MST Dock Connect results in crash v6.1-rc1

changed the description

added DC label

This is new from 6.0->6.1-rc1? Could you bisect it?

I have exactly this problem with Lenovo 40AY. Last time I tried an Aenzr DP 1.4 hub (2-lane, 4k 60Hz) and it worked fine (albeit on kernel 6.0.0). The new dock is also DP 1.4, also supports MST, but with added feature of DSC. At first I thought DSC was the culprit, but after setting amdgpu.dcdebugmask=0x04 the problem persists.

My journalctl -b -1 looks exactly like the one posted by the OP.

Computer: Asus UM5302TA, Ryzen 7 6800U

Monitor: Huawei MateView X 3840*2560 24bpp 60Hz over mini DP 1.2

Kernel: Linux 6.1.0-rc1, with .config from Ubuntu Mainline, compiled with a few patches related to ALSA

OS: Ubuntu 22.04

Mesa: kisak PPA

On kernel 6.0.0 (non-RC), it no longer crashes. It still doesn't work, though. This time, dmesg shows something different.

[  280.604096] [drm] DP Alt mode state on HPD: 1
[  280.672924] [drm] DM_MST: starting TM on aconnector: 00000000121db8a1 [id: 94]
[  280.935355] [drm] Downstream port present 1, type 0
[  281.075929] ------------[ cut here ]------------
[  281.075932] WARNING: CPU: 0 PID: 1457 at drivers/gpu/drm/display/drm_dp_mst_topology.c:4392 drm_dp_atomic_find_vcpi_slots.cold+0x0/0x25 [drm_display_helper]
[  281.075950] Modules linked in: uhid rfcomm ccm cmac algif_hash algif_skcipher af_alg bnep btusb btrtl btbcm btintel btmtk bluetooth r8153_ecm cdc_ether usbnet r8152 joydev ecdh_generic snd_ctl_led mii ecc snd_usb_audio intel_rapl_msr intel_rapl_common usbhid snd_usbmidi_lib amdgpu edac_mce_amd snd_soc_acp6x_mach snd_acp6x_pdm_dma snd_soc_dmic mt7921e mt7921_common snd_sof_amd_renoir snd_sof_amd_acp kvm_amd mt76_connac_lib snd_hda_codec_realtek mt76 snd_hda_scodec_cs35l41_spi iommu_v2 snd_sof_pci snd_hda_codec_generic snd_hda_codec_hdmi gpu_sched uvcvideo snd_sof kvm snd_sof_utils videobuf2_vmalloc videobuf2_memops drm_buddy mac80211 snd_hda_intel crct10dif_pclmul videobuf2_v4l2 snd_soc_core drm_ttm_helper ghash_clmulni_intel binfmt_misc ttm snd_intel_dspcfg videobuf2_common snd_intel_sdw_acpi drm_display_helper aesni_intel snd_seq_midi videodev snd_compress snd_seq_midi_event snd_hda_codec crypto_simd cec ac97_bus asus_nb_wmi cryptd snd_pcm_dmaengine asus_wmi rc_core
[  281.075991]  snd_hda_scodec_cs35l41_i2c nls_iso8859_1 input_leds rapl snd_rawmidi snd_hda_core snd_acp_pci mc platform_profile hid_multitouch serio_raw snd_hda_scodec_cs35l41 drm_kms_helper ledtrig_audio snd_hwdep wmi_bmof snd_pci_acp6x sparse_keymap snd_hda_cs_dsp_ctls cfg80211 i2c_algo_bit cs_dsp snd_seq snd_pci_acp5x snd_soc_cs35l41_lib fb_sys_fops snd_pcm ucsi_acpi snd_rn_pci_acp3x syscopyarea snd_acp_config sysfillrect typec_ucsi snd_seq_device snd_soc_acpi sysimgblt libarc4 ccp k10temp snd_timer snd_pci_acp3x typec snd mac_hid soundcore serial_multi_instantiate 8250_dw amd_pmc acpi_tad sch_fq_codel ipmi_devintf ipmi_msghandler msr parport_pc ppdev drm lp parport ramoops reed_solomon pstore_blk pstore_zone efi_pstore ip_tables x_tables autofs4 nvme hid_generic crc32_pclmul nvme_core i2c_piix4 xhci_pci xhci_pci_renesas wmi i2c_hid_acpi i2c_hid video hid
[  281.076036] CPU: 0 PID: 1457 Comm: gnome-shell Tainted: G        W          6.0.0-custom #1
[  281.076039] Hardware name: ASUSTeK COMPUTER INC. Zenbook UM5302TA_UM5302TA/UM5302TA, BIOS UM5302TA.310 08/10/2022
[  281.076040] RIP: 0010:drm_dp_atomic_find_vcpi_slots.cold+0x0/0x25 [drm_display_helper]
[  281.076048] Code: 45 0f b6 7c 24 0c 48 8b 55 c0 e9 51 4a ff ff 0f b6 f2 48 c7 c7 f0 e5 ab c0 88 55 c0 e8 ef 02 01 f4 0f b6 55 c0 e9 dc 4f ff ff <0f> 0b 49 8b 42 58 4c 89 fa 48 c7 c6 38 bb ab c0 48 8b 78 08 e8 0c
[  281.076050] RSP: 0018:ffffb110040eb6f0 EFLAGS: 00010246
[  281.076052] RAX: ffffa06d83f794d8 RBX: ffffa06d83f794c0 RCX: 0000000000000000
[  281.076053] RDX: ffffa06d83f79c40 RSI: ffffa06d86980530 RDI: ffffa06d85f07700
[  281.076054] RBP: ffffb110040eb730 R08: 000000000000001d R09: 0000000000000723
[  281.076055] R10: ffffa06d86980530 R11: 0000000000000000 R12: ffffa06d901a7050
[  281.076056] R13: 000000000000001d R14: ffffa06d901a7040 R15: ffffa06d81016800
[  281.076057] FS:  00007fe07e6795c0(0000) GS:ffffa0708e800000(0000) knlGS:0000000000000000
[  281.076058] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  281.076060] CR2: 00000f68f43ba028 CR3: 0000000148f6e000 CR4: 0000000000750ef0
[  281.076061] PKRU: 55555554
[  281.076062] Call Trace:
[  281.076063]  <TASK>
[  281.076065]  ? dc_link_should_enable_fec+0x87/0xa0 [amdgpu]
[  281.076274]  compute_mst_dsc_configs_for_link+0xb47/0x15f0 [amdgpu]
[  281.076465]  compute_mst_dsc_configs_for_state+0x1b8/0x2d0 [amdgpu]
[  281.076625]  ? drm_modeset_lock+0xd9/0xf0 [drm]
[  281.076643]  amdgpu_dm_atomic_check+0xfab/0x12f0 [amdgpu]
[  281.076772]  ? dm_crtc_duplicate_state+0x39/0x1c0 [amdgpu]
[  281.076897]  ? __kmalloc_track_caller+0x189/0x380
[  281.076903]  ? amdgpu_dm_connector_atomic_duplicate_state+0x2c/0x100 [amdgpu]
[  281.077026]  drm_atomic_check_only+0x657/0xb40 [drm]
[  281.077042]  ? __drm_mode_object_find+0x58/0xe0 [drm]
[  281.077058]  drm_atomic_commit+0x57/0xd0 [drm]
[  281.077071]  ? drm_plane_create_color_properties.cold+0x48/0x48 [drm]
[  281.077090]  drm_mode_atomic_ioctl+0xa0e/0xc00 [drm]
[  281.077107]  ? drm_atomic_set_property+0xb30/0xb30 [drm]
[  281.077120]  drm_ioctl_kernel+0xbb/0x150 [drm]
[  281.077134]  drm_ioctl+0x268/0x4b0 [drm]
[  281.077146]  ? drm_atomic_set_property+0xb30/0xb30 [drm]
[  281.077159]  ? amdgpu_drm_ioctl+0x6f/0x90 [amdgpu]
[  281.077245]  amdgpu_drm_ioctl+0x4e/0x90 [amdgpu]
[  281.077329]  __x64_sys_ioctl+0x95/0xd0
[  281.077333]  do_syscall_64+0x5c/0x90
[  281.077337]  ? syscall_exit_to_user_mode+0x26/0x50
[  281.077339]  ? do_syscall_64+0x69/0x90
[  281.077340]  ? do_syscall_64+0x69/0x90
[  281.077341]  ? do_syscall_64+0x69/0x90
[  281.077341]  ? sysvec_apic_timer_interrupt+0x4e/0x90
[  281.077343]  entry_SYSCALL_64_after_hwframe+0x63/0xcd
[  281.077347] RIP: 0033:0x7fe08191aaff
[  281.077348] Code: 00 48 89 44 24 18 31 c0 48 8d 44 24 60 c7 04 24 10 00 00 00 48 89 44 24 08 48 8d 44 24 20 48 89 44 24 10 b8 10 00 00 00 0f 05 <41> 89 c0 3d 00 f0 ff ff 77 1f 48 8b 44 24 18 64 48 2b 04 25 28 00
[  281.077349] RSP: 002b:00007ffd93a14380 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
[  281.077350] RAX: ffffffffffffffda RBX: 00007ffd93a14430 RCX: 00007fe08191aaff
[  281.077351] RDX: 00007ffd93a14430 RSI: 00000000c03864bc RDI: 0000000000000009
[  281.077352] RBP: 00000000c03864bc R08: 0000000000000015 R09: 0000000000000015
[  281.077352] R10: 0000556fecdd7df0 R11: 0000000000000246 R12: 0000556fecb95880
[  281.077353] R13: 0000000000000009 R14: 0000556fecf40400 R15: 0000556fec2add40
[  281.077354]  </TASK>
[  281.077355] ---[ end trace 0000000000000000 ]---
[  281.077358] amdgpu 0000:03:00.0: [drm] *ERROR* cannot allocate and release VCPI on [MST PORT:0000000017a05dfd] in the same state

