vanilla 2.6.0-test11 and CS4236 card

vanilla 2.6.0-test11 and CS4236 card

[Joseph Pingenot]

Howdy.

I'm having problems getting the CS4236+ driver to recognize my 
  CS4236B card.  pnp finds it on boot:
isapnp: Scanning for PnP cards...
isapnp: Card 'CS4236B'
isapnp: 1 Plug & Play card detected total

but the ALSA driver doesn't pick it up.
isapnp detection failed and probing for CS4236+ is not supported
CS4236+ soundcard not found or device busy

Furthermore, after fudging with manually setting it up via modprobe
  options, it's still not loading:
CS4236+ soundcard not found or device busy

This used to work in the 2.4 series kernel without any modprobe.conf
  settings; the OSS driver would pick it up.

Any assistance would be greatly appreciated; this is the only thing holding
  me back from 2.6 goodness.  ;)

[config, pnpdump, and other information available on request]
-Joseph
-- 
trelane@digitasaru.net--------------------------------------------------
"We continue to live in a world where all our know-how is locked into
 binary files in an unknown format. If our documents are our corporate
 memory, Microsoft still has us all condemned to Alzheimer's."
    --Simon Phipps, http://theregister.com/content/4/30410.html

Re: vanilla 2.6.0-test11 and CS4236 card

[Takashi Iwai]

[-- Attachment #1: Type: text/plain, Size: 965 bytes --]

At Tue, 2 Dec 2003 11:06:39 -0600,
Joseph Pingenot wrote:
> 
> Howdy.
> 
> I'm having problems getting the CS4236+ driver to recognize my 
>   CS4236B card.  pnp finds it on boot:
> isapnp: Scanning for PnP cards...
> isapnp: Card 'CS4236B'
> isapnp: 1 Plug & Play card detected total
> 
> but the ALSA driver doesn't pick it up.
> isapnp detection failed and probing for CS4236+ is not supported
> CS4236+ soundcard not found or device busy
> 
> Furthermore, after fudging with manually setting it up via modprobe
>   options, it's still not loading:
> CS4236+ soundcard not found or device busy
> 
> This used to work in the 2.4 series kernel without any modprobe.conf
>   settings; the OSS driver would pick it up.
> 
> Any assistance would be greatly appreciated; this is the only thing holding
>   me back from 2.6 goodness.  ;)

does the attached patch work?  (it's untested at all...)


--
Takashi Iwai <tiwai@suse.de>		ALSA Developer - www.alsa-project.org

[-- Attachment #2: Type: text/plain, Size: 765 bytes --]

--- linux-2.6.0-test11/drivers/pnp/card.c-dist	2003-12-02 18:14:21.000000000 +0100
+++ linux-2.6.0-test11/drivers/pnp/card.c	2003-12-02 18:29:20.000000000 +0100
@@ -26,8 +26,25 @@
 {
 	const struct pnp_card_device_id * drv_id = drv->id_table;
 	while (*drv_id->id){
-		if (compare_pnp_id(card->id,drv_id->id))
-			return drv_id;
+		if (compare_pnp_id(card->id,drv_id->id)) {
+			int i = 0;
+			for (;;) {
+				int found;
+				struct pnp_dev *dev;
+				if (i == PNP_MAX_DEVICES || ! *drv_id->devs[i].id)
+					return drv_id;
+				found = 0;
+				card_for_each_dev(card, dev) {
+					if (compare_pnp_id(dev->id, drv_id->devs[i].id)) {
+						found = 1;
+						break;
+					}
+				}
+				if (! found)
+					break;
+				i++;
+			}
+		}
 		drv_id++;
 	}
 	return NULL;

Re: vanilla 2.6.0-test11 and CS4236 card

[Joseph Pingenot]

>From Takashi Iwai on Tuesday, 02 December, 2003:

Thanks.  I'm trying to track it down via diagnostic printk statements;
  right now, it's failing at least somewhere in snd_cs4231_lib
  (am at cs4231_probe atm; is failing right there atm).
I'll test the patch when I can.

