[PATCH -V3 0/4] hugeltb: Fixes for hugetlb controller patches

[PATCH -V3 0/4] hugeltb: Fixes for hugetlb controller patches

[Aneesh Kumar K.V]

Hi Andrew,

This series contain fixes based on review feedback on top of the
hugetlb controller patches already in -mm. Please apply.

-aneesh

[PATCH -V3 0/4] hugeltb: Fixes for hugetlb controller patches

[Aneesh Kumar K.V]

Hi Andrew,

This series contain fixes based on review feedback on top of the
hugetlb controller patches already in -mm. Please apply.

-aneesh

--
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>

[PATCH -V3 1/4] hugeltb: Mark hugelb_max_hstate __read_mostly

[Aneesh Kumar K.V]

From: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>

We set this value only during boot.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 include/linux/hugetlb.h |    2 +-
 mm/hugetlb.c            |    2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
index 9650bb1..0f0877e 100644
--- a/include/linux/hugetlb.h
+++ b/include/linux/hugetlb.h
@@ -23,7 +23,7 @@ struct hugepage_subpool {
 };
 
 extern spinlock_t hugetlb_lock;
-extern int hugetlb_max_hstate;
+extern int hugetlb_max_hstate __read_mostly;
 #define for_each_hstate(h) \
 	for ((h) = hstates; (h) < &hstates[hugetlb_max_hstate]; (h)++)
 
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index a5a30bf..c57740b 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -37,7 +37,7 @@ const unsigned long hugetlb_zero = 0, hugetlb_infinity = ~0UL;
 static gfp_t htlb_alloc_mask = GFP_HIGHUSER;
 unsigned long hugepages_treat_as_movable;
 
-int hugetlb_max_hstate;
+int hugetlb_max_hstate __read_mostly;
 unsigned int default_hstate_idx;
 struct hstate hstates[HUGE_MAX_HSTATE];
 
-- 
1.7.10

[PATCH -V3 1/4] hugeltb: Mark hugelb_max_hstate __read_mostly

[Aneesh Kumar K.V]

From: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>

We set this value only during boot.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 include/linux/hugetlb.h |    2 +-
 mm/hugetlb.c            |    2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
index 9650bb1..0f0877e 100644
--- a/include/linux/hugetlb.h
+++ b/include/linux/hugetlb.h
@@ -23,7 +23,7 @@ struct hugepage_subpool {
 };
 
 extern spinlock_t hugetlb_lock;
-extern int hugetlb_max_hstate;
+extern int hugetlb_max_hstate __read_mostly;
 #define for_each_hstate(h) \
 	for ((h) = hstates; (h) < &hstates[hugetlb_max_hstate]; (h)++)
 
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index a5a30bf..c57740b 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -37,7 +37,7 @@ const unsigned long hugetlb_zero = 0, hugetlb_infinity = ~0UL;
 static gfp_t htlb_alloc_mask = GFP_HIGHUSER;
 unsigned long hugepages_treat_as_movable;
 
-int hugetlb_max_hstate;
+int hugetlb_max_hstate __read_mostly;
 unsigned int default_hstate_idx;
 struct hstate hstates[HUGE_MAX_HSTATE];
 
-- 
1.7.10

--
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>

[PATCH -V3 2/4] hugetlb: Move all the in use pages to active list

[Aneesh Kumar K.V]

From: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>

When we fail to allocate pages from the reserve pool, hugetlb
do try to allocate huge pages using alloc_buddy_huge_page.
Add these to the active list. We also need to add the huge
page we allocate when we soft offline the oldpage to active
list.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 mm/hugetlb.c |   11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index c57740b..ec7b86e 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -928,8 +928,14 @@ struct page *alloc_huge_page_node(struct hstate *h, int nid)
 	page = dequeue_huge_page_node(h, nid);
 	spin_unlock(&hugetlb_lock);
 
