Re: NUMA aware slab allocator V3

[Matthew Dobson <colpatch@us.ibm.com>]

Christoph Lameter wrote:
> On Fri, 20 May 2005, Matthew Dobson wrote:
> 
> 
>>Christoph, I'm getting the following errors building rc4-mm2 w/ GCC 2.95.4:
> 
> 
> Works fine here with gcc 2.95.4.ds15-22 but that is a debian gcc 
> 2.95.4 patched up to work correctly. If you need to address the pathology in pristine 
> gcc 2.95.4 by changing the source then declare the entry field with 0 
> members.
> 
> Index: linux-2.6.12-rc4/mm/slab.c
> ===================================================================
> --- linux-2.6.12-rc4.orig/mm/slab.c	2005-05-19 21:29:45.000000000 +0000
> +++ linux-2.6.12-rc4/mm/slab.c	2005-05-20 19:18:22.000000000 +0000
> @@ -267,7 +267,7 @@
>  #ifdef CONFIG_NUMA
>  	spinlock_t lock;
>  #endif
> -	void *entry[];
> +	void *entry[0];
>  };
>  
>  /* bootstrap: The caches do not work without cpuarrays anymore,
> 
> 
> 
> gcc 2.95 can produce proper code for ppc64?

Apparently...?


>>mm/slab.c:281: field `entry' has incomplete typemm/slab.c: In function
>>'cache_alloc_refill':
> 
> 
> See patch above?

Will do.


>>mm/slab.c:2497: warning: control reaches end of non-void function
> 
> 
> That is the end of cache_alloc_debug_check right? This is a void 
> function in my source.

Nope.  It's the end of this function:
static void *cache_alloc_refill(kmem_cache_t *cachep, unsigned int __nocast
flags)

Though I'm not sure why I'm getting this warning, since the function ends
like this:
	ac->touched = 1;
	return ac->entry[--ac->avail];
} <<--  Line 2497


>>mm/slab.c: In function `kmem_cache_alloc':
>>mm/slab.c:2567: warning: `objp' might be used uninitialized in this function
>>mm/slab.c: In function `kmem_cache_alloc_node':
>>mm/slab.c:2567: warning: `objp' might be used uninitialized in this function
>>mm/slab.c: In function `__kmalloc':
>>mm/slab.c:2567: warning: `objp' might be used uninitialized in this function
> 
> 
> There is a branch there and the object is initialized in either branch.

I agree.  Not sure why this warning is occurring, either.

I tried to build this twice on this particular box, to no avail.  3x == charm?

-Matt
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/