-Joseph
-- 
trelane@digitasaru.net--------------------------------------------------
"We continue to live in a world where all our know-how is locked into
 binary files in an unknown format. If our documents are our corporate
 memory, Microsoft still has us all condemned to Alzheimer's."
    --Simon Phipps, http://theregister.com/content/4/30410.html

Re: vanilla 2.6.0-test11 and CS4236 card

[Joseph Pingenot]

An update.  I've traced the problems down into snd_cs4231_probe().
After adding diagnostic printk statements, I get the following:
     *vcs4231_inb() returned 0xff
     *val is now 0x80
     *cs4231_inb & CS4231_INIT returned TRUE
     cs4231: port = 0x530, id = 0x0
     id is incorrect (id=0x00, but should be 0x0a)
    err in call to snd_cs4231_probe
err in call to snd_cs4231_create!
err running snd_cs4236_create!
CS4236+ soundcard not found or device busy
[all but the last line are mine].  the stuff prefixed by "*" comes from
  here in snd_cs4231_probe:

	for (i = 0; i < 50; i++) {
		mb();
		val = cs4231_inb(chip, CS4231P(REGSEL));
		printk("     *vcs4231_inb() returned 0x%0x\n", val);
		val &= CS4231_INIT;
		printk("     *val is now 0x%0x\n", val);
		/*if (cs4231_inb(chip, CS4231P(REGSEL)) & CS4231_INIT) {*/
		if (val) {
		  printk("     *cs4231_inb & CS4231_INIT returned TRUE\n");
			udelay(2000);
		} else {
			spin_lock_irqsave(&chip->reg_lock, flags);
			snd_cs4231_out(chip, CS4231_MISC_INFO, CS4231_MODE2);
			id = snd_cs4231_in(chip, CS4231_MISC_INFO) & 0x0f;
			printk("     *detecting id: id=%02x\n", id);
			spin_unlock_irqrestore(&chip->reg_lock, flags);
			if (id == 0x0a)
				break;	/* this is valid value */
		}
	}

Anyone know what the io port here does, and what the different return
  values are?  Pointers to documentation is more than welcome!  :)

Thanks!

-Joseph
-- 
trelane@digitasaru.net--------------------------------------------------
"We continue to live in a world where all our know-how is locked into
 binary files in an unknown format. If our documents are our corporate
 memory, Microsoft still has us all condemned to Alzheimer's."
    --Simon Phipps, http://theregister.com/content/4/30410.html

Re: vanilla 2.6.0-test11 and CS4236 card

[Adam Belay]

On Tue, Dec 02, 2003 at 06:31:57PM +0100, Takashi Iwai wrote:
> At Tue, 2 Dec 2003 11:06:39 -0600,
> Joseph Pingenot wrote:
> > 
> > Howdy.
> > 
> > I'm having problems getting the CS4236+ driver to recognize my 
> >   CS4236B card.  pnp finds it on boot:
> > isapnp: Scanning for PnP cards...
> > isapnp: Card 'CS4236B'
> > isapnp: 1 Plug & Play card detected total
> > 
> > but the ALSA driver doesn't pick it up.
> > isapnp detection failed and probing for CS4236+ is not supported
> > CS4236+ soundcard not found or device busy
> > 
> > Furthermore, after fudging with manually setting it up via modprobe
> >   options, it's still not loading:
> > CS4236+ soundcard not found or device busy
> > 
> > This used to work in the 2.4 series kernel without any modprobe.conf
> >   settings; the OSS driver would pick it up.
> > 
> > Any assistance would be greatly appreciated; this is the only thing holding
> >   me back from 2.6 goodness.  ;)
> 
> does the attached patch work?  (it's untested at all...)