-	if (!page)
+	if (!page) {
 		page = alloc_buddy_huge_page(h, nid);
+		if (page) {
+			spin_lock(&hugetlb_lock);
+			list_move(&page->lru, &h->hugepage_activelist);
+			spin_unlock(&hugetlb_lock);
+		}
+	}
 
 	return page;
 }
@@ -1155,6 +1161,9 @@ static struct page *alloc_huge_page(struct vm_area_struct *vma,
 			hugepage_subpool_put_pages(spool, chg);
 			return ERR_PTR(-ENOSPC);
 		}
+		spin_lock(&hugetlb_lock);
+		list_move(&page->lru, &h->hugepage_activelist);
+		spin_unlock(&hugetlb_lock);
 	}
 
 	set_page_private(page, (unsigned long)spool);
-- 
1.7.10

[PATCH -V3 2/4] hugetlb: Move all the in use pages to active list

[Aneesh Kumar K.V]

From: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>

When we fail to allocate pages from the reserve pool, hugetlb
do try to allocate huge pages using alloc_buddy_huge_page.
Add these to the active list. We also need to add the huge
page we allocate when we soft offline the oldpage to active
list.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 mm/hugetlb.c |   11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)

diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index c57740b..ec7b86e 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -928,8 +928,14 @@ struct page *alloc_huge_page_node(struct hstate *h, int nid)
 	page = dequeue_huge_page_node(h, nid);
 	spin_unlock(&hugetlb_lock);
 
-	if (!page)
+	if (!page) {
 		page = alloc_buddy_huge_page(h, nid);
+		if (page) {
+			spin_lock(&hugetlb_lock);
+			list_move(&page->lru, &h->hugepage_activelist);
+			spin_unlock(&hugetlb_lock);
+		}
+	}
 
 	return page;
 }
@@ -1155,6 +1161,9 @@ static struct page *alloc_huge_page(struct vm_area_struct *vma,
 			hugepage_subpool_put_pages(spool, chg);
 			return ERR_PTR(-ENOSPC);
 		}
+		spin_lock(&hugetlb_lock);
+		list_move(&page->lru, &h->hugepage_activelist);
+		spin_unlock(&hugetlb_lock);
 	}
 
 	set_page_private(page, (unsigned long)spool);
-- 
1.7.10

--
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>

[PATCH -V3 3/4] hugetlb/cgroup: Assign the page hugetlb cgroup when we move the page to active list.

[Aneesh Kumar K.V]

From: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>

page's hugetlb cgroup assign and moving to active list should happen with
hugetlb_lock held. Otherwise when we remove the hugetlb cgroup we would
iterate the active list and will find page with NULL hugetlb cgroup values.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 mm/hugetlb.c        |   22 ++++++++++------------
 mm/hugetlb_cgroup.c |    5 +++--
 2 files changed, 13 insertions(+), 14 deletions(-)

diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index ec7b86e..c39e4be 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -928,14 +928,8 @@ struct page *alloc_huge_page_node(struct hstate *h, int nid)
 	page = dequeue_huge_page_node(h, nid);
 	spin_unlock(&hugetlb_lock);
 
-	if (!page) {
+	if (!page)
 		page = alloc_buddy_huge_page(h, nid);
-		if (page) {
-			spin_lock(&hugetlb_lock);
-			list_move(&page->lru, &h->hugepage_activelist);
-			spin_unlock(&hugetlb_lock);
-		}
-	}
 
