kernel 2.6.32: How to enable ATI KMS?

cpuesser · 2009-12-10 19:58:19

A new bad message, when I boot, while my Notebook ist on battery, early kms does not work.
I get only a blank screen und nothing happens.

Last edited by cpuesser (2009-12-10 19:59:57)

Tapan · 2009-12-10 21:52:09

cpuesser wrote:

A new bad message, when I boot, while my Notebook ist on battery, early kms does not work.
I get only a blank screen und nothing happens.

My notebook boot fine on batery, when i use early KMS.

abraxas88 · 2009-12-12 13:19:05

Everything works fine, until I don't log in KDE and after a few seconds this is what happens:

[drm:radeon_fence_wait] *ERROR* fence(ffff880063116d80:0x000007FC) 510ms timeout going to reset GPU
[drm] CP reset succeed (RBBM_STATUS=0x00000140)
[drm] radeon: cp idle (0x10000000)
[drm] radeon: ring at 0x0000000080000000
[drm] ring test succeeded in 0 usecs
[drm] GPU reset succeed (RBBM_STATUS=0x00000140)
[drm:radeon_fence_wait] *ERROR* fence(ffff880063116d80:0x000007FC) 520ms timeout
[drm:radeon_fence_wait] *ERROR* last signaled fence(0x000007FC)
[drm:radeon_fence_wait] *ERROR* fence(ffff88006323fbc0:0x000007FD) 510ms timeout going to reset GPU
[drm] CP reset succeed (RBBM_STATUS=0x00000140)
[drm] radeon: cp idle (0x10000000)
[drm] radeon: ring at 0x0000000080000000
[drm] ring test succeeded in 0 usecs
[drm] GPU reset succeed (RBBM_STATUS=0x00000140)
[drm:radeon_fence_wait] *ERROR* fence(ffff88006323fbc0:0x000007FD) 520ms timeout
[drm:radeon_fence_wait] *ERROR* last signaled fence(0x000007FD)
[drm:radeon_fence_wait] *ERROR* fence(ffff880063116000:0x000007FE) 510ms timeout going to reset GPU
[drm] CP reset succeed (RBBM_STATUS=0x00000140)
[drm] radeon: cp idle (0x10000000)
[drm] radeon: ring at 0x0000000080000000
[drm] ring test succeeded in 0 usecs
[drm] GPU reset succeed (RBBM_STATUS=0x00000140)
[drm:radeon_fence_wait] *ERROR* fence(ffff880063116000:0x000007FE) 520ms timeout
[drm:radeon_fence_wait] *ERROR* last signaled fence(0x000007FE)
[drm:radeon_fence_wait] *ERROR* fence(ffff88006323f9c0:0x00000801) 510ms timeout going to reset GPU
[drm] CP reset succeed (RBBM_STATUS=0x00000140)
[drm] radeon: cp idle (0x10000000)
[drm] radeon: ring at 0x0000000080000000
[drm:r100_ring_test] *ERROR* radeon: ring test failed (sracth(0x15E4)=0xCAFEDEAD)
[drm:r100_cp_init] *ERROR* radeon: cp isn't working (-22).
[drm:r300_gpu_reset] *ERROR* Failed to reset GPU (RBBM_STATUS=0x80010140)
[drm:radeon_fence_wait] *ERROR* fence(ffff88006323f9c0:0x00000801) 666ms timeout
[drm:radeon_fence_wait] *ERROR* last signaled fence(0x00000801)
[drm:radeon_ib_schedule] *ERROR* radeon: couldn't schedule IB(3).
[drm:radeon_cs_ioctl] *ERROR* Faild to schedule IB !
[drm:radeon_ib_schedule] *ERROR* radeon: couldn't schedule IB(3).
[drm:radeon_cs_ioctl] *ERROR* Faild to schedule IB !
[drm:radeon_ib_schedule] *ERROR* radeon: couldn't schedule IB(3).
[drm:radeon_cs_ioctl] *ERROR* Faild to schedule IB !
[drm:radeon_ib_schedule] *ERROR* radeon: couldn't schedule IB(3).