> --- linux-2.6.0-test11/drivers/pnp/card.c-dist	2003-12-02 18:14:21.000000000 +0100
> +++ linux-2.6.0-test11/drivers/pnp/card.c	2003-12-02 18:29:20.000000000 +0100
> @@ -26,8 +26,25 @@
>  {
>  	const struct pnp_card_device_id * drv_id = drv->id_table;
>  	while (*drv_id->id){
> -		if (compare_pnp_id(card->id,drv_id->id))
> -			return drv_id;
> +		if (compare_pnp_id(card->id,drv_id->id)) {
> +			int i = 0;
> +			for (;;) {
> +				int found;
> +				struct pnp_dev *dev;
> +				if (i == PNP_MAX_DEVICES || ! *drv_id->devs[i].id)
> +					return drv_id;
> +				found = 0;
> +				card_for_each_dev(card, dev) {
> +					if (compare_pnp_id(dev->id, drv_id->devs[i].id)) {
> +						found = 1;
> +						break;
> +					}
> +				}
> +				if (! found)
> +					break;
> +				i++;
> +			}
> +		}
>  		drv_id++;
>  	}
>  	return NULL;
>


Here's a slight cleanup of the above patch -- tested with the cs4232 driver.
Any additional testing would be appreciated.

--- a/drivers/pnp/card.c	2003-11-26 20:45:38.000000000 +0000
+++ b/drivers/pnp/card.c	2003-12-02 23:06:55.000000000 +0000
@@ -26,8 +26,18 @@
 {
 	const struct pnp_card_device_id * drv_id = drv->id_table;
 	while (*drv_id->id){
-		if (compare_pnp_id(card->id,drv_id->id))
-			return drv_id;
+		if (compare_pnp_id(card->id,drv_id->id)) {
+			int i = 0;
+			for (;;i++) {
+				struct pnp_dev *dev;
+				if (i == PNP_MAX_DEVICES || ! *drv_id->devs[i].id)
+					return drv_id;
+				card_for_each_dev(card, dev) {
+					if (!compare_pnp_id(dev->id, drv_id->devs[i].id))
+						break;
+				}
+			}
+		}
 		drv_id++;
 	}
 	return NULL;

In this case we could also just continue if the MPU device isn't present.  It
would probably be a good convention to do so because if, for whatever reason
(3rd party driver, resource conflicts, etc), the MPU device is busy in any
matched card id set, the entire probe would fail.

How about something like this? (untested)

--- a/sound/isa/cs423x/cs4236.c	2003-11-26 20:43:05.000000000 +0000
+++ b/sound/isa/cs423x/cs4236.c	2003-12-02 23:36:59.000000000 +0000
@@ -294,13 +294,8 @@
 		kfree(cfg);
 		return -EBUSY;
 	}
-	if (id->devs[2].id[0]) {
+	if (id->devs[2].id[0])
 		acard->mpu = pnp_request_card_device(card, id->devs[2].id, NULL);
-		if (acard->mpu == NULL) {
-			kfree(cfg);
-			return -EBUSY;
-		}
-	}

 	/* WSS initialization */
 	pdev = acard->wss;



Thanks,
Adam

Re: vanilla 2.6.0-test11 and CS4236 card

[Takashi Iwai]

Hi Adam,

At Tue, 2 Dec 2003 23:44:32 +0000,
Adam Belay wrote:
> 
> Here's a slight cleanup of the above patch -- tested with the cs4232 driver.
> Any additional testing would be appreciated.
> 
> --- a/drivers/pnp/card.c	2003-11-26 20:45:38.000000000 +0000
> +++ b/drivers/pnp/card.c	2003-12-02 23:06:55.000000000 +0000
> @@ -26,8 +26,18 @@
>  {
>  	const struct pnp_card_device_id * drv_id = drv->id_table;
>  	while (*drv_id->id){
> -		if (compare_pnp_id(card->id,drv_id->id))
> -			return drv_id;
> +		if (compare_pnp_id(card->id,drv_id->id)) {
> +			int i = 0;
> +			for (;;i++) {
> +				struct pnp_dev *dev;
> +				if (i == PNP_MAX_DEVICES || ! *drv_id->devs[i].id)
> +					return drv_id;
> +				card_for_each_dev(card, dev) {
> +					if (!compare_pnp_id(dev->id, drv_id->devs[i].id))
> +						break;
> +				}
> +			}
> +		}
>  		drv_id++;
>  	}
>  	return NULL;
> 

i think the patch above doesn't work properly since it will continue
even if no devices match.  so, the loop after comparison of card id
makes no sense.


> In this case we could also just continue if the MPU device isn't present.  It
> would probably be a good convention to do so because if, for whatever reason
> (3rd party driver, resource conflicts, etc), the MPU device is busy in any
> matched card id set, the entire probe would fail.
> 
> How about something like this? (untested)
> 
> --- a/sound/isa/cs423x/cs4236.c	2003-11-26 20:43:05.000000000 +0000
> +++ b/sound/isa/cs423x/cs4236.c	2003-12-02 23:36:59.000000000 +0000
> @@ -294,13 +294,8 @@
>  		kfree(cfg);
>  		return -EBUSY;
>  	}
> -	if (id->devs[2].id[0]) {
> +	if (id->devs[2].id[0])
>  		acard->mpu = pnp_request_card_device(card, id->devs[2].id, NULL);
> -		if (acard->mpu == NULL) {
> -			kfree(cfg);
> -			return -EBUSY;
> -		}
> -	}
> 
>  	/* WSS initialization */
>  	pdev = acard->wss;

this looks fine to me.


ciao,

Takashi

Re: vanilla 2.6.0-test11 and CS4236 card

[wes schreiner]

I tested a combination of Takashi's version of the patch to 
drivers/pnp/card.c and Adam's patch to sound/isa/cs423x/cs4236.c applied 
to 2.6.0-test11 and report success. This is on a Dell Precision 
Workstation 410 which doesn't have a MPU device. The card has always 
been detected fine by ISAPNP and now when I modpobe snd-cs4236 the 
modules load just like they should. PCM output and input work fine. I 
exercised the audio system by running, at various times, xmms, aplay, 
alsamixer, and audacity and everything worked OK. No kernel-tainting 
modules were loaded. I would be happy to test the final version of these 
patches.

wes schreiner

Re: vanilla 2.6.0-test11 and CS4236 card

[Joseph Pingenot]

>From wes schreiner on Wednesday, 03 December, 2003:
>I tested a combination of Takashi's version of the patch to 
>drivers/pnp/card.c and Adam's patch to sound/isa/cs423x/cs4236.c applied 
>to 2.6.0-test11 and report success. This is on a Dell Precision 
>Workstation 410 which doesn't have a MPU device. The card has always 
>been detected fine by ISAPNP and now when I modpobe snd-cs4236 the 
>modules load just like they should. PCM output and input work fine. I 
>exercised the audio system by running, at various times, xmms, aplay, 
>alsamixer, and audacity and everything worked OK. No kernel-tainting 
>modules were loaded. I would be happy to test the final version of these 
>patches.

Sweet.  Stupid NVidia drivers.  Thanks for the help (I have exactly the
  same system).  I'll report back when I've recovered from NVidia driver's
  mess.  :)

-Joseph
-- 
Joseph===============================================trelane@digitasaru.net
"Asked by CollabNet CTO Brian Behlendorf whether Microsoft will enforce its
 patents against open source projects, Mundie replied, 'Yes, absolutely.'
 An audience member pointed out that many open source projects aren't
 funded and so can't afford legal representation to rival Microsoft's. 'Oh
 well,' said Mundie. 'Get your money, and let's go to court.' 
Microsoft's patents only defensive? http://swpat.ffii.org/players/microsoft

Re: vanilla 2.6.0-test11 and CS4236 card

[Joseph Pingenot]

Well, I applied the patch and it worked beautifully.

THE GOOD NEWS:
--------------

The patch worked great; the problem (as you had very likely surmised)
  was that it wasn't being activated by the Plug-n-Play system, and so the
  card wasn't talking.  [Sorry; I am new to ISA programming (well, systems
  programming in general.  At least I learned some cool stuff]  After I had
  run isapnp on the module sans your patch, or rebooted with your patch,
  there was a big POP! and the card was detected and ready.  It even played
  music!

THE BAD NEWS:
-------------

Both times (with your patch and with isapnp), I had the following problems:
  a) xmms wouldn't talk straight to the card even after selecting the alsa
       output plugin
  b) I started esd (the alsa version) and then told xmms to talk to esd.
       It worked out well for a couple of seconds, but then it crashed
       (both with your patch and with the manually activated (isapnp) one)
       
       It crashed HARD.

       It crashed so hard that it wiped out some files, like apt, I presume
         when it was cleaning "orphans" in /; the net result is that my
	 box is semi-hosed, since things like tr and apt-get and route
	 are now full of '\0' characters instead of their actual contents.

Sound played for about a second, then the machine hung hard (I don't think
  that the Magic SysRq keys even did anything), and the last couple of
  tenths of a second of sound just repeated over and over and over until
  I hit reset.  I don't have an oops; it was highly hosed.

When I can pull my system back together, I can try again.  Please help
  me find this new bug, so that the cs4236 driver is safe for other
  Dell P410 users!  :)

-Joseph

-- 
Joseph===============================================trelane@digitasaru.net
"Asked by CollabNet CTO Brian Behlendorf whether Microsoft will enforce its
 patents against open source projects, Mundie replied, 'Yes, absolutely.'
 An audience member pointed out that many open source projects aren't
 funded and so can't afford legal representation to rival Microsoft's. 'Oh
 well,' said Mundie. 'Get your money, and let's go to court.' 
Microsoft's patents only defensive? http://swpat.ffii.org/players/microsoft

Re: vanilla 2.6.0-test11 and CS4236 card

[Takashi Iwai]

Hi,

At Tue, 2 Dec 2003 21:17:51 -0600,
Joseph Pingenot wrote:
> 
> THE BAD NEWS:
> -------------
> 
> Both times (with your patch and with isapnp), I had the following problems:
>   a) xmms wouldn't talk straight to the card even after selecting the alsa
>        output plugin

what was the error exactly?

>   b) I started esd (the alsa version) and then told xmms to talk to esd.
>        It worked out well for a couple of seconds, but then it crashed
>        (both with your patch and with the manually activated (isapnp) one)
>        
>        It crashed HARD.

oh, that's bad.

>        It crashed so hard that it wiped out some files, like apt, I presume
>          when it was cleaning "orphans" in /; the net result is that my
> 	 box is semi-hosed, since things like tr and apt-get and route
> 	 are now full of '\0' characters instead of their actual contents.
> 
> Sound played for about a second, then the machine hung hard (I don't think
>   that the Magic SysRq keys even did anything), and the last couple of
>   tenths of a second of sound just repeated over and over and over until
>   I hit reset.  I don't have an oops; it was highly hosed.
> 
> When I can pull my system back together, I can try again.  Please help
>   me find this new bug, so that the cs4236 driver is safe for other
>   Dell P410 users!  :)

i'm puzzled because cs4236's pcm code is basically identical with
cs4231, and cs4232 seems working (at least the last time i tested).
anyway, please show the kernel config and check any other apps do
similar oops / crash.


--
Takashi Iwai <tiwai@suse.de>		ALSA Developer - www.alsa-project.org

Re: vanilla 2.6.0-test11 and CS4236 card

[Joseph Pingenot]

>From Takashi Iwai on Wednesday, 03 December, 2003:
>Hi,
>At Tue, 2 Dec 2003 21:17:51 -0600,
>Joseph Pingenot wrote:
>> THE BAD NEWS:
>> -------------
>> Both times (with your patch and with isapnp), I had the following problems:
>>   a) xmms wouldn't talk straight to the card even after selecting the alsa
>>        output plugin
>what was the error exactly?

