1.0.0pre1: au88x0 driver locks up kernel

1.0.0pre1: au88x0 driver locks up kernel

[Carolyn and Eric Hathaway]

I'm running RedHat 7.3 (RH kernel 2.4.20-20.7, gcc 2.96-113), and have a
generic Aureal Vortex 1 card (au8820 chipset).  Compiling and installing
the 1.0.0pre1 ALSA driver works fine, but if I do a 'modprobe
snd-au8820', the computer locks up instantly (can't even ssh in from
another computer).

Going back through recent CVS commits, I found that the problem is
caused by the changes described in the following alsa-cvslog message:

http://sourceforge.net/mailarchive/message.php?msg_id=6552838

Specifically, the problem is caused by the changes Takashi made to
au88x0.h and au88x0.c to eliminate calls to pci_request_regions() and
pci_release_regions().  If I revert those two files to their previous
versions, the driver compiles and works fine.  I'm not an ALSA driver
guru, but I didn't see anything immediately wrong with the changes
introduced by the above CVS commit that would cause such a drastic
kernel error.  Perhaps someone more familiar with the code in question 
could take another look at those changes?

Regards,
-Eric Hathaway




-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
Does SourceForge.net help you be more productive?  Does it
help you create better code?  SHARE THE LOVE, and help us help
YOU!  Click Here: http://sourceforge.net/donate/

Re: 1.0.0pre1: au88x0 driver locks up kernel

[Takashi Iwai]

At Thu, 20 Nov 2003 00:42:33 -0500,
Carolyn and Eric Hathaway wrote:
> 
> I'm running RedHat 7.3 (RH kernel 2.4.20-20.7, gcc 2.96-113), and have a
> generic Aureal Vortex 1 card (au8820 chipset).  Compiling and installing
> the 1.0.0pre1 ALSA driver works fine, but if I do a 'modprobe
> snd-au8820', the computer locks up instantly (can't even ssh in from
> another computer).
> 
> Going back through recent CVS commits, I found that the problem is
> caused by the changes described in the following alsa-cvslog message:
> 
> http://sourceforge.net/mailarchive/message.php?msg_id=6552838
> 
> Specifically, the problem is caused by the changes Takashi made to
> au88x0.h and au88x0.c to eliminate calls to pci_request_regions() and
> pci_release_regions().  If I revert those two files to their previous
> versions, the driver compiles and works fine.  I'm not an ALSA driver
> guru, but I didn't see anything immediately wrong with the changes
> introduced by the above CVS commit that would cause such a drastic
> kernel error.  Perhaps someone more familiar with the code in question 
> could take another look at those changes?

hmm, i don't see where hits the bug, but if the older version really
works, i'm willing to revert to it.  it'd be anyway better to add two
wrapper functions for 2.2.


thanks,

Takashi


-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
Does SourceForge.net help you be more productive?  Does it
help you create better code?  SHARE THE LOVE, and help us help
YOU!  Click Here: http://sourceforge.net/donate/

Re: 1.0.0pre1: au88x0 driver locks up kernel

[Sergey Vlasov]

[-- Attachment #1: Type: text/plain, Size: 1558 bytes --]

On Thu, Nov 20, 2003 at 11:21:43AM +0100, Takashi Iwai wrote:
> At Thu, 20 Nov 2003 00:42:33 -0500,
> Carolyn and Eric Hathaway wrote:
> > 
> > I'm running RedHat 7.3 (RH kernel 2.4.20-20.7, gcc 2.96-113), and have a
> > generic Aureal Vortex 1 card (au8820 chipset).  Compiling and installing
> > the 1.0.0pre1 ALSA driver works fine, but if I do a 'modprobe
> > snd-au8820', the computer locks up instantly (can't even ssh in from
> > another computer).
> > 
> > Going back through recent CVS commits, I found that the problem is
> > caused by the changes described in the following alsa-cvslog message:
> > 
> > http://sourceforge.net/mailarchive/message.php?msg_id=6552838
> > 
> > Specifically, the problem is caused by the changes Takashi made to
> > au88x0.h and au88x0.c to eliminate calls to pci_request_regions() and
> > pci_release_regions().  If I revert those two files to their previous
> > versions, the driver compiles and works fine.  I'm not an ALSA driver
> > guru, but I didn't see anything immediately wrong with the changes
> > introduced by the above CVS commit that would cause such a drastic
> > kernel error.  Perhaps someone more familiar with the code in question 
> > could take another look at those changes?
> 
> hmm, i don't see where hits the bug, but if the older version really
> works, i'm willing to revert to it.  it'd be anyway better to add two
> wrapper functions for 2.2.

Probably this part had broken it (all addresses in hwread()/hwwrite()
became wrong):

-       unsigned long *mmio;
+       unsigned long mmio;

[-- Attachment #2: Type: application/pgp-signature, Size: 189 bytes --]

Re: 1.0.0pre1: au88x0 driver locks up kernel

[Takashi Iwai]

At Thu, 20 Nov 2003 15:17:31 +0300,
Sergey Vlasov wrote:
> 
> [1  <text/plain; us-ascii (7bit)>]
> On Thu, Nov 20, 2003 at 11:21:43AM +0100, Takashi Iwai wrote:
> > At Thu, 20 Nov 2003 00:42:33 -0500,
> > Carolyn and Eric Hathaway wrote:
> > > 
> > > I'm running RedHat 7.3 (RH kernel 2.4.20-20.7, gcc 2.96-113), and have a
> > > generic Aureal Vortex 1 card (au8820 chipset).  Compiling and installing
> > > the 1.0.0pre1 ALSA driver works fine, but if I do a 'modprobe
> > > snd-au8820', the computer locks up instantly (can't even ssh in from
> > > another computer).
> > > 
> > > Going back through recent CVS commits, I found that the problem is
> > > caused by the changes described in the following alsa-cvslog message:
> > > 
> > > http://sourceforge.net/mailarchive/message.php?msg_id=6552838
> > > 
> > > Specifically, the problem is caused by the changes Takashi made to
> > > au88x0.h and au88x0.c to eliminate calls to pci_request_regions() and
> > > pci_release_regions().  If I revert those two files to their previous
> > > versions, the driver compiles and works fine.  I'm not an ALSA driver
> > > guru, but I didn't see anything immediately wrong with the changes
> > > introduced by the above CVS commit that would cause such a drastic
> > > kernel error.  Perhaps someone more familiar with the code in question 
> > > could take another look at those changes?
> > 
> > hmm, i don't see where hits the bug, but if the older version really
> > works, i'm willing to revert to it.  it'd be anyway better to add two
> > wrapper functions for 2.2.
> 
> Probably this part had broken it (all addresses in hwread()/hwwrite()
> became wrong):
> 
> -       unsigned long *mmio;
> +       unsigned long mmio;

yes, i understand now.  the calculation of offset value had to be
changed, too.


Takashi


-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
Does SourceForge.net help you be more productive?  Does it
help you create better code?  SHARE THE LOVE, and help us help
YOU!  Click Here: http://sourceforge.net/donate/