I'm getting oopses in /proc, I think as a result of running heavily
threaded programs.  I seem to see this in pairs; I'm not sure if they're
related or not:

Unable to handle kernel paging request at virtual address a55a5a5e
 printing eip:
c0209bcf
*pde = 00000000
Oops: 0000 [#1]
CPU:    0
EIP:    0060:[<c0209bcf>]    Not tainted
EFLAGS: 00010286
EIP is at tty_open+0x4bf/0x4e0
eax: a55a5a5a   ebx: c03e8080   ecx: fffffffa   edx: 00008802
esi: d293d3b8   edi: d293d3b8   ebp: cfd0fef8   esp: cfd0febc
ds: 007b   es: 007b   ss: 0068
Process soffice (pid: 4814, threadinfo=cfd0e000 task=d039f800)
Stack: d293d3b4 00000246 c0151fbb d293d3b8 00000098 00000006 00008803 00008802
       00500000 00000000 cfd0fef8 c017f584 c03e8080 d293d3b8 00000000 cfd0ff30
       c017eec6 c16b9dd4 d293d3b8 cfd0ff30 c0172da7 d293d3b8 00000000 00000098
Call Trace:
 [<c0151fbb>] check_poison_obj+0x2b/0x1a0
 [<c017f584>] cdev_get+0x34/0x70
 [<c017eec6>] chrdev_open+0x156/0x3e0
 [<c0172da7>] get_empty_filp+0x77/0x100
 [<c0170c8c>] dentry_open+0x12c/0x1c0
 [<c0170b56>] filp_open+0x66/0x70
 [<c01712a5>] sys_open+0x55/0x90
 [<c010a1ff>] syscall_call+0x7/0xb
 
Code: 8b 70 04 8b 40 08 81 ca 00 08 00 00 89 45 f0 89 57 18 e9 a8
 <1>Unable to handle kernel paging request at virtual address a55a5af2
 printing eip:
c01b31d1
*pde = 00000000
Oops: 0000 [#2]
CPU:    0
EIP:    0060:[<c01b31d1>]    Not tainted
EFLAGS: 00010286
EIP is at proc_pid_stat+0xe1/0x710
eax: dbbd3824   ebx: c030bf3c   ecx: a55a5a5a   edx: c03436f0
esi: 00000001   edi: d4672780   ebp: cfba7f3c   esp: cfba7e2c
ds: 007b   es: 007b   ss: 0068
Process ps (pid: 4822, threadinfo=cfba6000 task=d1cc3200)
Stack: d300833c 3f85c25c 04cadcb0 c0342d60 c0342d60 d0e2cba4 cfba7e74 c01b083d
       d0e2f190 d0e2cba4 0000000d c032ac97 00000004 d4672780 ffffffea fffffff4
       def284c4 def28444 cfba7e98 c0184cb3 def28444 d0e2f190 cec3bafc d0e2f190
Call Trace:
 [<c01b083d>] proc_pident_lookup+0xfd/0x270
 [<c0184cb3>] real_lookup+0xe3/0x110
 [<c018eb80>] dput+0x30/0x620
 [<c01850f5>] do_lookup+0x95/0xa0
 [<c014e78d>] buffered_rmqueue+0xad/0x230
 [<c0151fbb>] check_poison_obj+0x2b/0x1a0
 [<c014e9bc>] __alloc_pages+0xac/0x370
 [<c018655f>] may_open+0x5f/0x1d0
 [<c01ae78d>] proc_info_read+0x5d/0x170
 [<c0171ccf>] vfs_read+0xaf/0x120
 [<c01712c8>] sys_open+0x78/0x90
 [<c0171f5f>] sys_read+0x3f/0x60
 [<c010a1ad>] sysenter_past_esp+0x52/0x71
 
Code: 8b 91 98 00 00 00 89 55 c8 8b 59 04 8b 51 08 0f bf 43 5c 0f
  

I'm running with SLAB_DEBUG enabled, and there's clearly poison values
here.

I'm running soffice under Valgrind, which maintains a separate task for
each application task for running syscalls in.  Apparently soffice is
opening a tty, which oopses.  

When I later ps, it seems to oops as soon as it hits the initial
oowriter process, which is the initial perl script (run under Valgrind
in this case).  In particular, "cat /proc/4806/stat" causes the oops.  
"cat /proc/self/stat" is OK, so it is just specific to this task.. 
/proc/4806/tasks only contains 4806, which indicates the Valgrind proxy
task has terminated.

I also saw this is 2.6.0-test6-mm4.

	J