I'm seeing an intermittent oops on a heavily loaded SMP system (Compaq DL360 G2). I've read the messages suggesting this is related to the nvidia driver, but I don't have that loaded.
This is the vanilla 2.4.19 kernel, plus Intel's e1000.o driver and the ipvs patch.
Here's the ksymoops output:
kernel BUG at page_alloc.c:220!
invalid operand: 0000
CPU: 1
EIP: 0010:[rmqueue+509/592] Not tainted
EIP: 0010:[<c013425d>] Not tainted
Using defaults from ksymoops -t elf32-i386 -a i386
EFLAGS: 00010202
eax: 00000040 ebx: c118c2c0 ecx: 00001000 edx: 0000840e
esi: c027b174 edi: 0000eff0 ebp: c1000020 esp: cad79e3c
ds: 0018 es: 0018 ss: 0018
Process nagios (pid: 27141, stackpage=cad79000)
Stack: 00001000 0000740e 00000296 00000000 c027b174 c027b300
000001ff 00000000
c119af00 c01344f1 c027b174 c027b2fc 000001d2 ffffffff
00000000 0819afd4
00000001 c119af00 c0128fdb c012967c cf0490e0 0807a000
00000000 c3809b40
Call Trace: [__alloc_pages+65/384] [do_wp_page+187/672]
[do_no_page+92/528] [handle_mm_fault+141/208] [do_no
Call Trace: [<c01344f1>] [<c0128fdb>]
[<c012967c>] [<c01298bd>] [<c0123f36>]
[<c0116532>] [<c011d8b0>] [<c011dc9b>]
[<c0139d34>] [<c0116380>] [<c0108c0c>]
Code: 0f 0b dc 00 33 e8 23 c0 8b 43 18 a9 80 00 00 00 74 08 0f
0b
>>EIP; c013425d <rmqueue+1fd/250> <=====
Trace; c01344f1 <__alloc_pages+41/180>
Trace; c0128fdb <do_wp_page+bb/2a0>
Trace; c012967c <do_no_page+5c/210>
Trace; c01298bd <handle_mm_fault+8d/d0>
Trace; c0123f36 <do_notify_parent+a6/b0>
Trace; c0116532 <do_page_fault+1b2/4fb>
Trace; c011d8b0 <exit_notify+1f0/330>
Trace; c011dc9b <do_exit+2ab/2c0>
Trace; c0139d34 <filp_close+94/a0>
Trace; c0116380 <do_page_fault+0/4fb>
Trace; c0108c0c <error_code+34/3c>
Code; c013425d <rmqueue+1fd/250>
00000000 <_EIP>:
Code; c013425d <rmqueue+1fd/250> <=====
0: 0f 0b ud2a <=====
Code; c013425f <rmqueue+1ff/250>
2: dc 00 faddl (%eax)
Code; c0134261 <rmqueue+201/250>
4: 33 e8 xor %eax,%ebp
Code; c0134263 <rmqueue+203/250>
6: 23 c0 and %eax,%eax
Code; c0134265 <rmqueue+205/250>
8: 8b 43 18 mov 0x18(%ebx),%eax
Code; c0134268 <rmqueue+208/250>
b: a9 80 00 00 00 test $0x80,%eax
Code; c013426d <rmqueue+20d/250>
10: 74 08 je 1a <_EIP+0x1a>
c0134277 <rmqueue+217/250>
Code; c013426f <rmqueue+20f/250>
12: 0f 0b ud2a
kernel BUG at page_alloc.c:220!
invalid operand: 0000
Vijay Gadad
On 02:39, venerd? 18 ottobre 2002, Gadad, Vijay wrote:
> I'm seeing an intermittent oops on a heavily loaded SMP system (Compaq
> DL360 G2). I've read the messages suggesting this is related to the nvidia
> driver, but I don't have that loaded.
>
> This is the vanilla 2.4.19 kernel, plus Intel's e1000.o driver and the ipvs
> patch.
>
> Here's the ksymoops output:
>
>
>
>
> kernel BUG at page_alloc.c:220!
> invalid operand: 0000
> CPU: 1
> EIP: 0010:[rmqueue+509/592] Not tainted
this may be a patch:
[albert@yoda albert]$ diff -ruN page_alloc.c new_page_alloc.c
--- page_alloc.c
+++ new_page_alloc.c
@@ -167,7 +167,7 @@
#define MARK_USED(index, order, area) \
__change_bit((index) >> (1+(order)), (area)->map)
-static inline struct page * expand (zone_t *zone, struct page *page,
+static inline struct page * expand(zone_t *zone, struct page *page,
unsigned long index, int low, int high, free_area_t * area)
{
unsigned long size = 1 << high;
@@ -215,7 +215,6 @@
zone->free_pages -= 1UL << order;
page = expand(zone, page, index, order, curr_order,
area);
- spin_unlock_irqrestore(&zone->lock, flags);
set_page_count(page, 1);
if (BAD_RANGE(zone,page))
@@ -224,6 +223,7 @@
BUG();
if (PageActive(page))
BUG();
+ spin_unlock_irqrestore(&zone->lock, flags);
return page;
}
curr_order++;
__
Emiliano 'AlberT' Gabrielli