Standard thing when it can't talk to the sound card:

"Please check that:
  1. You have the correct output plugin selected
  2. No other programs is(sic) blocking the soundcard
  3. Your soundcard is configured properly
  
         [OK]"

FWIW, ALSA for the cs4281 chip on my home box works 100%, and xmms can talk
  to it without problem.  These are different distros (gentoo@home, debian@work)
  so it may also be a problem in Debian, but I don't think so.

>>   b) I started esd (the alsa version) and then told xmms to talk to esd.
>>        It worked out well for a couple of seconds, but then it crashed
>>        (both with your patch and with the manually activated (isapnp) one)
>>        
>>        It crashed HARD.
>oh, that's bad.

Ja.  I'm gonna have to sort it out whenever I can get my homework done.
  Either this afternoon or tomorrow.  Just gotta copy some files over by
  floppy (e.g. route) and then tar them over from another machine.

>>        It crashed so hard that it wiped out some files, like apt, I presume
>>          when it was cleaning "orphans" in /; the net result is that my
>> 	 box is semi-hosed, since things like tr and apt-get and route
>> 	 are now full of '\0' characters instead of their actual contents.
>> 
>> Sound played for about a second, then the machine hung hard (I don't think
>>   that the Magic SysRq keys even did anything), and the last couple of
>>   tenths of a second of sound just repeated over and over and over until
>>   I hit reset.  I don't have an oops; it was highly hosed.
>> 
>> When I can pull my system back together, I can try again.  Please help
>>   me find this new bug, so that the cs4236 driver is safe for other
>>   Dell P410 users!  :)
>i'm puzzled because cs4236's pcm code is basically identical with
>cs4231, and cs4232 seems working (at least the last time i tested).
>anyway, please show the kernel config and check any other apps do
>similar oops / crash.

Hmm.  I dunno.  The only odd thing atm is that the nvidia driver was loaded
  and had reported a couple of bugs.  When the machine's up next time,
  I'll try it sans nvidia and see what we get.  I don't suspect that's the
  problem, as everything but sound worked 100%, but we'll see.

Other than my diagnostic print statements or your pnp patch, everything
  at runtime was vanilla (except nvidia of course)

-Joseph

-- 
Joseph===============================================trelane@digitasaru.net
"Asked by CollabNet CTO Brian Behlendorf whether Microsoft will enforce its
 patents against open source projects, Mundie replied, 'Yes, absolutely.'
 An audience member pointed out that many open source projects aren't
 funded and so can't afford legal representation to rival Microsoft's. 'Oh
 well,' said Mundie. 'Get your money, and let's go to court.' 
Microsoft's patents only defensive? http://swpat.ffii.org/players/microsoft

Re: vanilla 2.6.0-test11 and CS4236 card

[Joseph Pingenot]

Status update.

I removed the nvida driver module (thus making it, contrary to what I
  had considered vanilla) *truly* vanilla.  [Sorry about that; I really
  didn't mean to mislead.]
Upon reboot, the cs4236B "card" works 100% (although, for some reason, the
  xmms alsa drivers can't talk to it directly, but that's survivable;
  I'm using esd right now anyway; it's likely a debian problem or a
  Joe Confusigration problem).
My compliments and gratitude to all those who helped; I have now moved
  my workstation to 2.6.  I'll buy a student copy of SuSE when my laptop
  gets here next week.  ;)

-Joseph

-- 
trelane@digitasaru.net--------------------------------------------------
"We continue to live in a world where all our know-how is locked into
 binary files in an unknown format. If our documents are our corporate
 memory, Microsoft still has us all condemned to Alzheimer's."
    --Simon Phipps, http://theregister.com/content/4/30410.html