Panic - no sound devices after upgrade

arch-general Tom Gundersen 3 participants 10 messages Jul 21, 2011 – Jul 25, 2011 11 messages not in the archive Sound

Share & tools

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): CAFEhRJf-5U=cS3iptpvGEVSHV+S=aBF_mWfWkC_QMRty_WnoPA@mail.gmail.com

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): CAFEhRJd9ZCSpx52S__vMdWK18mhMQrRoU9R5zNaZE=wksLRTvg@mail.gmail.com

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): CAFEhRJczj=_B7fK7GvBEnqpB2GJGnCnGkoA2w5vzbFfM2wAGZw@mail.gmail.com

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): CAG-2HqWqATSJ=oJTTvULXwPEG82yN+8e6GJFs759hPA-=krAOQ@mail.gmail.com

A message that is not in the archive (sent off-list or not yet received): [email protected]

A message that is not in the archive (sent off-list or not yet received): CAG-2HqX23C79vfccnu0tMqx+6K9Z5pS0_HekaLizQx7XfkPQ1Q@mail.gmail.com

A message that is not in the archive (sent off-list or not yet received): [email protected]

Tom Gundersen

author

2011-07-21 13:00 UTC #11 lore raw

Adrian, Takashi: I added you to cc, as you last touched this code, sorry if I should have cc'ed someone else. It appears that the HDSPM module does not recognise cards with firmware revision 0xCC:

Quoted 2 lines · Fons Adriaensen

Quoted 2 lines
Quoted 2 lines
Quoted 1 line
Quoted 1 line
* The driver for my card (RME HDSP MADI, snd_hdpsm) is loaded.
[ 48.057757] HDSPM: unknown firmware revision cc

[...]

Quoted 2 lines

Today I tried again downgrading the kernel to 2.6.37, to check if the firmware revision error would occur. It doesn't.

[...]

Quoted 1 line

Tentative conclusion: the snd_hdspm driver in 2.6.39 is not OK.

Ah, I see. Thanks for the clarification. This means that the kernel is not able to recognise your card. Pre-.39 it would load anyway, but since .39 it gives the warning you posted and refuses to load. Your firmware revision is 0xCC, which is not recognised. If your card is MADI, the kernel expects the firmware revision to be 0xD2 (or starting in 3.0, it also accepts 0xCF).

I guess this can be fixed easily by adding your firmware revision to the list of accepted revs.

Adrian: is any more info needed to add this rev?

Cheers,

Tom

Fons Adriaensen

2011-07-21 13:18 UTC #12 lore raw

Quoted 20 lines · Tom Gundersen

Adrian, Takashi: I added you to cc, as you last touched this code, sorry if I should have cc'ed someone else. It appears that the HDSPM module does not recognise cards with firmware revision 0xCC:
Quoted 2 lines · Fons Adriaensen
Quoted 2 lines
Quoted 2 lines
Quoted 1 line
Quoted 1 line
* The driver for my card (RME HDSP MADI, snd_hdpsm) is loaded.
[ 48.057757] HDSPM: unknown firmware revision cc
[...]
Quoted 2 lines
Today I tried again downgrading the kernel to 2.6.37, to check if the firmware revision error would occur. It doesn't.
[...]
Quoted 1 line
Tentative conclusion: the snd_hdspm driver in 2.6.39 is not OK.
Ah, I see. Thanks for the clarification. This means that the kernel is not able to recognise your card. Pre-.39 it would load anyway, but since .39 it gives the warning you posted and refuses to load. Your firmware revision is 0xCC, which is not recognised. If your card is MADI, the kernel expects the firmware revision to be 0xD2 (or starting in 3.0, it also accepts 0xCF).

Quoted 2 lines

I guess this can be fixed easily by adding your firmware revision to the list of accepted revs.

I have cards with CC and CF, and maybe also D2 (haven't checked them all).

Signature · 1 line

FA

Adrian Knoth

2011-07-21 14:07 UTC #13 lore raw

In reply to Tom Gundersen

On 07/21/11 15:00, Tom Gundersen wrote:

Hi!

Quoted 6 lines

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 1 line
Quoted 1 line
* The driver for my card (RME HDSP MADI, snd_hdpsm) is loaded.
[   48.057757] HDSPM: unknown firmware revision cc
firmware revision is 0xCC, which is not recognised. If your card is MADI, the kernel expects the firmware revision to be 0xD2 (or starting in 3.0, it also accepts 0xCF).

Exactly.

Quoted 2 lines

I guess this can be fixed easily by adding your firmware revision to the list of accepted revs.

I hope so, too. Please try the attached patch.

JFTR: You need at least ALSA-1.0.24 userland (or the corresponding hdspm.h respectively) for the new driver (2.6.39 and above). Likewise, if you intend to use hdspmixer, you need the yet unreleased 1.0.24.2 version of alsa-tools, so go for the git version:

http://git.alsa-project.org/?p=alsa-tools.git

HTH

sound/pci/rme9652/hdspm.c +2 -0

index 7222151..18d3424 100644
@@ -521,6 +521,7 @@ MODULE_SUPPORTED_DEVICE("{{RME HDSPM-MADI}}");
 #define HDSPM_DMA_AREA_KILOBYTES (HDSPM_DMA_AREA_BYTES/1024)
 
 /* revisions >= 230 indicate AES32 card */
+#define HDSPM_MADI_ANCIENT_REV	204
 #define HDSPM_MADI_OLD_REV	207
 #define HDSPM_MADI_REV		210
 #define HDSPM_RAYDAT_REV	211

@@ -6484,6 +6485,7 @@ static int __devinit snd_hdspm_create(struct snd_card *card,
 	switch (hdspm->firmware_rev) {
 	case HDSPM_MADI_REV:
 	case HDSPM_MADI_OLD_REV:
+	case HDSPM_MADI_ANCIENT_REV:
 		hdspm->io_type = MADI;
 		hdspm->card_name = "RME MADI";
 		hdspm->midiPorts = 3;

Tom Gundersen

author

2011-07-21 17:26 UTC #14 lore raw

Quoted 11 lines · Adrian Knoth

On 07/21/11 15:00, Tom Gundersen wrote:
Hi!
Quoted 6 lines
Quoted 2 lines
Quoted 2 lines
Quoted 2 lines
Quoted 1 line
Quoted 1 line
* The driver for my card (RME HDSP MADI, snd_hdpsm) is loaded.
[ 48.057757] HDSPM: unknown firmware revision cc
firmware revision is 0xCC, which is not recognised. If your card is MADI, the kernel expects the firmware revision to be 0xD2 (or starting in 3.0, it also accepts 0xCF).

Quoted 6 lines

Exactly.
Quoted 2 lines
I guess this can be fixed easily by adding your firmware revision to the list of accepted revs.
I hope so, too. Please try the attached patch.

Fons (or anyone else who could reproduce the problem),

If you want to try out Adrian's patch, I uploaded a kernel with the patch applied here:
<https://dev.archlinux.org/~tomegun/kernel26-2.6.39.1.3-1.1-i686.pkg.tar.xz>.

Cheers,

-t

Fons Adriaensen

2011-07-21 19:02 UTC #15 lore raw

Quoted 5 lines · Tom Gundersen

Fons (or anyone else who could reproduce the problem),
If you want to try out Adrian's patch, I uploaded a kernel with the patch applied here:
<https://dev.archlinux.org/~tomegun/kernel26-2.6.39.1.3-1.1-i686.pkg.tar.xz>.

Wow ! I'll have (hope to have) 5 machines to test it tomorrow !

Thanks for providing this !

Ciao,

Signature · 1 line

FA

Fons Adriaensen

2011-07-22 19:37 UTC #16 lore raw

In reply to Tom Gundersen

Quoted 5 lines · Tom Gundersen

Fons (or anyone else who could reproduce the problem),
If you want to try out Adrian's patch, I uploaded a kernel with the patch applied here:
<https://dev.archlinux.org/~tomegun/kernel26-2.6.39.1.3-1.1-i686.pkg.tar.xz>.

Installed on two machines so far, more tomorrow. Seems to have solved the problem :-) :-)

An unrelated question: I see references to the RME Raydat interface. Does that mean it's supported ? Also with period sizes < 1024 ?

