Resumes fails on AMD Ryzen 5 4500U with GPU Renoir (Acer Swift 3 SF314-42-R2UX)

As mentioned here: #1230 (comment 902370) I suspend isn't working on my machine either.

CPU: AMD Ryzen 5 4500U with Radeon Graphics (6) @ 2.375GHz
GPU: AMD ATI 03:00.0 Renoir
Model: Acer Swift 3 (SF314-42-R2UX)
OS: Ubuntu 21.04

I tried all patches mentioned in #1230 (closed) on Kernel 5.14 but with little success: Some times the machine will wake after first suspend but not always and at least I usually can SSH into my machine.

I was hoping for Kernel 5.15 and tried v5.15-rc2 but unfortunately the situation worsened: My machine wont wake up at all after suspend.

Trying to get on top off recent developments I found it quite hard to grasp, which patches on which kernel version I should try to solve the issue.

• Could somebody please summarize for me, what to try?

Here are some logs:

acpidump.txtacpidump.txt

Thanks for sharing the acpidump. In 5.15-rc1 https://github.com/torvalds/linux/commit/fa209644a7124b3f4cf811ced55daef49ae39ac6 was added but isn't needed for your system, which as you pointed out 5.15-rc2 doesn't help.

Can you please do the following:

• On your kernel command line for the following checks set up amd_pmc.dyndbg='+p' pm_debug_messages.
• Check 5.14.7 kernel. If it works here, can you please provide a dmesg of your full suspend and resume so we can compare 5.15-rcX.
• Check 5.15-rc3. The most notable change is https://github.com/torvalds/linux/commit/3c3c8e88c8712bfe06cd10d7ca77a94a33610cd6 but it may or may not help your system. If it doesn't help, please share more about how the failure occurs. Does it seem that it enters suspend but can't wake? Is the power LED "breathing"? Have you tried all the wake sources? Keyboard, touchpad, lid, power button?
• If all of those wake sources don't work, can you please try to plug in a USB keyboard before suspend and try to press a key to wake it? If this works, please provide a dmesg log.

One more thought - if external keyboard wakes the system you may need these patches: https://git.linaro.org/people/linus.walleij/linux-pinctrl.git/commit/?h=fixes&id=7e6f8d6f4a42ef9b693ff1b49267c546931d4619 and https://git.linaro.org/people/linus.walleij/linux-pinctrl.git/commit/?h=fixes&id=acd47b9f28e55b505aedb842131b40904e151d7c

added s0ix label

With Linux 5.15-rc3:

• System seems to go in some kind of suspend on lid close or manual suspend, as all ssh sessions freeze. The power LED however does not indicate suspend mode, as it does not »breath« but is constantly on.
• Opening of the lid or power button do not have any effect. All keyboard buttons (of the internal keyboard) however kind of wake the machine up:
  • Keyboard back light lights up.
  • Some times I can ssh into the machine but not always.
• Generally waking up the system** never works.**
• Here is the dmesg output (with amd_pmc.dyndbg='+p' pm_debug_messages): dmesg.linux15.rc3.wp.txt

With Linux 5.14.7:

• Shows the same behavior as 5.15-rc3.
• Here is the dmesg output (with amd_pmc.dyndbg='+p' pm_debug_messages): dmesg.linux14.7.txt

With Linux 5.14.0 and also with Linux 5.12.12 (the actual Ubuntu Kernel):

-System seems to go in some kind of suspend on lid close or manual suspend, as all ssh sessions freeze. The power LED however does not indicate suspend mode, as it does not »breath« but is constantly on.

• Opening of the lid or power button do not have any effect. All keyboard buttons (of the internal keyboard) however kind of wake the machine up:
  • Keyboard back light lights up.
  • Some times I can ssh into the machine but not always.
• Generally waking up the system works for the first time.
• Trying to wake the system for the second time after few key strokes it will reboot.
• Here is the dmesg output (with amd_pmc.dyndbg='+p' pm_debug_messages):dmesg.linux12.12.txt dmesg.linux14.0.txt
• Some times after suspend I can switch to console getting those messages:

Mi Sep 29 22:26:14 2021] amdgpu 0000:03:00.0: amdgpu: Msg issuing pre-check failed(0xff) and SMU may be not in the right state!
[Mi Sep 29 22:26:14 2021] amdgpu 0000:03:00.0: amdgpu: Attempt to get max GX frequency from SMC Failed !
[Mi Sep 29 22:26:14 2021] amdgpu 0000:03:00.0: amdgpu: Msg issuing pre-check failed(0xff) and SMU may be not in the right state!
[Mi Sep 29 22:26:20 2021] [drm:smu_v12_0_gfx_off_control [amdgpu]] *ERROR* disable gfxoff timeout and failed!
[Mi Sep 29 22:26:20 2021] amdgpu 0000:03:00.0: amdgpu: Failed to disable gfxoff!
[Mi Sep 29 22:26:20 2021] amdgpu 0000:03:00.0: amdgpu: Msg issuing pre-check failed(0xff) and SMU may be not in the right state!
[Mi Sep 29 22:26:20 2021] amdgpu 0000:03:00.0: amdgpu: Attempt to get max GX frequency from SMC Failed !

• Switching back to display will may show login screen but is frozen and may reboot after a while

With Linux 5.14-rc7:

• I here fans on »suspend« dmesg.linux14.rc7.txt

ps.: An external key board does not help.

So this bug report looks like it's the same as #1693 (closed) - it's got a different PEP ID than we support in the amd-pmc driver. That was fixed by a BIOS upgrade. Can you please check with Acer for an updated BIOS? I see updates for your model available here: https://www.acer.com/ac/en/US/content/support-product/8263?b=1

If the issue persists after BIOS upgrade, please provide a new acpidump.

assigned to @superm1

ps.: An external key board does not help.

As to be expected.

FWIW there are some patches that went out to the mailing list to add this ID. The BIOS upgrade should be your primary way to solve this, but if that doesn't help please give them a try.

https://patchwork.kernel.org/project/platform-driver-x86/list/?series=556591&state=%2A&archive=both

mentioned in issue #1738 (closed)

+1 I'm on same laptop with VoidLinux base image, I'd be happy to help with any information required.

For me I have different scenarios:

I have latest BIOS

>>> cat /sys/class/dmi/id/bios_version
V1.10

>>> uname -r
5.14.10_1

The resume freezes XOrg, after being in sleep (printf mem | sudo tee /sys/power/state) for more than 30s actually randomly. By freeze I mean the mouse moves, but nothing on the screen reacts, CPU seem to get hot.

I could recover by quitting the WM with keybind and re-entering the Xsession with startx. Attaching the dmesg.log.

I feel like the problem is due to rather an amdgpu fault (a short red part from the log):

[  422.925102] amdgpu 0000:03:00.0: amdgpu: [gfxhub0] retry page fault (src_id:0 ring:0 vmid:6 pasid:32769, for process Xorg pid 1503 thread Xorg:cs0 pid 1504)
[  422.925106] amdgpu 0000:03:00.0: amdgpu:   in page starting at address 0x0000000000000000 from IH client 0x1b (UTCL2)
[  422.925112] amdgpu 0000:03:00.0: amdgpu: VM_L2_PROTECTION_FAULT_STATUS:0x006C0071
[  422.925113] amdgpu 0000:03:00.0: amdgpu: 	Faulty UTCL2 client ID: CB (0x0)
[  422.925115] amdgpu 0000:03:00.0: amdgpu: 	MORE_FAULTS: 0x1
[  422.925117] amdgpu 0000:03:00.0: amdgpu: 	WALKER_ERROR: 0x0
[  422.925118] amdgpu 0000:03:00.0: amdgpu: 	PERMISSION_FAULTS: 0x7
[  422.925120] amdgpu 0000:03:00.0: amdgpu: 	MAPPING_ERROR: 0x0
[  422.925122] amdgpu 0000:03:00.0: amdgpu: 	RW: 0x1

...

[  424.420240] amdgpu 0000:03:00.0: [drm:amdgpu_ib_ring_tests [amdgpu]] *ERROR* IB test failed on gfx (-110).
[  424.420408] [drm:process_one_work] *ERROR* ib ring test failed (-110).

The system freezes.

Does this have any coincidence here?

@superm1 just for notification (I don't know if gitlab notifies without ping)

+1 I'm on same laptop with VoidLinux base image, I'd be happy to help with any information required. For me I have different scenarios: I have latest BIOS

If it's the same machine, please apply that set of patches https://patchwork.kernel.org/project/platform-driver-x86/list/?series=556591&state=%2A&archive=both to your kernel. If you continue to have a failure, please share your dmesg log from a boot with them applied.

How much time does building a kernel take (I'm just new to it, I guess I didn't specified multi-core use with -j, it took 1.5hr still building on net/ethernet directory). Is it near or should I restart with -j6?

Hmm, it does not fixes it. Here's the dmesg-patched5.14.log. Both errors are the same in here.

Plus I forgot to mention at time of boot there's two more error messages up (seen at tty login as well as in dmesg logs below 100lines approx):

[    3.254671] acp_pdm_mach acp_pdm_mach.0: snd_soc_register_card(acp) failed: -517
[    3.254728] acp_pdm_mach acp_pdm_mach.0: snd_soc_register_card(acp) failed: -517

(EDIT: bold these are at boot and not when xorg freezes)

Does it help anyhow?

EDIT: The difference I can see from the original error is only the status code in the line:

[   71.887691] amdgpu 0000:03:00.0: amdgpu: VM_L2_PROTECTION_FAULT_STATUS:0x005C0071

@superm1 Can I understand what those error status indicate, so I can also help in fixing that?

I'm not able to find meaning of any of those :(

I have a similar issue: When putting my Slim 7 Pro (14ARE05) to sleep and waking up again, my external monitor (connected via USB-C Dock / DP-alt-mode) will remain dark. This only happens when the refresh rate is set to 144hz. When setting it to 60hz, the issue does not occur. When plugging out the dock and reconnecting it, the monitor gets a signal again.

I'm running Fedora 34 (Gnome / Wayland) on 5.14.13-250.vanilla.1.fc34.x86_64 with all the latest drivers available from the official repositories installed.

Any clues on how to fix that? No issues with sleep (also deep state).

Any help is appreciated.

Plus I forgot to mention at time of boot there's two more error messages up (seen at tty login as well as in dmesg logs below 100lines approx): [ 3.254671] acp_pdm_mach acp_pdm_mach.0: snd_soc_register_card(acp) failed: -517

Those are red herrings and fixed by: https://github.com/torvalds/linux/commit/af7dc6f194a866cb3e991ef5248ffdeb3ef5c46a

[ 71.887691] amdgpu 0000:03:00.0: amdgpu: VM_L2_PROTECTION_FAULT_STATUS:0x005C0071 @superm1 Can I understand what those error status indicate, so I can also help in fixing that?

Can you try with iommu=off to see that this helps?

Any clues on how to fix that? No issues with sleep (also deep state).

@qdrop we need more details, and more importantly we need to be sure that you have the same issue as is reported. Marking a duplicate (after we confirm it) is cheap, but tracking 5 issues in a single thread is expensive. Please open your own issue and attach your dmesg and acpidump to the issue.

@superm1 No, it again changes the error code: amdgpu: VM_L2_PROTECTION_FAULT_STATUS:0x004C0071 but have no effect.

can you please give your acpidump?

Here we go acpidump.txt

Can we backtrack the stack (something like stacktrace in java?) to see where did this arrived? I haven't did debugging in C so I'm unfamiliar.

confirmed your acpi dump for sure also requires the AMD0005. One more debugging idea is to try to update to latest linux-firmware.git if you haven't already. If it still happens there to figure out what was happening differently that caused the crash I think we'll need to look at an STB dump comparing your good and failed result. Please apply https://patchwork.kernel.org/project/platform-driver-x86/patch/20211005155641.2308-1-Sanket.Goswami@amd.com/

You might have a few dependencies to use it from linux-next, so look for those if it doesn't apply cleanly.

Capture the stb from debugfs (/sys/kernel/debug/amd_pmc/stb_read) both from a successful resume and a failed one (and label them accordingly). Attach them to this bug and one of us will need to compare them for anything notably different.

Hmm, few hunks failed to register from the patch, which version I have to test the patch with, I'm on linux5.14-5.14.13_1 currently. Is that patch extracted somewhere from the kernel v5.15?

I was using 20210818, I'll try with latest 20210919 shortly. I'll include patch in a bit as well. I'll notify when I'm done with both :)

Just a question should I start with the topmost all defaults or those kernel-parameters and kernel patches you gave before applied (along with updated firmware with current patches)?