* kernel BUG at rmap.c:409! with 2.5.31 and akpm patches.
@ 2002-08-19 20:54 Steven Cole
2002-08-19 21:21 ` Andrew Morton
0 siblings, 1 reply; 6+ messages in thread
From: Steven Cole @ 2002-08-19 20:54 UTC (permalink / raw)
To: Andrew Morton; +Cc: linux-mm, Rik van Riel, Steven Cole
Here's a new one.
With this patch applied to 2.5.31,
http://www.zip.com.au/~akpm/linux/patches/2.5/2.5.31/stuff-sent-to-linus/everything.gz
I got this BUG:
kernel BUG at rmap.c:409!
while running dbench 40 as a stress test.
The filesystem on which dbench was being run was mounted as ext3.
The box is dual p3, scsi.
This test box got its root filesystem (then ext2) destroyed during
testing last week, so I loaded RH 7.3 on it this morning and made all
partitions ext3. Now it's ready for more abuse.
Steven
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: kernel BUG at rmap.c:409! with 2.5.31 and akpm patches.
2002-08-19 20:54 kernel BUG at rmap.c:409! with 2.5.31 and akpm patches Steven Cole
@ 2002-08-19 21:21 ` Andrew Morton
2002-08-19 22:04 ` Steven Cole
0 siblings, 1 reply; 6+ messages in thread
From: Andrew Morton @ 2002-08-19 21:21 UTC (permalink / raw)
To: Steven Cole; +Cc: linux-mm, Rik van Riel, Steven Cole
Steven Cole wrote:
>
> Here's a new one.
>
> With this patch applied to 2.5.31,
> http://www.zip.com.au/~akpm/linux/patches/2.5/2.5.31/stuff-sent-to-linus/everything.gz
>
> I got this BUG:
> kernel BUG at rmap.c:409!
> while running dbench 40 as a stress test.
>
OK, ext3's habit of leaving buffers attached to truncated pages
seems to have tripped us up:
if (page->pte.chain && !page->mapping && !PagePrivate(page)) {
...
}
if (page->pte.chain) {
switch (try_to_unmap(page)) {
So if the page has a pte_chain, and no ->mapping, but has buffers
we go blam.
vmscan.c | 5 +++--
1 files changed, 3 insertions, 2 deletions
--- 2.5.31/mm/vmscan.c~foo Mon Aug 19 14:20:55 2002
+++ 2.5.31-akpm/mm/vmscan.c Mon Aug 19 14:21:11 2002
@@ -146,11 +146,13 @@ shrink_list(struct list_head *page_list,
rmap_lock = lock_rmap(page);
}
+ mapping = page->mapping;
+
/*
* The page is mapped into the page tables of one or more
* processes. Try to unmap it here.
*/
- if (page->pte.chain) {
+ if (page->pte.chain && mapping) {
switch (try_to_unmap(page)) {
case SWAP_ERROR:
case SWAP_FAIL:
@@ -164,7 +166,6 @@ shrink_list(struct list_head *page_list,
}
}
unlock_rmap(rmap_lock);
- mapping = page->mapping;
/*
* FIXME: this is CPU-inefficient for shared mappings.
.
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: kernel BUG at rmap.c:409! with 2.5.31 and akpm patches.
2002-08-19 21:21 ` Andrew Morton
@ 2002-08-19 22:04 ` Steven Cole
2002-08-20 13:39 ` Steven Cole
0 siblings, 1 reply; 6+ messages in thread
From: Steven Cole @ 2002-08-19 22:04 UTC (permalink / raw)
To: Andrew Morton; +Cc: linux-mm, Rik van Riel
On Mon, 2002-08-19 at 15:21, Andrew Morton wrote:
> Steven Cole wrote:
> >
> > Here's a new one.
> >
> > With this patch applied to 2.5.31,
> > http://www.zip.com.au/~akpm/linux/patches/2.5/2.5.31/stuff-sent-to-linus/everything.gz
> >
> > I got this BUG:
> > kernel BUG at rmap.c:409!
> > while running dbench 40 as a stress test.
> >
>
> OK, ext3's habit of leaving buffers attached to truncated pages
> seems to have tripped us up:
>
> if (page->pte.chain && !page->mapping && !PagePrivate(page)) {
> ...
> }
>
> if (page->pte.chain) {
> switch (try_to_unmap(page)) {
>
> So if the page has a pte_chain, and no ->mapping, but has buffers
> we go blam.
[patch snipped]
Patch applied, running dbench 1..128. Up to 52 clients so far, and no
blam yet. I'll run this test several times overnight and let you know
if anything else falls out.
Thanks,
Steven
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: kernel BUG at rmap.c:409! with 2.5.31 and akpm patches.
