swsusp in 2.5.53 BUG on kernel/suspend.c line 718

DISCUSSION bert hubert 2 participants 3 messages Dec 27, 2002

Subsystems (2)

Memory management Page allocator

lore

Share & tools

lore

bert hubert

author

2002-12-27 14:12 UTC #0 lore raw

Hi!

I wanted to try software suspend again in Linux as 2.5 is doing almost everything pretty well for me already.

I boot my uniprocessor Pentium III laptop with:

kernel (hd0,0)/boot/vmlinuz-2.5.53 root=/dev/hda1 resume=/dev/hda2

# swapon -s
Filename				Type		Size	Used Priority
/dev/hda2                               partition	489972	0    -1

$ cat /proc/meminfo 
MemTotal:       191240 kB

When I suspend, things proceed swimmingly, I see a lot of dots printed and processes entering the refrigerator, until line 718 is hit in kernel/suspend.c:

   if (nr_copy_pages != count_and_copy_data_pages(pagedir_nosave)) /* copy */
       BUG();

When I aded some printks, it turns out that count_and_copy_data pages returns 5440 (decimal) and that nr_copy_pages is 5458, 18 more. Before this function is called, the address c034c000 was printed twice prefixed with 'nosave', once during each call of count_and_copy_data_pages it appears.

So it appears some pages were freed in the critical section!

Another interesting note is that pdflush reported 'Bogus wakeup' twice during the refrigeration phase. I also see two pdflushes running.

If I remove the BUG();, on resume it crashes on an unhandled NULL pointer, the EIP is in a function aptly named do_magic() at +0x9e.

Compiler is gcc 3.2.1. Anything I can do to help, just let me know!

Regards,

bert

Signature · 3 lines

http://www.PowerDNS.com      Open source, database driven DNS Software 
http://lartc.org           Linux Advanced Routing & Traffic Control HOWTO
http://netherlabs.nl                         Consulting

Pavel Machek

2002-12-27 15:01 UTC #1 lore raw

You need one-liner to fix this, search mailing lists.

Pavel

Quoted 42 lines

Hi!
I wanted to try software suspend again in Linux as 2.5 is doing almost everything pretty well for me already.
I boot my uniprocessor Pentium III laptop with:
kernel (hd0,0)/boot/vmlinuz-2.5.53 root=/dev/hda1 resume=/dev/hda2
# swapon -s
Filename				Type		Size	Used Priority
/dev/hda2                               partition	489972	0    -1
$ cat /proc/meminfo 
MemTotal:       191240 kB
When I suspend, things proceed swimmingly, I see a lot of dots printed and processes entering the refrigerator, until line 718 is hit in kernel/suspend.c:
   if (nr_copy_pages != count_and_copy_data_pages(pagedir_nosave)) /* copy */
       BUG();
When I aded some printks, it turns out that count_and_copy_data pages returns 5440 (decimal) and that nr_copy_pages is 5458, 18 more. Before this function is called, the address c034c000 was printed twice prefixed with 'nosave', once during each call of count_and_copy_data_pages it appears.
So it appears some pages were freed in the critical section!
Another interesting note is that pdflush reported 'Bogus wakeup' twice during the refrigeration phase. I also see two pdflushes running.
If I remove the BUG();, on resume it crashes on an unhandled NULL pointer, the EIP is in a function aptly named do_magic() at +0x9e.
Compiler is gcc 3.2.1. Anything I can do to help, just let me know!
Regards,
bert

Signature · 2 lines

Casualities in World Trade Center: ~3k dead inside the building,
cryptography in U.S.A. and free speech in Czech Republic.

bert hubert

author

2002-12-27 18:26 UTC #2 lore raw

Quoted 1 line · Pavel Machek

You need one-liner to fix this, search mailing lists.

Your patch below indeed works, except for my network adaptor which needs 'ifconfig eth0 down', 'ifconfig eth0 up' before it works again.

It says:

NETDEV WATCHDOG: eth0: transmit timed out
eth0: Transmit timeout, status 00000000 00000240
00:01.1 Ethernet controller: Silicon Integrated Systems [SiS] SiS900 10/100 Ethernet (rev 82)

Can a non-guru add the magic handlers to network drivers to make them wake up again properly?

Thanks!

mm/page_alloc.c 2002-12-18 22:30:47.000000000 +0100 renamed from mm/page_alloc.c 2002-12-18 22:21:13.000000000 +0100 +1 -1

@@ -389,7 +389,7 @@
        unsigned long flags;
        struct page *page = NULL;
 
-       if (order == 0) {
+       if ((order == 0) && !cold) {
                struct per_cpu_pages *pcp;
 
                pcp = &zone->pageset[get_cpu()].pcp[cold];

Signature · 3 lines

http://www.PowerDNS.com      Open source, database driven DNS Software 
http://lartc.org           Linux Advanced Routing & Traffic Control HOWTO
http://netherlabs.nl                         Consulting