At first I thought it was a DSC issue, after setting amdgpu.dcdebugmask=0x04 with no avail, I started questioning if triple display MST was the fault -- both 40AS and 40AY came with triple stream MST, and my working Aenzer is dual stream MST only. Later I bought a CalDigit SOHO, which is a dual stream MST hub, and to my surprise it still doesn't work.

The dmesg error is the same as my last post, with the difference being this time the external monitor would work, just for once, with very wrong resolution and over-scanning (start and end position). I hotplugged it again and it never worked.

Below is the first dmesg error after I first plugged it in (after gdm login, before entering desktop). Kernel 6.0.0-release.

[    7.064900] ------------[ cut here ]------------
[    7.064904] WARNING: CPU: 6 PID: 951 at drivers/gpu/drm/display/drm_dp_mst_topology.c:4392 drm_dp_atomic_find_vcpi_slots.cold+0x0/0x25 [drm_display_helper]
[    7.064924] Modules linked in: cmac algif_hash algif_skcipher af_alg cdc_ether snd_usb_audio usbnet r8152 snd_ctl_led mii usbhid snd_usbmidi_lib snd_soc_dmic snd_soc_acp6x_mach snd_acp6x_pdm_dma snd_sof_amd_renoir amdgpu snd_sof_amd_acp snd_hda_codec_realtek snd_sof_pci iommu_v2 gpu_sched drm_buddy snd_hda_codec_generic snd_sof drm_ttm_helper ttm snd_sof_utils snd_hda_codec_hdmi drm_display_helper snd_soc_core cec snd_hda_intel snd_intel_dspcfg snd_compress rc_core snd_intel_sdw_acpi ac97_bus bnep snd_hda_codec snd_pcm_dmaengine snd_acp_pci drm_kms_helper snd_pci_acp6x snd_hda_core snd_pci_acp5x i2c_algo_bit snd_rn_pci_acp3x snd_hwdep fb_sys_fops uas snd_acp_config syscopyarea sysfillrect snd_soc_acpi usb_storage snd_pcm sysimgblt snd_pci_acp3x joydev intel_rapl_msr intel_rapl_common snd_hda_scodec_cs35l41_spi edac_mce_amd kvm_amd snd_hda_scodec_cs35l41_i2c snd_hda_scodec_cs35l41 snd_hda_cs_dsp_ctls cs_dsp snd_soc_cs35l41_lib ccp snd_seq_midi snd_seq_midi_event binfmt_misc snd_rawmidi
[    7.064977]  mt7921e kvm mt7921_common mt76_connac_lib uvcvideo snd_seq mt76 videobuf2_vmalloc videobuf2_memops crct10dif_pclmul videobuf2_v4l2 snd_seq_device nls_iso8859_1 ghash_clmulni_intel asus_nb_wmi snd_timer videobuf2_common mac80211 aesni_intel asus_wmi platform_profile input_leds snd btusb crypto_simd videodev cryptd btrtl ledtrig_audio rapl hid_multitouch wmi_bmof sparse_keymap mc serio_raw soundcore btbcm ucsi_acpi btintel cfg80211 typec_ucsi k10temp btmtk libarc4 typec mac_hid serial_multi_instantiate 8250_dw amd_pmc acpi_tad bluetooth ecdh_generic ecc sch_fq_codel ipmi_devintf ipmi_msghandler msr parport_pc ppdev lp parport drm ramoops pstore_blk reed_solomon pstore_zone efi_pstore ip_tables x_tables autofs4 nvme hid_generic crc32_pclmul xhci_pci i2c_piix4 nvme_core xhci_pci_renesas wmi i2c_hid_acpi video i2c_hid hid
[    7.065035] CPU: 6 PID: 951 Comm: gnome-shell Not tainted 6.0.0-custom #1
[    7.065038] Hardware name: ASUSTeK COMPUTER INC. Zenbook UM5302TA_UM5302TA/UM5302TA, BIOS UM5302TA.310 08/10/2022
[    7.065039] RIP: 0010:drm_dp_atomic_find_vcpi_slots.cold+0x0/0x25 [drm_display_helper]