Any suggestions? Video card is ATI (of course) Radeon Xpress 200M (ven 5A62)

zodmaner · 2009-12-16 16:29:14

Tobias have created a proper solution to this problem, which required testing by us users.

For everyone who could spare time, please visit this bug report and follow the instructions there to help test the solution out.

Last edited by zodmaner (2009-12-16 16:34:06)

abraxas88 · 2009-12-16 18:57:34

I'll try later and I'll post the results

jarryson · 2009-12-17 07:12:33

the random-initrd do work, thanks

but the graphics perfomance reduced a lot when using kernel 2.6.32. E.g i will meet a lot of pause when using opera and even using mplayer to watch a movie. which i can't quite stand. Glxgears score reduced from 1600fps(kernel 2.6.31) to 18fps, but the 2D and 3D acceleration seems all right.

my card is X700 mobility.

abraxas88 · 2009-12-17 14:58:23

No, it doesn't work I don't know why, that's the dmesg output

[drm:radeon_fence_wait] *ERROR* fence(ffff88006458da00:0x00000865) 510ms timeout going to reset GPU
[drm:r300_ga_reset] *ERROR* VAP & CP still busy (RBBM_STATUS=0x84000140)
[drm] GA reset succeed (RBBM_STATUS=0x00000140)
[drm] radeon: cp idle (0x10000000)
[drm] radeon: ring at 0x0000000080000000
[drm:r100_ring_test] *ERROR* radeon: ring test failed (sracth(0x15E4)=0xCAFEDEAD)
[drm:r100_cp_init] *ERROR* radeon: cp isn't working (-22).
[drm] CP reset succeed (RBBM_STATUS=0x00000140)
[drm] GPU reset succeed (RBBM_STATUS=0x00000140)
[drm:radeon_fence_wait] *ERROR* fence(ffff88006458da00:0x00000865) 663ms timeout
[drm:radeon_fence_wait] *ERROR* last signaled fence(0x00000865)
[drm:radeon_ib_schedule] *ERROR* radeon: couldn't schedule IB(11).
[drm:radeon_cs_ioctl] *ERROR* Faild to schedule IB !
[drm:radeon_ib_schedule] *ERROR* radeon: couldn't schedule IB(11).
[drm:radeon_cs_ioctl] *ERROR* Faild to schedule IB !
[drm:radeon_ib_schedule] *ERROR* radeon: couldn't schedule IB(11).
[drm:radeon_cs_ioctl] *ERROR* Faild to schedule IB !

zodmaner · 2009-12-17 16:03:28

abraxas88 wrote:

No, it doesn't work I don't know why, that's the dmesg output

[drm:radeon_fence_wait] *ERROR* fence(ffff88006458da00:0x00000865) 510ms timeout going to reset GPU
[drm:r300_ga_reset] *ERROR* VAP & CP still busy (RBBM_STATUS=0x84000140)
[drm] GA reset succeed (RBBM_STATUS=0x00000140)
[drm] radeon: cp idle (0x10000000)
[drm] radeon: ring at 0x0000000080000000
[drm:r100_ring_test] *ERROR* radeon: ring test failed (sracth(0x15E4)=0xCAFEDEAD)
[drm:r100_cp_init] *ERROR* radeon: cp isn't working (-22).
[drm] CP reset succeed (RBBM_STATUS=0x00000140)
[drm] GPU reset succeed (RBBM_STATUS=0x00000140)
[drm:radeon_fence_wait] *ERROR* fence(ffff88006458da00:0x00000865) 663ms timeout
[drm:radeon_fence_wait] *ERROR* last signaled fence(0x00000865)
[drm:radeon_ib_schedule] *ERROR* radeon: couldn't schedule IB(11).
[drm:radeon_cs_ioctl] *ERROR* Faild to schedule IB !
[drm:radeon_ib_schedule] *ERROR* radeon: couldn't schedule IB(11).
[drm:radeon_cs_ioctl] *ERROR* Faild to schedule IB !
[drm:radeon_ib_schedule] *ERROR* radeon: couldn't schedule IB(11).
[drm:radeon_cs_ioctl] *ERROR* Faild to schedule IB !