 	return page;
 }
@@ -1150,9 +1144,13 @@ static struct page *alloc_huge_page(struct vm_area_struct *vma,
 	}
 	spin_lock(&hugetlb_lock);
 	page = dequeue_huge_page_vma(h, vma, addr, avoid_reserve);
-	spin_unlock(&hugetlb_lock);
-
-	if (!page) {
+	if (page) {
+		/* update page cgroup details */
+		hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h),
+					     h_cg, page);
+		spin_unlock(&hugetlb_lock);
+	} else {
+		spin_unlock(&hugetlb_lock);
 		page = alloc_buddy_huge_page(h, NUMA_NO_NODE);
 		if (!page) {
 			hugetlb_cgroup_uncharge_cgroup(idx,
@@ -1162,6 +1160,8 @@ static struct page *alloc_huge_page(struct vm_area_struct *vma,
 			return ERR_PTR(-ENOSPC);
 		}
 		spin_lock(&hugetlb_lock);
+		hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h),
+					     h_cg, page);
 		list_move(&page->lru, &h->hugepage_activelist);
 		spin_unlock(&hugetlb_lock);
 	}
@@ -1169,8 +1169,6 @@ static struct page *alloc_huge_page(struct vm_area_struct *vma,
 	set_page_private(page, (unsigned long)spool);
 
 	vma_commit_reservation(h, vma, addr);
-	/* update page cgroup details */
-	hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h), h_cg, page);
 	return page;
 }
 
diff --git a/mm/hugetlb_cgroup.c b/mm/hugetlb_cgroup.c
index 8e7ca0a..55e109a 100644
--- a/mm/hugetlb_cgroup.c
+++ b/mm/hugetlb_cgroup.c
@@ -218,6 +218,7 @@ done:
 	return ret;
 }
 
+/* Should be called with hugetlb_lock held */
 void hugetlb_cgroup_commit_charge(int idx, unsigned long nr_pages,
 				  struct hugetlb_cgroup *h_cg,
 				  struct page *page)
@@ -225,9 +226,7 @@ void hugetlb_cgroup_commit_charge(int idx, unsigned long nr_pages,
 	if (hugetlb_cgroup_disabled() || !h_cg)
 		return;
 
-	spin_lock(&hugetlb_lock);
 	set_hugetlb_cgroup(page, h_cg);
-	spin_unlock(&hugetlb_lock);
 	return;
 }
 
@@ -391,6 +390,7 @@ int __init hugetlb_cgroup_file_init(int idx)
 void hugetlb_cgroup_migrate(struct page *oldhpage, struct page *newhpage)
 {
 	struct hugetlb_cgroup *h_cg;
+	struct hstate *h = page_hstate(oldhpage);
 
 	if (hugetlb_cgroup_disabled())
 		return;
@@ -403,6 +403,7 @@ void hugetlb_cgroup_migrate(struct page *oldhpage, struct page *newhpage)
 
 	/* move the h_cg details to new cgroup */
 	set_hugetlb_cgroup(newhpage, h_cg);
+	list_move(&newhpage->lru, &h->hugepage_activelist);
 	spin_unlock(&hugetlb_lock);
 	cgroup_release_and_wakeup_rmdir(&h_cg->css);
 	return;
-- 
1.7.10

[PATCH -V3 3/4] hugetlb/cgroup: Assign the page hugetlb cgroup when we move the page to active list.

[Aneesh Kumar K.V]

From: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>

page's hugetlb cgroup assign and moving to active list should happen with
hugetlb_lock held. Otherwise when we remove the hugetlb cgroup we would
iterate the active list and will find page with NULL hugetlb cgroup values.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 mm/hugetlb.c        |   22 ++++++++++------------
 mm/hugetlb_cgroup.c |    5 +++--
 2 files changed, 13 insertions(+), 14 deletions(-)

diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index ec7b86e..c39e4be 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -928,14 +928,8 @@ struct page *alloc_huge_page_node(struct hstate *h, int nid)
 	page = dequeue_huge_page_node(h, nid);
 	spin_unlock(&hugetlb_lock);
 
