include/linux/slab_def.h:135:27: warning: array subscript

Hi Christoph,

FYI, there are new compile warnings show up in

tree:   git://git.kernel.org/pub/scm/linux/kernel/git/glommer/memcg.git slab-common/kmalloc
head:   f957c1ead131972db578b945e25982f4607da6ba
commit: 55a6b9f968a877c728ed6e438fa68741a7711d89 [28/30] CK1 [12/13] Common names for the array of kmalloc caches
config: blackfin-BF533-EZKIT_defconfig

All warnings:

In file included from include/linux/slab.h:314:0,

                 from include/linux/proc_fs.h:4,
                 from fs/proc/stat.c:6:

fs/proc/stat.c: In function 'stat_open':
include/linux/slab_def.h:135:27: warning: array subscript is above array bounds [-Warray-bounds]

vim +135 include/linux/slab_def.h

36555751 (Eduard - Gabriel Munteanu 2008-08-10  119) 	struct kmem_cache *cachep;
36555751 (Eduard - Gabriel Munteanu 2008-08-10  120) 	void *ret;
36555751 (Eduard - Gabriel Munteanu 2008-08-10  121) 
2e892f43 (Christoph Lameter         2006-12-13  122) 	if (__builtin_constant_p(size)) {
9e396407 (Christoph Lameter         2012-09-26  123) 		int i;
6cb8f913 (Christoph Lameter         2007-07-17  124) 
6cb8f913 (Christoph Lameter         2007-07-17  125) 		if (!size)
6cb8f913 (Christoph Lameter         2007-07-17  126) 			return ZERO_SIZE_PTR;
6cb8f913 (Christoph Lameter         2007-07-17  127) 
9e396407 (Christoph Lameter         2012-09-26  128) 		i = kmalloc_index(size);
9e396407 (Christoph Lameter         2012-09-26  129) 
4b51d669 (Christoph Lameter         2007-02-10  130) #ifdef CONFIG_ZONE_DMA
4b51d669 (Christoph Lameter         2007-02-10  131) 		if (flags & GFP_DMA)
24e75516 (Christoph Lameter         2012-09-26  132) 			cachep = kmalloc_dma_caches[i];
36555751 (Eduard - Gabriel Munteanu 2008-08-10  133) 		else
4b51d669 (Christoph Lameter         2007-02-10  134) #endif
24e75516 (Christoph Lameter         2012-09-26 @135) 			cachep = kmalloc_caches[i];
36555751 (Eduard - Gabriel Munteanu 2008-08-10  136) 
85beb586 (Steven Rostedt            2010-11-24  137) 		ret = kmem_cache_alloc_trace(size, cachep, flags);
36555751 (Eduard - Gabriel Munteanu 2008-08-10  138) 
36555751 (Eduard - Gabriel Munteanu 2008-08-10  139) 		return ret;
2e892f43 (Christoph Lameter         2006-12-13  140) 	}
2e892f43 (Christoph Lameter         2006-12-13  141) 	return __kmalloc(size, flags);
2e892f43 (Christoph Lameter         2006-12-13  142) }
2e892f43 (Christoph Lameter         2006-12-13  143) 

---
0-DAY kernel build testing backend Open Source Technology Centre
Fengguang Wu, Yuanhan Liu Intel Corporation

Fixed by this patch:

Index: linux/mm/slab.c
=================================--- linux.orig/mm/slab.c 2012-09-28 12:20:41.504733528 -0500
+++ linux/mm/slab.c 2012-09-28 12:20:12.732133191 -0500
@@ -2406,7 +2406,7 @@ __kmem_cache_create (struct kmem_cache *
size += BYTES_PER_WORD;
}
#if FORCED_DEBUG && defined(CONFIG_DEBUG_PAGEALLOC)
- if (size >= malloc_sizes[INDEX_L3 + 1].cs_size
+ if (size >= kmalloc_size(INDEX_L3 + 1))
&& cachep->object_size > cache_line_size() && ALIGN(size, align) < PAGE_SIZE) {
cachep->obj_offset += PAGE_SIZE - ALIGN(size, align);
size = PAGE_SIZE;

Will be folded into the patchset for the next release.

Is this issue really caused by the specified commit? I do not see how the patch can cause such a problem.

It should be a wrong bisection. Sorry. I checked manually and find that the previous commit stopped early at this error:

In file included from /c/kernel-tests/src/next/include/linux/slab.h:309:0,

                 from /c/kernel-tests/src/next/include/linux/proc_fs.h:4,
                 from /c/kernel-tests/src/next/fs/proc/stat.c:6:

/c/kernel-tests/src/next/include/linux/slab_def.h: In function 'kmalloc':
/c/kernel-tests/src/next/include/linux/slab_def.h:132:13: error: 'kmalloc_dma_caches' undeclared (first use in this function) /c/kernel-tests/src/next/include/linux/slab_def.h:132:13: note: each undeclared identifier is reported only once for each function it appears in /c/kernel-tests/src/next/include/linux/slab_def.h:135:13: error: 'kmalloc_caches' undeclared (first use in this function)

Thus the warning was "hide" until the current commit:

In file included from /c/kernel-tests/src/next/include/linux/slab.h:314:0,

                 from /c/kernel-tests/src/next/include/linux/proc_fs.h:4,
                 from /c/kernel-tests/src/next/fs/proc/stat.c:6:

/c/kernel-tests/src/next/fs/proc/stat.c: In function 'stat_open':
/c/kernel-tests/src/next/include/linux/slab_def.h:135:27: warning: array subscript is above array bounds [-Warray-bounds]

Thanks,
Fengguang

Ok there were some fixes in that area in V2. Could we try again and see if the new version fixes it?

I plan to upload a version with your new patches shortly. Sorry, I am trying to get a new version of my memcg-slab series with your feedback incorporated as soon as I can, so I was a bit slow with review yesterday in general.

Well looks things are moving already much faster then in the past. So do not worry about small delays like that.