Did you follow the instructions in bug report properly? How do you re-create the initrd image?

I followed the instructions but instead of using "mkinitcpio -g test.img" I just went ahead and do a "mkinitcpio -p kernel26".

abraxas88 · 2009-12-17 16:40:28

Well, that could be an interesting alternative, I've just downgraded to 2.6.31... I'll re-upgrade to 32 and'll rebuild the image using mkinitcpio -p kernel26

abraxas88 · 2009-12-17 17:38:02

Ok, that's what I did: I left the files posted in the bug report and I installed 2.6.32 again, so the image went built with that configuration.
Nothing new, that's the dmesg output (again):

[drm:radeon_fence_wait] *ERROR* fence(ffff88006a8a5080:0x00000115) 510ms timeout going to reset GPU
[drm] RB2D reset succeed (RBBM_STATUS=0x80010140)
[drm] CP reset succeed (RBBM_STATUS=0x00000140)
[drm] radeon: cp idle (0x10000000)
[drm] radeon: ring at 0x0000000080000000
[drm] ring test succeeded in 0 usecs
[drm] GPU reset succeed (RBBM_STATUS=0x00000140)
[drm:radeon_fence_wait] *ERROR* fence(ffff88006a8a5080:0x00000115) 520ms timeout
[drm:radeon_fence_wait] *ERROR* last signaled fence(0x00000115)
[drm:radeon_fence_wait] *ERROR* fence(ffff88006a8a5a40:0x00000116) 510ms timeout going to reset GPU
[drm] CP reset succeed (RBBM_STATUS=0x00000140)
[drm] radeon: cp idle (0x10000000)
[drm] radeon: ring at 0x0000000080000000
[drm] ring test succeeded in 0 usecs
[drm] GPU reset succeed (RBBM_STATUS=0x00000140)
[drm:radeon_fence_wait] *ERROR* fence(ffff88006a8a5a40:0x00000116) 520ms timeout
[drm:radeon_fence_wait] *ERROR* last signaled fence(0x00000116)
[drm:radeon_fence_wait] *ERROR* fence(ffff88006a8a5c80:0x00000117) 510ms timeout going to reset GPU
[drm] CP reset succeed (RBBM_STATUS=0x00000140)
[drm] radeon: cp idle (0x10000000)
[drm] radeon: ring at 0x0000000080000000
[drm] ring test succeeded in 0 usecs
[drm] GPU reset succeed (RBBM_STATUS=0x00000140)
[drm:radeon_fence_wait] *ERROR* fence(ffff88006a8a5c80:0x00000117) 520ms timeout
[drm:radeon_fence_wait] *ERROR* last signaled fence(0x00000117)
[drm:radeon_fence_wait] *ERROR* fence(ffff88006a8a5840:0x00000118) 510ms timeout going to reset GPU
[drm] CP reset succeed (RBBM_STATUS=0x00000140)
[drm] radeon: cp idle (0x10000000)
[drm] radeon: ring at 0x0000000080000000
[drm] ring test succeeded in 0 usecs
[drm] GPU reset succeed (RBBM_STATUS=0x00000140)
[drm:radeon_fence_wait] *ERROR* fence(ffff88006a8a5840:0x00000118) 520ms timeout
[drm:radeon_fence_wait] *ERROR* last signaled fence(0x00000118)
[drm:radeon_fence_wait] *ERROR* fence(ffff88006a8a5800:0x00000119) 510ms timeout going to reset GPU
[drm] CP reset succeed (RBBM_STATUS=0x00000140)
[drm] radeon: cp idle (0x10000000)
[drm] radeon: ring at 0x0000000080000000
[drm] ring test succeeded in 0 usecs
[drm] GPU reset succeed (RBBM_STATUS=0x00000140)
[drm:radeon_fence_wait] *ERROR* fence(ffff88006a8a5800:0x00000119) 520ms timeout
[drm:radeon_fence_wait] *ERROR* last signaled fence(0x00000119)
[drm:radeon_fence_wait] *ERROR* fence(ffff88006a8a5880:0x0000011A) 510ms timeout going to reset GPU
[drm] CP reset succeed (RBBM_STATUS=0x00000140)
[drm] radeon: cp idle (0x10000000)
[drm] radeon: ring at 0x0000000080000000
[drm] ring test succeeded in 0 usecs
[drm] GPU reset succeed (RBBM_STATUS=0x00000140)
[drm:radeon_fence_wait] *ERROR* fence(ffff88006a8a5880:0x0000011A) 520ms timeout
[drm:radeon_fence_wait] *ERROR* last signaled fence(0x0000011A)
[drm:radeon_fence_wait] *ERROR* fence(ffff88007413d440:0x00000120) 510ms timeout going to reset GPU
[drm] CP reset succeed (RBBM_STATUS=0x00000140)
[drm] radeon: cp idle (0x10000000)
[drm] radeon: ring at 0x0000000080000000
[drm:r100_ring_test] *ERROR* radeon: ring test failed (sracth(0x15E4)=0xCAFEDEAD)
[drm:r100_cp_init] *ERROR* radeon: cp isn't working (-22).
[drm:r300_gpu_reset] *ERROR* Failed to reset GPU (RBBM_STATUS=0x80010140)
[drm:radeon_fence_wait] *ERROR* fence(ffff88007413d440:0x00000120) 663ms timeout
[drm:radeon_fence_wait] *ERROR* last signaled fence(0x00000120)
[drm:radeon_ib_schedule] *ERROR* radeon: couldn't schedule IB(14).
[drm:radeon_cs_ioctl] *ERROR* Faild to schedule IB !
[drm:radeon_ib_schedule] *ERROR* radeon: couldn't schedule IB(14).
[drm:radeon_cs_ioctl] *ERROR* Faild to schedule IB !
[drm:radeon_ib_schedule] *ERROR* radeon: couldn't schedule IB(14).
[drm:radeon_cs_ioctl] *ERROR* Faild to schedule IB !
[drm:radeon_ib_schedule] *ERROR* radeon: couldn't schedule IB(14).
[drm:radeon_cs_ioctl] *ERROR* Faild to schedule IB !
[drm:radeon_ib_schedule] *ERROR* radeon: couldn't schedule IB(14).
[drm:radeon_cs_ioctl] *ERROR* Faild to schedule IB !

