openafs module cannot be built on PPC64

for info, i tried earlier versions of spca5xx from
http://mxhaard.free.fr/spca50x/Download/?M=D and found out that the last
version compiling on -powerpc64 is spca5xx-20050601

piem@calabaza:~/spca5xx-20050601$ make
   Building SPCA5XX driver for 2.5/2.6 kernel.
   Remember: you must have read/write access to your kernel source tree.
make -C /lib/modules/`uname -r`/build SUBDIRS=/home/piem/spca5xx-20050601 modules
make[1]: Entering directory `/usr/src/linux-headers-2.6.12-1-powerpc64'
  CC [M] /home/piem/spca5xx-20050601/drivers/usb/spca5xx.o
/home/piem/spca5xx-20050601/drivers/usb/spca5xx.c: In function ‘spca50x_configure_sensor’:
/home/piem/spca5xx-20050601/drivers/usb/spca5xx.c:5690: warning: ISO C90 forbids mixed declarations and code
/home/piem/spca5xx-20050601/drivers/usb/spca5xx.c: In function ‘spca5xx_probe’:
/home/piem/spca5xx-20050601/drivers/usb/spca5xx.c:5855: warning: ‘defaultpipe’ may be used uninitialized in this function
/home/piem/spca5xx-20050601/drivers/usb/spca5xx.c:5854: warning: ‘defaultrows’ may be used uninitialized in this function
/home/piem/spca5xx-20050601/drivers/usb/spca5xx.c:5853: warning: ‘defaultcols’ may be used uninitialized in this function
  CC [M] /home/piem/spca5xx-20050601/drivers/usb/spcadecoder.o
  LD [M] /home/piem/spca5xx-20050601/spca5xx.o
  Building modules, stage 2.
  MODPOST
  CC /home/piem/spca5xx-20050601/spca5xx.mod.o
  LD [M] /home/piem/spca5xx-20050601/spca5xx.ko
make[1]: Leaving directory `/usr/src/linux-headers-2.6.12-1-powerpc64'

so far, the camera (041e:4034 Creative instant) is working great in
pd-pdp, but i can't get camorama, gphoto2 or gtkam to detect it. i found
the following in dmesg, with similar lines for each usb devices
(including the camera):

ioctl32(gtkam:5245): Unknown cmd fd(6) cmd(c00c5512){00} arg(fff990c8) on /proc/bus/usb/003/008

ciao, piem

Hi Michel:

You may already be aware of this but spca5xx has now made it into
Debian's testing repository. Kel and I are working to get the
toolchain (spcaview etc) into Debian as well, that should happen
fairly soon.

Some bad news though, we have received a bug report from a power PC
user (Paul Brossier), it looks like the driver doesnt compile on
powerpc. Do you have a co-maintainer versed on powerpc that can look
into this for us by any chance?

Also, we have set up a mail list for discussing debian packaging
aspects of your driver. I dont expect it to be a high traffic site
but if you are interested in joining the list point your browser here:

http://lists.alioth.debian.org/mailman/listinfo/pkg-spca5xx-devel

The alioth site also has a forum but we (Kel and I) are planning to
shut that down, we dont need a list server *and* a forum :-)

Steve

Paul Brossier(<email address hidden>)@2005-10-17 17:31:
> Package: spca5xx-source
> Version: 20051001-1
> Severity: normal
>
> Hi,
>
> thanks for packaging these modules. the module works fine on an i386
> box, but I have troubles compiling it on 2.6.12-1-powerpc64. Attached is
> the compilation log.
>
>
> piem@calabaza:/usr/src/modules/spca5xx$ make
> Building SPCA5XX driver for 2.5/2.6 kernel.
> Remember: you must have read/write access to your kernel source tree.
> make -C /lib/modules/`uname -r`/build SUBDIRS=/usr/src/modules/spca5xx CC=cc modules
> make[1]: Entering directory `/usr/src/linux-headers-2.6.12-1-powerpc64'
> CC [M] /usr/src/modules/spca5xx/drivers/usb/spca5xx.o
> In file included from include/linux/thread_info.h:20,
> from include/linux/spinlock.h:12,
> from include/linux/capability.h:45,
> from include/linux/sched.h:7,
> from include/linux/module.h:10,
> from /usr/src/modules/spca5xx/drivers/usb/spca5xx.c:39:
> include/linux/bitops.h: In function ???generic_hweight64???:
> include/linux/bitops.h:123: warning: integer constant is too large for ???unsigned long??? type
> include/linux/bitops.h:123: warning: integer constant is too large for ???unsigned long??? type
> include/linux/bitops.h:124: warning: integer constant is too large for ???unsigned long??? type
> include/linux/bitops.h:124: warning: integer constant is too large for ???unsigned long??? type
> include/linux/bitops.h:125: warning: integer constant is too large for ???unsigned long??? type
> include/linux/bitops.h:125: warning: integer constant is too large for ???unsigned long??? type
> include/linux/bitops.h:126: warning: integer constant is too large for ???unsigned long??? type
> include/linux/bitops.h:126: warning: integer constant is too large for ???unsigned long??? type
> include/linux/bitops.h:127: warning: integer constant is too large for ???unsigned long??? type
> include/linux/bitops.h:127: warning: integer constant is too large for ???unsigned long??? type
> include/linux/bitops.h:128: warning: right shift count >= width of type
> In file included from include/linux/byteorder/big_endian.h:12,
> from include/asm/byteorder.h:94,
> from include/linux/kernel.h:16,
> ...

