[PATCH next/mmotm] slab: fix DEBUG_SLAB build

[PATCH next/mmotm] slab: fix DEBUG_SLAB build

[Hugh Dickins]

[-- Attachment #1: Type: TEXT/PLAIN, Size: 1571 bytes --]

Fix CONFIG_SLAB=y CONFIG_DEBUG_SLAB=y build error and warnings.

Now that ARCH_SLAB_MINALIGN defaults to __alignof__(unsigned long long),
it is always defined (when slab.h included), but cannot be used in #if:
mm/slab.c: In function ‘cache_alloc_debugcheck_after’:
mm/slab.c:3156:5: warning: "__alignof__" is not defined
mm/slab.c:3156:5: error: missing binary operator before token "("
make[1]: *** [mm/slab.o] Error 1

So just remove the #if and #endif lines, but then 64-bit build warns:
mm/slab.c: In function ‘cache_alloc_debugcheck_after’:
mm/slab.c:3156:6: warning: cast from pointer to integer of different size
mm/slab.c:3158:10: warning: format ‘%d’ expects type ‘int’, but argument
                            3 has type ‘long unsigned int’
Fix those with casts, whatever the actual type of ARCH_SLAB_MINALIGN.

Signed-off-by: Hugh Dickins <hughd@google.com>
---
 mm/slab.c |    6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

--- mmotm/mm/slab.c	2011-07-05 12:36:54.294718254 -0700
+++ linux/mm/slab.c	2011-07-05 14:19:24.942328542 -0700
@@ -3153,12 +3153,10 @@ static void *cache_alloc_debugcheck_afte
 	objp += obj_offset(cachep);
 	if (cachep->ctor && cachep->flags & SLAB_POISON)
 		cachep->ctor(objp);
-#if ARCH_SLAB_MINALIGN
-	if ((u32)objp & (ARCH_SLAB_MINALIGN-1)) {
+	if ((unsigned long)objp & (ARCH_SLAB_MINALIGN-1)) {
 		printk(KERN_ERR "0x%p: not aligned to ARCH_SLAB_MINALIGN=%d\n",
-		       objp, ARCH_SLAB_MINALIGN);
+		       objp, (int)ARCH_SLAB_MINALIGN);
 	}
-#endif
 	return objp;
 }
 #else

[PATCH next/mmotm] slab: fix DEBUG_SLAB build

[Hugh Dickins]

Fix CONFIG_SLAB=y CONFIG_DEBUG_SLAB=y build error and warnings.

Now that ARCH_SLAB_MINALIGN defaults to __alignof__(unsigned long long),
it is always defined (when slab.h included), but cannot be used in #if:
mm/slab.c: In function `cache_alloc_debugcheck_after':
mm/slab.c:3156:5: warning: "__alignof__" is not defined
mm/slab.c:3156:5: error: missing binary operator before token "("
make[1]: *** [mm/slab.o] Error 1

So just remove the #if and #endif lines, but then 64-bit build warns:
mm/slab.c: In function `cache_alloc_debugcheck_after':
mm/slab.c:3156:6: warning: cast from pointer to integer of different size
mm/slab.c:3158:10: warning: format `%d' expects type `int', but argument
                            3 has type `long unsigned int'
Fix those with casts, whatever the actual type of ARCH_SLAB_MINALIGN.

Signed-off-by: Hugh Dickins <hughd@google.com>
---
Sorry, as akpm points out, the patch I sent first was mime-mangled:
compiler uses non-ASCII quotes in its messages, which mislead my mailer.

 mm/slab.c |    6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

--- mmotm/mm/slab.c	2011-07-05 12:36:54.294718254 -0700
+++ linux/mm/slab.c	2011-07-05 14:19:24.942328542 -0700
@@ -3153,12 +3153,10 @@ static void *cache_alloc_debugcheck_afte
 	objp += obj_offset(cachep);
 	if (cachep->ctor && cachep->flags & SLAB_POISON)
 		cachep->ctor(objp);
-#if ARCH_SLAB_MINALIGN
-	if ((u32)objp & (ARCH_SLAB_MINALIGN-1)) {
+	if ((unsigned long)objp & (ARCH_SLAB_MINALIGN-1)) {
 		printk(KERN_ERR "0x%p: not aligned to ARCH_SLAB_MINALIGN=%d\n",
-		       objp, ARCH_SLAB_MINALIGN);
+		       objp, (int)ARCH_SLAB_MINALIGN);
 	}
-#endif
 	return objp;
 }
 #else

Re: [PATCH next/mmotm] slab: fix DEBUG_SLAB build

[Christoph Lameter]

On Mon, 11 Jul 2011, Hugh Dickins wrote:

> -#if ARCH_SLAB_MINALIGN
> -	if ((u32)objp & (ARCH_SLAB_MINALIGN-1)) {
> +	if ((unsigned long)objp & (ARCH_SLAB_MINALIGN-1)) {
>  		printk(KERN_ERR "0x%p: not aligned to ARCH_SLAB_MINALIGN=%d\n",

Change %d to %ul for consistencies sake and drop the cast of
ARCH_SLAB_MINALIGN?