[    7.065049] Code: 45 0f b6 7c 24 0c 48 8b 55 c0 e9 51 4a ff ff 0f b6 f2 48 c7 c7 f0 25 04 c1 88 55 c0 e8 ef c2 68 c3 0f b6 55 c0 e9 dc 4f ff ff <0f> 0b 49 8b 42 58 4c 89 fa 48 c7 c6 38 fb 03 c1 48 8b 78 08 e8 0c
[    7.065052] RSP: 0018:ffffa9ef01d13700 EFLAGS: 00010246
[    7.065054] RAX: ffff945dbd7af598 RBX: ffff945dbd7af580 RCX: 0000000000000000
[    7.065056] RDX: ffff945dbd7af500 RSI: ffff945d83adc530 RDI: ffff945d85643500
[    7.065057] RBP: ffffa9ef01d13740 R08: 000000000000001d R09: 0000000000000723
[    7.065058] R10: ffff945d83adc530 R11: 0000000000000000 R12: ffff945dbd7afd50
[    7.065060] R13: 000000000000001d R14: ffff945dbd7afd40 R15: ffff945d90477000
[    7.065061] FS:  00007f3a7112c5c0(0000) GS:ffff94608e780000(0000) knlGS:0000000000000000
[    7.065062] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[    7.065064] CR2: 000055f7de6a999c CR3: 0000000124b4e000 CR4: 0000000000750ee0
[    7.065065] PKRU: 55555554
[    7.065066] Call Trace:
[    7.065068]  <TASK>
[    7.065070]  ? dc_link_should_enable_fec+0x87/0xa0 [amdgpu]
[    7.065360]  compute_mst_dsc_configs_for_link+0xb47/0x15f0 [amdgpu]
[    7.065578]  pre_validate_dsc+0x4c9/0x700 [amdgpu]
[    7.065780]  amdgpu_dm_atomic_check+0x1da/0x12f0 [amdgpu]
[    7.065979]  ? dm_crtc_duplicate_state+0x39/0x1c0 [amdgpu]
[    7.066178]  ? allocate_slab+0x239/0x490
[    7.066184]  drm_atomic_check_only+0x657/0xb40 [drm]
[    7.066212]  ? __drm_mode_object_find+0x58/0xe0 [drm]
[    7.066238]  drm_atomic_commit+0x57/0xd0 [drm]
[    7.066259]  ? drm_plane_create_color_properties.cold+0x48/0x48 [drm]
[    7.066287]  drm_mode_atomic_ioctl+0xa0e/0xc00 [drm]
[    7.066313]  ? drm_atomic_set_property+0xb30/0xb30 [drm]
[    7.066336]  drm_ioctl_kernel+0xbb/0x150 [drm]
[    7.066362]  drm_ioctl+0x268/0x4b0 [drm]
[    7.066385]  ? drm_atomic_set_property+0xb30/0xb30 [drm]
[    7.066409]  amdgpu_drm_ioctl+0x4e/0x90 [amdgpu]
[    7.066573]  __x64_sys_ioctl+0x95/0xd0
[    7.066579]  do_syscall_64+0x5c/0x90
[    7.066585]  ? exc_page_fault+0x87/0x180
[    7.066588]  entry_SYSCALL_64_after_hwframe+0x63/0xcd
[    7.066593] RIP: 0033:0x7f3a7451aaff
[    7.066595] Code: 00 48 89 44 24 18 31 c0 48 8d 44 24 60 c7 04 24 10 00 00 00 48 89 44 24 08 48 8d 44 24 20 48 89 44 24 10 b8 10 00 00 00 0f 05 <41> 89 c0 3d 00 f0 ff ff 77 1f 48 8b 44 24 18 64 48 2b 04 25 28 00
[    7.066597] RSP: 002b:00007ffc87ab4940 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
[    7.066599] RAX: ffffffffffffffda RBX: 00007ffc87ab49f0 RCX: 00007f3a7451aaff
[    7.066600] RDX: 00007ffc87ab49f0 RSI: 00000000c03864bc RDI: 0000000000000009
[    7.066601] RBP: 00000000c03864bc R08: 0000000000000015 R09: 0000000000000015
[    7.066602] R10: 00007f3a7461a190 R11: 0000000000000246 R12: 000055f7e0873310
[    7.066603] R13: 0000000000000009 R14: 000055f7e0fa2300 R15: 000055f7e086a560
[    7.066606]  </TASK>
[    7.066607] ---[ end trace 0000000000000000 ]---
[    7.066610] amdgpu 0000:03:00.0: [drm] *ERROR* cannot allocate and release VCPI on [MST PORT:000000001dcd3887] in the same state
[    7.066616] [drm] pre_compute_mst_dsc_configs_for_state() failed