michel Xhaard(<email address hidden>)@2005-10-18 10:02:
> Steve,
> Ok i wil look I have also forward the bug report to Tomas Groth working on
> spca5xx ppc support.

Excellent, thanks Michel

Steve

Paul Brossier wrote:

>for info, i tried earlier versions of spca5xx from
>http://mxhaard.free.fr/spca50x/Download/?M=D and found out that the last
>version compiling on -powerpc64 is spca5xx-20050601
>
>piem@calabaza:~/spca5xx-20050601$ make
> Building SPCA5XX driver for 2.5/2.6 kernel.
> Remember: you must have read/write access to your kernel source tree.
>make -C /lib/modules/`uname -r`/build SUBDIRS=/home/piem/spca5xx-20050601 modules
>make[1]: Entering directory `/usr/src/linux-headers-2.6.12-1-powerpc64'
> CC [M] /home/piem/spca5xx-20050601/drivers/usb/spca5xx.o
>/home/piem/spca5xx-20050601/drivers/usb/spca5xx.c: In function ‘spca50x_configure_sensor’:
>/home/piem/spca5xx-20050601/drivers/usb/spca5xx.c:5690: warning: ISO C90 forbids mixed declarations and code
>/home/piem/spca5xx-20050601/drivers/usb/spca5xx.c: In function ‘spca5xx_probe’:
>/home/piem/spca5xx-20050601/drivers/usb/spca5xx.c:5855: warning: ‘defaultpipe’ may be used uninitialized in this function
>/home/piem/spca5xx-20050601/drivers/usb/spca5xx.c:5854: warning: ‘defaultrows’ may be used uninitialized in this function
>/home/piem/spca5xx-20050601/drivers/usb/spca5xx.c:5853: warning: ‘defaultcols’ may be used uninitialized in this function
> CC [M] /home/piem/spca5xx-20050601/drivers/usb/spcadecoder.o
> LD [M] /home/piem/spca5xx-20050601/spca5xx.o
> Building modules, stage 2.
> MODPOST
> CC /home/piem/spca5xx-20050601/spca5xx.mod.o
> LD [M] /home/piem/spca5xx-20050601/spca5xx.ko
>make[1]: Leaving directory `/usr/src/linux-headers-2.6.12-1-powerpc64'
>
>so far, the camera (041e:4034 Creative instant) is working great in
>pd-pdp, but i can't get camorama, gphoto2 or gtkam to detect it. i found
>the following in dmesg, with similar lines for each usb devices
>(including the camera):
>
>ioctl32(gtkam:5245): Unknown cmd fd(6) cmd(c00c5512){00} arg(fff990c8) on /proc/bus/usb/003/008
>
>ciao, piem
>
>
>
Thanks for your time piem, very appreciated!

The upstream authours have been informed and are now investigating.

Thanks, Kel.

Hi all,

On Tue, Oct 18, 2005 at 12:16:27AM -0700, Stephen Birch wrote:
> Hi Michel:
>
> You may already be aware of this but spca5xx has now made it into
> Debian's testing repository. Kel and I are working to get the
> toolchain (spcaview etc) into Debian as well, that should happen
> fairly soon.

great. thanks for your work.

> Some bad news though, we have received a bug report from a power PC
> user (Paul Brossier), it looks like the driver doesnt compile on
> powerpc. Do you have a co-maintainer versed on powerpc that can look
> into this for us by any chance?

i just got my ibook back from repair. spca5xx 20051001 compiled fine
there, so the issue only applies to powerpc64 (a 64 bit kernel with a
biarch toolchain), as suggested by gcc errors.

cheers, paul

Paul Brossier <email address hidden> writes:

>> Some bad news though, we have received a bug report from a power PC
>> user (Paul Brossier), it looks like the driver doesnt compile on
>> powerpc. Do you have a co-maintainer versed on powerpc that can look
>> into this for us by any chance?
>
> i just got my ibook back from repair. spca5xx 20051001 compiled fine
> there, so the issue only applies to powerpc64 (a 64 bit kernel with a
> biarch toolchain), as suggested by gcc errors.

