fglrx with radeon xpress 200m on 2.6.15 causes kernel errors when used

cwhalen · 01-28-2006, 08:39 PM

Hi, thanks for checking this out. I'm really stuck on this one.

My setup is:
Slackware 10.2
vanilla kernel 2.6.15 patched for verify_area support
ATI proprietary driver 8.20.8-i386 for ATI Radeon XPress 200M (IGP)
Xorg 6.8.2
KDE 3.5
gcc 3.4.5
glibc 2.3.5

I've gotten the ATI driver to compile and start with DRI enabled. All my programs that use OpenGL are properly accellerated, but they are unable to close, and in some cases even fail to respond to the KILL signal.

Also, OpenGL causes the kernel to report errors, such as the following when I ran glxinfo:

[fglrx] Maximum main memory to use for locked dma buffers: 312 MBytes.
ACPI: PCI Interrupt 0000:01:05.0[A] -> Link [LNKB] -> GSI 10 (level, low) -> IRQ
10
[fglrx] module loaded - fglrx 8.20.8 [Dec 6 2005] on minor 0
[fglrx] ACPI power management is initialized.

...

Bad page state at free_hot_cold_page (in process 'X', page c129cd40)
flags:0x80000414 mapping:00000000 mapcount:0 count:0
Backtrace:
[<c013cd9f>] bad_page+0x5f/0xa0
[<c013d5df>] free_hot_cold_page+0x3f/0x120
[<c014750a>] zap_pte_range+0x17a/0x260
[<c0147690>] unmap_page_range+0xa0/0x150
[<c014780d>] unmap_vmas+0xcd/0x1d0
[<c014b911>] unmap_region+0x81/0x110
[<c014bbf5>] do_munmap+0xc5/0x110
[<c014bc7b>] sys_munmap+0x3b/0x60
[<c0103039>] syscall_call+0x7/0xb
Trying to fix it up, but a reboot is needed
[fglrx] free PCIe = 54804480
[fglrx] max PCIe = 54804480
[fglrx] free LFB = 120254464
[fglrx] max LFB = 120254464
[fglrx] free Inv = 0
[fglrx] max Inv = 0
[fglrx] total Inv = 0
[fglrx] total TIM = 0
[fglrx] total FB = 0
[fglrx] total PCIe = 16384
------------[ cut here ]------------
kernel BUG at mm/swap.c:49!
invalid operand: 0000 [#1]
PREEMPT
Modules linked in: vmnet vmmon fglrx verify_area ndiswrapper
CPU: 0
EIP: 0060:[<c0142df0>] Tainted: P B VLI
EFLAGS: 00210256 (2.6.15-x86)
EIP is at put_page+0x60/0x80
eax: 00000000 ebx: c88cd6d4 ecx: c129cd40 edx: c129cd40
esi: b7db5000 edi: c129cd40 ebp: c9b21f18 esp: c9b21e88
ds: 007b es: 007b ss: 0068
Process glxinfo (pid: 2986, threadinfo=c9b20000 task=c9b11a90)
Stack: c014750a c129cd40 14e6a067 00000000 ffffffff d6b0cac0 c922c284 c90d8b80
b7db6000 c9b21f18 c0147690 c054b0c0 c922c284 c90d8b7c b7db5000 b7db6000
c9b21f18 00000000 b7db5fff c90d8b7c c90d8b7c 00000001 c922c284 b7db5000
Call Trace:
[<c014750a>] zap_pte_range+0x17a/0x260
[<c0147690>] unmap_page_range+0xa0/0x150
[<c014780d>] unmap_vmas+0xcd/0x1d0
[<c014b911>] unmap_region+0x81/0x110
[<c014bbf5>] do_munmap+0xc5/0x110
[<c014bc7b>] sys_munmap+0x3b/0x60
[<c0103039>] syscall_call+0x7/0xb
Code: 00 00 8b 52 0c 89 d1 8b 02 f6 c4 40 74 03 8b 4a 0c 8b 41 04 40 74 1e 83 42
04 ff 0f 98 c0 84 c0 74 ce 8b 4a 30 89 54 24 04 ff e1 <0f> 0b 31 00 16 47 43 c0
eb b0 0f 0b 29 00 16 47 43 c0 eb d8 8d
<6>note: glxinfo[2986] exited with preempt_count 2
scheduling while atomic: glxinfo/0x00000002/2986
[<c0415603>] schedule+0x5d3/0x6a0
[<c0118005>] vprintk+0x2c5/0x300
[<c0118005>] vprintk+0x2c5/0x300
[<c0416439>] rwsem_down_read_failed+0x99/0x180
[<c010358a>] show_trace+0x2a/0x70
[<c011b5c1>] .text.lock.exit+0x27/0x86
[<c011a193>] do_exit+0xf3/0x460
[<c0103c70>] do_invalid_op+0x0/0x90
[<c01039e0>] die+0x160/0x170
[<c0103cea>] do_invalid_op+0x7a/0x90
[<c038f764>] pci_mmcfg_write+0x64/0x150
[<c0142df0>] put_page+0x60/0x80
[<d8ca6f8a>] __ke_spin_unlock+0xa/0x20 [fglrx]
[<d8ca98c7>] drm_free+0xb7/0x150 [fglrx]
[<c02586e6>] pci_bus_write_config_dword+0x36/0x70
[<d8cb3bf4>] firegl_aperture_free+0x1e4/0x220 [fglrx]
[<d8ca6f8a>] __ke_spin_unlock+0xa/0x20 [fglrx]
[<c010330f>] error_code+0x4f/0x60
[<c0142df0>] put_page+0x60/0x80
[<c014750a>] zap_pte_range+0x17a/0x260
[<c0147690>] unmap_page_range+0xa0/0x150
[<c014780d>] unmap_vmas+0xcd/0x1d0
[<c014b911>] unmap_region+0x81/0x110
[<c014bbf5>] do_munmap+0xc5/0x110
[<c014bc7b>] sys_munmap+0x3b/0x60
[<c0103039>] syscall_call+0x7/0xb