The strange thing is that it happens as I log in kdm, so could it maybe an Xorg error? /var/log/Xorg.0.log has not errors, though.

jocheem67 · 2009-12-17 17:53:41

I've got kms working ( early start ), with "firmware" in the hooks section. No "radeon", gave me a module not found-error...

System is pretty snappy now. Bit better tha before, especially 2d.

germandev · 2009-12-18 06:54:54

Hi!

I am trying to get KMS early start to work. My graphic card is a HD 3200 and I am using the current kernel 2.6.32 from the testing repo.

I tried a lot of workarounds because I got that "firmware load - 1 minute delay" problem. After Tobias' solution everything seemed to work. X and kdm start but after trying to login I get the following messages in the error log:

Dec 17 10:20:40 myhost kernel: [drm:radeon_cs_ioctl] *ERROR* Invalid command stream !
Dec 17 10:20:41 myhost kdm[4397]: X server for display :0 terminated unexpectedly
Dec 17 10:20:41 myhost kernel: [drm:drm_mode_getfb] *ERROR* invalid framebuffer id
Dec 17 10:20:49 myhost kernel: [drm:r600_cs_packet_next_reloc_mm] *ERROR* No packet3 for relocation for packet at 47.
Dec 17 10:20:49 myhost kernel: [drm:r600_packet3_check] *ERROR* bad SET_CONTEXT_REG 0x28014
Dec 17 10:20:49 myhost kernel: [drm:radeon_cs_ioctl] *ERROR* Invalid command stream !
Dec 17 10:20:49 myhost kdm[4397]: X server for display :0 terminated unexpectedly
Dec 17 10:20:49 myhost kernel: [drm:drm_mode_getfb] *ERROR* invalid framebuffer id

It's quite annyoing because I have to use Windows atm instead

Do you have any idea what the problem might be?