Do you or someone else can provide an access to a powerpc64 machine to
me to test and fix the problem?

--
        O T A V I O S A L V A D O R
---------------------------------------------
 E-mail: <email address hidden> UIN: 5906116
 GNU/Linux User: 239058 GPG ID: 49A5F855
 Home Page: http://www.freedom.ind.br/otavio
---------------------------------------------
"Microsoft gives you Windows ... Linux gives
 you the whole house."

# Automatically generated email from bts, devscripts version 2.9.8
retitle 334392 does not compile in powerpc64

Hello,

I uploaded a new version of spca5xx yestarday to Debian (20051101-1)
and it should enters archive today. Would be good if you could test it
and check if this problem still exist.

If you can confirm that, I'll try to catch access to any machine of
powerpc64 for testing.

Thanks,

--
        O T A V I O S A L V A D O R
---------------------------------------------
 E-mail: <email address hidden> UIN: 5906116
 GNU/Linux User: 239058 GPG ID: 49A5F855
 Home Page: http://www.freedom.ind.br/otavio
---------------------------------------------
"Microsoft gives you Windows ... Linux gives
 you the whole house."

Le Mardi 1 Novembre 2005 12:15, Otavio Salvador a écrit :
> Hello,
>
> I uploaded a new version of spca5xx yestarday to Debian (20051101-1)
> and it should enters archive today. Would be good if you could test it
> and check if this problem still exist.
>
> If you can confirm that, I'll try to catch access to any machine of
> powerpc64 for testing.
>
> Thanks,
I did not change the MR97311 code so i think the ppc64 bug is here too, i will
try to change this part for the next release.
Regards
--
Michel Xhaard
http://mxhaard.free.fr

On Tue, Nov 01, 2005 at 03:08:50PM +0100, michel Xhaard wrote:
> Le Mardi 1 Novembre 2005 12:15, Otavio Salvador a écrit :
> > Hello,
> >
> > I uploaded a new version of spca5xx yestarday to Debian (20051101-1)
> > and it should enters archive today. Would be good if you could test it
> > and check if this problem still exist.
> >
> > If you can confirm that, I'll try to catch access to any machine of
> > powerpc64 for testing.
> > Thanks,

Otavio, i can't give you access to my machine easily, but debian should
have a few of these available. Not sure which they are though. I guess
Sven Luther is the one to ask.

> I did not change the MR97311 code so i think the ppc64 bug is here too, i will
> try to change this part for the next release.

Ready for testing :-)

bye, piem

Paul Brossier <email address hidden> writes:

>> > If you can confirm that, I'll try to catch access to any machine of
>> > powerpc64 for testing.
>> > Thanks,
>
> Otavio, i can't give you access to my machine easily, but debian should
> have a few of these available. Not sure which they are though. I guess
> Sven Luther is the one to ask.

I'll try to get access to one machine ASAP.

--
        O T A V I O S A L V A D O R
---------------------------------------------
 E-mail: <email address hidden> UIN: 5906116
 GNU/Linux User: 239058 GPG ID: 49A5F855
 Home Page: http://www.freedom.ind.br/otavio
---------------------------------------------
"Microsoft gives you Windows ... Linux gives
 you the whole house."

Hello,

I tested current code and it doesn't fail to build anymore but we have
another bug: How to detect ppc64 arch and kernel and then to use -m64
as gcc option to compile?

Since I don't have root permission on test machine, I cannot test
module-assistant with spca5xx-source package from sid but would be
good if you could do a test and check if it's working.

--
        O T A V I O S A L V A D O R
---------------------------------------------
 E-mail: <email address hidden> UIN: 5906116
 GNU/Linux User: 239058 GPG ID: 49A5F855
 Home Page: http://www.freedom.ind.br/otavio
---------------------------------------------
"Microsoft gives you Windows ... Linux gives
 you the whole house."

reassign 334392 module-assistant
thanks

----
Hello folks,

I got module-assistant to work in a test machine and it fail to detect
the need of -m64 gcc param. So this isn't our fault anymore and I'm
reassign this to it.

--
        O T A V I O S A L V A D O R
---------------------------------------------
 E-mail: <email address hidden> UIN: 5906116
 GNU/Linux User: 239058 GPG ID: 49A5F855
 Home Page: http://www.freedom.ind.br/otavio
---------------------------------------------
"Microsoft gives you Windows ... Linux gives
 you the whole house."

