Skip to contentSkip to conversation

Pain and anguish: ALSA CVS and Linux-2.4.20

DISCUSSION Chris Rankin 2 participants 4 messages Nov 30, 2002 – Dec 1, 2002

lore
Share & tools
lore
#0 lore raw

Hi,

Remind me to NEVER try and update two systems at the same time again! I have compiled Linux 2.4.20 for my SMP box with very latest ALSA CVS modules. However, 2.4.20 deadlocked when I rebooted it and needed rescuing by the NMI watchdog.

Unfortunately, I have not tried these ALSA modules with my trusty 2.4.19 kernel and so don't know whether ALSA is to blame or whether its the new devfs. The balance of opinion on linus-kernel is that it's ALSA:

ksymoops 2.4.8 on i686 2.4.20. Options used
-V (default)
-k /proc/ksyms (default)
-l /proc/modules (default)
-o /lib/modules/2.4.20/ (default)
-m /boot/System.map-2.4.20 (specified)

Nov 30 18:26:23 (none) kernel: NMI Watchdog detected
LOCKUP on CPU0, eip c01062f
Nov 30 18:26:23 (none) kernel: CPU:    0
Nov 30 18:26:23 (none) kernel: EIP:   
0010:[__write_lock_failed+15/32]    Not t
Nov 30 18:26:23 (none) kernel: EFLAGS: 00000087
Nov 30 18:26:23 (none) kernel: eax: f770cf4c   ebx:
f78fbd20   ecx: 00000286   e
Nov 30 18:26:23 (none) kernel: esi: f770ce00   edi:
f7e4a220   ebp: f7e4a220   e
Nov 30 18:26:23 (none) kernel: ds: 0018   es: 0018  
ss: 0018
Nov 30 18:26:23 (none) kernel: Process alsactl (pid:
391, stackpage=f76e1000)
Nov 30 18:26:23 (none) kernel: Stack: f88e3010
00000038 000001f0 f88e5c60 f88e13
Nov 30 18:26:23 (none) kernel:        f7e4a220
00000000 f7714a40 f7e4a220 f77139
Nov 30 18:26:23 (none) kernel:        f7e4a220
f76e1f84 c014e7d9 ffffffeb f7e4a2
Nov 30 18:26:23 (none) kernel: Call Trace:   
[<f88e3010>] [<f88e5c60>] [<f88e13
Nov 30 18:26:23 (none) kernel: Code: 75 f6 f0 81 28 00

00 00 01 0f 85 e2 ff ff f
Using defaults from ksymoops -t elf32-i386 -a i386

Quoted 7 lines
Quoted 7 lines

eax; f770cf4c <_end+373d8adc/38511bf0>
ebx; f78fbd20 <_end+375c78b0/38511bf0>
edx; f78fbd38 <_end+375c78c8/38511bf0>
esi; f770ce00 <_end+373d8990/38511bf0>
edi; f7e4a220 <_end+37b15db0/38511bf0>
ebp; f7e4a220 <_end+37b15db0/38511bf0>
esp; f76e1ee8 <_end+373ada78/38511bf0>

Trace; f88e3010 <[snd].text.lock.control+5/135> Trace; f88e5c60 <[snd]snd_fops+0/48>
Trace; f88e1370 <[snd]snd_ctl_open+0/130> Trace; f88de22f <[snd]snd_open+df/160>

Code; 00000000 Before first symbol
00000000 <_EIP>:
Code; 00000000 Before first symbol
0: 75 f6 jne fffffff8
<_EIP+0xfffffff8> fffffff8 <EN
Code; 00000002 Before first symbol
2: f0 81 28 00 00 00 01 lock subl
$0x1000000,(%eax)
Code; 00000009 Before first symbol
9: 0f 85 e2 ff ff ff jne fffffff1
<_EIP+0xfffffff1> fffffff1 <EN
Code; 0000000f Before first symbol
f: c3 ret
Code; 00000010 Before first symbol
10: 90 nop
Code; 00000011 Before first symbol
11: f0 ff 00 lock incl (%eax)

To this end, I have been examining recent changes in the alsa-kernel/core directory. In particular, the snd_ctl_release() function in the file control.c:

Line 126:

        write_lock_irqsave(&card->control_rwlock,
flags);
        list_del(&ctl->list);
        list_for_each(list, &card->controls) {
                control = snd_kcontrol(list);
                if (control->owner == ctl)
                        control->owner = NULL;
        }
        write_unlock(&card->control_owner_lock);

Am I reading this wrong? It looks as if it's taking one lock but releasing another.

Cheers,
Chris

__________________________________________________
Do You Yahoo!?
Everything you'll ever need on one web page from News and Sport to Email and Music Charts
http://uk.my.yahoo.com

-------------------------------------------------------
This SF.net email is sponsored by: Get the new Palm Tungsten T handheld. Power & Color in a compact size!
http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0002en

#1 lore raw
Quoted 7 lines · On Sat, 30 Nov 2002, [iso-8859-1] Chris Rankin wrote:

Hi,

Remind me to NEVER try and update two systems at the same time again! I have compiled Linux 2.4.20 for my SMP box with very latest ALSA CVS modules. However, 2.4.20 deadlocked when I rebooted it and needed rescuing by the NMI watchdog.

Could you try the current CVS? Hopefully, I fixed the problem.

Jaroslav

-----
Jaroslav Kysela <[email protected]>
Linux Kernel Sound Maintainer
ALSA Project, SuSE Labs

-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf

#2 lore raw
Quoted 2 lines · --- Jaroslav Kysela

Could you try the current CVS? Hopefully, I fixed
the problem.

Just one small spelling mistake ;-)

Index: alsa-kernel/core/control.c
===================================================================
RCS file: /cvsroot/alsa/alsa-kernel/core/control.c,v retrieving revision 1.21
diff -u -u -r1.21 control.c
--- alsa-kernel/core/control.c 1 Dec 2002 18:20:16
-0000 1.21
+++ alsa-kernel/core/control.c 1 Dec 2002 18:32:25 -0000
@@ -132,7 +132,7 @@

                        control->owner = NULL;
        }
        write_unlock(&card->control_owner_lock);
-       write_lock_irqrestore(&card->control_rwlock,
flags);
+       write_unlock_irqrestore(&card->control_rwlock,
flags);
        snd_ctl_empty_read_queue(ctl);
        snd_magic_kfree(ctl);
        dec_mod_count(card->module);

Cheers,
Chris

__________________________________________________
Do You Yahoo!?
Everything you'll ever need on one web page
from News and Sport to Email and Music Charts
http://uk.my.yahoo.com

-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf

#3 lore raw
Quoted 5 lines · On Sun, 1 Dec 2002, [iso-8859-1] Chris Rankin wrote:
Quoted 2 lines · --- Jaroslav Kysela

Could you try the current CVS? Hopefully, I fixed
the problem.

Just one small spelling mistake ;-)

Yep. I discovered it in the verification phase ;-)

Jaroslav

-----
Jaroslav Kysela <[email protected]>
Linux Kernel Sound Maintainer
ALSA Project, SuSE Labs

-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf