lkml.org 
[lkml]   [2008]   [Jul]   [11]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: SL*B: drop kmem cache argument from constructor
Andrew Morton wrote:
> On Thu, 10 Jul 2008 11:44:16 -0500 Jon Tollefson <kniht@linux.vnet.ibm.com> wrote:
>
>
>> Alexey Dobriyan wrote:
>>
>>> Kmem cache passed to constructor is only needed for constructors that are
>>> themselves multiplexeres. Nobody uses this "feature", nor does anybody uses
>>> passed kmem cache in non-trivial way, so pass only pointer to object.
>>>
>>> Non-trivial places are:
>>> arch/powerpc/mm/init_64.c
>>> arch/powerpc/mm/hugetlbpage.c
>>>
>>>
>> ...<snip>...
>>
>>> --- a/arch/powerpc/mm/hugetlbpage.c
>>> +++ b/arch/powerpc/mm/hugetlbpage.c
>>> @@ -595,9 +595,9 @@ static int __init hugepage_setup_sz(char *str)
>>> }
>>> __setup("hugepagesz=", hugepage_setup_sz);
>>>
>>> -static void zero_ctor(struct kmem_cache *cache, void *addr)
>>> +static void zero_ctor(void *addr)
>>> {
>>> - memset(addr, 0, kmem_cache_size(cache));
>>> + memset(addr, 0, HUGEPTE_TABLE_SIZE);
>>>
>>>
>> This isn't going to work with the multiple huge page size support. The
>> HUGEPTE_TABLE_SIZE macro now takes a parameter with of the mmu psize
>> index to indicate the size of page.
>>
>>
>
> hrm. I suppose we could hold our noses and use ksize(), assuming that
> we're ready to use ksize() at this stage in the object's lifetime.
>
> Better would be to just use kmem_cache_zalloc()?
>
> --- a/arch/powerpc/mm/hugetlbpage.c~slb-drop-kmem-cache-argument-from-constructor-fix
> +++ a/arch/powerpc/mm/hugetlbpage.c
> @@ -113,7 +113,7 @@ static inline pte_t *hugepte_offset(huge
> static int __hugepte_alloc(struct mm_struct *mm, hugepd_t *hpdp,
> unsigned long address, unsigned int psize)
> {
> - pte_t *new = kmem_cache_alloc(huge_pgtable_cache(psize),
> + pte_t *new = kmem_cache_zalloc(huge_pgtable_cache(psize),
> GFP_KERNEL|__GFP_REPEAT);
>
> if (! new)
> @@ -730,11 +730,6 @@ static int __init hugepage_setup_sz(char
> }
> __setup("hugepagesz=", hugepage_setup_sz);
>
> -static void zero_ctor(void *addr)
> -{
> - memset(addr, 0, HUGEPTE_TABLE_SIZE);
> -}
> -
> static int __init hugetlbpage_init(void)
> {
> unsigned int psize;
> @@ -756,7 +751,7 @@ static int __init hugetlbpage_init(void)
> HUGEPTE_TABLE_SIZE(psize),
> HUGEPTE_TABLE_SIZE(psize),
> 0,
> - zero_ctor);
> + NULL);
> if (!huge_pgtable_cache(psize))
> panic("hugetlbpage_init(): could not create %s"\
> "\n", HUGEPTE_CACHE_NAME(psize));
> _
>
>
> btw, Nick, what's with that dopey
>
> huge_pgtable_cache(psize) = kmem_cache_create(...
>
> trick? The result of a function call is not an lvalue, and writing a
> macro which pretends to be a function and then using it in some manner
> in which a function cannot be used is seven ways silly :(
>
That silliness came from me.
It came from my simplistic translation of the existing code to handle
multiple huge page sizes. I would agree it would be easier to read and
more straight forward to just have the indexed array directly on the
left side instead of a macro. I can send out a patch that makes that
change if desired.
Something such as

+#define HUGE_PGTABLE_INDEX(psize) (HUGEPTE_CACHE_NUM + psize - 1)

-huge_pgtable_cache(psize) = kmem_cache_create(...
+pgtable_cache[HUGE_PGTABLE_INDEX(psize)] = kmem_cache_create(...


or if there is a more accepted way of handling this situation I can
amend it differently.

Jon



\
 
 \ /
  Last update: 2008-07-11 23:43    [W:0.158 / U:0.108 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site