-	if (!page) {
+	if (!page)
 		page = alloc_buddy_huge_page(h, nid);
-		if (page) {
-			spin_lock(&hugetlb_lock);
-			list_move(&page->lru, &h->hugepage_activelist);
-			spin_unlock(&hugetlb_lock);
-		}
-	}
 
 	return page;
 }
@@ -1150,9 +1144,13 @@ static struct page *alloc_huge_page(struct vm_area_struct *vma,
 	}
 	spin_lock(&hugetlb_lock);
 	page = dequeue_huge_page_vma(h, vma, addr, avoid_reserve);
-	spin_unlock(&hugetlb_lock);
-
-	if (!page) {
+	if (page) {
+		/* update page cgroup details */
+		hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h),
+					     h_cg, page);
+		spin_unlock(&hugetlb_lock);
+	} else {
+		spin_unlock(&hugetlb_lock);
 		page = alloc_buddy_huge_page(h, NUMA_NO_NODE);
 		if (!page) {
 			hugetlb_cgroup_uncharge_cgroup(idx,
@@ -1162,6 +1160,8 @@ static struct page *alloc_huge_page(struct vm_area_struct *vma,
 			return ERR_PTR(-ENOSPC);
 		}
 		spin_lock(&hugetlb_lock);
+		hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h),
+					     h_cg, page);
 		list_move(&page->lru, &h->hugepage_activelist);
 		spin_unlock(&hugetlb_lock);
 	}
@@ -1169,8 +1169,6 @@ static struct page *alloc_huge_page(struct vm_area_struct *vma,
 	set_page_private(page, (unsigned long)spool);
 
 	vma_commit_reservation(h, vma, addr);
-	/* update page cgroup details */
-	hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h), h_cg, page);
 	return page;
 }
 
diff --git a/mm/hugetlb_cgroup.c b/mm/hugetlb_cgroup.c
index 8e7ca0a..55e109a 100644
--- a/mm/hugetlb_cgroup.c
+++ b/mm/hugetlb_cgroup.c
@@ -218,6 +218,7 @@ done:
 	return ret;
 }
 
+/* Should be called with hugetlb_lock held */
 void hugetlb_cgroup_commit_charge(int idx, unsigned long nr_pages,
 				  struct hugetlb_cgroup *h_cg,
 				  struct page *page)
@@ -225,9 +226,7 @@ void hugetlb_cgroup_commit_charge(int idx, unsigned long nr_pages,
 	if (hugetlb_cgroup_disabled() || !h_cg)
 		return;
 
-	spin_lock(&hugetlb_lock);
 	set_hugetlb_cgroup(page, h_cg);
-	spin_unlock(&hugetlb_lock);
 	return;
 }
 
@@ -391,6 +390,7 @@ int __init hugetlb_cgroup_file_init(int idx)
 void hugetlb_cgroup_migrate(struct page *oldhpage, struct page *newhpage)
 {
 	struct hugetlb_cgroup *h_cg;
+	struct hstate *h = page_hstate(oldhpage);
 
 	if (hugetlb_cgroup_disabled())
 		return;
@@ -403,6 +403,7 @@ void hugetlb_cgroup_migrate(struct page *oldhpage, struct page *newhpage)
 
 	/* move the h_cg details to new cgroup */
 	set_hugetlb_cgroup(newhpage, h_cg);
+	list_move(&newhpage->lru, &h->hugepage_activelist);
 	spin_unlock(&hugetlb_lock);
 	cgroup_release_and_wakeup_rmdir(&h_cg->css);
 	return;
-- 
1.7.10

--
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>

[PATCH -V3 4/4] hugetlb/cgroup: Remove exclude and wakeup rmdir calls from migrate

[Aneesh Kumar K.V]

From: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>

We already hold the hugetlb_lock. That should prevent a parallel
cgroup rmdir from touching page's hugetlb cgroup. So remove
the exclude and wakeup calls.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 mm/hugetlb_cgroup.c |    6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/mm/hugetlb_cgroup.c b/mm/hugetlb_cgroup.c
index 55e109a..a7a0a79 100644
--- a/mm/hugetlb_cgroup.c
+++ b/mm/hugetlb_cgroup.c
@@ -387,6 +387,10 @@ int __init hugetlb_cgroup_file_init(int idx)
 	return 0;
 }
 
