SiS/Trident 4DWave sound driver oops
I just grabbed 2.4.12-ac6 and the OOPS continues. This is with soundcore, ac97_codec, and trident all built as modules.
Again, like my previous report (which was on Linus' 2.4.12), I can 'modprobe soundcore' and 'modprobe ac97_codec' with no problems. Once I do the 'modprobe trident', I get the oops. After the oops, the system is still responsive, and just like Stuart Young reported, the trident module is stuck in "initalizing".
See below for the details including the oops report itself.
Mike Robbins
[email protected]
(Please also cc your reply to me.)
cat /proc/modules after the oops:
-----------
trident 26800 (initializing)
ac97_codec 9120 0 [trident] soundcore 3568 2 [trident]
NVdriver 713104 17
8139too 12896 1 (autoclean)
-----------
Relevant part of /proc/pci:
----------
Bus 0, device 3, function 0:
Multimedia audio controller: PCI device 10b9:5451 (Acer Laboratories
Inc. [ALi]) (rev 2).
IRQ 10.
Master Capable. Latency=64. Min Gnt=2.Max Lat=24.
I/O at 0xc400 [0xc4ff].
Non-prefetchable 32 bit memory at 0xdfffe000 [0xdfffefff].----------
Part of /proc/iomem:
dfffe000-dfffefff : PCI device 10b9:5451 (Acer Laboratories Inc. [ALi])
Part of /proc/ioports:
c400-c4ff : PCI device 10b9:5451 (Acer Laboratories Inc. [ALi])
SYSLOG LEADING UP TO OOPS:
(modprobe trident executed here, and this is the result:)
----------
Oct 25 08:35:55 tbird kernel: Trident 4DWave/SiS 7018/ALi 5451,Tvia CyberPro 5050 PCI Audio, version 0.14.9d, 08:28:09 Oct 25 2001 Oct 25 08:35:55 tbird kernel: PCI: Assigned IRQ 10 for device 00:03.0 Oct 25 08:35:55 tbird kernel: trident: ALi Audio Accelerator found at IO 0xc400, IRQ 10 Oct 25 08:35:55 tbird kernel: ac97_codec: AC97 Audio codec, id: 0x414c:0x4326 (Unknown) Oct 25 08:35:55 tbird kernel: ali: AC97 CODEC read timed out. Oct 25 08:35:55 tbird kernel: ali: AC97 CODEC write timed out. Oct 25 08:35:55 tbird kernel: ali: AC97 CODEC read timed out. Oct 25 08:35:55 tbird last message repeated 2 times Oct 25 08:35:55 tbird kernel: ac97_codec: AC97 codec, id: 0x0000:0x0000 (Unknown) Oct 25 08:35:55 tbird kernel: ali: AC97 CODEC read timed out. Oct 25 08:35:55 tbird kernel: ali: AC97 CODEC write timed out. Oct 25 08:35:55 tbird kernel: ali: AC97 CODEC read timed out. Oct 25 08:35:55 tbird kernel: Unable to handle kernel NULL pointer
dereference at virtual address 00000000
---------- (the rest of the oops follows this)
KSYMOOPS REPORT:
-----------
Unable to handle kernel NULL pointer dereference at virtual address 00000000 e5991d7e *pde = 00000000 Oops: 0000 CPU: 0 EIP: 0010:[<e5991d7e>] Not tainted
Using defaults from ksymoops -t elf32-i386 -a i386
EFLAGS: 00010206
eax: 00000000 ebx: dc4c7cc0 ecx: df408000 edx: 00000006
esi: 00000000 edi: dc4c7d70 ebp: 00000001 esp: dcd01e98
ds: 0018 es: 0018 ss: 0018
Process modprobe (pid: 885, stackpage=dcd01000)
Stack: dc4c7cc0 02000000 dc4c7d70 00000001 e5999fcb dc4c7cc0 00000024
00000003
dee5c000 dfff5000 e599b120 dfff6400 e599a410 dee5c000 ffffffed
0000c400
02000000 e599b068 e599b1e0 dfff6400 00000000 c01b22a4 dfff6400
e599b068
Call Trace: [<e5999fcb>] [<e599b120>] [<e599a410>] [<e599b068>][<e599b1e0>]
[<c01b22a4>] [<e599b068>] [<e599b1e0>] [<c01b2302>] [<e599b1e0>]
[<e599a634>]
[<e599b1e0>] [<e599aee0>] [sys_init_module+1317/1504] [<e599b860>]
[<e5995060>] [system_call+51/56]
[<e599b1e0>] [<e599aee0>] [<c01142c5>] [<e599b860>] [<e5995060>]
[<c0106ecb>]
Code: 8b 00 85 c0 74 04 53 ff d0 5a 31 ed 83 3d e0 31 99 e5 ff ba
Quoted 1 line
Quoted 1 line
EIP;
e5991d7e<[ac97_codec]ac97_init_mixer+7e/e0> <=====
Trace; e5999fcb <[trident]trident_ac97_init+22b/2f0> Trace; e599b120 <[trident]trident_audio_fops+0/60> Trace; e599a410 <[trident]trident_probe+380/4a0> Trace; e599b068 <[trident]trident_pci_tbl+54/a8> Trace; e599b1e0 <[trident]trident_pci_driver+0/3f> Trace; c01b22a4 <pci_announce_device+34/50>
Trace; e599b068 <[trident]trident_pci_tbl+54/a8> Trace; e599b1e0 <[trident]trident_pci_driver+0/3f> Trace; c01b2302 <pci_register_driver+42/60>
Trace; e599b1e0 <[trident]trident_pci_driver+0/3f> Trace; e599a634 <[trident]trident_init_module+24/50> Trace; e599b1e0 <[trident]trident_pci_driver+0/3f> Trace; e599aee0 <[trident]__module_pci_device_size+614/693> Trace; e599b1e0 <[trident]trident_pci_driver+0/3f> Trace; e599aee0 <[trident]__module_pci_device_size+614/693> Trace; c01142c5 <sys_init_module+525/5e0>
Trace; e599b860 <.bss.end+1/????>
Trace; e5995060 <[trident]trident_enable_loop_interrupts+0/80> Trace; c0106ecb <system_call+33/38>
Code; e5991d7e <[ac97_codec]ac97_init_mixer+7e/e0> 00000000 <_EIP>:
Code; e5991d7e <[ac97_codec]ac97_init_mixer+7e/e0> <=====
0: 8b 00 mov (%eax),%eax <=====
Code; e5991d80 <[ac97_codec]ac97_init_mixer+80/e0>
2: 85 c0 test %eax,%eax
Code; e5991d82 <[ac97_codec]ac97_init_mixer+82/e0>
4: 74 04 je a <_EIP+0xa> e5991d88
<[ac97_codec]ac97_init_mixer+88/e0>
Code; e5991d84 <[ac97_codec]ac97_init_mixer+84/e0>
6: 53 push %ebx
Code; e5991d85 <[ac97_codec]ac97_init_mixer+85/e0>
7: ff d0 call *%eax
Code; e5991d87 <[ac97_codec]ac97_init_mixer+87/e0>
9: 5a pop %edx
Code; e5991d88 <[ac97_codec]ac97_init_mixer+88/e0>
a: 31 ed xor %ebp,%ebp
Code; e5991d8a <[ac97_codec]ac97_init_mixer+8a/e0>
c: 83 3d e0 31 99 e5 ff cmpl $0xffffffff,0xe59931e0
Code; e5991d91 <[ac97_codec]ac97_init_mixer+91/e0>
13: ba 00 00 00 00 mov $0x0,%edx
----------
Hello,
At 25 Oct 2001 09:24:21 -0400,
Quoted 10 lines · Michael F. Robbins
I just grabbed 2.4.12-ac6 and the OOPS continues. This is with soundcore,
ac97_codec, and trident all built as modules.Again, like my previous report (which was on Linus' 2.4.12), I can 'modprobe soundcore' and 'modprobe
ac97_codec' with no problems. Once I do the 'modprobe trident', I get the oops. After the oops, the system is still responsive, and just like Stuart Young reported, the trident module is stuck in "initalizing".
Quoted 5 lines
See below for the details including the oops report itself.
Mike Robbins
[email protected]
(Please also cc your reply to me.)
Following patch may fix your oops. Please try.
drivers/sound/ac97_codec.c +0 -1
@@ -143,7 +143,6 @@ {0x83847656, "SigmaTel STAC9756/57", &sigmatel_9744_ops}, {0x83847684, "SigmaTel STAC9783/84?", &null_ops}, {0x57454301, "Winbond 83971D", &null_ops}, - {0,} }; static const char *ac97_stereo_enhancements[] =
Quoted 1 line · Tachino Nobuhiro
Following patch may fix your oops. Please try.
Hm, I don't think so. The last area is marked zero so code can know when it ends. This is common practice.
drivers/sound/ac97_codec.c line 148
Quoted 11 lines
diff -u -r
linux-2.4.12-ac5.org/drivers/sound/ac97_codec.clinux-2.4.12-ac5/drivers/sound/ac97_codec.cdrivers/sound/ac97_codec.c +0
-1@@ -143,7 +143,6 @@ {0x83847656, "SigmaTel STAC9756/57", &sigmatel_9744_ops}, {0x83847684, "SigmaTel STAC9783/84?", &null_ops}, {0x57454301, "Winbond 83971D", &null_ops}, - {0,} }; static const char *ac97_stereo_enhancements[] =
Robert Love
Hello,
At 25 Oct 2001 21:45:58 -0400,
Quoted 6 lines · Robert Love
Quoted 1 line · Tachino Nobuhiro
Following patch may fix your oops. Please try.
Hm, I don't think so. The last area is marked zero so code can know when it ends. This is common practice.
But the code does not use the last area. this is the code in ac97_probe_codec().
id1 = codec->codec_read(codec, AC97_VENDOR_ID1);
id2 = codec->codec_read(codec, AC97_VENDOR_ID2);
for (i = 0; i < ARRAY_SIZE(ac97_codec_ids); i++) {
if (ac97_codec_ids[i].id == ((id1 << 16) | id2)) {
codec->type = ac97_codec_ids[i].id;
codec->name = ac97_codec_ids[i].name;
codec->codec_ops = ac97_codec_ids[i].ops;
break;
}
}If id1 and id2 happen to be 0, it matches the last entry and codec_ops is set to uncertain value(maybe 0). it may cause the oops in ac97_init_mixer().
Quoted 6 lines · Tachino Nobuhiro
Quoted 2 lines · Robert Love
Hm, I don't think so. The last area is marked zero so code can know when it ends. This is common practice.
But the code does not use the last area. this is the code in
ac97_probe_codec().
ARRAY_SIZE(x) returns the number of elements in x, but since everything is 0-referenced going from 0 to i < ARRAY_SIZE isn't a problem.
ie int x[3];ARRAY_SIZE(x) = 3;
but x[2] is last element... so no issue here.
Quoted 11 lines
id1 = codec->codec_read(codec, AC97_VENDOR_ID1);
id2 = codec->codec_read(codec, AC97_VENDOR_ID2);
for (i = 0; i < ARRAY_SIZE(ac97_codec_ids); i++) {
if (ac97_codec_ids[i].id == ((id1 << 16) | id2)) {
codec->type = ac97_codec_ids[i].id;
codec->name = ac97_codec_ids[i].name;
codec->codec_ops = ac97_codec_ids[i].ops;
break;
}
}Quoted 2 lines
If id1 and id2 happen to be 0, it matches the last entry and
codec_opsis set to uncertain value(maybe 0). it may cause the oops inac97_init_mixer().
Robert Love
At 25 Oct 2001 22:02:20 -0400,
Quoted 9 lines · Robert Love
Quoted 6 lines · Tachino Nobuhiro
Quoted 2 lines · Robert Love
Hm, I don't think so. The last area is marked zero so code can know when it ends. This is common practice.
But the code does not use the last area. this is the code in
ac97_probe_codec().
Quoted 6 lines
ARRAY_SIZE(x)returns the number of elements in x, but since everything is 0-referenced going from 0 to i <ARRAY_SIZEisn't a problem.ie int x[3];
ARRAY_SIZE(x)= 3;
but x[2] is last element... so no issue here.
No. {0, } is the last elemnet of ac97_codec_ids[] and that index isARRAY_SIZE(ac97_code_ids) - 1. So this element which should be used as a loop terminator is used as a valid entry in for loop incorrectly.
Please read ac97_codec.c
Quoted 5 lines · Tachino Nobuhiro
No. {0, } is the last elemnet of ac97_codec_ids[] and that index is
ARRAY_SIZE(ac97_code_ids)- 1. So this element which should be used as a loop terminator is used as a valid entry in for loop incorrectly.Please read
ac97_codec.c
You are right; I apologize.
Robert Love
Quoted 8 lines · Robert Love
Quoted 5 lines · Tachino Nobuhiro
No. {0, } is the last elemnet of ac97_codec_ids[] and that index is
ARRAY_SIZE(ac97_code_ids)- 1. So this element which should be used as a loop terminator is used as a valid entry in for loop incorrectly.Please read
ac97_codec.cYou are right; I apologize.
Implemented the patch suggested, and the module no longer oops's, and I get codec id's listed as 0x0000:0x0000 (Unknown). I still get no sound like I did with 2.4.7 (using mpg123 as a test, with a known working mp3 file), and output to the device is blocked (nothing gets written).
How can I find out the ac97 codec ID for this chipset (if there is one) so it can be added to the ac97_codec_ids array? From what I can tell, it's as though the codec->codec_read(codec, AC97_VENDOR_ID#) isn't returning the codec value for this system at all.
Any suggestions?
AMC Enterprises P/L - Stuart Young First Floor - Network and Systems Admin 3 Chesterville Rd - [email protected] Cheltenham Vic 3192 - Ph: (03) 9584-2700
http://www.amc.com.au/ - Fax: (03) 9584-2755
Quoted 4 lines
How can I find out the ac97 codec ID for this chipset (if there is one) so it can be added to the
ac97_codec_idsarray? From what I can tell, it's as though thecodec->codec_read(codec, AC97_VENDOR_ID#)isn't returning the codec value for this system at all.
Something is failing to bring up the AC97 codec bus and/or set it up properly. Can you find exactly which patch broke that for you (you'll possibly want to keep fixing the codec table as you test older ones)
Quoted 8 lines · Alan Cox
Quoted 4 lines
How can I find out the ac97 codec ID for this chipset (if there is one) so it can be added to the
ac97_codec_idsarray? From what I can tell, it's as though thecodec->codec_read(codec, AC97_VENDOR_ID#)isn't returning the codec value for this system at all.Something is failing to bring up the AC97 codec bus and/or set it up properly. Can you find exactly which patch broke that for you (you'll possibly want to keep fixing the codec table as you test older ones)
Other than my system works as a module, I have the same problems. I bet you that 2.4.8 is the last kernel that worked for him. (It was for me.) I do not know about pre-patches. If you want, I will find that out for my case.
Trever Adams
Re: SiS/Trident 4DWave sound driver
In reply to Alan CoxQuoted 6 lines · Alan Cox
Quoted 2 lines
though the
codec->codec_read(codec, AC97_VENDOR_ID#)isn't returning the codec value for this system at all.Something is failing to bring up the AC97 codec bus and/or set it up properly. Can you find exactly which patch broke that for you (you'll possibly want to keep fixing the codec table as you test older ones)
Well the thing is this system has never had working sound. I've got 2 of these machines, both with the array fix, and neither get sound. I've tried 2.4.7 (which didn't have the array fix), 2.4.9, 2.4.12, 2.4.13, and I'm about to try some earlier kernels, but since this chip is "notoriously new" I would not be suprised if it's not returning anything for what should be valid codec calls. Will take a peek and get back to you.
PS: This is one of these SiS630S all in one integrated things, and also one of the offenders in that SiS FrameBuffer stuff we've been discussing, so I wouldn't be suprised in anomalies at all.
AMC Enterprises P/L - Stuart Young First Floor - Network and Systems Admin 3 Chesterville Rd - [email protected] Cheltenham Vic 3192 - Ph: (03) 9584-2700
http://www.amc.com.au/ - Fax: (03) 9584-2755
Re: SiS/Trident 4DWave sound driver (Update)
In reply to Alan CoxQuoted 6 lines · Alan Cox
Quoted 2 lines
though the
codec->codec_read(codec, AC97_VENDOR_ID#)isn't returning the codec value for this system at all.Something is failing to bring up the AC97 codec bus and/or set it up properly. Can you find exactly which patch broke that for you (you'll possibly want to keep fixing the codec table as you test older ones)
Update: - 2.4.3 appears to have the same issue of no-find the codec.
2.2.19pre17 reports the wrong h/ware address for the sound device (the same one as the IDE bus - urghy!), so I can't even load the driver to test if it'll work (which I doubt).
Upon looking at the code between 2.4.0 and 2.4.13, in particular at trident_ac97_get() and trident_ac97_set() there is practically no difference between them, except for the addition of another option in the switch statement for another card. Almost all the additions and changes between versions have been specifically ALi or similar chipsets, and don't seem to affect the SiS stuff.
So where to now?
Thinking mebbe I should hook this machine up to the net outside the firewall, plug in the webcam and point it at it, give you an ssh account on it, and connect an oscilloscope to the audio and let you fiddle. Useful for the SiS FrameBuffer thing as well I'd guess. *grin*
Talk about remote debugging eh?
AMC Enterprises P/L - Stuart Young First Floor - Network and Systems Admin 3 Chesterville Rd - [email protected] Cheltenham Vic 3192 - Ph: (03) 9584-2700
http://www.amc.com.au/ - Fax: (03) 9584-2755
Re: SiS drivers (more)
In reply to Alan CoxAnd more on the SiS7018/Trident audio driver...
Using 2.4.13-ac4, I get...
trident: can't allocate I/O space at 0x1000
.. and then insmod fails with...
/lib/modules/2.4.13-ac4/kernel/drivers/sound/trident.o: init_module: No such device
.. which doesn't bode well I guess, least for my situation.
On the SiS FrameBuffer driver front, with 2.4.13-ac4 it comes up with the same thing (blank/glowing display), except it flickers a bit more than the last driver. Once again, the console and the rest of the machine works, just no display output.
Going over the code, this looks a lot more like the core of the driver in XFree86 4.1.0, which might prove useful. In fact, I'd guess it's pretty much the same code, given some of the names of the functions, etc.
AMC Enterprises P/L - Stuart Young First Floor - Network and Systems Admin 3 Chesterville Rd - [email protected] Cheltenham Vic 3192 - Ph: (03) 9584-2700
http://www.amc.com.au/ - Fax: (03) 9584-2755
Quick summary: sound now works on the ALi 5451 chip, but has some non-fatal error messages.
drivers/sound/ac97_codec.c line 148
Quoted 11 lines · Tachino Nobuhiro
diff -u -r
linux-2.4.12-ac5.org/drivers/sound/ac97_codec.clinux-2.4.12-ac5/drivers/sound/ac97_codec.cdrivers/sound/ac97_codec.c +0
-1@@ -143,7 +143,6 @@ {0x83847656, "SigmaTel STAC9756/57", &sigmatel_9744_ops}, {0x83847684, "SigmaTel STAC9783/84?", &null_ops}, {0x57454301, "Winbond 83971D", &null_ops}, - {0,} }; static const char *ac97_stereo_enhancements[] =
I just removed that final data set and the preceeding comma and recompiled 2.4.12-ac6. I'm pleased to report that the sound now works just fine! The module loads, and I do get some (non-fatal) error messages, but once its loaded it runs like a champ.
BTW, I was also getting these error messages on my old 2.4.7 install. This is my 2.4.12-ac6 log after modprobe trident. Take a look:
----------
Trident 4DWave/SiS 7018/ALi 5451,Tvia CyberPro 5050 PCI Audio, version 0.14.9d, 19:15:41 Oct 26 2001
PCI: Assigned IRQ 10 for device 00:03.0
trident: ALi Audio Accelerator found at IO 0xc400, IRQ 10ac97_codec: AC97 Audio codec, id: 0x414c:0x4326 (Unknown)
ali: AC97 CODEC read timed out.
ali: AC97 CODEC write timed out.
ali: AC97 CODEC read timed out.
last message repeated 2 timesac97_codec: AC97 codec, id: 0x0000:0x0000 (Unknown)
ali: AC97 CODEC write timed out.
ali: AC97 CODEC read timed out.
ali: AC97 CODEC write timed out.
last message repeated 10 times
ali: AC97 CODEC read timed out.
last message repeated 127 times
----------
This hangs the system totally for about 5 seconds. After this, it's all fine and the sound works great. One strange thing I notice from the above log is the different codec IDs showing up...
Thanks,
Mike Robbins
[email protected]
(Please also cc your reply to me.)
Quoted 8 lines · Robert Love
Quoted 5 lines · Tachino Nobuhiro
No. {0, } is the last elemnet of ac97_codec_ids[] and that index is
ARRAY_SIZE(ac97_code_ids)- 1. So this element which should be used as a loop terminator is used as a valid entry in for loop incorrectly.Please read
ac97_codec.cYou are right; I apologize.
I think the way this is coded stinks anyway. the {0,} should be used as a loop-terminator, not ARRAY_SIZE(blaha) - 1. Yes, using 0-termination wastes space. But it's cleaner and in line with what most other code does.
/David Weinehall
_ _
// David Weinehall <[email protected]> /> Northern lights wander \\
// Project MCA Linux hacker // Dance across the winter sky // \> http://www.acc.umu.se/~tao/ </ Full colour fire </
Quoted 4 lines · David Weinehall
I think the way this is coded stinks anyway. the {0,} should be used as a loop-terminator, not
ARRAY_SIZE(blaha)- 1. Yes, using 0-termination wastes space. But it's cleaner and in line with what most other code does.
Agreed. Also, I didn't check if other ac97 code uses the {0,} as a terminator. Removing it may break that.
The patch below accomplishes this.
However, now that I am actually looking at the code <g>, I don't see why this would be a problem either way. Even though the loop reads the "terminal" entry, it just checks whether it equals the specified id. It is equal to 0 so I assume it never will...we aren't dereferencing it or anything.
drivers/sound/ac97_codec.c +3 -2
@@ -669,7 +669,7 @@ {u16 id1, id2;u16 audio, modem; -int i; +int i =0 ;/* probing AC97 codec, AC97 2.0 says that bit 15 of register 0x00 (reset) should * be read zero.
@@ -700,13 +700,14 @@ id1 = codec->codec_read (codec,AC97_VENDOR_ID1 ); id2 = codec->codec_read (codec,AC97_VENDOR_ID2 ); -for (i =0 ; i <ARRAY_SIZE (ac97_codec_ids); i++) { +while (a97_codec_ids[i].id !=0 ) {if (ac97_codec_ids[i].id == ((id1 <<16 ) | id2)) { codec->type = ac97_codec_ids[i].id; codec->name = ac97_codec_ids[i].name; codec->codec_ops = ac97_codec_ids[i].ops;break ; } + i++; }if (codec->name ==NULL ) codec->name ="Unknown" ;
Robert Love