So, here are some more test results on the SOHO:

MacBook Pro 14 with Apple M1 and macOS Monterey: works perfectly at 3840 x 2160 x 24bpp x 60Hz.
Huawei MateBook E 2022 with i7-1160G7 with Windows 11 Pro: works perfectly at 3840 x 2160 x 24bpp x 60Hz.
Huawei Mate 40 Pro with Harmony 2, kernel 4.14: kinda works, but with visual artifacts in the lower 1/10.
One Netbook One X Pro with i7-1160G7 and Ubuntu 20.04 LTS HWE: recognizes, but not displaying anything.

It seems to be a Linux DRM MST issue, not essentially an amdgpu issue. Strangely enough, both my Aenzer and SOHO are based on VMM5210, but the former apparently works fine. Probably the Aenzer has a different firmware version than the SOHO, and I think we might need external help from Synaptics, who made this chip and reference firmware.

Note: as I was typing, my monitor lost connection once again, this time plugged directly to my computer over 4-lane DP over type C (USB 2.0-only). This has happened, to date, only once. Below is the dmesg for this event:

[ 1122.524185] amdgpu 0000:03:00.0: amdgpu: 00000000ba7f8667 pin failed
[ 1122.524196] [drm:dm_plane_helper_prepare_fb [amdgpu]] *ERROR* Failed to pin framebuffer with error -12

This is probably an isolated one-time oddity, but this might reveal something useful to the AMD guys.

I'm glad to report that my problem seems to be fixed by removing amdgpu.dcdebugmask=0x04. With both DSC and MST enabled, my system now works happily with SOHO. The SOHO has been upgraded to its latest 2021/12/31 firmware.

To recap my experience:

Kernel	DSC	MST	Experience	Device
6.0	OFF	Double	Not working, recover	CalDigit SOHO
6.0	OFF	Triple	Not working, recover	Lenovo 40AY
6.0	ON	Double	Working	CalDigit SOHO
6.0	ON	Triple	Not tested	N/A
6.1	OFF	Double	Not working, crash	CalDigit SOHO
6.1	OFF	Triple	Not working, crash	Lenovo 40AY
6.1	ON	Double	Working	Aenzer 12-in-1
6.1	ON	Triple	Not working, crash	Lenovo 40AY

This seems to be a triple MST issue certainly on kernel 6.1, and maybe on kernel 6.0 as well. Also, the DRM code on kernel 6.1 seems to favor kicking amdgpu out of the kernel when something goes wrong, while amdgpu was able to regain control of GPU on kernel 6.0.

Lessons learned: do not mess with debug features, without messing with DSC-off my SOHO would have been working since yesterday.

On a side note: amdgpu in kernel 6.1 is no longer able to output sound over type C, with or without an MST hub. This, compounded with the inability to sleep in S0 mode has kept me from keep using it.

On a second note, despite VMM5210 being capable of DSC and MST, neither the Aenzer nor the CalDigit are able to output dual 4K 60Hz. Theoretically, (3840 x 2560 x 24bpp x 60Hz + 3840 x 2160 x 24bpp x 60Hz)/3=8.7Gbps, which is way within what 2 lanes of DP1.4 can carry. Test subjects: MateView X (3840 x 2560 x 59.98Hz) and C-Force CF011 (3840 x 2160 x 60Hz).

This means the CalDigit SOHO is working fine with 2x 4K? Thinking of buying that one too because the Lenovo Docks are just a pain..

No. As I've written, "neither the Aenzer not the CalDigit are able to output dual 4K 60Hz". A single output works on a dual MST hub, but not on a triple MST hub. As a matter of fact, my dual 4K (4K and 4k plus) setup does NOT work simultaneously on the SOHO. Both are recognized, but dmesg shows the following error:

[ 7388.869738] [drm:dm_plane_helper_prepare_fb [amdgpu]] *ERROR* Failed to pin framebuffer with error -12

And the newly added display gets recognized in Gnome Settings, but doe NOT output anything.

Ah, I understand, thanks for clarifying.

Something weird I noticed: In earlier Versions (Not sure if just the Kernel matters here). I was able to use 2x 4K, 1x FHD and 1x FHD Internal Laptop monitor simultaneously. I'm not able to get this working any more ATM.

mentioned in issue #2216 (closed)

Seeing a similar issue in 6.0.3 and 6.1-rc1 and not 6.0.2.

Hangs after gdm launches with [ 23.624482] amdgpu 0000:6e:00.0: [drm] *ERROR* [CRTC:67:crtc-0] flip_done timed out and only if an MST hub is connected.

Found an adapter to hook up the same exact monitors without MST, confirmed seeing a regression between 6.0.2 and 6.0.3 that makes this MST setup unusable, it's not specific to having the set of monitors attached. The MST hub is a Club 3D 3 port hub.

The obvious MST related commit drm/dp_mst: fix drm_dp_dpcd_read return value checks is not the cause.

Could you bisect from 6.0.2 to 6.0.3?

probably not until the weekend

@superm1 I have now bisected the crash issue

➜  linux-mainline git:(01ad1d9c2888) ✗ git bisect good
4d07b0bc403403438d9cf88450506240c5faf92f is the first bad commit
commit 4d07b0bc403403438d9cf88450506240c5faf92f
Author: Lyude Paul <lyude@redhat.com>
Date:   Wed Aug 17 15:38:46 2022 -0400

    drm/display/dp_mst: Move all payload info into the atomic state

--

Looks like there are two additional issues:

Starting with 6.0-rc1 2x4k + 1xFHD monitors over MST doesn't work anymore. Just the 1xFHD monitor is running
Somewhere starting with 6.1-rc1 the MST monitor shows red lines, this was introduces before the MST crash issue

Will bisect them as soon as I find time and open separate issues..

Thanks for bisecting!

@lyudess are you aware of this already? Not sure if it's the same thing you and KH were looking at already in #2171 (closed).

Looks like it! And yes - the patch there hopefully should fix things, although I messed up the first few times writing it

Great! Let's fold them together then.

marked this issue as a duplicate of #2171 (closed)

closed

marked this issue as related to #2171 (closed)

mentioned in issue #2181

mentioned in issue #2171 (closed)

mentioned in issue #2279 (closed)

MST Dock Connect results in crash v6.1-rc1

Designs

Child items ...

Activity

Admin message

Admin message

MST Dock Connect results in crash v6.1-rc1

Activity