Oops in cs4281_suspend
I'm using alsa-0.9.0_rc5 with 2.4.20-pre10-mjc2 and I get a kernel oops in cs4281_suspend when suspeding the machine. Result of ksymoops is attached.
Matching the disassembly with the assembly output of gcc with the same options (also attached), it shows that the problem is in the line:
chip->suspend_regs[i] = snd_cs4281_peekBA0(chip, saved_regs[i]);
more specifically, in the inlined snd_cs4281_peekBA0() function:
return readl(chip->ba0 + offset);
I tried to add some more debugging output in cs4281_suspend to see why this happens, but I can't get anything out of it. One thing is sure - removing the snd-cs4281 module before suspending avoids the problem completely.
I hope this enables you to fix the problem. Please tell me if there is anything more I can do.
José Fonseca
ksymoops 2.4.5 on i686 2.4.20-pre10-mjc2. Options used
-V (default)
-k /proc/ksyms (default)
-l /proc/modules (default)
-o /lib/modules/2.4.20-pre10-mjc2/ (default) -m /boot/System.map (specified)
cs: cb_alloc(bus 2): vendor 0x10ec, device 0x8139
8139too Fast Ethernet driver 0.9.26
Unable to handle kernel paging request at virtual address 991b6920 cc8cc437 *pde = 00000000 Oops: 0000 CPU: 0 EIP: 0010:[<cc8cc437>] Not tainted
Using defaults from ksymoops -t elf32-i386 -a i386
EFLAGS: 00010282
eax: cc8e7920 ebx: cc8ccf40 ecx: cc8cf000 edx: 00000018
esi: cbabc800 edi: 00000003 ebp: c1487ed4 esp: c1487eac
ds: 0018 es: 0018 ss: 0018
Process apmd (pid: 4914, stackpage=c1487000)
Stack: cbabcc00 cba88008 cba850c0 cc8cc512 cbabc800 cbaae460 c1487ed4 c01b4002
cba88000 00000003 c1487ef0 c01b40e1 cba88000 00000003 cba850c0 00000003
00000003 c1487f08 c01b41ce cba850c0 00000003 cba7a2e0 00000000 c1487f14
Call Trace: [<cc8cc512>] [<c01b4002>] [<c01b40e1>] [<c01b41ce>] [<c01b425f>]
[<c012404d>] [<c012410b>] [<c0110f6f>] [<c0111612>] [<c0144d76>] [<c0106fab>]
Code: 8b 04 08 89 84 96 64 01 00 00 42 eb ec ff 80 78 01 00 00 0f Quoted 8 lines
Quoted 8 lines
EIP;
cc8cc437<[snd-cs4281]cs4281_suspend+67/90> <=====eax;
cc8e7920<[8139too]rtl8139_pci_driver+0/27>
ebx;cc8ccf40<[snd-cs4281]saved_regs+0/60>
ecx;cc8cf000<[snd-cs4281].bss.end+1fe1/15041>
esi;cbabc800<_end+b7fd648/c583ea8>
ebp;c1487ed4<_end+11c8d1c/c583ea8>
esp;c1487eac<_end+11c8cf4/c583ea8>
Trace; cc8cc512 <[snd-cs4281]snd_cs4281_suspend+12/20> Trace; c01b4002 <pci_pm_suspend_device+22/26> Trace; c01b40e1 <pci_pm_suspend_bus+4b/66> Trace; c01b41ce <pci_pm_suspend+22/46>
Trace; c01b425f <pci_pm_callback+2f/38> Trace; c012404d <pm_send+61/80>
Trace; c012410b <pm_send_all+5f/8c>
Trace; c0110f6f <suspend+d/ec>
Trace; c0111612 <do_ioctl+f2/140>
Trace; c0144d76 <sys_ioctl+e4/264>
Trace; c0106fab <system_call+33/38>
Code; cc8cc437 <[snd-cs4281]cs4281_suspend+67/90> 00000000 <_EIP>:
Code; cc8cc437 <[snd-cs4281]cs4281_suspend+67/90> <=====
0: 8b 04 08 mov (%eax,%ecx,1),%eax <=====
Code; cc8cc43a <[snd-cs4281]cs4281_suspend+6a/90>
3: 89 84 96 64 01 00 00 mov %eax,0x164(%esi,%edx,4)
Code; cc8cc441 <[snd-cs4281]cs4281_suspend+71/90>
a: 42 inc %edx
Code; cc8cc442 <[snd-cs4281]cs4281_suspend+72/90>
b: eb ec jmp fffffff9 <_EIP+0xfffffff9>
Code; cc8cc444 <[snd-cs4281]cs4281_suspend+74/90>
d: ff 80 78 01 00 00 incl 0x178(%eax)
Code; cc8cc44a <[snd-cs4281]cs4281_suspend+7a/90>
13: 0f 00 00 sldtl (%eax)
[...]
.file 159 "../alsa-kernel/pci/cs4281.c" [...] .text .align 2 .p2align 4,,15 .type cs4281_suspend,@function cs4281_suspend: .LFB47: .loc 159 2093 0 pushl %esi .LCFI129: pushl %ebx .LCFI130: subl $4, %esp .LCFI131: movl 16(%esp), %esi .loc 159 2094 0 .LBB292: movl 44(%esi), %eax .loc 30 115 0 .LBB293: .LBB294: leal 376(%eax), %ecx .loc 30 120 0 #APP # atomic down operation decl 376(%eax) js 2f 1: .subsection 1 .ifndef .text.lock.cs4281 .text.lock.cs4281: .endif 2: call __down_failed jmp 1b .previous
.loc 159 2099 0 #NO_APP .LBE294: .LBE293: cmpl $768, 372(%eax) je .L985 .loc 159 2102 0 movl 48(%esi), %eax .loc 159 2112 0 movl $saved_regs, %ebx .loc 159 2102 0 movl %eax, (%esp) call snd_pcm_suspend_all_R9cd5838a .loc 159 2112 0 xorl %edx, %edx .loc 159 578 0 .LBB295: movl 4(%esi), %ecx movl 1024(%ecx), %eax .loc 159 2105 0 .LBE295: orl $65536, %eax .loc 159 573 0 .LBB296: movl %eax, 1024(%ecx) .LBE296: .LBB297: movl $2, 8(%ecx) .p2align 4,,15 .L995: .loc 159 2113 0 .LBE297: movl (%ebx,%edx,4), %eax testl %eax, %eax je .L991 .loc 159 578 0 ; cs4281.c:578 .LBB298: movl (%eax,%ecx), %eax ; <===== .loc 159 577 0 .LBE298: movl %eax, 356(%esi,%edx,4) .loc 159 2112 0 .L991: incl %edx jmp .L995 .L985: .loc 30 201 0 .LBB299: .LBB300: #APP # atomic up operation incl 376(%eax) jle 2f 1: .subsection 1 .ifndef .text.lock.cs4281 .text.lock.cs4281: .endif 2: call __up_wakeup jmp 1b .previous .subsection 0
.loc 159 2135 0
#NO_APP
popl %eax
popl %ebx
popl %esi
ret
.LBE300:
.LBE299:
.LBE292:
.LFE47:
.Lfe47:
.size cs4281_suspend,.Lfe47-cs4281_suspend
[...]
Quoted 3 lines
options (also attached), it shows that the problem is in the line:
chip->suspend_regs[i] = snd_cs4281_peekBA0(chip, saved_regs[i]);
Very good report :-)
I think the following patch should solve this problem. But since that suspend/resume seems to not have been tested at all for that driver, may be some other things will show up.
Index: alsa-kernel/pci/cs4281.c
===================================================================
RCS file: /cvsroot/alsa/alsa-kernel/pci/cs4281.c,v retrieving revision 1.29
pci/cs4281.c +2 -2
@@ -2109,7 +2109,7 @@snd_cs4281_pokeBA0 (chip,BA0_HICR ,BA0_HICR_CHGM );/* remember the status registers */ -for (i =0 ;number_of (saved_regs); i++) +for (i =0 ; i <number_of (saved_regs); i++)if (saved_regs[i]) chip->suspend_regs[i] =snd_cs4281_peekBA0 (chip, saved_regs[i]);
@@ -2153,7 +2153,7 @@snd_cs4281_chip_init (chip,0 );/* restore the status registers */ -for (i =0 ;number_of (saved_regs); i++) +for (i =0 ; i <number_of (saved_regs); i++)if (saved_regs[i])snd_cs4281_pokeBA0 (chip, saved_regs[i], chip->suspend_regs[i]);
Best regards,
--
Tomasz Motylewski
BFAD GmbH & Co. KG
-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
Quoted 6 lines · tomasz motylewski
Quoted 3 lines
options (also attached), it shows that the problem is in the line:
chip->suspend_regs[i] = snd_cs4281_peekBA0(chip, saved_regs[i]);Very good report :-)
Well I'm still beating myself for not noticing what was the bug even after looking so many times to those lines! 8O
Quoted 3 lines
I think the following patch should solve this problem. But since that suspend/resume seems to not have been tested at all for that driver, may be some other things will show up.
It solved indeed. And surprisingly suspend/resume worked flawlessly after.
Thanks,
José Fonseca
__________________________________________________
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