linux-next: arm allmodconfig

linux-next: arm allmodconfig

[Andrew Morton]

> arch/arm/mm/dma-mapping.c: In function `dma_sync_sg_for_cpu':
> arch/arm/mm/dma-mapping.c:588: warning: statement with no effect

Something in here:

                dmabounce_sync_for_cpu(dev, sg_dma_address(s), 0,
                                        sg_dma_len(s), dir);                    

> kernel/sched.c: In function `add_preempt_count':
> kernel/sched.c:4294: warning: unsupported arg to `__builtin_return_address'
> kernel/sched.c:4296: warning: unsupported arg to `__builtin_return_address'
> kernel/sched.c:4298: warning: unsupported arg to `__builtin_return_address'
> kernel/sched.c: In function `sub_preempt_count':
> kernel/sched.c:4294: warning: unsupported arg to `__builtin_return_address'
> kernel/sched.c:4296: warning: unsupported arg to `__builtin_return_address'
> kernel/sched.c:4298: warning: unsupported arg to `__builtin_return_address'

Related to CALLER_ADDR2, etc.

> drivers/atm/zatm.c: In function `refill_pool':
> drivers/atm/zatm.c:226: warning: `virt_to_bus' is deprecated (declared at /usr/src/devel/arch/arm/include/asm/memory.h:184)
> 
<50ish similar warnings snipped>

> drivers/atm/ambassador.h:257:1: warning: "FLASH_BASE" redefined
> In file included from arch/arm/mach-versatile/include/mach/irqs.h:22,
>                  from /usr/src/devel/arch/arm/include/asm/irq.h:4,
>                  from /usr/src/devel/arch/arm/include/asm/hardirq.h:6,
>                  from include/linux/hardirq.h:7,
>                  from include/asm-generic/local.h:5,
>                  from /usr/src/devel/arch/arm/include/asm/local.h:1,
>                  from include/linux/module.h:20,
>                  from drivers/atm/ambassador.c:25:
> arch/arm/mach-versatile/include/mach/platform.h:443:1: warning: this is the location of the previous definition
> In file included from drivers/atm/ambassador.c:44:
> drivers/atm/ambassador.h:258:1: warning: "FLASH_SIZE" redefined
> In file included from arch/arm/mach-versatile/include/mach/irqs.h:22,
>                  from /usr/src/devel/arch/arm/include/asm/irq.h:4,
>                  from /usr/src/devel/arch/arm/include/asm/hardirq.h:6,
>                  from include/linux/hardirq.h:7,
>                  from include/asm-generic/local.h:5,
>                  from /usr/src/devel/arch/arm/include/asm/local.h:1,
>                  from include/linux/module.h:20,
>                  from drivers/atm/ambassador.c:25:
> arch/arm/mach-versatile/include/mach/platform.h:444:1: warning: this is the location of the previous definition

This has been there for ages.  FLASH_BASE and FLASH_SIZE are silly
identifier for any code to use anywhere, really.

> drivers/atm/ambassador.c: In function `check_area':
> drivers/atm/ambassador.c:425: warning: `virt_to_bus' is deprecated (declared at /usr/src/devel/arch/arm/include/asm/memory.h:184)
<zillions more of these>

> In file included from drivers/atm/iphase.c:65:
> drivers/atm/iphase.h:332:1: warning: "EPROM_SIZE" redefined
> In file included from arch/arm/mach-versatile/include/mach/irqs.h:22,
>                  from /usr/src/devel/arch/arm/include/asm/irq.h:4,
>                  from /usr/src/devel/arch/arm/include/asm/hardirq.h:6,
>                  from include/linux/hardirq.h:7,
>                  from include/asm-generic/local.h:5,
>                  from /usr/src/devel/arch/arm/include/asm/local.h:1,
>                  from include/linux/module.h:20,
>                  from drivers/atm/iphase.c:43:
> arch/arm/mach-versatile/include/mach/platform.h:453:1: warning: this is the location of the previous definition

EPROM_SIZE namespace clash.

> crypto/testmgr.c: In function `alg_test_comp':
> crypto/testmgr.c:829: warning: 'ret' might be used uninitialized in this function

There's no way for the compiler to know that this code isn't buggy.

Suggest that this be fixed via uninitialized_var() or, better, convert
to a do{}while() loop.

I notice that test_comp() also has two locals called `ret' - one
shadowing the other.

> mm/mlock.c: In function `__mlock_vma_pages_range':
> mm/mlock.c:169: warning: 'ret' might be used uninitialized in this function

Again, gcc can't know this is unbuggy.  a do{}while() conversion would
fix it, I guess.

> In file included from drivers/atm/eni.c:29:
> drivers/atm/midway.h:19:1: warning: "EPROM_SIZE" redefined
> In file included from arch/arm/mach-versatile/include/mach/irqs.h:22,
>                  from /usr/src/devel/arch/arm/include/asm/irq.h:4,
>                  from /usr/src/devel/arch/arm/include/asm/hardirq.h:6,
>                  from include/linux/hardirq.h:7,
>                  from include/asm-generic/local.h:5,
>                  from /usr/src/devel/arch/arm/include/asm/local.h:1,
>                  from include/linux/module.h:20,
>                  from drivers/atm/eni.c:6:
> arch/arm/mach-versatile/include/mach/platform.h:453:1: warning: this is the location of the previous definition

Zzzz..

> drivers/atm/firestream.c: In function `submit_qentry':
> drivers/atm/firestream.c:608: warning: `bus_to_virt' is deprecated (declared at /usr/src/devel/arch/arm/include/asm/memory.h:189)

zillions more snipped

> include/asm-generic/cmpxchg-local.h:15: warning: 'prev' might be used uninitialized in this function
> drivers/gpu/drm/drm_lock.c: In function `drm_lock_free':
> include/asm-generic/cmpxchg-local.h:15: warning: 'prev' might be used uninitialized in this function
> include/asm-generic/cmpxchg-local.h:15: warning: 'prev' might be used uninitialized in this function
> drivers/gpu/drm/drm_lock.c: In function `drm_notifier':
> include/asm-generic/cmpxchg-local.h:15: warning: 'prev' might be used uninitialized in this function
> drivers/gpu/drm/drm_lock.c: In function `drm_idlelock_release':
> include/asm-generic/cmpxchg-local.h:15: warning: 'prev' might be used uninitialized in this function

didn't look.

> drivers/gpu/drm/drm_agpsupport.c:36:21: asm/agp.h: No such file or directory
> make[3]: *** [drivers/gpu/drm/drm_agpsupport.o] Error 1
> make[2]: *** [drivers/gpu/drm] Error 2
> make[1]: *** [drivers/gpu] Error 2
> make[1]: *** Waiting for unfinished jobs....

DRM has been breaking arm allmodconfig for at least a year.

> include/asm-generic/cmpxchg-local.h:15: warning: 'prev' might be used uninitialized in this function
> kernel/trace/trace_irqsoff.c: In function `start_critical_timings':
> kernel/trace/trace_irqsoff.c:185: warning: unsupported arg to `__builtin_return_address'
> kernel/trace/trace_irqsoff.c: In function `stop_critical_timings':
> kernel/trace/trace_irqsoff.c:222: warning: unsupported arg to `__builtin_return_address'
> sound/soc/codecs/tlv320aic23.c: In function `tlv320aic23_set_dai_sysclk':
> sound/soc/codecs/tlv320aic23.c:424: warning: unused variable `codec'
> kernel/trace/trace_sched_wakeup.c: In function `probe_wakeup_sched_switch':
> kernel/trace/trace_sched_wakeup.c:163: warning: unsupported arg to `__builtin_return_address'
> kernel/trace/trace_sched_wakeup.c:163: warning: unsupported arg to `__builtin_return_address'
> kernel/trace/trace_sched_wakeup.c: In function `probe_wakeup':
> kernel/trace/trace_sched_wakeup.c:265: warning: unsupported arg to `__builtin_return_address'
> kernel/trace/trace_sched_wakeup.c:265: warning: unsupported arg to `__builtin_return_address'

More CALLER_ADDR stuff.