2002-08-19 22:04 ` Steven Cole
@ 2002-08-20 13:39 ` Steven Cole
2002-08-20 17:07 ` Andrew Morton
0 siblings, 1 reply; 6+ messages in thread
From: Steven Cole @ 2002-08-20 13:39 UTC (permalink / raw)
To: Andrew Morton; +Cc: linux-mm, Rik van Riel
On Mon, 2002-08-19 at 16:04, Steven Cole wrote:
> On Mon, 2002-08-19 at 15:21, Andrew Morton wrote:
> > Steven Cole wrote:
> > >
> > > Here's a new one.
> > >
> > > With this patch applied to 2.5.31,
> > > http://www.zip.com.au/~akpm/linux/patches/2.5/2.5.31/stuff-sent-to-linus/everything.gz
> > >
[earlier problem snipped]
>
> [patch snipped]
>
> Patch applied, running dbench 1..128. Up to 52 clients so far, and no
> blam yet. I'll run this test several times overnight and let you know
> if anything else falls out.
Something else fell out. I got kernel BUG at page_alloc.c:98! three
times. I ran the output of dmesg through ksymoops, and here is the
result. The traceback for the third occurance was identical to that of
the first, so that was snipped.
Steven
ksymoops 2.4.4 on i686 2.5.31. Options used
-v linux-2.5.31-akpm/vmlinux (specified)
-K (specified)
-L (specified)
-O (specified)
-m linux-2.5.31-akpm/System.map (specified)
kernel BUG at page_alloc.c:98!
invalid operand: 0000
CPU: 1
EIP: 0010:[<c0132733>] Not tainted
Using defaults from ksymoops -t elf32-i386 -a i386
EFLAGS: 00010282
eax: f78b4fe0 ebx: c19bf730 ecx: 00000000 edx: 00000000
esi: d9f03e70 edi: 00000005 ebp: 00000010 esp: d9f03e14
ds: 0018 es: 0018 ss: 0018
Stack: 00007ff0 c100001c c1a79c88 c19a001c c02a710c 00000203 fffffffe 000013cc
00000000 00000001 d9f03e70 00000005 00000010 c013315a c1926414 00000002
c0131059 d9f03e6c f7944c2b f7944ce0 f72a6ac0 00000000 00000002 c19d69ec
Call Trace: [<c013315a>] [<c0131059>] [<c0134190>] [<c0134355>] [<c0129ff1>]
[<c0115cd0>] [<c0117ab8>] [<c011adbf>] [<c0115c75>] [<c0108ce3>]
Code: 0f 0b 62 00 a5 f7 26 c0 8b 03 ba 04 00 00 00 83 e0 10 74 1d
>>EIP; c0132733 <__free_pages_ok+93/310> <=====
Trace; c013315a <__pagevec_free+1a/20>
Trace; c0131059 <__pagevec_release+f9/110>
Trace; c0134190 <swap_free+20/40>
Trace; c0134355 <remove_exclusive_swap_page+d5/110>
Trace; c0129ff1 <exit_mmap+1a1/280>
Trace; c0115cd0 <default_wake_function+0/40>
Trace; c0117ab8 <mmput+48/70>
Trace; c011adbf <do_exit+df/2c0>
Trace; c0115c75 <schedule+325/380>
Trace; c0108ce3 <syscall_call+7/b>
Code; c0132733 <__free_pages_ok+93/310>
00000000 <_EIP>:
Code; c0132733 <__free_pages_ok+93/310> <=====
0: 0f 0b ud2a <=====
Code; c0132735 <__free_pages_ok+95/310>
2: 62 00 bound %eax,(%eax)
Code; c0132737 <__free_pages_ok+97/310>
4: a5 movsl %ds:(%esi),%es:(%edi)
Code; c0132738 <__free_pages_ok+98/310>
5: f7 26 mull (%esi)
Code; c013273a <__free_pages_ok+9a/310>
7: c0 8b 03 ba 04 00 00 rorb $0x0,0x4ba03(%ebx)
Code; c0132741 <__free_pages_ok+a1/310>
e: 00 83 e0 10 74 1d add %al,0x1d7410e0(%ebx)
kernel BUG at page_alloc.c:98!
