Kernel bug error at pageattr.c

[startail]

Hi,

During the last few hours I've had a problem with my kernel.
I think it started after recompiling it to Preempt Model Desktop and Timer Frequency to 1000 HZ

I use ati-drivers 8.20.8 and kernel 2.6.15-gentoo

My ATI drivers seem to crash for some reason when I log out from an X server (and it has also crashen while just switching themes in Gnome)

The error I get is:

Code: Select all

Jan  7 22:23:24 tailz kernel BUG at arch/i386/mm/pageattr.c:137!
Jan  7 22:23:24 tailz invalid operand: 0000 [#1]
Jan  7 22:23:24 tailz Modules linked in: fglrx agpgart nfsd exportfs lockd sunrpc snd_seq snd_seq_device snd_intel8x0 snd_ac97_codec snd_ac97_bus snd_pcm snd_timer snd soundcore snd_page_alloc smsc47m1 eeprom adm1025 hwmon_vid hwmon i2c_isa i2c_i801 i2c_core rtc usbhid ehci_hcd ohci_hcd uhci_hcd usbcore cdemu 8139too mii ide_cd cdrom loop vfat fat unix
Jan  7 22:23:24 tailz CPU:    0
Jan  7 22:23:24 tailz EIP:    0060:[<c010fc8e>]    Tainted: P      VLI
Jan  7 22:23:24 tailz EFLAGS: 00013082   (2.6.15-gentoo)
Jan  7 22:23:24 tailz EIP is at __change_page_attr+0xba/0x15a
Jan  7 22:23:24 tailz eax: 19c001e3   ebx: 19c80000   ecx: c10061c0   edx: d9c80000
Jan  7 22:23:24 tailz esi: c030ed9c   edi: c1000000   ebp: 00000163   esp: db2a1e78
Jan  7 22:23:24 tailz ds: 007b   es: 007b   ss: 0068
Jan  7 22:23:24 tailz Process X (pid: 7069, threadinfo=db2a0000 task=db278a90)
Jan  7 22:23:24 tailz Stack: d9c80000 c1339000 00000010 00003246 00000011 c010fd4f c1339000 00000163
Jan  7 22:23:24 tailz dad49300 e1780000 e180b520 00000000 c010fa4b c1338e00 00000011 00000163
Jan  7 22:23:24 tailz e180b520 00000004 e17a8cf5 e1780000 c1338e00 00000010 00000163 e180b520
Jan  7 22:23:24 tailz Call Trace:
Jan  7 22:23:24 tailz [<c010fd4f>] change_page_attr+0x21/0x36
Jan  7 22:23:24 tailz [<c010fa4b>] iounmap+0x90/0xc5
Jan  7 22:23:24 tailz [<e17a8cf5>] agp_generic_free_gatt_table+0x62/0xce [fglrx]
Jan  7 22:23:24 tailz [<e17aece4>] agp_backend_cleanup+0xc/0x4a [fglrx]
Jan  7 22:23:24 tailz [<e17b1de4>] __ke_agp_uninit+0xe/0x24 [fglrx]
Jan  7 22:23:24 tailz [<e17c8d95>] _firegl_release_agp+0x15/0x140 [fglrx]
Jan  7 22:23:24 tailz [<e17b72a5>] firegl_takedown+0x335/0xb90 [fglrx]
Jan  7 22:23:24 tailz [<e17b664f>] firegl_release+0x12f/0x190 [fglrx]
Jan  7 22:23:24 tailz [<e17af0cf>] ip_firegl_release+0xd/0x10 [fglrx]
Jan  7 22:23:24 tailz [<c014413b>] __fput+0x83/0x10a
Jan  7 22:23:24 tailz [<c0142df2>] filp_close+0x4c/0x55
Jan  7 22:23:24 tailz [<c0142e49>] sys_close+0x4e/0x5a
Jan  7 22:23:24 tailz [<c0102a29>] syscall_call+0x7/0xb
Jan  7 22:23:24 tailz Code: 24 04 56 e8 26 ff ff ff 89 d9 83 c4 0c 8b 01 89 ca f6 c4 40 74 03 8b 51 0c ff 42 04 eb 15 84 c0 78 09 09 eb 89 1e ff 49 04 eb 08 <0f> 0b 89 00 eb d2 27 c0 8b 01 f6 c4 04 0f 85 85 00 00 00 8b 01
Jan  7 22:23:24 tailz gdm[7065]: Error reinitilizing server

I have now set back the settings to Preemept Model Server and 100HZ, but I don't know if this is the real problem here.

[startail]

This bug is probebly fixed in the next version of the kernel release...

http://www.kernel.org/pub/linux/kernel/ ... 5-git3.log

[PATCH] x86: change_page_attr() fix

The 'make rodata read-only' patch in -mm exposes a latent bug in the 32-bit
change_page_attr() function, which causes certain CPUs (Those with NX
basically) to reboot instantly after pages are marked read-only.

The same bug got fixed a while back on x86-64, but never got propagated to
i386.

Stuart Hayes from Dell also picked up on this last June, but it never got
fixed, as the only thing affected by it aparently was the nvidia driver.

Blatantly stealing description from his post..

"It doesn't appear to be fixed (in the i386 arch). The
change_page_attr()/split_large_page() code will still still set all the
4K PTEs to PAGE_KERNEL (setting the _PAGE_NX bit) when a large page
needs to be split.

This wouldn't be a problem for the bulk of the kernel memory, but there
are pages in the lower 4MB of memory that's free, and are part of large
executable pages that also contain kernel code. If change_page_attr()
is called on these, it will set the _PAGE_NX bit on the whole 2MB region
that was covered by the large page, causing a large chunk of kernel code
to be non-executable."

Signed-off-by: Arjan van de Ven <arjan@infradead.org>
Signed-off-by: Dave Jones <davej@redhat.com>
Cc: <Stuart_Hayes@Dell.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[Schnacki]

This bug is probebly fixed in the next version of the kernel release...

http://www.kernel.org/pub/linux/kernel/ ... 5-git3.log

[PATCH] x86: change_page_attr() fix

The 'make rodata read-only' patch in -mm exposes a latent bug in the 32-bit
change_page_attr() function, which causes certain CPUs (Those with NX
basically) to reboot instantly after pages are marked read-only.
[...]

Well. I have this pageattr-Error, too. It manifests immediately on boot. I never get to see the X-Server and can't use the console either. ctrl-alt-del works to reboot and login via ssh works too.

The thing is, I've got an Athlon XP 2400+ which doesn't have any NX-capabilities. So I'm not sure whether there's any connection with this report. But the error seems to be similar.

I'm recompiling my kernel with a single possibly relevant change to the faulty config. I had the CC_OPTIMIZE_FOR_SIZE option set to "y" (default) and changed it to "n" now. I'll see if that might help. Every other setting (where applicable) was reused from 2.6.14 with "make oldconfig". :/

Ciao, Jan

[mushroom blue]

hi there.
this can be solved by setting "UseInternalAGPGart" to "no" in your xorg.conf file, and building the appropriate agpgart module in the kernel (I use via-agp for my board). with the latest ati-drivers, 8.22.5 at the time I write this, all is wonderful.

[ponzio]

not for me, nothing is beautiful except my kernel panic...
could you please post your lsmod related to agp, this is mine

Code: Select all

via_agp                 7936  1
agpgart                28848  1 via_agp

and fglrx when in runlevel 5