page_alloc bug in 2.4.17-pre8

page_alloc bug in 2.4.17-pre8

[Mike Black]

This machine had been up for 2-1/2 days and had run this backup (afio) twice successfully.

Here's line 108 of page_alloc.c:
        if (PageLRU(page))
                BUG();

Hopefully this doesn't indicate a CPU problem?  The power supply on this thing blew Saturday but has run OK until now.

May 22 00:51:01 picard kernel: kernel BUG at page_alloc.c:108!
May 22 00:51:01 picard kernel: invalid operand: 0000
May 22 00:51:01 picard kernel: CPU:    1
May 22 00:51:01 picard kernel: EIP:    0010:[swap_duplicate+82/192]    Not tainted
May 22 00:51:01 picard kernel: EFLAGS: 00010286
May 22 00:51:01 picard kernel: eax: 01000009   ebx: c176aed0   ecx: c176aed0   edx: cfd67690
May 22 00:51:01 picard kernel: esi: 00000000   edi: cfd67690   ebp: 000000e3   esp: ce55fef8
May 22 00:51:01 picard kernel: ds: 0018   es: 0018   ss: 0018
May 22 00:51:01 picard kernel: Process afio (pid: 15467, stackpage=ce55f000)
May 22 00:51:01 picard kernel: Stack: c176aed0 00001000 cfd67690 000000e3 00008000 f09fb2ac 00015f90 00000000
May 22 00:51:01 picard kernel:        c176aed0 c176aed0 efd5ae28 cfd67690 000000e2 c0131858 c012a04d ce55ff8c
May 22 00:51:01 picard kernel:        c176aed0 00000000 00001000 00000000 dd3318a0 ffffffea 00008000 00001000
May 22 00:51:01 picard kernel: Call Trace: [<f09fb2ac>] [shmem_truncate+172/524] [generic_file_write+1781/1856] [mprotect_fixup+270/
1176] [change_protection+268/348]
May 22 00:51:01 picard kernel:    [create_empty_buffers+75/80] [probe_irq_on+15/292]
May 22 00:51:01 picard kernel:
May 22 00:51:01 picard kernel: Code: 0f 0b 6c 00 93 b3 24 c0 89 d8 2b 05 f0 70 30 c0 69 c0 ab aa

Michael D. Black mblack@csihq.com
http://www.csihq.com/
http://www.csihq.com/~mike
321-676-2923, x203
Melbourne FL

Re: page_alloc bug in 2.4.17-pre8

[Hugh Dickins]

On Thu, 23 May 2002, Mike Black wrote:
> This machine had been up for 2-1/2 days and had run this backup (afio) twice successfully.
> 
> Here's line 108 of page_alloc.c:
>         if (PageLRU(page))
>                 BUG();
> 
> Hopefully this doesn't indicate a CPU problem?  The power supply on this thing blew Saturday but has run OK until now.
> 
> May 22 00:51:01 picard kernel: kernel BUG at page_alloc.c:108!
> May 22 00:51:01 picard kernel: invalid operand: 0000
> May 22 00:51:01 picard kernel: CPU:    1
> May 22 00:51:01 picard kernel: EIP:    0010:[swap_duplicate+82/192]    Not tainted

There were quite a number of reports of those PageLRU BUGs on 2.4.17.
No idea what fixed them, but 2.4.18 (and 2.4.19-pre) has seemed free
of them (Ben LaHaise made a plausible change, but closer analysis
suggested it couldn't really be the fix).  Suggest you upgrade.

Your oops report, by the way, must have been using the wrong System.map:
page_alloc.c:108 is in __free_pages_ok(), swap_duplicate() is over in
swapfile.c.  But no matter, page_alloc.c:108 identifies it well enough.

Hugh 

Re: page_alloc bug in 2.4.17-pre8

[Mike Black]

Oops -- got the wrong version (head was in the wrong kernel space).
This occurred on 2.4.19-pre8

----- Original Message ----- 
From: "Hugh Dickins" <hugh@veritas.com>
To: "Mike Black" <mblack@csihq.com>
Cc: "linux-kernel" <linux-kernel@vger.kernel.org>
Sent: Thursday, May 23, 2002 8:12 AM
Subject: Re: page_alloc bug in 2.4.17-pre8


> On Thu, 23 May 2002, Mike Black wrote:
> > This machine had been up for 2-1/2 days and had run this backup (afio) twice successfully.
> > 
> > Here's line 108 of page_alloc.c:
> >         if (PageLRU(page))
> >                 BUG();
> > 
> > Hopefully this doesn't indicate a CPU problem?  The power supply on this thing blew Saturday but has run OK until now.
> > 
> > May 22 00:51:01 picard kernel: kernel BUG at page_alloc.c:108!
> > May 22 00:51:01 picard kernel: invalid operand: 0000
> > May 22 00:51:01 picard kernel: CPU:    1
> > May 22 00:51:01 picard kernel: EIP:    0010:[swap_duplicate+82/192]    Not tainted
> 
> There were quite a number of reports of those PageLRU BUGs on 2.4.17.
> No idea what fixed them, but 2.4.18 (and 2.4.19-pre) has seemed free
> of them (Ben LaHaise made a plausible change, but closer analysis
> suggested it couldn't really be the fix).  Suggest you upgrade.
> 
> Your oops report, by the way, must have been using the wrong System.map:
> page_alloc.c:108 is in __free_pages_ok(), swap_duplicate() is over in
> swapfile.c.  But no matter, page_alloc.c:108 identifies it well enough.
> 
> Hugh 
> 

Re: page_alloc bug in 2.4.17-pre8

[Hugh Dickins]

On Thu, 23 May 2002, Mike Black wrote:
> Oops -- got the wrong version (head was in the wrong kernel space).
> This occurred on 2.4.19-pre8

Then page_alloc.c:108 is a different check from what you showed, it's

	if (page->mapping)
		BUG();

I know nothing of that one, sorry.

Hugh