lkml.org 
[lkml]   [2018]   [Oct]   [5]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
Patch in this message
/
SubjectRe: [PATCH 1/4] mm/hugetlb: Enable PUD level huge page migration
From
Date


On 10/03/2018 07:06 PM, Michal Hocko wrote:
> On Wed 03-10-18 18:36:39, Anshuman Khandual wrote:
> [...]
>> So we have two checks here
>>
>> 1) platform specific arch_hugetlb_migration -> In principle go ahead
>>
>> 2) huge_movable() during allocation
>>
>> - If huge page does not have to be placed on movable zone
>>
>> - Allocate any where successfully and done !
>>
>> - If huge page *should* be placed on a movable zone
>>
>> - Try allocating on movable zone
>>
>> - Successfull and done !
>>
>> - If the new page could not be allocated on movable zone
>>
>> - Abort the migration completely
>>
>> OR
>>
>> - Warn and fall back to non-movable
>
> I guess you are still making it more complicated than necessary. The
> later is really only about __GFP_MOVABLE at this stage. I would just
> make it simple for now. We do not have to implement any dynamic
> heuristic right now. All that I am asking for is to split the migrate
> possible part from movable part.
>
> I should have been more clear about that I guess from my very first
> reply. I do like how you moved the current coarse grained
> hugepage_migration_supported to be more arch specific but I merely
> wanted to point out that we need to do some other changes before we can
> go that route and that thing is to distinguish movable from migration
> supported.
>
> See my point?

Does the following sound close enough to what you are looking for ?

diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
index 9df1d59..070c419 100644
--- a/include/linux/hugetlb.h
+++ b/include/linux/hugetlb.h
@@ -504,6 +504,13 @@ static inline bool hugepage_migration_supported(struct hstate *h)
return arch_hugetlb_migration_supported(h);
}

+static inline bool hugepage_movable_required(struct hstate *h)
+{
+ if (hstate_is_gigantic(h))
+ return true;
+ return false;
+}
+
static inline spinlock_t *huge_pte_lockptr(struct hstate *h,
struct mm_struct *mm, pte_t *pte)
{
@@ -600,6 +607,11 @@ static inline bool hugepage_migration_supported(struct hstate *h)
return false;
}

+static inline bool hugepage_movable_required(struct hstate *h)
+{
+ return false;
+}
+
static inline spinlock_t *huge_pte_lockptr(struct hstate *h,
struct mm_struct *mm, pte_t *pte)
{
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index 3c21775..8b0afdc 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -1635,6 +1635,9 @@ struct page *alloc_huge_page_node(struct hstate *h, int nid)
if (nid != NUMA_NO_NODE)
gfp_mask |= __GFP_THISNODE;

+ if (hugepage_movable_required(h))
+ gfp_mask |= __GFP_MOVABLE;
+
spin_lock(&hugetlb_lock);
if (h->free_huge_pages - h->resv_huge_pages > 0)
page = dequeue_huge_page_nodemask(h, gfp_mask, nid, NULL);
@@ -1652,6 +1655,9 @@ struct page *alloc_huge_page_nodemask(struct hstate *h, int preferred_nid,
{
gfp_t gfp_mask = htlb_alloc_mask(h);

+ if (hugepage_movable_required(h))
+ gfp_mask |= __GFP_MOVABLE;
+
spin_lock(&hugetlb_lock);
if (h->free_huge_pages - h->resv_huge_pages > 0) {
struct page *page;
\
 
 \ /
  Last update: 2018-10-05 09:35    [W:0.103 / U:4.216 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site