778] pstate: 61400005 (nZCv daif +PAN -UAO -TCO +DIT -SSBS BTYPE=--)
> [   43.640787] pc : CD_RUN+0x20c/0x594 [msm]
> [   43.640972] lr : CD_RUN+0x208/0x594 [msm]
> [   43.641155] sp : ffff8000891c3990
> [   43.641161] x29: ffff8000891c3990 x28: ffffc3ee34aef850 x27: ffff000809253000
> [   43.641180] x26: 0000000000000001 x25: 0000000000000001 x24: 0000000000000020
> [   43.641198] x23: 0000000000000000 x22: 0000000000000000 x21: 0000000000002114
> [   43.641216] x20: 0000000a28796b5c x19: ffff000813d7f000 x18: ffff8000891ad078
> [   43.641233] x17: 3a74756f2064656d x16: 6974207265706d75 x15: 646873617263205d
> [   43.641250] x14: 3fffffffffffefff x13: 0000000000000003 x12: ffffc3ee5c6649c0
> [   43.641267] x11: 3fffffffffffefff x10: 0000000000000003 x9 : 9e14538c77bb2600
> [   43.641284] x8 : 9e14538c77bb2600 x7 : 2e30303031306561 x6 : 207570645f6d736d
> [   43.641301] x5 : ffffc3ee5cb061bf x4 : 0000000000000345 x3 : 0000000000000000
> [   43.641318] x2 : 0000000000000000 x1 : ffff8000891c3790 x0 : 00000000ffffff92
> [   43.641336] Call trace:
> [   43.641343]  CD_RUN+0x20c/0x594 [msm] (P)
> [   43.641530]  a8xx_gpu_state_get+0xdc4/0x2698 [msm]
> [   43.641714]  msm_gpu_open+0xc0/0x134 [msm]
> [   43.641898]  full_proxy_open_regular+0xe0/0x1e8
> [   43.641913]  do_dentry_open+0x154/0x4d8
> [   43.641927]  vfs_open_consume+0x40/0x114
> [   43.641939]  path_openat+0x440/0x6a0
> [   43.641950]  do_file_open+0xdc/0x198
> [   43.641960]  do_sys_openat2+0x88/0x118
> [   43.641972]  __arm64_sys_openat+0x78/0xac
> [   43.641985]  invoke_syscall+0x6c/0x12c
> [   43.641997]  el0_svc_common+0x98/0xf0
> [   43.642007]  do_el0_svc+0x28/0x38
> [   43.642016]  el0_svc+0x4c/0x1ac
> [   43.642031]  el0t_64_sync_handler+0x6c/0xe4
> [   43.642043]  el0t_64_sync+0x1a0/0x1a4
> [   43.642056] ---[ end trace 0000000000000000 ]---

Hmm, not sure about this, other than maybe whatever is causing GMU
fence errors is also causing the timeout?

If you run something that keeps the gpu active (or disable ifpc), does
that help?

BR,
-R