> net/netlabel/netlabel_mgmt.c: In function `netlbl_mgmt_listentry':
> net/netlabel/netlabel_mgmt.c:268: warning: 'ret_val' might be used uninitialized in this function

Assumes that netlbl_af4list_foreach_rcu() will never iterate zero times.

Assumes that entry->type can never have an unexpected value.

gcc cannot get this right, and C-level changes will be needed to shut
this up.


> net/phonet/pn_dev.c: In function `phonet_device_get':
> net/phonet/pn_dev.c:72: warning: 'dev' might be used uninitialized in this function

Assumes that pndevs.list can enver be empty.  Only fixable at the C
level.  uninitialized_var(), or initialize it to NULL or whatever.

> include/asm-generic/xor.h:696: warning: 'xor_block_8regs_p' defined but not used
> include/asm-generic/xor.h:704: warning: 'xor_block_32regs_p' defined but not used

Statically defined storage in a header file.  Puke.

> crypto/async_tx/async_tx.c: In function `async_trigger_callback':
> crypto/async_tx/async_tx.c:584: warning: 'chan' might be used uninitialized in this function

I suppose that if gcc was genius enough they might make this go away. 
But I expect the code could be reorganised to fix this.


> sound/pci/hda/patch_sigmatel.c: In function `stac92xx_smux_enum_put':
> sound/pci/hda/patch_sigmatel.c:571: warning: 'val' might be used uninitialized in this function

This looks like a bug.

> fs/ntfs/compress.c: In function `ntfs_read_compressed_block':
> fs/ntfs/compress.c:521: warning: 'pages' might be used uninitialized in this function

Well that's weird.

> include/linux/slub_def.h:231: warning: 'ret' might be used uninitialized in this function
> include/linux/slub_def.h:231: warning: 'ret' might be used uninitialized in this function
> fs/xfs/quota/xfs_qm_syscalls.c: In function `xfs_qm_dqrele_inodes_ag':
> fs/xfs/quota/xfs_qm_syscalls.c:1039: warning: 'inode_refed' might be used uninitialized in this function
> fs/xfs/xfs_rtalloc.c: In function `xfs_growfs_rt':
> fs/xfs/xfs_rtalloc.c:1875: warning: 'tp' might be used uninitialized in this function
> fs/udf/super.c: In function `udf_process_sequence':
> fs/udf/super.c:1240: warning: 'map' might be used uninitialized in this function

I lost interest here.

> fs/quota_v2.c: In function `v2_read_file_info':
> fs/quota_v2.c:119: warning: integer constant is too large for "unsigned long" type
> fs/quota_v2.c:120: warning: integer constant is too large for "unsigned long" type

That's the new quota stuff.  I thought I fixed that...

Re: linux-next: arm allmodconfig

[Takashi Iwai]

At Tue, 28 Oct 2008 17:56:04 -0700,
Andrew Morton wrote:
> 
> > sound/soc/codecs/tlv320aic23.c: In function `tlv320aic23_set_dai_sysclk':
> > sound/soc/codecs/tlv320aic23.c:424: warning: unused variable `codec'

This will be either removed or used with other codes.
Waiting for a patch of the original author right now.

> > sound/pci/hda/patch_sigmatel.c: In function `stac92xx_smux_enum_put':
> > sound/pci/hda/patch_sigmatel.c:571: warning: 'val' might be used uninitialized in this function
> 
> This looks like a bug.

Yep, indeed.  Fixed now on sound git tree.
(Wondering why this warning doesn't hit on my machine with
 gcc-4.3.3...)


Thanks!

