Re: [patch] do_no_pfn()

[Jes Sorensen <jes@sgi.com>]

Andrew Morton wrote:

> How does this followup look?
> 
> We don't want the rarely-used do_no_pfn() to get inlined in the oft-used
> handle_pte_fault(), using up icache.  Mark it noinline and unlikely.


I'd say it looks good - will give a microscopic slowdown for do_no_pfn
but compared to the overall benefit I think thats more than acceptable.

Acked-by: Jes Sorensen <jes@sgi.com>

Cheers,
Jes

> --- a/mm/memory.c~do_no_pfn-tweaks
> +++ a/mm/memory.c
> @@ -2276,8 +2276,10 @@ oom:
>   *
>   * It is expected that the ->nopfn handler always returns the same pfn
>   * for a given virtual mapping.
> + *
> + * Mark this `noinline' to prevent it from bloating the main pagefault code.
>   */
> -static int do_no_pfn(struct mm_struct *mm, struct vm_area_struct *vma,
> +static noinline int do_no_pfn(struct mm_struct *mm, struct vm_area_struct *vma,
>  		     unsigned long address, pte_t *page_table, pmd_t *pmd,
>  		     int write_access)
>  {
> @@ -2376,7 +2378,7 @@ static inline int handle_pte_fault(struc
>  					return do_no_page(mm, vma, address,
>  							  pte, pmd,
>  							  write_access);
> -				if (vma->vm_ops->nopfn)
> +				if (unlikely(vma->vm_ops->nopfn))
>  					return do_no_pfn(mm, vma, address, pte,
>  							 pmd, write_access);
>  			}
> _