Re: [PATCH next/mmotm] slab: fix DEBUG_SLAB build

[Hugh Dickins]

On Tue, 12 Jul 2011, Christoph Lameter wrote:
> On Mon, 11 Jul 2011, Hugh Dickins wrote:
> 
> > -#if ARCH_SLAB_MINALIGN
> > -	if ((u32)objp & (ARCH_SLAB_MINALIGN-1)) {
> > +	if ((unsigned long)objp & (ARCH_SLAB_MINALIGN-1)) {
> >  		printk(KERN_ERR "0x%p: not aligned to ARCH_SLAB_MINALIGN=%d\n",
> 
> Change %d to %ul for consistencies sake and drop the cast of
> ARCH_SLAB_MINALIGN?

I don't think we can drop the cast of ARCH_SLAB_MINALIGN, or not without
a wander through all the architectures: it could well be defined as a bare
integer in some of them, even if the default definition as __alignof__
comes out as an unsigned long (which itself surprised me).

I could make it an unsigned long cast, and change the %d to %lu,
but there doesn't seem to be any real point to doing so.

Hugh

Re: [PATCH next/mmotm] slab: fix DEBUG_SLAB build

[Christoph Lameter]

On Tue, 12 Jul 2011, Hugh Dickins wrote:

> On Tue, 12 Jul 2011, Christoph Lameter wrote:
> > On Mon, 11 Jul 2011, Hugh Dickins wrote:
> >
> > > -#if ARCH_SLAB_MINALIGN
> > > -	if ((u32)objp & (ARCH_SLAB_MINALIGN-1)) {
> > > +	if ((unsigned long)objp & (ARCH_SLAB_MINALIGN-1)) {
> > >  		printk(KERN_ERR "0x%p: not aligned to ARCH_SLAB_MINALIGN=%d\n",
> >
> > Change %d to %ul for consistencies sake and drop the cast of
> > ARCH_SLAB_MINALIGN?
>
> I don't think we can drop the cast of ARCH_SLAB_MINALIGN, or not without
> a wander through all the architectures: it could well be defined as a bare
> integer in some of them, even if the default definition as __alignof__
> comes out as an unsigned long (which itself surprised me).

A bare integer can be handled by a %lu in a printk without the need for a
cast.

Re: [PATCH next/mmotm] slab: fix DEBUG_SLAB build

[Hugh Dickins]

On Wed, 13 Jul 2011, Christoph Lameter wrote:
> On Tue, 12 Jul 2011, Hugh Dickins wrote:
> > On Tue, 12 Jul 2011, Christoph Lameter wrote:
> > > On Mon, 11 Jul 2011, Hugh Dickins wrote:
> > >
> > > > -#if ARCH_SLAB_MINALIGN
> > > > -	if ((u32)objp & (ARCH_SLAB_MINALIGN-1)) {
> > > > +	if ((unsigned long)objp & (ARCH_SLAB_MINALIGN-1)) {
> > > >  		printk(KERN_ERR "0x%p: not aligned to ARCH_SLAB_MINALIGN=%d\n",
> > >
> > > Change %d to %ul for consistencies sake and drop the cast of
> > > ARCH_SLAB_MINALIGN?
> >
> > I don't think we can drop the cast of ARCH_SLAB_MINALIGN, or not without
> > a wander through all the architectures: it could well be defined as a bare
> > integer in some of them, even if the default definition as __alignof__
> > comes out as an unsigned long (which itself surprised me).
> 
> A bare integer can be handled by a %lu in a printk without the need for a
> cast.

Really?  Here if I try that on x86_64 with #define ARCH_SLAB_MINALIGN 8, I get
mm/slab.c: In function `cache_alloc_debugcheck_after':
mm/slab.c:3160:10: warning: format `%lu' expects type `long unsigned int', but argument 3 has type `int'

Hugh

Re: [PATCH next/mmotm] slab: fix DEBUG_SLAB build

[Christoph Lameter]

On Wed, 13 Jul 2011, Hugh Dickins wrote:

> Really?  Here if I try that on x86_64 with #define ARCH_SLAB_MINALIGN 8, I get
> mm/slab.c: In function `cache_alloc_debugcheck_after':
> mm/slab.c:3160:10: warning: format `%lu' expects type `long unsigned int', but argument 3 has type `int'

Gosh. I thought integers (especially constants) are automatically
converted to the right format. Guess that is not possible here since the
format strings is not always available for checking.

Re: [PATCH next/mmotm] slab: fix DEBUG_SLAB build

[Christoph Lameter]

Acked-by: Christoph Lameter <cl@linux.com>