Ciao,

Signature · 1 line

FA

Adrian Knoth

2011-07-22 23:14 UTC #17 lore raw

Quoted 8 lines · Fons Adriaensen

Quoted 5 lines
Fons (or anyone else who could reproduce the problem),
If you want to try out Adrian's patch, I uploaded a kernel with the patch applied here:
<https://dev.archlinux.org/~tomegun/kernel26-2.6.39.1.3-1.1-i686.pkg.tar.xz>.
Installed on two machines so far, more tomorrow. Seems to have solved the problem :-) :-)

Good to know. I'll wait for your final confirmation and then commit it.

Quoted 3 lines

An unrelated question: I see references to the RME Raydat interface. Does that mean it's supported ? Also with period sizes < 1024 ?

It is. It's actually the card I'm developing on. So far, all period sizes from 64 to 8k (iirc) work. I don't know if 32 is supposed to work, at least on my machine, it doesn't.

HTH

Signature · 1 line

mail: [email protected]  	http://adi.thur.de	PGP/GPG: key via keyserver

Fons Adriaensen

2011-07-23 14:21 UTC #18 lore raw

Quoted 8 lines · Adrian Knoth

Quoted 6 lines
Quoted 3 lines
If you want to try out Adrian's patch, I uploaded a kernel with the patch applied here:
<https://dev.archlinux.org/~tomegun/kernel26-2.6.39.1.3-1.1-i686.pkg.tar.xz>.
Installed on two machines so far, more tomorrow.
Seems to have solved the problem :-) :-)
Good to know. I'll wait for your final confirmation and then commit it.

One problem: the gain settings have changed, now having a range of 0..64. How do these values map to the internal ones (0..0xffff) ? I *hope* there is value corresponding to EXACTLY 0x8000 (gain = 1) !!! The reason is that MADI is bit-exact and is used to tranport other values than just audio samples. This works of course only with exact unity gain.

BTW, why was this change made ? For pro users (are there other of this card) this is a regression.

Ciao,

Signature · 1 line

FA

Adrian Knoth

2011-07-23 15:08 UTC #19 lore raw

On 07/23/11 16:21, Fons Adriaensen wrote:

Note: I suggest to narrow the CC list a bit and better move the discussion to alsa-devel. At least, let's drop LKML and arch.

Quoted 6 lines

Quoted 2 lines
Quoted 1 line
Seems to have solved the problem :-) :-)
Good to know. I'll wait for your final confirmation and then commit it.
One problem: the gain settings have changed, now having a range of 0..64. How do these values map to the internal ones (0..0xffff) ? I *hope* there is value corresponding to EXACTLY 0x8000 (gain = 1) !!!

For the simple ALSA mixer, the driver does

   gain = ucontrol->value.integer.value[0]*UNITY_GAIN/64;

UNITY_GAIN is 32768 (0x8000), so if you select 64 on the ALSA side, it gives you exactly what you need. Or in other words: faders to 100% in alsamixer, that's it.

With hdspmixer or any other software that's using the full mixer, you can still use the fine-grained 65k steps.

Quoted 1 line

BTW, why was this change made?

No idea. I reckon faberman was tired of stepping through 65k mixer levels in alsamixer. ;) (disclaimer: I never used the old driver, so I don't know how it was before)

Cheers

Fons Adriaensen

2011-07-25 15:26 UTC #20 lore raw

In reply to Adrian Knoth

Quoted 10 lines · Adrian Knoth

Quoted 8 lines
Quoted 5 lines
Fons (or anyone else who could reproduce the problem),
If you want to try out Adrian's patch, I uploaded a kernel with the patch applied here:
<https://dev.archlinux.org/~tomegun/kernel26-2.6.39.1.3-1.1-i686.pkg.tar.xz>.
Installed on two machines so far, more tomorrow.
Seems to have solved the problem :-) :-)
Good to know. I'll wait for your final confirmation and then commit it.

More news, not all good. After trying out all combinations and doing it again to be sure it seems that:

* kernel26-2.6.39.1.3-1.1-i686.pkg.tar.xz and its hdspm driver work well on the five machines that have cards with firmware revision CC.

* With newer cards (they arrived just two weeks ago) with revision CF things seem to work but there are crackles in the sound - not loud but clear enough. Sounds like single samples being modified every now and then.

- Older cards on the same machine with the same kernel work fine.
- Kernel 2.6.37 on the same machine with the new cards works fine as well.

So there must have been some regression.

For now I have to revert two machines (those using the new cards) to 2.6.37.

Ciao,

Signature · 1 line

FA