Hi,

On Thu, Nov 03, 2005 at 06:29:33PM -0200, Otavio Salvador wrote:
> reassign 334392 module-assistant
> thanks
>
> ----
> Hello folks,
>
> I got module-assistant to work in a test machine and it fail to detect
> the need of -m64 gcc param. So this isn't our fault anymore and I'm
> reassign this to it.

indeed. as a workaround, i could compile on 2.6.14 changing the line 54
in the Makefile of spca5xx-source 20051101-1 to

        $(MAKE) -C $(KERNELDIR) SUBDIRS=$(PWD) CC="$(CC) -m64" modules

and the module seems to work fine now, sorry for the noise :)

thanks, piem

Paul Brossier <email address hidden> writes:

>> I got module-assistant to work in a test machine and it fail to detect
>> the need of -m64 gcc param. So this isn't our fault anymore and I'm
>> reassign this to it.
>
> indeed. as a workaround, i could compile on 2.6.14 changing the line 54
> in the Makefile of spca5xx-source 20051101-1 to
>
> $(MAKE) -C $(KERNELDIR) SUBDIRS=$(PWD) CC="$(CC) -m64" modules
>
> and the module seems to work fine now, sorry for the noise :)

No problem. :-D

We now need to identify a way to discover if a kernel was built using
64 or 32 bits mode. So we can fix it on module-assistant.

--
        O T A V I O S A L V A D O R
---------------------------------------------
 E-mail: <email address hidden> UIN: 5906116
 GNU/Linux User: 239058 GPG ID: 49A5F855
 Home Page: http://www.freedom.ind.br/otavio
---------------------------------------------
"Microsoft gives you Windows ... Linux gives
 you the whole house."

Hi,

Shouldn't the following code snippet be enough to make module-assistant
work on PPC64 and hopefully still work on all machines with a 64-bit
kernel. Maybe it should also check if the arch is powerpc just to be on
the safe side?

--- module-assistant-0.10.2/modass/include/generic.make.orig 2006-04-05 10:22:36.000000000 +0200
+++ module-assistant-0.10.2/modass/include/generic.make 2006-04-05 10:23:08.000000000 +0200
@@ -48,6 +48,11 @@
 endif
 endif

+# Check if we are running a 64bit kernel
+ifneq "$(wildcard $(KSRC)/include/linux/config/64bit.h)" ""
+ CFLAGS += -m64
+endif
+
 # Special case gcc 2.7.2
 ifeq ($(CC),gcc-2.7)
 CC = gcc272

It is completely untested though as I don't have access to any PPC64 machine.

/torkel

Hello,

I would like to get more competent comments from concerned kernel
packages on the issue quoted below. There is already a hack for sparc64:

ifeq ($(ARCH),)
SPARCH=$(shell grep 'CONFIG_SPARC..=y' "$(KSRC)/.config" 2>/dev/null| cut -d= -f1)
#maybe a different ARCH on sparc
ifeq (CONFIG_SPARC32,$(SPARCH))
ARCH :=sparc
export ARCH
endif
ifeq (CONFIG_SPARC64,$(SPARCH))
ARCH :=sparc64
export ARCH
endif
endif

* Björn Torkelsson [Wed, Apr 05 2006, 10:29:41AM]:

> Shouldn't the following code snippet be enough to make module-assistant
> work on PPC64 and hopefully still work on all machines with a 64-bit
> kernel. Maybe it should also check if the arch is powerpc just to be on
> the safe side?

> --- module-assistant-0.10.2/modass/include/generic.make.orig 2006-04-05 10:22:36.000000000 +0200
> +++ module-assistant-0.10.2/modass/include/generic.make 2006-04-05 10:23:08.000000000 +0200
> @@ -48,6 +48,11 @@
> endif
> endif
>
> +# Check if we are running a 64bit kernel
> +ifneq "$(wildcard $(KSRC)/include/linux/config/64bit.h)" ""
> + CFLAGS += -m64
> +endif
> +
> # Special case gcc 2.7.2
> ifeq ($(CC),gcc-2.7)
> CC = gcc272
>
> It is completely untested though as I don't have access to any PPC64 machine.
>
> /torkel
>
>

--
<krid> Hi, gibts einen Kazaa client für GNU?
<panthera> krid: falsche frage
<krid> panthera: Gibt es einen Kazaa client für Debian GNU/Linux? :-)
<panthera> krid: falsche frage ;)
<krid> panthera: Hm. Welche Frage ist die richtige?
<panthera> krid: gibt es einen freien ftp-client fuer linux?

Eduard Bloch <email address hidden> writes:

> Hello,
>
> I would like to get more competent comments from concerned kernel
> packages on the issue quoted below. There is already a hack for sparc64:

I'm not sure this answeres your problem but in recent kernels the
kernel adds the -m64 flag for amd64 automaticaly. The same should
happen for all other multiarch archs for both -m32 and -m64 as needed.

I also send in a patch for make-kpkg (applied in sid) for multiarch
cross-compile support using --arch <arch> --cross-compile '-'. This
tells make-kpkg to build for the arch but still use the normal "gcc"
instead of defaulting to "arch-os-gnu-gcc" as with other
cross-compiles.

Those two combined make building a kernel for "the other" arch of a
system simple.

MfG
        Goswin