+/*
+ * hugetlb_lock will make sure a parallel cgroup rmdir won't happen
+ * when we migrate hugepages
+ */
 void hugetlb_cgroup_migrate(struct page *oldhpage, struct page *newhpage)
 {
 	struct hugetlb_cgroup *h_cg;
@@ -399,13 +403,11 @@ void hugetlb_cgroup_migrate(struct page *oldhpage, struct page *newhpage)
 	spin_lock(&hugetlb_lock);
 	h_cg = hugetlb_cgroup_from_page(oldhpage);
 	set_hugetlb_cgroup(oldhpage, NULL);
-	cgroup_exclude_rmdir(&h_cg->css);
 
 	/* move the h_cg details to new cgroup */
 	set_hugetlb_cgroup(newhpage, h_cg);
 	list_move(&newhpage->lru, &h->hugepage_activelist);
 	spin_unlock(&hugetlb_lock);
-	cgroup_release_and_wakeup_rmdir(&h_cg->css);
 	return;
 }
 
-- 
1.7.10

[PATCH -V3 4/4] hugetlb/cgroup: Remove exclude and wakeup rmdir calls from migrate

[Aneesh Kumar K.V]

From: "Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>

We already hold the hugetlb_lock. That should prevent a parallel
cgroup rmdir from touching page's hugetlb cgroup. So remove
the exclude and wakeup calls.

Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
---
 mm/hugetlb_cgroup.c |    6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/mm/hugetlb_cgroup.c b/mm/hugetlb_cgroup.c
index 55e109a..a7a0a79 100644
--- a/mm/hugetlb_cgroup.c
+++ b/mm/hugetlb_cgroup.c
@@ -387,6 +387,10 @@ int __init hugetlb_cgroup_file_init(int idx)
 	return 0;
 }
 
+/*
+ * hugetlb_lock will make sure a parallel cgroup rmdir won't happen
+ * when we migrate hugepages
+ */
 void hugetlb_cgroup_migrate(struct page *oldhpage, struct page *newhpage)
 {
 	struct hugetlb_cgroup *h_cg;
@@ -399,13 +403,11 @@ void hugetlb_cgroup_migrate(struct page *oldhpage, struct page *newhpage)
 	spin_lock(&hugetlb_lock);
 	h_cg = hugetlb_cgroup_from_page(oldhpage);
 	set_hugetlb_cgroup(oldhpage, NULL);
-	cgroup_exclude_rmdir(&h_cg->css);
 
 	/* move the h_cg details to new cgroup */
 	set_hugetlb_cgroup(newhpage, h_cg);
 	list_move(&newhpage->lru, &h->hugepage_activelist);
 	spin_unlock(&hugetlb_lock);
-	cgroup_release_and_wakeup_rmdir(&h_cg->css);
 	return;
 }
 
-- 
1.7.10

--
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: [PATCH -V3 0/4] hugeltb: Fixes for hugetlb controller patches

[Michal Hocko]

On Fri 15-06-12 18:11:18, Aneesh Kumar K.V wrote:
> Hi Andrew,
> 
> This series contain fixes based on review feedback on top of the
> hugetlb controller patches already in -mm. Please apply.
> 
> -aneesh
> 

You can add my to all 4 patches.
Reviewed-by: Michal Hocko <mhocko@suse.cz>

-- 
Michal Hocko
SUSE Labs
SUSE LINUX s.r.o.
Lihovarska 1060/12
190 00 Praha 9    
Czech Republic

Re: [PATCH -V3 0/4] hugeltb: Fixes for hugetlb controller patches

[Michal Hocko]

On Fri 15-06-12 18:11:18, Aneesh Kumar K.V wrote:
> Hi Andrew,
> 
> This series contain fixes based on review feedback on top of the
> hugetlb controller patches already in -mm. Please apply.
> 
> -aneesh
> 

You can add my to all 4 patches.
Reviewed-by: Michal Hocko <mhocko@suse.cz>

-- 
Michal Hocko
SUSE Labs
SUSE LINUX s.r.o.
Lihovarska 1060/12
190 00 Praha 9    
Czech Republic

--
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>