Thank you in advance!

ConnorBehan · 2009-12-18 07:32:20

tomk wrote:

you don't need radeon in the MODULES line, if that's what you mean, You do need radeon in the HOOKS line, as detailed above. I don't know what you mean by "it's already built-in".

Really? I do not have radeon or firmware anywhere in my mkinitcpio.conf. All I do is load radeon from the MODULES array in rc.conf and according the dmesg, it finds the R300_cp.bin perfectly. A framebuffer device is created before I start X (which radeon only does when KMS is working) and Xorg.0.log shows the following:

cat /var/log/Xorg.0.log | grep KMS
(II) [KMS] Kernel modesetting enabled.
(II) RADEON(0): KMS Color Tiling: disabled

Is the "early" start somehow critical to getting *good* radeon KMS performance? Because right now my framerate is horrible.

tomk · 2009-12-18 08:20:52

ConnorBehan - I was referring to the original workaround for the early-start method i.e. before the fix suggested in the bug report. Loading from rc.conf is the late-start method. There should be no appreciable difference in performance between the two.

linuxgod · 2009-12-18 17:12:59

still doesn't work

rusty99 · 2009-12-18 21:10:31

linuxgod wrote:

still doesn't work

What doesn't? You don't mention anywhere on this page what you've tried or were going to try next, a bit more info would be helpful pal.

germandev · 2009-12-20 16:01:44

doesn't anyone have any idea what my problem might be?

hungerfish · 2009-12-22 16:06:28

Ok, I have found a new problem...

On a second machine,using stock 2.6.32 kernel from arch, BOTH the 'original' workaround (radeon hook) and the fix from the bug report result in a segfault during loading the radeon module.
I found this:
http://www.mail-archive.com/dri-devel@l … 45422.html
which is what I am getting. This is obviously an old agp-machine ... using 'nomodeset' option resolves the segfault, but also disables kms

linuxgod · 2009-12-28 16:14:04

Wont Work Until You Buy Intel Or Nvidia

TemplarGR · 2009-12-28 16:44:28

linuxgod wrote:

Wont Work Until You Buy Intel Or Nvidia

Well it is only a matter of time until opensource ati drivers surpass in quality even nvidia's drivers. But i shouldn't answer to you since you are obviously a troll...

R00KIE · 2009-12-28 17:31:28

Ati KMS isn't working for me at all now with a Mobility Radeon HD 2400 [1002:94c9] (with kernel 2.6.32.2-2), by this I mean early or late KMS (kernel 2.6.32.1 seemed to work fine though). It just not ready for prime time yet I guess.

TemplarGR wrote:

linuxgod wrote:
Wont Work Until You Buy Intel Or Nvidia
Well it is only a matter of time until opensource ati drivers surpass in quality even nvidia's drivers. But i shouldn't answer to you since you are obviously a troll...

I guess you are right, either troll or a very happy costumer .... until it suddenly stops working with no good cause that is

linuxgod · 2009-12-28 18:38:20

yo man, Im da happiest customer having ati chips in the world!!
im so happy to have chips that do not work at all, oh yeah!
3D dont work, oh yeah!
And even Propriatire drivers dont work (even with old kernel and old Xserver), oh yeah!
And Im so happy, COZ im ATI MEGAPATRIOT!!
love ati - be da happy fool

hungerfish · 2009-12-28 19:40:19

So is there anyone out there with agp getting segfaults with ati-kms?

wantilles · 2009-12-28 20:40:25

TemplarGR wrote:

linuxgod wrote:
Wont Work Until You Buy Intel Or Nvidia
Well it is only a matter of time until opensource ati drivers surpass in quality even nvidia's drivers.