After I ran glxinfo, it froze and would not return to the shell. glxgears also gives me errors and will not close properly.

If anyone knows what's wrong or could give me a hand I would greatly appreciate it!

Cheers,
Chris

comprookie2000 · 01-29-2006, 12:52 PM

This is for the drm part in the kernel .config is yours like this

Code:

#
# Ftape, the floppy tape device driver
#
# CONFIG_FTAPE is not set
CONFIG_AGP=m
# CONFIG_AGP_ALI is not set
CONFIG_AGP_ATI=m
# CONFIG_AGP_AMD is not set
# CONFIG_AGP_AMD64 is not set
# CONFIG_AGP_INTEL is not set
# CONFIG_AGP_NVIDIA is not set
# CONFIG_AGP_SIS is not set
# CONFIG_AGP_SWORKS is not set
# CONFIG_AGP_VIA is not set
# CONFIG_AGP_EFFICEON is not set
CONFIG_DRM=m
# CONFIG_DRM_TDFX is not set
# CONFIG_DRM_R128 is not set
CONFIG_DRM_RADEON=m
# CONFIG_DRM_MGA is not set
# CONFIG_DRM_SIS is not set
# CONFIG_DRM_VIA is not set
# CONFIG_DRM_SAVAGE is not set

It needs those modules to work right.

cwhalen · 01-29-2006, 01:13 PM

Yes, mine's basically the same, except that AGP and AGP_ATI are builtin (=y) rather than modules. DRM and DRM_RADEON are both configured as modules.

The error that I get is always the same, it says Bad page state at free_hot_cold_page

Do you think that downgrading to an older driver version would work? I tried it again with the very latest driver version (8.21.7), but it still happens. Other people have reported that things have gotton broken and have actually found it better to downgrade...

sw67 · 01-29-2006, 04:54 PM

I had this same error when I used fglrx with a 2.6.15 or above kernel.
From what I can tell, it's a disagreement between the memory management of ATI and the kernel causing X and other things to crash when they try and use the fglrx module.
I figured this because when I told X to use the vesa or vga drivers, it worked perfectly.
In the end, I gave up and tried a 2.6.14.4 kernel, and everything works great.
I think it's just down to ATI not being compatible with the new kernel - I hear there were a few changes in it.
Have to play the waiting game for ATI to release a new driver.

Sam

cwhalen · 01-30-2006, 01:17 PM

I downgraded to 2.6.13.1, and verify_area() is still built in. It now kind of works; the driver doesn't report errors, but the performance is actually *worse* with direct rendering! I'm going to open a new thread becuase the problem has changed from the original post.

Thanks for all your help!

RedShirt · 01-30-2006, 01:57 PM

Have you tried the new kernel with the new script from ati? 8.21.7 is supposed to go with the new Kernel, perhaps the errors will go away?

EDIT: you may also wish to upgrade to Xorg 6.9 or 7.0 to see about any potential errors between X, the Kernel, and ATI there.

cwhalen · 01-30-2006, 02:08 PM

Yeah, I've tried the versions available on ATI's website as of last night, including 8.21.7, and I still had problems with compiling. I've got a temporary solution to this problem now by downgrading, so it's ok. More important is my current framerate issue.

I was thinking about upgrading X, just never tried it yet. I suppose I'll give that a shot before I post again.
Cheers!

cwhalen · 01-30-2006, 02:50 PM

I don't like replying to my own posts but I fixed the problem, it was an issue with shared memory.
http://www.linuxquestions.org/questi...d.php?t=404979
(Thanks to 43r05p4c3 for raising that point.)

Thanks to everyone that helped me out!