XFS oops on loading the module

XFS oops on loading the module

[martin f krafft]

[-- Attachment #1: Type: text/plain, Size: 1893 bytes --]

I just tried to mount an XFS filesystem on this AMD K6 machine,
booted with the 2.6.8 kernel for FAI
(http://www.informatik.uni/koeln.de/fai) (let me know if you need
any information about it), and modprobe segfaults with a kernel bug.
Have you seen this before? Thanks!

sh-2.05b# modprobe xfs
Segmentation fault
------------[ cut here ]------------
kernel BUG at kernel/module.c:264!
invalid operand: 0000 [#1]
SMP 
Modules linked in: ext3 jbd mbcache sr_mod sd_mod scsi_mod ide_generic usbmouse usbhid ide_cd cdrom usbkbd floppy rtc via82cxxx slc90e66 sis5513 siimage serverworks rz1000 piix pdc202xx_old pdc202xx_new hpt366 ide_disk hpt34x cs5530 cmd64x amd74xx alim15x3 aec62xx ide_core uhci_hcd usbcore
CPU:    0
EIP:    0060:[<c012bcf6>]    Not tainted
EFLAGS: 00010202   (2.6.8-fai) 
EIP is at percpu_modalloc+0xe/0xf8
eax: 0000004b   ebx: e09f9400   ecx: e09f940c   edx: 0000000f
esi: e09f84c4   edi: 00000258   ebp: e09fa9c8   esp: d9963f0c
ds: 007b   es: 007b   ss: 0068
Process modprobe (pid: 1272, threadinfo=d9962000 task=dfb50650)
Stack: e09f9400 e09f84c4 00000258 e09fa9c8 c012d8be e098f000 c012d8ed 00000148 
       00000020 40157000 080509b8 c031f504 d9962000 c416c0a0 00000044 00000060 
       df0ad4e0 00000000 00000000 e09f9400 0000000f 00000000 00000000 00000000 
Call Trace:
 [<c012d8be>] load_module+0x3c6/0x904
 [<c012d8ed>] load_module+0x3f5/0x904
 [<c012de59>] sys_init_module+0x5d/0x200
 [<c0105d5f>] syscall_call+0x7/0xb
Code: 0f 0b 08 01 6f a5 2b c0 89 f6 bd a0 10 40 c0 31 f6 a1 2c 64 

-- 
martin;              (greetings from the heart of the sun.)
  \____ echo mailto: !#^."<*>"|tr "<*> mailto:" net@madduck
 
invalid/expired pgp subkeys? use subkeys.pgp.net as keyserver!
spamtraps: madduck.bogus@madduck.net
 
because light travels faster than sound,
some people appear to be intelligent,
until you hear them speak.

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: XFS oops on loading the module

[Nathan Scott]

Hi there,

On Sat, Oct 16, 2004 at 06:50:58PM +0200, martin f krafft wrote:
> I just tried to mount an XFS filesystem on this AMD K6 machine,
> booted with the 2.6.8 kernel for FAI
> (http://www.informatik.uni/koeln.de/fai) (let me know if you need
> any information about it), and modprobe segfaults with a kernel bug.
> Have you seen this before? Thanks!
> 
> sh-2.05b# modprobe xfs
> Segmentation fault
> ------------[ cut here ]------------
> kernel BUG at kernel/module.c:264!

IOW... this, I guess:

        for (i = 0; i < pcpu_num_used; ptr += block_size(pcpu_size[i]), i++) {
                /* Extra for alignment requirement. */
                extra = ALIGN((unsigned long)ptr, align) - (unsigned long)ptr;
                BUG_ON(i == 0 && extra != 0);

> Modules linked in: ext3 jbd mbcache sr_mod sd_mod scsi_mod ide_generic usbmouse usbhid ide_cd cdrom usbkbd floppy rtc via82cxxx slc90e66 sis5513 siimage serverworks rz1000 piix pdc202xx_old pdc202xx_new hpt366 ide_disk hpt34x cs5530 cmd64x amd74xx alim15x3 aec62xx ide_core uhci_hcd usbcore
> CPU:    0
> EIP:    0060:[<c012bcf6>]    Not tainted
> EFLAGS: 00010202   (2.6.8-fai) 
> EIP is at percpu_modalloc+0xe/0xf8
> eax: 0000004b   ebx: e09f9400   ecx: e09f940c   edx: 0000000f
> esi: e09f84c4   edi: 00000258   ebp: e09fa9c8   esp: d9963f0c
> ds: 007b   es: 007b   ss: 0068
> Process modprobe (pid: 1272, threadinfo=d9962000 task=dfb50650)
> Stack: e09f9400 e09f84c4 00000258 e09fa9c8 c012d8be e098f000 c012d8ed 00000148 
>        00000020 40157000 080509b8 c031f504 d9962000 c416c0a0 00000044 00000060 
>        df0ad4e0 00000000 00000000 e09f9400 0000000f 00000000 00000000 00000000 
> Call Trace:
>  [<c012d8be>] load_module+0x3c6/0x904
>  [<c012d8ed>] load_module+0x3f5/0x904
>  [<c012de59>] sys_init_module+0x5d/0x200
>  [<c0105d5f>] syscall_call+0x7/0xb
> Code: 0f 0b 08 01 6f a5 2b c0 89 f6 bd a0 10 40 c0 31 f6 a1 2c 64 
> 

I haven't come across this before - the only percpu variables
in XFS are the performance stats.  I've CC'd a couple of people
who may have more of a clue as to whats gone wrong here.

cheers.

-- 
Nathan

Re: XFS oops on loading the module

[Rusty Russell]

On Tue, 2004-10-19 at 10:49, Nathan Scott wrote:
> Hi there,
> 
> On Sat, Oct 16, 2004 at 06:50:58PM +0200, martin f krafft wrote:
> > I just tried to mount an XFS filesystem on this AMD K6 machine,
> > booted with the 2.6.8 kernel for FAI
> > (http://www.informatik.uni/koeln.de/fai) (let me know if you need
> > any information about it), and modprobe segfaults with a kernel bug.
> > Have you seen this before? Thanks!
> > 
> > sh-2.05b# modprobe xfs
> > Segmentation fault
> > ------------[ cut here ]------------
> > kernel BUG at kernel/module.c:264!
> 
> IOW... this, I guess:
> 
>         for (i = 0; i < pcpu_num_used; ptr += block_size(pcpu_size[i]), i++) {
>                 /* Extra for alignment requirement. */
>                 extra = ALIGN((unsigned long)ptr, align) - (unsigned long)ptr;
>                 BUG_ON(i == 0 && extra != 0);

Somehow, your per-cpu section was not aligned sufficiently for this
alignment request.  Since you passed the BUG_ON(align > SMP_CACHE_BYTES)
above, it implies that your per-cpu section isn't correctly aligned.

Does this fix it?

Name: Align per-cpu Section Correctly
Status: Untested
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

The per-cpu section must be aligned to a cacheline: it's currently
aligned to a hardcoded 32 bytes on x86, which is wrong in some
configs.  On x86 the .data.cacheline_aligned is also 32-byte aligned,
which seems wrong too.

We can't use SMP_CACHE_BYTES because we can't include linux/cache.h
from asm.

Other archs need similar changes.

diff -urpN --exclude TAGS -X /home/rusty/devel/kernel/kernel-patches/current-dontdiff --minimal .26520-linux-2.6.9-rc4-bk2/arch/i386/kernel/vmlinux.lds.S .26520-linux-2.6.9-rc4-bk2.updated/arch/i386/kernel/vmlinux.lds.S
--- .26520-linux-2.6.9-rc4-bk2/arch/i386/kernel/vmlinux.lds.S	2004-10-11 15:16:58.000000000 +1000
+++ .26520-linux-2.6.9-rc4-bk2.updated/arch/i386/kernel/vmlinux.lds.S	2004-10-19 12:45:54.000000000 +1000
@@ -5,6 +5,7 @@
 #include <asm-generic/vmlinux.lds.h>
 #include <asm/thread_info.h>
 #include <asm/page.h>
+#include <asm/cache.h>
 
 OUTPUT_FORMAT("elf32-i386", "elf32-i386", "elf32-i386")
 OUTPUT_ARCH(i386)
@@ -47,7 +48,7 @@ SECTIONS
   . = ALIGN(4096);
   .data.page_aligned : { *(.data.idt) }
 
-  . = ALIGN(32);
+  . = ALIGN(L1_CACHE_BYTES);
   .data.cacheline_aligned : { *(.data.cacheline_aligned) }
 
   _edata = .;			/* End of data section */
@@ -96,7 +97,7 @@ SECTIONS
   __initramfs_start = .;
   .init.ramfs : { *(.init.ramfs) }
   __initramfs_end = .;
-  . = ALIGN(32);
+  . = ALIGN(L1_CACHE_BYTES);
   __per_cpu_start = .;
   .data.percpu  : { *(.data.percpu) }
   __per_cpu_end = .;

-- 
Anyone who quotes me in their signature is an idiot -- Rusty Russell