Takashi

Re: [xfs-masters] linux-next: arm allmodconfig

[Dave Chinner]

On Tue, Oct 28, 2008 at 05:56:04PM -0700, Andrew Morton wrote:
> > fs/xfs/quota/xfs_qm_syscalls.c: In function `xfs_qm_dqrele_inodes_ag':
> > fs/xfs/quota/xfs_qm_syscalls.c:1039: warning: 'inode_refed' might be used uninitialized in this function

I haven't seen this one before. And it's a real bug, too.
gcc-4.3.1 on x86_64 is not picking this warning up:

$ gcc -v
Using built-in specs.
Target: x86_64-linux-gnu
Configured with: ../src/configure -v --with-pkgversion='Debian 4.3.1-9' --with-bugurl=file:///usr/share/doc/gcc-4.3/README.Bugs --enable-languages=c,c++,fortran,objc,obj-c++ --prefix=/usr --enable-shared --with-system-zlib --libexecdir=/usr/lib --without-included-gettext --enable-threads=posix --enable-nls --with-gxx-include-dir=/usr/include/c++/4.3 --program-suffix=-4.3 --enable-clocale=gnu --enable-libstdcxx-debug --enable-objc-gc --enable-mpfr --enable-cld --enable-checking=release --build=x86_64-linux-gnu --host=x86_64-linux-gnu --target=x86_64-linux-gnu
Thread model: posix
gcc version 4.3.1 (Debian 4.3.1-9)
$
$
$ touch fs/xfs/quota/xfs_qm_syscalls.c
$ make -j2 ARCH=um
  SYMLINK arch/um/include/shared/kern_constants.h
make[1]: `arch/um/sys-x86_64/user-offsets.s' is up to date.
  CHK     include/linux/version.h
  CHK     include/linux/utsrelease.h
  CALL    scripts/checksyscalls.sh
  CHK     include/linux/compile.h
  CC      fs/xfs/quota/xfs_qm_syscalls.o
  LD      fs/xfs/xfs.o
  LD      fs/xfs/built-in.o
  LD      fs/built-in.o

What compiler version are you using?

I'll send out a patch to fix this in a minute.

> > fs/xfs/xfs_rtalloc.c: In function `xfs_growfs_rt':
> > fs/xfs/xfs_rtalloc.c:1875: warning: 'tp' might be used uninitialized in this function

False positive, and I don't get this reported, either.

Hold on - the above gcc binary only emits a warning for the xfs_growfs_rt issue
when CONFIG_CC_OPTIMIZE_FOR_SIZE=y is set. It still doesn't catch the quota
bug, though. This is so fucked up....

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

Re: [xfs-masters] linux-next: arm allmodconfig

[Ingo Molnar]

* Dave Chinner <david@fromorbit.com> wrote:

> > > fs/xfs/xfs_rtalloc.c: In function `xfs_growfs_rt':
> > > fs/xfs/xfs_rtalloc.c:1875: warning: 'tp' might be used uninitialized in this function
> 
> False positive, and I don't get this reported, either.
> 
> Hold on - the above gcc binary only emits a warning for the 
> xfs_growfs_rt issue when CONFIG_CC_OPTIMIZE_FOR_SIZE=y is set. It 
> still doesn't catch the quota bug, though. This is so fucked up....

FYI, i've got 3 XFS warnings mapped in tip/auto-warnings-next:

 earth4:~/tip> gll linus..auto-warnings-next | grep xfs

 ec2f37c: work around warning in fs/xfs/xfs_mount.c
 8501db3: work around warning in fs/xfs/xfs_rtalloc.c
 8077af8: warnings: fix xfs posix acl

you can find those commits in the auto-warnings-next branch of -tip:

  http://people.redhat.com/mingo/tip.git/README

the rtalloc annotation is below. I went through the flow and the code 
seems to be correct and GCC is wrong.

	Ingo

------------->
From 8501db35588df4f35d67d8ba207422006a214ae7 Mon Sep 17 00:00:00 2001
From: Ingo Molnar <mingo@elte.hu>
Date: Mon, 18 Aug 2008 15:21:19 +0200
Subject: [PATCH] work around warning in fs/xfs/xfs_rtalloc.c
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 8bit

fix warning:

  fs/xfs/xfs_rtalloc.c: In function ‘xfs_growfs_rt’:
  fs/xfs/xfs_rtalloc.c:1875: warning: ‘tp’ may be used uninitialized in this function

This is a spurious gcc warning - it does not realize the correct/bug-free
flow of logic regarding the 'error' and 'tp' variables.