Really?

I just broke the back of my chair from the uncontrollable laughter.

Because I have been running Linux since May 2004 (5.5 years).

And I have been hearing the same argument since then.

But with no results, no improvement, until now.

The man is right.

If you want to do anything else than gaming, and run any OS except Windows, your only option is nVidia (or Intel). Not ATi.

Last edited by wantilles (2009-12-28 20:40:56)

TemplarGR · 2009-12-29 22:02:48

wantilles wrote:

TemplarGR wrote:
linuxgod wrote:
Wont Work Until You Buy Intel Or Nvidia
Well it is only a matter of time until opensource ati drivers surpass in quality even nvidia's drivers.
Really?
I just broke the back of my chair from the uncontrollable laughter.
Because I have been running Linux since May 2004 (5.5 years).
And I have been hearing the same argument since then.
But with no results, no improvement, until now.
The man is right.
If you want to do anything else than gaming, and run any OS except Windows, your only option is nVidia (or Intel). Not ATi.

Just because in the past support was poor doesn't mean it will always be this way. Ati open source drivers are quickly maturing.

Arch Linux

#51 2009-12-10 19:58:19

Re: kernel 2.6.32: How to enable ATI KMS?

#52 2009-12-10 21:52:09

Re: kernel 2.6.32: How to enable ATI KMS?

#53 2009-12-12 13:19:05

Re: kernel 2.6.32: How to enable ATI KMS?

#54 2009-12-16 16:29:14

Re: kernel 2.6.32: How to enable ATI KMS?

#55 2009-12-16 18:57:34

Re: kernel 2.6.32: How to enable ATI KMS?

#56 2009-12-17 07:12:33

Re: kernel 2.6.32: How to enable ATI KMS?

#57 2009-12-17 14:58:23

Re: kernel 2.6.32: How to enable ATI KMS?

#58 2009-12-17 16:03:28

Re: kernel 2.6.32: How to enable ATI KMS?

#59 2009-12-17 16:40:28

Re: kernel 2.6.32: How to enable ATI KMS?

#60 2009-12-17 17:38:02

Re: kernel 2.6.32: How to enable ATI KMS?

#61 2009-12-17 17:53:41

Re: kernel 2.6.32: How to enable ATI KMS?

#62 2009-12-18 06:54:54

Re: kernel 2.6.32: How to enable ATI KMS?

#63 2009-12-18 07:32:20

Re: kernel 2.6.32: How to enable ATI KMS?

#64 2009-12-18 08:20:52

Re: kernel 2.6.32: How to enable ATI KMS?

#65 2009-12-18 17:12:59

Re: kernel 2.6.32: How to enable ATI KMS?

#66 2009-12-18 21:10:31

Re: kernel 2.6.32: How to enable ATI KMS?

#67 2009-12-20 16:01:44

Re: kernel 2.6.32: How to enable ATI KMS?

#68 2009-12-22 16:06:28

Re: kernel 2.6.32: How to enable ATI KMS?

#69 2009-12-28 16:14:04

Re: kernel 2.6.32: How to enable ATI KMS?

#70 2009-12-28 16:44:28

Re: kernel 2.6.32: How to enable ATI KMS?

#71 2009-12-28 17:31:28

Re: kernel 2.6.32: How to enable ATI KMS?

#72 2009-12-28 18:38:20

Re: kernel 2.6.32: How to enable ATI KMS?

#73 2009-12-28 19:40:19

Re: kernel 2.6.32: How to enable ATI KMS?

#74 2009-12-28 20:40:25

Re: kernel 2.6.32: How to enable ATI KMS?

#75 2009-12-29 22:02:48

Re: kernel 2.6.32: How to enable ATI KMS?

Board footer