invalid operand: 0000
CPU: 1
EIP: 0010:[<c0132733>] Not tainted
EFLAGS: 00010286
eax: f7933860 ebx: c1a308dc ecx: 00000000 edx: 00000000
esi: e841fde0 edi: 00000005 ebp: 00000010 esp: e841fd84
ds: 0018 es: 0018 ss: 0018
Stack: c74cf1a0 c0308e40 c1b1e120 c19a001c c02a7100 00000202 ffffffff 0000289e
00000000 00000000 e841fde0 00000005 00000010 c013315a c1a812c8 00000010
c0131059 e841fddc dd0e2904 c0173df6 c013e3db de4ce960 00000010 c1a308dc
Call Trace: [<c013315a>] [<c0131059>] [<c0173df6>] [<c013e3db>] [<c012a32f>]
[<c012a783>] [<c016c6a9>] [<c016c701>] [<c012a90d>] [<c015242d>] [<c015268d>]
[<c0150f46>] [<c014719d>] [<c0149723>] [<c01482e2>] [<c01497c9>] [<c013c89d>]
[<c0108ce3>]
Code: 0f 0b 62 00 a5 f7 26 c0 8b 03 ba 04 00 00 00 83 e0 10 74 1d
>>EIP; c0132733 <__free_pages_ok+93/310> <=====
Trace; c013315a <__pagevec_free+1a/20>
Trace; c0131059 <__pagevec_release+f9/110>
Trace; c0173df6 <journal_unmap_buffer+106/190>
Trace; c013e3db <wake_up_buffer+b/30>
Trace; c012a32f <remove_from_page_cache+2f/40>
Trace; c012a783 <truncate_list_pages+2b3/350>
Trace; c016c6a9 <ext3_do_update_inode+2c9/350>
Trace; c016c701 <ext3_do_update_inode+321/350>
Trace; c012a90d <truncate_inode_pages+8d/d0>
Trace; c015242d <generic_delete_inode+5d/140>
Trace; c015268d <iput+5d/60>
Trace; c0150f46 <d_delete+66/c0>
Trace; c014719d <permission+3d/50>
Trace; c0149723 <vfs_unlink+1b3/1d0>
Trace; c01482e2 <lookup_hash+42/90>
Trace; c01497c9 <sys_unlink+89/f0>
Trace; c013c89d <sys_close+5d/70>
Trace; c0108ce3 <syscall_call+7/b>
Code; c0132733 <__free_pages_ok+93/310>
00000000 <_EIP>:
Code; c0132733 <__free_pages_ok+93/310> <=====
0: 0f 0b ud2a <=====
Code; c0132735 <__free_pages_ok+95/310>
2: 62 00 bound %eax,(%eax)
Code; c0132737 <__free_pages_ok+97/310>
4: a5 movsl %ds:(%esi),%es:(%edi)
Code; c0132738 <__free_pages_ok+98/310>
5: f7 26 mull (%esi)
Code; c013273a <__free_pages_ok+9a/310>
7: c0 8b 03 ba 04 00 00 rorb $0x0,0x4ba03(%ebx)
Code; c0132741 <__free_pages_ok+a1/310>
e: 00 83 e0 10 74 1d add %al,0x1d7410e0(%ebx)
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: kernel BUG at rmap.c:409! with 2.5.31 and akpm patches.
2002-08-20 13:39 ` Steven Cole
@ 2002-08-20 17:07 ` Andrew Morton
2002-08-20 17:17 ` Steven Cole
0 siblings, 1 reply; 6+ messages in thread
From: Andrew Morton @ 2002-08-20 17:07 UTC (permalink / raw)
To: Steven Cole; +Cc: linux-mm, Rik van Riel
Steven Cole wrote:
>
> On Mon, 2002-08-19 at 16:04, Steven Cole wrote:
> > On Mon, 2002-08-19 at 15:21, Andrew Morton wrote:
> > > Steven Cole wrote:
> > > >
> > > > Here's a new one.
> > > >
> > > > With this patch applied to 2.5.31,
> > > > http://www.zip.com.au/~akpm/linux/patches/2.5/2.5.31/stuff-sent-to-linus/everything.gz
> > > >
> [earlier problem snipped]
> >
> > [patch snipped]
> >
> > Patch applied, running dbench 1..128. Up to 52 clients so far, and no
> > blam yet. I'll run this test several times overnight and let you know
> > if anything else falls out.
>
> Something else fell out. I got kernel BUG at page_alloc.c:98! three
> times.
That's the infamous non-NULL page->pte.chain.
> ...
>
> >>EIP; c0132733 <__free_pages_ok+93/310> <=====
> Trace; c013315a <__pagevec_free+1a/20>
> Trace; c0131059 <__pagevec_release+f9/110>
> Trace; c0134190 <swap_free+20/40>
> Trace; c0134355 <remove_exclusive_swap_page+d5/110>
> Trace; c0129ff1 <exit_mmap+1a1/280>
> Trace; c0115cd0 <default_wake_function+0/40>
> Trace; c0117ab8 <mmput+48/70>
> Trace; c011adbf <do_exit+df/2c0>
> Trace; c0115c75 <schedule+325/380>
OK, it was mapped.
> ...