No code changed:
   7c10fd959065115c8e252ff5a861a01e  xfs_rtalloc.o.before.asm
   7c10fd959065115c8e252ff5a861a01e  xfs_rtalloc.o.after.asm

Signed-off-by: Ingo Molnar <mingo@elte.hu>
---
 fs/xfs/xfs_rtalloc.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/fs/xfs/xfs_rtalloc.c b/fs/xfs/xfs_rtalloc.c
index e2f68de..fe5de08 100644
--- a/fs/xfs/xfs_rtalloc.c
+++ b/fs/xfs/xfs_rtalloc.c
@@ -1872,7 +1872,7 @@ xfs_growfs_rt(
 	xfs_extlen_t	rsumblocks;	/* current number of rt summary blks */
 	xfs_sb_t	*sbp;		/* old superblock */
 	xfs_fsblock_t	sumbno;		/* summary block number */
-	xfs_trans_t	*tp;		/* transaction pointer */
+	xfs_trans_t	*uninitialized_var(tp);	/* transaction pointer */
 
 	sbp = &mp->m_sb;
 	cancelflags = 0;

Re: linux-next: arm allmodconfig

[Russell King - ARM Linux]

On Tue, Oct 28, 2008 at 05:56:04PM -0700, Andrew Morton wrote:
> > arch/arm/mm/dma-mapping.c: In function `dma_sync_sg_for_cpu':
> > arch/arm/mm/dma-mapping.c:588: warning: statement with no effect
> 
> Something in here:
> 
>                 dmabounce_sync_for_cpu(dev, sg_dma_address(s), 0,
>                                         sg_dma_len(s), dir);                    

That's because dmabounce_sync_for_xxx() functions return whether they
did anything, so in the case where the buffer isn't bounced, we fall
through to the standard DMA cache maintainence methods.

When there's no dmabounce, we define these functions to (1), which
of course causes the warning.  I'll eventually convert them to inline
functions instead.

> > kernel/sched.c: In function `add_preempt_count':
> > kernel/sched.c:4294: warning: unsupported arg to `__builtin_return_address'
> > kernel/sched.c:4296: warning: unsupported arg to `__builtin_return_address'
> > kernel/sched.c:4298: warning: unsupported arg to `__builtin_return_address'
> > kernel/sched.c: In function `sub_preempt_count':
> > kernel/sched.c:4294: warning: unsupported arg to `__builtin_return_address'
> > kernel/sched.c:4296: warning: unsupported arg to `__builtin_return_address'
> > kernel/sched.c:4298: warning: unsupported arg to `__builtin_return_address'
> 
> Related to CALLER_ADDR2, etc.

Later ARM toolchains seem to have dropped support for non-zero arguments
to __builtin_return_address() - presumably to support this new fangled
EABI stuff.  Not much can be done about that without screaming at the
toolchain people.

> > drivers/atm/zatm.c: In function `refill_pool':
> > drivers/atm/zatm.c:226: warning: `virt_to_bus' is deprecated (declared at /usr/src/devel/arch/arm/include/asm/memory.h:184)
> > 
> <50ish similar warnings snipped>

We really really want to see the end of virt_to_bus() in the kernel -
since ARM is a non-cache coherent architecture, every usage of it is
a bug since there'll be no cache maintainence.

So, essentially, these warnings are saying that this driver is buggy
on ARM.

> > drivers/atm/ambassador.h:257:1: warning: "FLASH_BASE" redefined
> > In file included from arch/arm/mach-versatile/include/mach/irqs.h:22,
> >                  from /usr/src/devel/arch/arm/include/asm/irq.h:4,
> >                  from /usr/src/devel/arch/arm/include/asm/hardirq.h:6,
> >                  from include/linux/hardirq.h:7,
> >                  from include/asm-generic/local.h:5,
> >                  from /usr/src/devel/arch/arm/include/asm/local.h:1,
> >                  from include/linux/module.h:20,
> >                  from drivers/atm/ambassador.c:25:
> > arch/arm/mach-versatile/include/mach/platform.h:443:1: warning: this is the location of the previous definition
> > In file included from drivers/atm/ambassador.c:44:
> > drivers/atm/ambassador.h:258:1: warning: "FLASH_SIZE" redefined

Here we go again...

I do wish that people wouldn't include "platform.h" headers defining lots
of generically named constants into a header file used by most of the
kernel build.  It's stupid for several reasons:

1. see the above warnings.
2. if you change anything in that header, you're going to get a needless
   full kernel rebuild
3. and that causes additional strain on kautobuild's build caches

We've had this in the past with sa1100-regs.h and similar, and we solved
it there.

Unfortunately, people persist in including their platform includes into
lots of header files rather than having just the relevent C files include
them.  Eg, arch/arm/plat-omap/include/mach/hardware.h:

/*
 * ---------------------------------------------------------------------------
 * Board specific defines
 * ---------------------------------------------------------------------------
 */

#ifdef CONFIG_MACH_OMAP_INNOVATOR
#include "board-innovator.h"
#endif

#ifdef CONFIG_MACH_OMAP_H2
#include "board-h2.h"
#endif
...
#ifdef CONFIG_MACH_SX1
#include "board-sx1.h"
#endif

which is included via asm/io.h, asm/irq.h, asm/pci.h, asm/vga.h.

As long as this idiocracy, we're going to see these kinds of clashes.
This stuff needs to die.  I've a good idea to do a sweep of all ARM
includes, and commit a patch to my devel tree for linux-next to remove
such includes as an encouragement to fixing this stuff properly.

People will then have a couple of months to sort out their mess.

> > crypto/testmgr.c: In function `alg_test_comp':
> > crypto/testmgr.c:829: warning: 'ret' might be used uninitialized in this function
> 
> There's no way for the compiler to know that this code isn't buggy.
> 
> Suggest that this be fixed via uninitialized_var() or, better, convert
> to a do{}while() loop.
> 
> I notice that test_comp() also has two locals called `ret' - one
> shadowing the other.

Here's a question: Why do we have to have the test manager built in if
we build some crypto modules?  What if you're building an embedded target
and don't want to test the crypto modules on every boot?  (Think code
size, boot time, etc.)

> > include/asm-generic/cmpxchg-local.h:15: warning: 'prev' might be used uninitialized in this function
> > drivers/gpu/drm/drm_lock.c: In function `drm_lock_free':
> > include/asm-generic/cmpxchg-local.h:15: warning: 'prev' might be used uninitialized in this function
> > include/asm-generic/cmpxchg-local.h:15: warning: 'prev' might be used uninitialized in this function
> > drivers/gpu/drm/drm_lock.c: In function `drm_notifier':
> > include/asm-generic/cmpxchg-local.h:15: warning: 'prev' might be used uninitialized in this function
> > drivers/gpu/drm/drm_lock.c: In function `drm_idlelock_release':
> > include/asm-generic/cmpxchg-local.h:15: warning: 'prev' might be used uninitialized in this function
> 
> didn't look.

No idea.

> > drivers/gpu/drm/drm_agpsupport.c:36:21: asm/agp.h: No such file or directory
> > make[3]: *** [drivers/gpu/drm/drm_agpsupport.o] Error 1
> > make[2]: *** [drivers/gpu/drm] Error 2
> > make[1]: *** [drivers/gpu] Error 2
> > make[1]: *** Waiting for unfinished jobs....
> 
> DRM has been breaking arm allmodconfig for at least a year.

I've no idea what asm/agp.h is supposed to have in it, no idea what
DRM actually is other than "Digital Rights Management" ;)

I suspect the right answer is to make DRM depend on !ARM - I know of
no ARM platform with AGP.

