| Bug ID | 100979 |
|---|---|
| Summary | Radeon r4 on a6-6310(BEEMA) APU hard lockup on hibernate and = on second resume from suspend |
| Product | DRI |
| Version | DRI git |
| Hardware | x86-64 (AMD64) |
| OS | Linux (All) |
| Status | NEW |
| Severity | major |
| Priority | medium |
| Component | DRM/AMDgpu |
| Assignee | dri-devel@lists.freedesktop.org |
| Reporter | soprwa@gmail.com |
I'm using kernel 4.11 on gentoo with SI/CIK enabled on Lenovo = G50-45 Notebook. Machine has AMD APU A6 6130 with Radeon r4 graphics card (Beema/Mullins). T= his CPU supports olny AMD IOMMU v1. There is no discrete graphic card on it, APU only. When I try to hibernate this notebook it doesn't turning off. I have to pre= ss power button to reboot the machine. Similar situation is on "radeon&qu= ot; driver, and I had submitted bug report about this on kernel's bugzilla: https://bu= gzilla.kernel.org/show_bug.cgi?id=3D191571 But in this situation I'm unable to bisect because as I remember correctly problem always occur on amdgpu driver, so I think those two can correlate w= ith each other. (dmesg form hibernation process attached). As for suspend. I can suspend/resume machine successfully only once in a row. Second time machine suspends correctly, but on resume I have hard lockup, fans are spin= ning on full rpm's and I cannot do anything but pressing power button to reboot(= cold boot) the netbook. Moreover in dmesg after first suspend I've got error messages: [drm:amdgpu_atombios_dp_link_train [amdgpu]] *ERROR* displayport link status failed [drm:amdgpu_atombios_dp_link_train [amdgpu]] *ERROR* clock recovery failed Attachments: 1. Log after clean start. 2. Kernel's config file. 3. Log after hibernation process. 4. Log after first suspend/resume. 5. Log after second suspend/resume.
Created attachment 1=
31281 [details]
system log after clean boot
Created at=
tachment 131282 [details]
system log after performing hibernate
Created attachmen=
t 131284 [details]
system log after first suspend
Created attachment 131285 [deta=
ils]
system log after second suspend and hard lockup
The situation over here is maybe related: Hibernating/suspend-to-disk/s4 on my new AMD Carrizo box fails. Upon hibernate, the system just reboots. No hints in the logs. Suspend-to-RAM/S3 works fine. When I disable amdgpu, hibernate & resume work fine. Linux-4.11.3-gentoo-x86_64-AMD_A12-9800_RADEON_R7,_12_COMPUTE_CORES_4C+8G-w= ith-gentoo-2.3, sys-kernel/linux-firmware-20170519 Cheers, Daniel
Now on kernel 4.11.7 and linux-firmware-20170622, still no goo= d.
Created attachment 132411 [details]
dmesg after first suspend/resume process kernel 4.12
Situation still persist on kernel 4.12. After second "suspend" ma=
chine cannot
resume.
I've attached dmesg after first suspend/resume process.
If I could help/test patches please let me know.
Thanks for your effort,
Przemek.
I have just upgraded kernel to 4.15. There is a big progress. Laptop can now successfully suspend (S3) and resum= e=20 many times in a row. _Thank you very much for your hard work_. But unfortunately hibernate to disk (S4) still does not work as expected. Process is causing hard lockup (system freeze) just on the first attempt.=20 Display goes black (backlight is on), cpu is getting hot (fans are working = 100% rpms), and I can do noting more than press "power button" to hard= reset the machine. There is no more "amdgpu_atombios_dp_link_train" message in dmesg= instead there are mesages related to "swiotlb buffer is full" and "swiotlb= : coherent allocation failed" as in the bug: https://bugs.freedesktop.org/show_bug.= cgi?id=3D104082. Thanks, Przemek
After some research I think that messages "swiotlb buffer= is full" and "swiotlb: coherent allocation failed" are not related to this bug: https://lkml.org/lkml/2018/= 1/16/106
Created attachment 1=
37085 [details]
kernel log during hibernate
Kernel log taken during hibernate process. Netbook was booted up with comma=
nd
line "initcall_debug" and "no_console_suspend".
The valid mailing list post, when it comes to messages "s= wiotlb buffer is full" and "swiotlb: coherent allocation failed", is: https://lkml.org/lkml/2018/= 1/10/132. Thanks to Alex Deucher correcting me in another bug report.
Using the opportunity that I was working on another bug report= with AMD DC kernel driver, and have git cloned agd5f drm-next-4.17-wip branch, I have t= ried to debug the situation on hibernate process with this experimental kernel a= lso. This time hibernation image is created, and laptop turns off after that. But - during the hibernation process screen goes totally white, process take approximately 15-20 seconds on ssd drive before netbook gets power-off. During resume hibernation image is read (it is possible to see percentage) = and then screen goes white, and machine is in a locked-up state. I can do nothi= ng but press power button to hard-reset the netbook. Moreover I foud only once an error message in kernel log: "Mar 28 01:36:54 eclipse kernel: [ 86.665473] [drm:drm_atomic_helper_wait_for_flip_done] *ERROR* [CRTC:39:crtc-0] flip_do= ne timed out Mar 28 01:36:54 eclipse kernel: [ 86.828685] [drm:gfx_v7_0_ring_test_ring] *ERROR* amdgpu: ring 0 test failed (scratch(0xC040)=3D0xCAFEDEAD) Mar 28 01:36:54 eclipse kernel: [ 86.828698] [drm:amdgpu_device_ip_resume_phase2] *ERROR* resume of IP block <gfx_v7_= 0> failed -22 Mar 28 01:36:54 eclipse kernel: [ 86.828703] [drm:amdgpu_device_resume] *ERROR* amdgpu_device_ip_resume failed (-22). Mar 28 01:36:54 eclipse kernel: [ 86.828718] dpm_run_callback(): pci_pm_restore+0x0/0xa0 returns -22 Mar 28 01:36:54 eclipse kernel: [ 86.828742] PM: Device 0000:00:01.0 fail= ed to restore async: error -22" of course I am unable to reproduce it. I am not sure how much it is related= but this could be usefull, tough. Kernel was booted up with amdgpu.dc=3D1. Any help is appreciated. Thanks, Przemek.
| What | Removed | Added |
|---|---|---|
| Status | NEW | RESOLVED |
| Resolution | --- | FIXED |
I have tested hibernation on amd-staging-drm-next (git status = - commit=20 fa16d1eb6a78b265480bd4c2b8739c1ea261cdd8 ) and it's working as it should (w= ith minor glitch). Moreover I can suspend and resume the machine many times in a row without freeze/lockup.=20 I nave no idea which commit made things work again, because haven't checked this future lately. The only problem is that the second monitor connected to hdmi output is tur= ning of after suspend/hibernate, and eDP screen brightness level is maxed out af= ter resume from hibernate but this is not the case of this report. Given the above I'm closing this bug report as RESOLVED/FIXED. Thank you very much, Przemek.
Just for the rectification, after resume from hibernate both screens lights up as they should (eDP and HDMI).