DMA_AUTOINIT missing in asm-sparc64/dma.h

DMA_AUTOINIT missing in asm-sparc64/dma.h

[Jeremy Huddleston]

[-- Attachment #1: Type: text/plain, Size: 2744 bytes --]

I was going to test out various PCI sound cards I've had lying around
when I noticed that DMA_AUTOINIT isn't defined in asm-sparc64/dma.h like
it is in other archs... most archs defined it as 2 while arm defines it
as 4.


  CC [M]  sound/oss/dmabuf.o
sound/oss/dmabuf.c: In function `sound_start_dma':
sound/oss/dmabuf.c:155: warning: implicit declaration of function
`disable_dma'
sound/oss/dmabuf.c:156: warning: implicit declaration of function
`clear_dma_ff'
sound/oss/dmabuf.c:157: warning: implicit declaration of function
`set_dma_mode'
sound/oss/dmabuf.c:158: warning: implicit declaration of function
`set_dma_addr'
sound/oss/dmabuf.c:159: warning: implicit declaration of function
`set_dma_count'
sound/oss/dmabuf.c:160: warning: implicit declaration of function
`enable_dma'
sound/oss/dmabuf.c: In function `DMAbuf_get_buffer_pointer':
sound/oss/dmabuf.c:653: warning: implicit declaration of function
`get_dma_residue'
sound/oss/dmabuf.c: In function `local_start_dma':
sound/oss/dmabuf.c:943: error: `DMA_AUTOINIT' undeclared (first use in
this function)
sound/oss/dmabuf.c:943: error: (Each undeclared identifier is reported
only once
sound/oss/dmabuf.c:943: error: for each function it appears in.)
make[2]: *** [sound/oss/dmabuf.o] Error 1
make[1]: *** [sound/oss] Error 2
make: *** [sound] Error 2

/usr/src/linux $ grep -nR DMA_AUTOINIT .
./arch/arm/kernel/dma-isa.c:31:#define ISA_DMA_AUTOINIT 0x10
./arch/arm/kernel/dma-isa.c:121:                if (dma->dma_mode &
DMA_AUTOINIT)
./arch/arm/kernel/dma-isa.c:122:                        mode |=
ISA_DMA_AUTOINIT;
./include/asm-ppc64/dma.h:150:#define DMA_AUTOINIT       0x10
./include/asm-m68k/apollodma.h:120:#define DMA_AUTOINIT 0x10
./include/asm-parisc/dma.h:43:#define DMA_AUTOINIT     0x10
./include/asm-mips/dma.h:151:#define DMA_AUTOINIT       0x10
./include/asm-ppc/dma.h:176:#define DMA_AUTOINIT                0x10
./include/asm-x86_64/dma.h:135:#define DMA_AUTOINIT     0x10
./include/asm-arm/dma.h:23:#define DMA_AUTOINIT  4
./include/asm-alpha/dma.h:197:#define DMA_AUTOINIT      0x10
./include/asm-i386/dma.h:135:#define DMA_AUTOINIT       0x10
./include/asm-arm26/dma.h:77:#define DMA_AUTOINIT        4
./include/asm-sh/mpc1211/dma.h:135:#define DMA_AUTOINIT 0x10
./include/asm/dma.h:135:#define DMA_AUTOINIT    0x10
./sound/oss/dmabuf.c:943:       sound_start_dma(dmap,
dmap->raw_buf_phys, dmap->bytes_in_use, dma_mode | DMA_AUTOINIT);
./sound/oss/os.h:48:#undef PSEUDO_DMA_AUTOINIT
./sound/pci/ice1712/ice1712.h:210:#define ICE1712_DMA_AUTOINIT         
0x10
./sound/pci/ice1712/ice1712.c:595:      outb(ICE1712_DMA_MODE_WRITE |
ICE1712_DMA_AUTOINIT, ice->ddma_port + 0x0b);
<./sound/isa snipped>

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: DMA_AUTOINIT missing in asm-sparc64/dma.h

[David S. Miller]

On Sat, 26 Jun 2004 03:11:23 -0700
Jeremy Huddleston <eradicator@gentoo.org> wrote:

> I was going to test out various PCI sound cards I've had lying around
> when I noticed that DMA_AUTOINIT isn't defined in asm-sparc64/dma.h like
> it is in other archs... most archs defined it as 2 while arm defines it
> as 4.

These pieces of code you are trying to compile are attempting
to do ISA DMA controller programming.  Such things do not
exist on Sparc.

I never get these errors and I have a large number of PCI sound
cards enabled in my config, which is arch/sparc64/defconfig BTW.

Which card are you enabling in the build?

Re: DMA_AUTOINIT missing in asm-sparc64/dma.h

[Jeremy Huddleston]

[-- Attachment #1: Type: text/plain, Size: 1006 bytes --]

On Sat, 2004-06-26 at 10:02, David S. Miller wrote:
> On Sat, 26 Jun 2004 03:11:23 -0700
> Jeremy Huddleston <eradicator@gentoo.org> wrote:
> 
> > I was going to test out various PCI sound cards I've had lying around
> > when I noticed that DMA_AUTOINIT isn't defined in asm-sparc64/dma.h like
> > it is in other archs... most archs defined it as 2 while arm defines it
> > as 4.
> 
> These pieces of code you are trying to compile are attempting
> to do ISA DMA controller programming.  Such things do not
> exist on Sparc.
> 
> I never get these errors and I have a large number of PCI sound
> cards enabled in my config, which is arch/sparc64/defconfig BTW.

arch/sparc64/defconfig does not enable OSS which is why you haven't
noticed this problem.

> Which card are you enabling in the build?

dmabuf.o (which has the reference to DMA_AUTOINIT) is compiled
regardless of which oss modules you select.  I caught this with just
emu10k1, es1370, and es1371 selected under OSS.



[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: DMA_AUTOINIT missing in asm-sparc64/dma.h

[David S. Miller]

On Sat, 26 Jun 2004 12:17:37 -0700
Jeremy Huddleston <eradicator@gentoo.org> wrote:

> > Which card are you enabling in the build?
> 
> dmabuf.o (which has the reference to DMA_AUTOINIT) is compiled
> regardless of which oss modules you select.  I caught this with just
> emu10k1, es1370, and es1371 selected under OSS.

There are perfectly compilable ALSA drivers for these cards,
why not use those?

Yes, the OSS layer does a lot of PC-centric stuff, better to
stick with the ALSA drivers as I do watch over them for
portability when changes are made.

Re: DMA_AUTOINIT missing in asm-sparc64/dma.h

[Jeremy Huddleston]

[-- Attachment #1: Type: text/plain, Size: 1334 bytes --]

On Sat, 2004-06-26 at 14:37, David S. Miller wrote:
> On Sat, 26 Jun 2004 12:17:37 -0700
> Jeremy Huddleston <eradicator@gentoo.org> wrote:
> 
> > > Which card are you enabling in the build?
> > 
> > dmabuf.o (which has the reference to DMA_AUTOINIT) is compiled
> > regardless of which oss modules you select.  I caught this with just
> > emu10k1, es1370, and es1371 selected under OSS.
> 
> There are perfectly compilable ALSA drivers for these cards,
> why not use those?

Because I was having trouble with the alsa drivers, so I thought I'd
give the oss drivers a try...

> Yes, the OSS layer does a lot of PC-centric stuff, better to
> stick with the ALSA drivers as I do watch over them for
> portability when changes are made.

Yeah, I figure that's a safe bet... I just wanted to check it out...
perhaps the Kconfig should be updated to exclude OSS as a selectable
option on sparcs since it won't even compile because of the PC-centric
stuff.

And an odd thing I've noticed with my es1371 (Ensoniq 5880 AudioPCI (rev
02)) card... I can use it through the OSS emulation layer, but I can't
use alsa-specific applications with it... ie. I can use aumix/sox but
not alsamixer/aplay.  I'm going to try updating alsa-{lib,utils} from
1.0.4 to 1.0.5 right now, but I don't think that's the problem...


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: DMA_AUTOINIT missing in asm-sparc64/dma.h

[David S. Miller]

On Sat, 26 Jun 2004 15:22:58 -0700
Jeremy Huddleston <eradicator@gentoo.org> wrote:

> And an odd thing I've noticed with my es1371 (Ensoniq 5880 AudioPCI (rev
> 02)) card... I can use it through the OSS emulation layer, but I can't
> use alsa-specific applications with it... ie. I can use aumix/sox but
> not alsamixer/aplay.  I'm going to try updating alsa-{lib,utils} from
> 1.0.4 to 1.0.5 right now, but I don't think that's the problem...

Probably errors in the 32-bit compat layer.  I bet if you build the
alsa apps native 64-bit they work.

Re: DMA_AUTOINIT missing in asm-sparc64/dma.h

[Jason Wever]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Sat, 26 Jun 2004, David S. Miller wrote:

> Yes, the OSS layer does a lot of PC-centric stuff, better to
> stick with the ALSA drivers as I do watch over them for
> portability when changes are made.

On a side note, has cs4231 ever been at a point in the 2.5 or 2.6 kernels 
where it has been usable?  It'll compile, build and load fine at least for 
PCI systems (SBUS is hit and miss whether it'll build but that's more 
typically core ALSA components) but the audio produced is never  
understandable.  On PCI I have tried turning on all of the debugging 
options but it doesn't seem to generate any helpful errors.

Thanks,
- -- 
Jason Wever
Gentoo/Sparc Co-Team Lead
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)

iD8DBQFA3gXmdKvgdVioq28RAqiAAJ4mHIPxGAwWh8DWeKSb2AFup+fszACdFPl9
4TAie6t4aH1gnqfdmX80svo=sqKP
-----END PGP SIGNATURE-----

Re: DMA_AUTOINIT missing in asm-sparc64/dma.h

[David S. Miller]

On Sat, 26 Jun 2004 17:25:24 -0600 (MDT)
Jason Wever <weeve@gentoo.org> wrote:

> On a side note, has cs4231 ever been at a point in the 2.5 or 2.6 kernels 
> where it has been usable?  It'll compile, build and load fine at least for 
> PCI systems (SBUS is hit and miss whether it'll build but that's more 
> typically core ALSA components) but the audio produced is never  
> understandable.  On PCI I have tried turning on all of the debugging 
> options but it doesn't seem to generate any helpful errors.

I had it %90 working mid-2.5.x, but I simply haven't fiddled with it
since.  I also lack any interesting in working on it as I have my
hands full with enough other things at the moment. :)

Re: DMA_AUTOINIT missing in asm-sparc64/dma.h

[Jason Wever]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Sat, 26 Jun 2004, David S. Miller wrote:

> I had it %90 working mid-2.5.x, but I simply haven't fiddled with it
> since.  I also lack any interesting in working on it as I have my
> hands full with enough other things at the moment. :)

Gotcha.  Are there any resources on-line or otherwise that could be used 
by others to help get this working?  I'm interested in getting it working 
but I'm thinking my enthusiasm outweighs my coding skills at the moment :)  

Also were the ALSA folks been much involved with the work on the driver in 
the 2.5 days and if so, is there someone good to follow-up with over 
there?

Thanks,
- -- 
Jason Wever
Gentoo/Sparc Co-Team Lead
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)

iD8DBQFA3gfPdKvgdVioq28RAmNcAJ9tkPq6+Irw9qaX6aCPodmE80n+rgCgn7v9
J2NwowMQscep7+xpv78rCBo=+AoH
-----END PGP SIGNATURE-----

Re: DMA_AUTOINIT missing in asm-sparc64/dma.h

[David S. Miller]

On Sat, 26 Jun 2004 17:33:33 -0600 (MDT)
Jason Wever <weeve@gentoo.org> wrote:

> Gotcha.  Are there any resources on-line or otherwise that could be used 
> by others to help get this working?  I'm interested in getting it working 
> but I'm thinking my enthusiasm outweighs my coding skills at the moment :)  

The best reference is the 2.4.x driver.  The netbsd driver is not a
bad refernce either.

The EBUS DMA engine is documented at:

http://www.sun.com/oem/products/manuals/802-7837.pdf

And that's likely where the problem are, the DMA controller and
the CS4321 chip aren't sync'ing up correctly.  That was the most
difficult part of the driver to get right.

> Also were the ALSA folks been much involved with the work on the driver in 
> the 2.5 days and if so, is there someone good to follow-up with over 
> there?

None at all, I wrote it all by myself using the ALSA ISA cs432x driver as
a guide.

Re: DMA_AUTOINIT missing in asm-sparc64/dma.h

[Jason Wever]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Sat, 26 Jun 2004, David S. Miller wrote:

> The best reference is the 2.4.x driver.  The netbsd driver is not a
> bad refernce either.
> 
> The EBUS DMA engine is documented at:
> 
> http://www.sun.com/oem/products/manuals/802-7837.pdf

Cool, thanks for the pointers!

- -- 
Jason Wever
Gentoo/Sparc Co-Team Lead
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.4 (GNU/Linux)

iD8DBQFA3gtFdKvgdVioq28RAipTAJ0ZEnHQfTBTdM84qS+kJTpxQiEjJgCeOiJh
tVP4fhKs+PjhhocN7eVS7WwlLO
-----END PGP SIGNATURE-----