Re: linux-next: arm allmodconfig

[Pekka J Enberg]

On Tue, 28 Oct 2008, Andrew Morton wrote:
> > include/linux/slub_def.h:231: warning: 'ret' might be used uninitialized in this function
> > include/linux/slub_def.h:231: warning: 'ret' might be used uninitialized in this function
> 
> I lost interest here.

Looks like a bogus warning to me. But we might as well clean it up a bit 
and maybe the compiler is less confused that way.

		Pekka

>From 48724e2995a0fd292d326f79caebe61f4f115021 Mon Sep 17 00:00:00 2001
From: Pekka Enberg <penberg@cs.helsinki.fi>
Date: Wed, 29 Oct 2008 12:13:37 +0200
Subject: [PATCH] slub: avoid bogus warning on arm

Move the ret variable to where we actually need it in an attempt to avoid a
bugs warning when compiling for ARM.

Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
---
 include/linux/slub_def.h |    3 +--
 1 files changed, 1 insertions(+), 2 deletions(-)

diff --git a/include/linux/slub_def.h b/include/linux/slub_def.h
index dc28432..4d3de0c 100644
--- a/include/linux/slub_def.h
+++ b/include/linux/slub_def.h
@@ -228,14 +228,13 @@ static __always_inline void *kmalloc_large(size_t size, gfp_t flags)
 
 static __always_inline void *kmalloc(size_t size, gfp_t flags)
 {
-	void *ret;
-
 	if (__builtin_constant_p(size)) {
 		if (size > PAGE_SIZE)
 			return kmalloc_large(size, flags);
 
 		if (!(flags & SLUB_DMA)) {
 			struct kmem_cache *s = kmalloc_slab(size);
+			void *ret;
 
 			if (!s)
 				return ZERO_SIZE_PTR;
-- 
1.5.4.3

Re: linux-next: arm allmodconfig

[Paul Moore]

On Tuesday 28 October 2008 8:56:04 pm Andrew Morton wrote:
> > net/netlabel/netlabel_mgmt.c: In function `netlbl_mgmt_listentry':
> > net/netlabel/netlabel_mgmt.c:268: warning: 'ret_val' might be used
> > uninitialized in this function
>
> Assumes that netlbl_af4list_foreach_rcu() will never iterate zero
> times.
>
> Assumes that entry->type can never have an unexpected value.
>
> gcc cannot get this right, and C-level changes will be needed to shut
> this up.

No idea why this didn't come up when I was compiling this ... oh well, 
good thing I'm not a proud guy :)  I'll get this fixed and post a patch 
later today.

Thanks.

-- 
paul moore
linux @ hp

Re: linux-next: arm allmodconfig

[Tony Lindgren]

* Russell King - ARM Linux <linux@arm.linux.org.uk> [081029 02:59]:
> On Tue, Oct 28, 2008 at 05:56:04PM -0700, Andrew Morton wrote:

<snip>

> 
> > > drivers/atm/ambassador.h:257:1: warning: "FLASH_BASE" redefined
> > > In file included from arch/arm/mach-versatile/include/mach/irqs.h:22,
> > >                  from /usr/src/devel/arch/arm/include/asm/irq.h:4,
> > >                  from /usr/src/devel/arch/arm/include/asm/hardirq.h:6,
> > >                  from include/linux/hardirq.h:7,
> > >                  from include/asm-generic/local.h:5,
> > >                  from /usr/src/devel/arch/arm/include/asm/local.h:1,
> > >                  from include/linux/module.h:20,
> > >                  from drivers/atm/ambassador.c:25:
> > > arch/arm/mach-versatile/include/mach/platform.h:443:1: warning: this is the location of the previous definition
> > > In file included from drivers/atm/ambassador.c:44:
> > > drivers/atm/ambassador.h:258:1: warning: "FLASH_SIZE" redefined
> 
> Here we go again...
> 
> I do wish that people wouldn't include "platform.h" headers defining lots
> of generically named constants into a header file used by most of the
> kernel build.  It's stupid for several reasons:
> 
> 1. see the above warnings.
> 2. if you change anything in that header, you're going to get a needless
>    full kernel rebuild
> 3. and that causes additional strain on kautobuild's build caches
> 
> We've had this in the past with sa1100-regs.h and similar, and we solved
> it there.
> 
> Unfortunately, people persist in including their platform includes into
> lots of header files rather than having just the relevent C files include
> them.  Eg, arch/arm/plat-omap/include/mach/hardware.h:
> 
> /*
>  * ---------------------------------------------------------------------------
>  * Board specific defines
>  * ---------------------------------------------------------------------------
>  */
> 
> #ifdef CONFIG_MACH_OMAP_INNOVATOR
> #include "board-innovator.h"
> #endif
> 
> #ifdef CONFIG_MACH_OMAP_H2
> #include "board-h2.h"
> #endif
> ...
> #ifdef CONFIG_MACH_SX1
> #include "board-sx1.h"
> #endif
> 
> which is included via asm/io.h, asm/irq.h, asm/pci.h, asm/vga.h.
> 
> As long as this idiocracy, we're going to see these kinds of clashes.
> This stuff needs to die.  I've a good idea to do a sweep of all ARM
> includes, and commit a patch to my devel tree for linux-next to remove
> such includes as an encouragement to fixing this stuff properly.
> 
> People will then have a couple of months to sort out their mess.

I'll remove those omap includes, I have few other patches pending before
I get to that though.

Cheers,

Tony

Re: linux-next: arm allmodconfig

[Paul Moore]

On Wednesday 29 October 2008 9:53:02 am Paul Moore wrote:
> On Tuesday 28 October 2008 8:56:04 pm Andrew Morton wrote:
> > > net/netlabel/netlabel_mgmt.c: In function
> > > `netlbl_mgmt_listentry': net/netlabel/netlabel_mgmt.c:268:
> > > warning: 'ret_val' might be used uninitialized in this function
> >
> > Assumes that netlbl_af4list_foreach_rcu() will never iterate zero
> > times.
> >
> > Assumes that entry->type can never have an unexpected value.
> >
> > gcc cannot get this right, and C-level changes will be needed to
> > shut this up.
>
> No idea why this didn't come up when I was compiling this ... oh
> well, good thing I'm not a proud guy :)  I'll get this fixed and post
> a patch later today.

I updated my lblnet-2.6_next tree with a fix as well as another patch 
from Roel Kluin to fix a signed/unsigned mistake.  I'll post the 
patches to the relevant lists but I'm waiting one more patch from 
someone first.

In the meantime, interested parties can find the git tree here:

 * git://git.infradead.org/users/pcmoore/lblnet-2.6_next

-- 
paul moore
linux @ hp

Re: [xfs-masters] Re: linux-next: arm allmodconfig

[Dave Chinner]

On Wed, Oct 29, 2008 at 09:24:59AM +0100, Ingo Molnar wrote:
> 
> * Dave Chinner <david@fromorbit.com> wrote:
> 
> > > > fs/xfs/xfs_rtalloc.c: In function `xfs_growfs_rt':
> > > > fs/xfs/xfs_rtalloc.c:1875: warning: 'tp' might be used uninitialized in this function
> > 
> > False positive, and I don't get this reported, either.
> > 
> > Hold on - the above gcc binary only emits a warning for the 
> > xfs_growfs_rt issue when CONFIG_CC_OPTIMIZE_FOR_SIZE=y is set. It 
> > still doesn't catch the quota bug, though. This is so fucked up....
> 
> FYI, i've got 3 XFS warnings mapped in tip/auto-warnings-next:
> 
>  earth4:~/tip> gll linus..auto-warnings-next | grep xfs
> 
>  ec2f37c: work around warning in fs/xfs/xfs_mount.c

Can't say I've ever seen that warning. What gcc version is
generating it?

>  8501db3: work around warning in fs/xfs/xfs_rtalloc.c

Yeah, that's bogus.

>  8077af8: warnings: fix xfs posix acl

Should be fixed in the -next tree.

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com