linux-next: wireless-current tree build failure

linux-next: wireless-current tree build failure

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 624 bytes --]

Hi John,

Today's linux-next build (powerpc allyesconfig) failed like this:

drivers/net/wireless/ath9k/hw.c: In function 'ath9k_hw_9280_spur_mitigate':
drivers/net/wireless/ath9k/hw.c:4660: internal compiler error: in expand_expr_real_1, at expr.c:9199

Clearly not our fault! :-) I have informed our toolchain person.  This
was using the Debian gcc-4.3.  It also fails using the Debian 4.2
compiler but not the 4.1 one.

For now, I will apply the following patch to just stop it building on
PowerPC.

-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

Re: linux-next: wireless-current tree build failure

[Adrian Bunk]

On Tue, Aug 05, 2008 at 04:21:52PM +1000, Stephen Rothwell wrote:
> Hi John,
> 
> Today's linux-next build (powerpc allyesconfig) failed like this:
> 
> drivers/net/wireless/ath9k/hw.c: In function 'ath9k_hw_9280_spur_mitigate':
> drivers/net/wireless/ath9k/hw.c:4660: internal compiler error: in expand_expr_real_1, at expr.c:9199
> 
> Clearly not our fault! :-) I have informed our toolchain person.  This
> was using the Debian gcc-4.3.  It also fails using the Debian 4.2
> compiler but not the 4.1 one.
>...

Alexandros Couloumbis already opened a bug for this issue on mips, which 
also seems to be affected:
  http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37014

It is a gcc bug, but we'll have to work around it in the ath9k driver.

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed

Re: linux-next: wireless-current tree build failure

[John W. Linville]

On Tue, Aug 05, 2008 at 06:37:57PM +0300, Adrian Bunk wrote:
> On Tue, Aug 05, 2008 at 04:21:52PM +1000, Stephen Rothwell wrote:
> > Hi John,
> > 
> > Today's linux-next build (powerpc allyesconfig) failed like this:
> > 
> > drivers/net/wireless/ath9k/hw.c: In function 'ath9k_hw_9280_spur_mitigate':
> > drivers/net/wireless/ath9k/hw.c:4660: internal compiler error: in expand_expr_real_1, at expr.c:9199
> > 
> > Clearly not our fault! :-) I have informed our toolchain person.  This
> > was using the Debian gcc-4.3.  It also fails using the Debian 4.2
> > compiler but not the 4.1 one.
> >...
> 
> Alexandros Couloumbis already opened a bug for this issue on mips, which 
> also seems to be affected:
>   http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37014
> 
> It is a gcc bug, but we'll have to work around it in the ath9k driver.

I imagine that we would be happy to do that.  Anyone have a clue as to how?

John
-- 
John W. Linville
linville@tuxdriver.com

Re: linux-next: wireless-current tree build failure

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 475 bytes --]

Hi Adrian,

On Tue, 5 Aug 2008 18:37:57 +0300 Adrian Bunk <bunk@kernel.org> wrote:
>
> Alexandros Couloumbis already opened a bug for this issue on mips, which 
> also seems to be affected:
>   http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37014
> 
> It is a gcc bug, but we'll have to work around it in the ath9k driver.

Thanks for fishing that out Adrian.
-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

Re: linux-next: wireless-current tree build failure

[Stephen Rothwell]

[-- Attachment #1: Type: text/plain, Size: 501 bytes --]

Hi John,

On Tue, 5 Aug 2008 12:00:26 -0400 "John W. Linville" <linville@tuxdriver.com> wrote:
>
> I imagine that we would be happy to do that.  Anyone have a clue as to how?

I think that "not yet" is the answer :-(

If we don't get a useful answer soon, we may just have to merge the
driver with appropriately commented negative "depends on" in Kconfig for
the arches it breaks.

-- 
Cheers,
Stephen Rothwell                    sfr@canb.auug.org.au
http://www.canb.auug.org.au/~sfr/

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

[2.6 patch] ath9k: work around gcc ICEs

[Adrian Bunk]

On Tue, Aug 05, 2008 at 12:00:26PM -0400, John W. Linville wrote:
> On Tue, Aug 05, 2008 at 06:37:57PM +0300, Adrian Bunk wrote:
> > On Tue, Aug 05, 2008 at 04:21:52PM +1000, Stephen Rothwell wrote:
> > > Hi John,
> > > 
> > > Today's linux-next build (powerpc allyesconfig) failed like this:
> > > 
> > > drivers/net/wireless/ath9k/hw.c: In function 'ath9k_hw_9280_spur_mitigate':
> > > drivers/net/wireless/ath9k/hw.c:4660: internal compiler error: in expand_expr_real_1, at expr.c:9199
> > > 
> > > Clearly not our fault! :-) I have informed our toolchain person.  This
> > > was using the Debian gcc-4.3.  It also fails using the Debian 4.2
> > > compiler but not the 4.1 one.
> > >...
> > 
> > Alexandros Couloumbis already opened a bug for this issue on mips, which 
> > also seems to be affected:
> >   http://gcc.gnu.org/bugzilla/show_bug.cgi?id=37014
> > 
> > It is a gcc bug, but we'll have to work around it in the ath9k driver.
> 
> I imagine that we would be happy to do that.  Anyone have a clue as to how?

Patch below, verified to fix the issue with the following compilers:
- powerpc 4.2.4
- powerpc latest 4.3 SVN
- powerpc latest 4.4 SVN
- mips 4.2.4
- mips 4.3.1

The patch has to be applied twice for fixing both copies of the
same code.  :-)

> John

cu
Adrian


<--  snip  -->


This patch works around an internal compiler error (gcc bug #37014) in 
all gcc 4.2 compilers and the gcc 4.3 series up to at least 4.3.1
on at least powerpc and mips.

Many thanks to Andrew Pinski for analyzing the gcc bug.

Signed-off-by: Adrian Bunk <bunk@kernel.org>

---
d1ec5b9fba6c50c7e98a85776a3b95d3a3a3179b 
diff --git a/drivers/net/wireless/ath9k/hw.c b/drivers/net/wireless/ath9k/hw.c
index 1f6f393..5edc56d 100644
--- a/drivers/net/wireless/ath9k/hw.c
+++ b/drivers/net/wireless/ath9k/hw.c
@@ -4801,7 +4801,11 @@
 
 	for (i = 0; i < 123; i++) {
 		if ((cur_vit_mask > lower) && (cur_vit_mask < upper)) {
-			if ((abs(cur_vit_mask - bin)) < 75)
+
+			/* workaround for gcc bug #37014 */
+			volatile int tmp = abs(cur_vit_mask - bin);
+
+			if (tmp < 75)
 				mask_amt = 1;
 			else
 				mask_amt = 0;