> >>EIP; c0132733 <__free_pages_ok+93/310> <=====
> Trace; c013315a <__pagevec_free+1a/20>
> Trace; c0131059 <__pagevec_release+f9/110>
> Trace; c0173df6 <journal_unmap_buffer+106/190>
> Trace; c013e3db <wake_up_buffer+b/30>
> Trace; c012a32f <remove_from_page_cache+2f/40>
> Trace; c012a783 <truncate_list_pages+2b3/350>
> Trace; c016c6a9 <ext3_do_update_inode+2c9/350>
> Trace; c016c701 <ext3_do_update_inode+321/350>
> Trace; c012a90d <truncate_inode_pages+8d/d0>
> Trace; c015242d <generic_delete_inode+5d/140>
> Trace; c015268d <iput+5d/60>
> Trace; c0150f46 <d_delete+66/c0>
> Trace; c014719d <permission+3d/50>
> Trace; c0149723 <vfs_unlink+1b3/1d0>
> Trace; c01482e2 <lookup_hash+42/90>
> Trace; c01497c9 <sys_unlink+89/f0>
> Trace; c013c89d <sys_close+5d/70>
Odd. Was this just running dbench? If so, odd. dbench
doesn't mmap files, yet here we seem to have a truncated,
mapped page.
I wonder where that came from?
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/
^ permalink raw reply [flat|nested] 6+ messages in thread
* Re: kernel BUG at rmap.c:409! with 2.5.31 and akpm patches.
2002-08-20 17:07 ` Andrew Morton
@ 2002-08-20 17:17 ` Steven Cole
0 siblings, 0 replies; 6+ messages in thread
From: Steven Cole @ 2002-08-20 17:17 UTC (permalink / raw)
To: Andrew Morton; +Cc: linux-mm, Rik van Riel
On Tue, 2002-08-20 at 11:07, Andrew Morton wrote:
> > ...
> > >>EIP; c0132733 <__free_pages_ok+93/310> <=====
> > Trace; c013315a <__pagevec_free+1a/20>
> > Trace; c0131059 <__pagevec_release+f9/110>
> > Trace; c0173df6 <journal_unmap_buffer+106/190>
> > Trace; c013e3db <wake_up_buffer+b/30>
> > Trace; c012a32f <remove_from_page_cache+2f/40>
> > Trace; c012a783 <truncate_list_pages+2b3/350>
> > Trace; c016c6a9 <ext3_do_update_inode+2c9/350>
> > Trace; c016c701 <ext3_do_update_inode+321/350>
> > Trace; c012a90d <truncate_inode_pages+8d/d0>
> > Trace; c015242d <generic_delete_inode+5d/140>
> > Trace; c015268d <iput+5d/60>
> > Trace; c0150f46 <d_delete+66/c0>
> > Trace; c014719d <permission+3d/50>
> > Trace; c0149723 <vfs_unlink+1b3/1d0>
> > Trace; c01482e2 <lookup_hash+42/90>
> > Trace; c01497c9 <sys_unlink+89/f0>
> > Trace; c013c89d <sys_close+5d/70>
>
> Odd. Was this just running dbench? If so, odd. dbench
> doesn't mmap files, yet here we seem to have a truncated,
> mapped page.
>
> I wonder where that came from?
Don't know. httpd, smb, and nfs are all off.
Only dbench as far as I can tell.
The clock was off by 8 hours (I just fixed that), but
it happened last night when the system thought it was 0530 this morning,
which was 2130 last night, so nothing else was happening.
[root@spc5 log]# grep BUG messages
Aug 20 01:24:15 spc5 kernel: kernel BUG at page_alloc.c:98!
Aug 20 05:30:13 spc5 kernel: kernel BUG at page_alloc.c:98!
Aug 20 05:30:16 spc5 kernel: kernel BUG at page_alloc.c:98!
crontab is standard RH 7.3:
[root@spc5 etc]# cat crontab
SHELL=/bin/bash
PATH=/sbin:/bin:/usr/sbin:/usr/bin
MAILTO=root
HOME=/
# run-parts
01 * * * * root run-parts /etc/cron.hourly
02 4 * * * root run-parts /etc/cron.daily
22 4 * * 0 root run-parts /etc/cron.weekly
42 4 1 * * root run-parts /etc/cron.monthly
0-59/5 * * * * root /usr/bin/mrtg /etc/mrtg/mrtg.cfg
Steven
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org. For more info on Linux MM,
see: http://www.linux-mm.org/
^ permalink raw reply [flat|nested] 6+ messages in thread
end of thread, other threads:[~2002-08-20 17:17 UTC | newest]
Thread overview: 6+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2002-08-19 20:54 kernel BUG at rmap.c:409! with 2.5.31 and akpm patches Steven Cole
2002-08-19 21:21 ` Andrew Morton
2002-08-19 22:04 ` Steven Cole
2002-08-20 13:39 ` Steven Cole
2002-08-20 17:07 ` Andrew Morton
2002-08-20 17:17 ` Steven Cole
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox