[OpenAFS] OpenAFS 1.2.10 + kernel 2.4.23

[Holger Naundorf]

Hi,
I have seen the following kernel message on my box after upgrading to the
newest 2.4.23 kernel (needed after the root exploit was announced):

Dec  5 13:38:56 xeon17 kernel: Unable to handle kernel paging request at
virtual address f8a96000
Dec  5 13:38:56 xeon17 kernel:  printing eip:
Dec  5 13:38:56 xeon17 kernel: f8a1e030
Dec  5 13:38:56 xeon17 kernel: *pde =3D 37e75067
Dec  5 13:38:56 xeon17 kernel: *pte =3D 00000000
Dec  5 13:38:56 xeon17 kernel: Oops: 0000
Dec  5 13:38:56 xeon17 kernel: CPU:    0
Dec  5 13:38:56 xeon17 kernel: EIP:    0010:[<f8a1e030>]    Tainted: PF
Dec  5 13:38:56 xeon17 kernel: EFLAGS: 00010206
Dec  5 13:38:56 xeon17 kernel: eax: 00000000   ebx: 00000000
ecx: fffff82b   edx: 11a40000
Dec  5 13:38:56 xeon17 kernel: esi: f8b261b0   edi: f8a96000
ebp: f8a95820   esp: e6813ed8
Dec  5 13:38:56 xeon17 kernel: ds: 0018   es: 0018   ss: 0018
Dec  5 13:38:56 xeon17 kernel: Process tcsh (pid: 4282,
stackpage=3De6813000)
Dec  5 13:38:56 xeon17 kernel: Stack: e6813f48 f8a1d451 f8b261b0 61ae3033
00000241 00000241 f8e2b750 00000000=20
Dec  5 13:38:56 xeon17 kernel:        000007d4 f8b261b0 080b2000 c01180e6
00000096 200011a4 300a6666 63300a31=20
Dec  5 13:38:56 xeon17 kernel:        41cf42d3 00000001 e6812000 e6812000
f6c24c80 080b2000 00030002 d441b880=20
Dec  5 13:38:56 xeon17 kernel: Call Trace:    [<f8a1d451>]
[do_page_fault+422/1243] [filldir64+0/320] [vfs_readdir+132/208]
[filldir64+0/320]
Dec  5 13:38:56 xeon17 kernel: Call Trace:    [<f8a1d451>] [<c01180e6>]
[<c014de80>] [<c014d844>] [<c014de80>]
Dec  5 13:38:56 xeon17 kernel:   [sys_getdents64+79/188] [filldir64+0/320]
[system_call+51/56]
Dec  5 13:38:56 xeon17 kernel:   [<c014e00f>] [<c014de80>] [<c0108c23>]
Dec  5 13:38:56 xeon17 kernel:=20
Dec  5 13:38:56 xeon17 kernel: Code: f2 ae f7 d1 49 89 4c 24 20 8b 86 44
01 00 00 89 44 24 30 8b=20

The address points to somewhere within the AFS module

/proc/ksyms:

f8a16f20 __insmod_libafs-2.4.23.mp_S.rodata_L2232       [libafs-2.4.23.mp]

is the closest I see...

The symptoms are that the machine grinds to a near complete stop for
anything trying to start new processes (I could still surf inside an open
Mozilla, but opening new windows, running top, trying to log into the box
as root on the console failed...) and the load of the box conmtinued to
creep up.

I am not enough of a kernel/AFS hacker to say what this is (ant I haven't
been able to reproduce it in the last few hours) so if anyone has seen
something like this before and could comment on this - I would be really
grateful...

Thanks in advance

Holger N.

--=20
Holger Naundorf                            naundorf@aei.mpg.de
MPI f=FCr Gravitationsphysik                 Phone (++49)(+331)567-7364=20
Albert-Einstein-Institut                   Fax   (++49)(+331)567-7298=20