Re: 答复: Regression with kernel 4.18 - AMD RX 550 fails IB ring test on power-up

Luís Mendes <luis.p.mendes@xxxxxxxxx> · Sun, 1 Sep 2019 13:29:15 +0100

Good News!! I've re-tested this with Linux Kernel 5.3.0-rc6 with Ubuntu 19.4 and Mesa 19.0.2 and Mesa 19.0.8 with the same Polaris RX460 and the same TYAN S7002 and TYAN S7025 and now it is working properly, being able to enter the desktop and running glmark2 as well as OpenCL.

On Sat, Oct 20, 2018 at 6:58 PM Luís Mendes <luis.p.mendes@xxxxxxxxx> wrote:
The problems remains with Linux 4.18 and Linux 4.19 kernels. I am unable to use AMD RX 460 and AMD RX 550 on my x64 Linux platforms.

I've installed Windows 10 in the same machine along with win10-64bit-radeon-software-adrenalin-edition-18.10.1-oct18.exe and under Windows the same RX 460 card *works fine* and I am able to run OpenCL applications.

The driver is hanging since kernel 4.15, I am getting:
[   33.504100] [drm:drm_atomic_helper_wait_for_flip_done [drm_kms_helper]] *ERROR* [CRTC:42:crtc-0] flip_done timed out
[   43.744094] [drm:drm_atomic_helper_wait_for_dependencies [drm_kms_helper]] *ERROR* [CRTC:42:crtc-0] flip_done timed out                                                      
[   53.984089] [drm:drm_atomic_helper_wait_for_dependencies [drm_kms_helper]] *ERROR* [CONNECTOR:54:HDMI-A-1] flip_done timed out                                               
[   64.224036] [drm:drm_atomic_helper_wait_for_dependencies [drm_kms_helper]] *ERROR* [PLANE:40:plane-4] flip_done timed out                                                    
[   64.224141] [drm:amdgpu_dm_atomic_commit_tail [amdgpu]] *ERROR* amdgpu_dm_commit_planes: acrtc 0, already busy    

And after commit: drm/amdgpu: defer test IBs on the
          rings at boot (V3)

                2c773de2ecb8c327f2448bd1eecad224e9227087
I get with kernels 4.18 and 4.19 as well as Ubuntu 18.10 stock kernel (can't even install Ubuntu 18.10 because it hangs with amdgpu):
[drm:gfx_v8_0_ring_test_ib [amdgpu]] *ERROR* amdgpu: IB test timed out.
[drm_amdgpu_ib_ring_tests [amdgpu]] *ERROR* amdgpu: failed testing IB on GFX ring (-110).
[drm:amdgpu_device_ip_late_init_func_handler [amdgpu]] *ERROR* ib ring test failed (-110).
[drm:amdgpu_job_timedout [amdgpu]] *ERROR* ring sdma1 timeout, last signalled seq=0, last emitted seq=1
and the kernel blocks indefinitely:
task plymouthd:449 blocked for more than 120 seconds.

Is there any hope on getting this fixed?

On Thu, Jul 12, 2018 at 2:56 PM Luís Mendes <luis.p.mendes@xxxxxxxxx> wrote:
Hi Jim,

Replies in between.

Regards,
Luís

On Thu, Jul 12, 2018 at 3:16 AM, jimqu <jimqu@xxxxxxx> wrote:

    On 2018年07月12日 05:27, Luís Mendes
      wrote:

        Hi Jim,

        I followed your suggestion and was able to bisect the
          kernel patches.
        The offending patch is: drm/amdgpu: defer test IBs on the
          rings at boot (V3)

        commit:

                2c773de2ecb8c327f2448bd1eecad224e9227087

        After reverting this patch the IB test succeeded with
          kernel v4.18-rc4 on both systems and the amdgpu driver was
          correctly loaded both on SAPPHIRE RX550 4GB and on SAPPHIRE
          RX460 2GB.

    Alex, Christian, What do you think about the patch?

        The GPU hang remains, however.

         I will try to configure a remote IPMI connection to see
          what is happening with the kernel boot or setup a serial
          console for the Kernel.

    You can set up remote connection by ssh, and also you can add amdgpu
    to blacklist first, and manually modprobe amdgpu.
R: I was able to setup a remote serial console with console=ttyS0,11520n8 kernel parameter. 
Boot log follows attached as file kernel_bisected_v4.18-rc4_log.txt. 
First noticeable issue seems to be:
[    6.131989] amdgpu: [powerplay] 
[    6.131989]  last message was failed ret is 65535
...
and later hangs with:
[   33.504100] [drm:drm_atomic_helper_wait_for_flip_done [drm_kms_helper]] *ERROR* [CRTC:42:crtc-0] flip_done timed out
[   43.744094] [drm:drm_atomic_helper_wait_for_dependencies [drm_kms_helper]] *ERROR* [CRTC:42:crtc-0] flip_done timed out                                                      
[   53.984089] [drm:drm_atomic_helper_wait_for_dependencies [drm_kms_helper]] *ERROR* [CONNECTOR:54:HDMI-A-1] flip_done timed out                                               
[   64.224036] [drm:drm_atomic_helper_wait_for_dependencies [drm_kms_helper]] *ERROR* [PLANE:40:plane-4] flip_done timed out                                                    
[   64.224141] [drm:amdgpu_dm_atomic_commit_tail [amdgpu]] *ERROR* amdgpu_dm_commit_planes: acrtc 0, already busy                                                               

    What's about xinit? What is MESA driver version on your platform?
R: I am running Ubuntu 18.04 with bisected kernel 4.18-rc4 using libdrm-2.4.92 and mesa-18.1.0.
xinit output follows attached as xinit_log.txt

        Thanks & Regards,
        Luís

        On Wed, Jul 11, 2018 at 10:56 AM, jimqu
          <jimqu@xxxxxxx>
          wrote:

              HI Luis,

              Let us trace the issue one by one.

              IB test fail:
              This should be regression issue on 4.18, you can bisect
                the kernel patches.
              GPU hang:
              Fix IB test fail first.

              Thanks

                  JimQu

                  On
                    2018年07月11日 17:34, Luís Mendes wrote:

                      Hi Jim,

                      Thanks for your interest in this issue.
                        Actually this is a multiple issue... not only
                        the IB ring test is failing... as I am having
                        quite some trouble getting the cards SAPPHIRE RX
                        550 4GB on a Tyan S7025 and SAPPHIRE RX 460 2GB
                        on a TYAN S7002 to work, both systems using same
                        Ubuntu 18.04 with vanilla kernel.

                      1. May you also test earlier kernel? v4.17
                          or v4.16.

                      I've tested kernels v4.17.5 and v4.16.6 with
                        same system and both are able to pass the IB
                        ring test and system boots into X using NVIDIA
                        as the display connected card.
                      dmesg log attached for kernel 4.17.5, file
                        TYAN_S7025_kernelv4.17.5_amdgpu_IB_ring_test_OK.txt.

                      2. May you test the issue only with
                          amdgpu?

                        - I've tested on a TYAN S7002 system with a
                          single SAPPHIRE RX 460 2GB, on-board VGA
                          enabled and used as primary display.
                        Kernel v4.18-rc4 fails the IB ring test,
                          system is able to enter X through the on-board
                          VGA. 

                        dmesg log attached for kernel 4.18-rc4,
                          file TYAN_S7002_kernel_v4.18-rc4_IB_ring_test_fail.txt.

                        - Same TYAN S7002 system, but now with
                          on-board VGA disabled and using RX 460 as
                          display connected card.

                          Kernels v4.17.5 and v4.16.6 are able to
                            pass the IB ring test, but GPU hangs before
                            entering X. Don't have logs for these yet.

                          Regards,
                          Luís Mendes
                          Aparapi contributor and MSc Researcher

                      On Wed, Jul 11, 2018 at
                        3:49 AM, Qu, Jim <Jim.Qu@xxxxxxx>
                        wrote:

                        Hi Luis,

                          1. May you also test earlier kernel? v4.17 or
                          v4.16.

                          2. May you test the issue only with amdgpu?

                          Thanks

                          JimQu

                          ________________________________________

                          发件人: amd-gfx <amd-gfx-bounces@xxxxxxxxxxxxxxxxxxxxx>
                          代表 Luís Mendes <luis.p.mendes@xxxxxxxxx>

                          发送时间: 2018年7月11日 6:04:00

                          收件人: Michel Dänzer; Koenig, Christian; amd-gfx
                          list

                          主题: Re: Regression with kernel 4.18 - AMD RX
                          550 fails IB ring test on power-up

                            Hi,

                            Issue remains in kernel 4.18-rc4 using
                            SAPPHIRE RX 550 4GB.

                            Logs follow attached.

                            Regards,

                            Luis

                            On
                              Tue, Jun 26, 2018 at 10:08 AM, Luís Mendes
                              <luis.p.mendes@xxxxxxxxx<mailto:luis.p.mendes@xxxxxxxxx>>
                              wrote:

                              Hi,

                              I've tried kernel 4.18-rc2 on a system
                              with a NVIDIA GTX 1050 Ti and an AMD RX
                              550 4GB and the RX 550 card is failing the
                              IB ring test.

                              [    5.033217] [drm:gfx_v8_0_ring_test_ib
                              [amdgpu]] *ERROR* amdgpu: ib test failed
                              (scratch(0xC040)=0xFFFFFFFF)

                              [    5.033264] [drm:amdgpu_ib_ring_tests
                              [amdgpu]] *ERROR* amdgpu: failed testing
                              IB on ring 6 (-22).

                              Please see the attached log.

                              Regards,

                              Luís

_______________________________________________
amd-gfx mailing list
amd-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/amd-gfx