Announcement

Collapse
No announcement yet.

Crash hunting in Radeon KMS

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • #21
    Latest drm-radeon-testing has GPU recovery code. I just tested it with an app known to lock-up the GPU here, and it works quite nicely. No more rebooting or even restarting X. Maybe fglrx team should steal some of that code

    Comment


    • #22
      Originally posted by zika View Post
      After I went back from xorg-edgers to "Lucid-offical" set of drivers etc. I did not have a single crash in KMS. So, there is something in that difference-set that makes Lucid to crash, at least on my machine...
      When I said "official", I meant, of course, open-source-drivers... I'm back on xorg-edgers drivers just to see if something changed and if crashes are gone...

      Comment


      • #23
        Originally posted by zika View Post
        When I said "official", I meant, of course, open-source-drivers... I'm back on xorg-edgers drivers just to see if something changed and if crashes are gone...
        No, crashed several hours after I upgraded to xorg-edgers. I'm staying with it but in UMS...

        Comment


        • #24
          Originally posted by zika View Post
          No, crashed several hours after I upgraded to xorg-edgers. I'm staying with it but in UMS...
          This bug makes much of the efforts in making Lucid a great LTS tainted. Specially with the exclusion of fsync in unpack part in dpkg. I would like to do an upgrade, of kernel, for example, with those two lurking around my ext4...

          Comment


          • #25
            Originally posted by zika View Post
            This bug makes much of the efforts in making Lucid a great LTS tainted. Specially with the exclusion of fsync in unpack part in dpkg. I would like to do an upgrade, of kernel, for example, with those two lurking around my ext4...
            I owe You guys (and girls) an apology. I've got a crash today with KMS and "official" set of packages. So, the ball, as it seems, is not in Your court. Sorry... Back to UMS and xorg-edgers...

            Comment


            • #26
              i have the same random crashes.
              Karmic + 2.6.33 mainline + edgers + KMS with a HD3650 AGP
              Could it be the rv635 (same as zika, except maybe the AGP)?
              PY

              Comment


              • #27
                The first good advice on where to look for a troublemaker: http://ubuntuforums.org/showpost.php...7&postcount=47 Thank You dino99!!!!

                Comment


                • #28
                  GPU lockup detection rocks

                  Hello,

                  I tried the following configuration:
                  - Ubuntu karmic + Xorg/drm/mesa from lucid
                  - drm-radeon-testing kernel (thank God somebody invented 'make localmodconfig')
                  everything from today.

                  I am currently running glxgears, firefox, mplayer with composition. "Interesting" stuff appears in dmesg:

                  Code:
                  [  107.475152] You have old & broken userspace please consider updating mesa
                  [  812.630956] radeon 0000:01:00.0: GPU lockup CP stall for more than 1000msec
                  [  812.630965] ------------[ cut here ]------------
                  [  812.631020] WARNING: at drivers/gpu/drm/radeon/radeon_fence.c:234 radeon_fence_wait+0x35d/0x3c0 [radeon]()
                  [  812.631026] Hardware name: ESPRIMO Mobile V5545           
                  [  812.631031] GPU lockup (waiting for 0x000ED360 last fence id 0x000ED357)
                  [  812.631036] Modules linked in: binfmt_misc snd_hda_codec_conexant bridge snd_hda_intel stp ppdev snd_hda_codec bnep snd_hwdep snd_pcm_oss snd_mixer_oss snd_pcm snd_seq_dummy snd_seq_oss snd_seq_midi arc4 snd_rawmidi snd_seq_midi_event iwlagn snd_seq iwlcore snd_timer iptable_filter snd_seq_device joydev ip_tables mac80211 snd soundcore lp acer_wmi psmouse snd_page_alloc cfg80211 x_tables parport led_class serio_raw btusb sha256_generic aes_x86_64 aes_generic dm_crypt usbhid fbcon tileblit font bitblit softcursor radeon ttm drm_kms_helper r8169 usb_storage mii drm i2c_algo_bit intel_agp video output
                  [  812.631139] Pid: 1430, comm: Xorg Not tainted 2.6.33 #1
                  [  812.631144] Call Trace:
                  [  812.631160]  [<ffffffff81057e68>] warn_slowpath_common+0x78/0xb0
                  [  812.631169]  [<ffffffff81057efc>] warn_slowpath_fmt+0x3c/0x40
                  [  812.631204]  [<ffffffffa00f233d>] radeon_fence_wait+0x35d/0x3c0 [radeon]
                  [  812.631215]  [<ffffffff81077460>] ? autoremove_wake_function+0x0/0x40
                  [  812.631249]  [<ffffffffa00f25ab>] ? radeon_fence_create+0xcb/0x130 [radeon]
                  [  812.631288]  [<ffffffffa0108ba3>] radeon_ib_get+0x123/0x230 [radeon]
                  [  812.631326]  [<ffffffffa010a0bb>] radeon_cs_ioctl+0x8b/0x1c0 [radeon]
                  [  812.631365]  [<ffffffffa0108245>] ? radeon_gem_wait_idle_ioctl+0xd5/0x130 [radeon]
                  [  812.631391]  [<ffffffffa002d1da>] drm_ioctl+0x32a/0x4b0 [drm]
                  [  812.631430]  [<ffffffffa010a030>] ? radeon_cs_ioctl+0x0/0x1c0 [radeon]
                  [  812.631441]  [<ffffffff811272f2>] ? do_sync_read+0xd2/0x110
                  [  812.631450]  [<ffffffff8107abd2>] ? enqueue_hrtimer+0x82/0xd0
                  [  812.631460]  [<ffffffff81136ab8>] vfs_ioctl+0x38/0xd0
                  [  812.631468]  [<ffffffff8113703a>] do_vfs_ioctl+0x8a/0x5b0
                  [  812.631477]  [<ffffffff8112798f>] ? vfs_read+0x12f/0x1a0
                  [  812.631485]  [<ffffffff811375e1>] sys_ioctl+0x81/0xa0
                  [  812.631495]  [<ffffffff81009f42>] system_call_fastpath+0x16/0x1b
                  [  812.631501] ---[ end trace bdf1d2605ae35df2 ]---
                  [  812.631515] [drm] Disabling audio support
                  [  812.631569] radeon 0000:01:00.0: GPU softreset 
                  [  812.631576] radeon 0000:01:00.0:   R_008010_GRBM_STATUS=0xA0003030
                  [  812.631582] radeon 0000:01:00.0:   R_008014_GRBM_STATUS2=0x00000003
                  [  812.631589] radeon 0000:01:00.0:   R_000E50_SRBM_STATUS=0x200000C0
                  [  812.631603] radeon 0000:01:00.0:   R_008020_GRBM_SOFT_RESET=0x00007FEE
                  [  812.646616] radeon 0000:01:00.0: R_008020_GRBM_SOFT_RESET=0x00000001
                  [  812.662638] radeon 0000:01:00.0:   R_008010_GRBM_STATUS=0xA0003030
                  [  812.662645] radeon 0000:01:00.0:   R_008014_GRBM_STATUS2=0x00000003
                  [  812.662651] radeon 0000:01:00.0:   R_000E50_SRBM_STATUS=0x200080C0
                  [  812.663658] radeon 0000:01:00.0: GPU reset succeed
                  [  813.098987] [drm] Clocks initialized !
                  [  813.133255] [drm] ring test succeeded in 1 usecs
                  [  813.133274] [drm] ib test succeeded in 1 usecs
                  [  813.133280] [drm] Enabling audio support
                  Good news GPU lockup is working pretty well. Some rendering is screwed (especially text), but by doing VT switch the situation is restored.

                  Bad news GPU lockup is used. What should I do so that the user-space stops locking the GPU?

                  Cheers,
                  Cristi.

                  Comment

                  Working...
                  X