[RFC patch] mm: hugetlb: fix __unmap_hugepage_range

[RFC patch] mm: hugetlb: fix __unmap_hugepage_range

[Hillf Danton]

First, after flushing TLB, we have no need to scan pte from start again.
Second, before bail out loop, the address is forwarded one step.

Signed-off-by: Hillf Danton <hillf.zj@alibaba-inc.com>
---

--- a/mm/hugetlb.c	Fri Oct 31 11:47:25 2014
+++ b/mm/hugetlb.c	Fri Oct 31 11:52:42 2014
@@ -2641,8 +2641,9 @@ void __unmap_hugepage_range(struct mmu_g
 
 	tlb_start_vma(tlb, vma);
 	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);
+	address = start;
 again:
-	for (address = start; address < end; address += sz) {
+	for (; address < end; address += sz) {
 		ptep = huge_pte_offset(mm, address);
 		if (!ptep)
 			continue;
@@ -2689,6 +2690,7 @@ again:
 		page_remove_rmap(page);
 		force_flush = !__tlb_remove_page(tlb, page);
 		if (force_flush) {
+			address += sz;
 			spin_unlock(ptl);
 			break;
 		}
--


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [RFC patch] mm: hugetlb: fix __unmap_hugepage_range

[Michal Hocko]

[CCing people involved in 24669e58477e2]

On Fri 31-10-14 12:22:12, Hillf Danton wrote:
> First, after flushing TLB, we have no need to scan pte from start again.
> Second, before bail out loop, the address is forwarded one step.

I can imagine a more comprehensive wording here. It is not immediately
clear whether this is just an optimization or a bug fix as well
(especially the second part).

Anyway the optimization looks good to me.

> Signed-off-by: Hillf Danton <hillf.zj@alibaba-inc.com>

Reviewed-by: Michal Hocko <mhocko@suse.cz>

> ---
> 
> --- a/mm/hugetlb.c	Fri Oct 31 11:47:25 2014
> +++ b/mm/hugetlb.c	Fri Oct 31 11:52:42 2014
> @@ -2641,8 +2641,9 @@ void __unmap_hugepage_range(struct mmu_g
>  
>  	tlb_start_vma(tlb, vma);
>  	mmu_notifier_invalidate_range_start(mm, mmun_start, mmun_end);
> +	address = start;
>  again:
> -	for (address = start; address < end; address += sz) {
> +	for (; address < end; address += sz) {
>  		ptep = huge_pte_offset(mm, address);
>  		if (!ptep)
>  			continue;
> @@ -2689,6 +2690,7 @@ again:
>  		page_remove_rmap(page);
>  		force_flush = !__tlb_remove_page(tlb, page);
>  		if (force_flush) {
> +			address += sz;
>  			spin_unlock(ptl);
>  			break;
>  		}
> --
> 
> 

-- 
Michal Hocko
SUSE Labs

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>

Re: [RFC patch] mm: hugetlb: fix __unmap_hugepage_range

[David Rientjes]

On Fri, 31 Oct 2014, Hillf Danton wrote:

> First, after flushing TLB, we have no need to scan pte from start again.
> Second, before bail out loop, the address is forwarded one step.
> 
> Signed-off-by: Hillf Danton <hillf.zj@alibaba-inc.com>

Acked-by: David Rientjes <rientjes@google.com>

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>