linux-kselftest.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH v9 1/8] hugetlb_cgroup: Add hugetlb_cgroup reservation counter
@ 2019-12-17 23:16 Mina Almasry
  2019-12-17 23:16 ` [PATCH v9 2/8] hugetlb_cgroup: add interface for charge/uncharge hugetlb reservations Mina Almasry
                   ` (9 more replies)
  0 siblings, 10 replies; 27+ messages in thread
From: Mina Almasry @ 2019-12-17 23:16 UTC (permalink / raw)
  To: mike.kravetz
  Cc: shuah, almasrymina, rientjes, shakeelb, gthelen, akpm,
	linux-kernel, linux-mm, linux-kselftest, cgroups, aneesh.kumar,
	mkoutny, Hillf Danton

These counters will track hugetlb reservations rather than hugetlb
memory faulted in. This patch only adds the counter, following patches
add the charging and uncharging of the counter.

This is patch 1 of an 8 patch series.

Problem:
Currently tasks attempting to allocate more hugetlb memory than is available get
a failure at mmap/shmget time. This is thanks to Hugetlbfs Reservations [1].
However, if a task attempts to allocate hugetlb memory only more than its
hugetlb_cgroup limit allows, the kernel will allow the mmap/shmget call,
but will SIGBUS the task when it attempts to fault the memory in.

We have developers interested in using hugetlb_cgroups, and they have expressed
dissatisfaction regarding this behavior. We'd like to improve this
behavior such that tasks violating the hugetlb_cgroup limits get an error on
mmap/shmget time, rather than getting SIGBUS'd when they try to fault
the excess memory in.

The underlying problem is that today's hugetlb_cgroup accounting happens
at hugetlb memory *fault* time, rather than at *reservation* time.
Thus, enforcing the hugetlb_cgroup limit only happens at fault time, and
the offending task gets SIGBUS'd.

Proposed Solution:
A new page counter named hugetlb.xMB.reservation_[limit|usage]_in_bytes. This
counter has slightly different semantics than
hugetlb.xMB.[limit|usage]_in_bytes:

- While usage_in_bytes tracks all *faulted* hugetlb memory,
reservation_usage_in_bytes tracks all *reserved* hugetlb memory and
hugetlb memory faulted in without a prior reservation.

- If a task attempts to reserve more memory than limit_in_bytes allows,
the kernel will allow it to do so. But if a task attempts to reserve
more memory than reservation_limit_in_bytes, the kernel will fail this
reservation.

This proposal is implemented in this patch series, with tests to verify
functionality and show the usage. We also added cgroup-v2 support to
hugetlb_cgroup so that the new use cases can be extended to v2.

Alternatives considered:
1. A new cgroup, instead of only a new page_counter attached to
   the existing hugetlb_cgroup. Adding a new cgroup seemed like a lot of code
   duplication with hugetlb_cgroup. Keeping hugetlb related page counters under
   hugetlb_cgroup seemed cleaner as well.

2. Instead of adding a new counter, we considered adding a sysctl that modifies
   the behavior of hugetlb.xMB.[limit|usage]_in_bytes, to do accounting at
   reservation time rather than fault time. Adding a new page_counter seems
   better as userspace could, if it wants, choose to enforce different cgroups
   differently: one via limit_in_bytes, and another via
   reservation_limit_in_bytes. This could be very useful if you're
   transitioning how hugetlb memory is partitioned on your system one
   cgroup at a time, for example. Also, someone may find usage for both
   limit_in_bytes and reservation_limit_in_bytes concurrently, and this
   approach gives them the option to do so.

Testing:
- Added tests passing.
- Used libhugetlbfs for regression testing.

[1]: https://www.kernel.org/doc/html/latest/vm/hugetlbfs_reserv.html

Signed-off-by: Mina Almasry <almasrymina@google.com>
Acked-by: Hillf Danton <hdanton@sina.com>

---
 include/linux/hugetlb.h |   4 +-
 mm/hugetlb_cgroup.c     | 116 +++++++++++++++++++++++++++++++++++-----
 2 files changed, 106 insertions(+), 14 deletions(-)

diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
index 1e897e4168ac1..dea6143aa0685 100644
--- a/include/linux/hugetlb.h
+++ b/include/linux/hugetlb.h
@@ -432,8 +432,8 @@ struct hstate {
 	unsigned int surplus_huge_pages_node[MAX_NUMNODES];
 #ifdef CONFIG_CGROUP_HUGETLB
 	/* cgroup control files */
-	struct cftype cgroup_files_dfl[5];
-	struct cftype cgroup_files_legacy[5];
+	struct cftype cgroup_files_dfl[7];
+	struct cftype cgroup_files_legacy[9];
 #endif
 	char name[HSTATE_NAME_LEN];
 };
diff --git a/mm/hugetlb_cgroup.c b/mm/hugetlb_cgroup.c
index e434b05416c68..35415af9ed26f 100644
--- a/mm/hugetlb_cgroup.c
+++ b/mm/hugetlb_cgroup.c
@@ -36,6 +36,11 @@ struct hugetlb_cgroup {
 	 */
 	struct page_counter hugepage[HUGE_MAX_HSTATE];

+	/*
+	 * the counter to account for hugepage reservations from hugetlb.
+	 */
+	struct page_counter reserved_hugepage[HUGE_MAX_HSTATE];
+
 	atomic_long_t events[HUGE_MAX_HSTATE][HUGETLB_NR_MEMORY_EVENTS];
 	atomic_long_t events_local[HUGE_MAX_HSTATE][HUGETLB_NR_MEMORY_EVENTS];

@@ -55,6 +60,14 @@ struct hugetlb_cgroup {

 static struct hugetlb_cgroup *root_h_cgroup __read_mostly;

+static inline struct page_counter *
+hugetlb_cgroup_get_counter(struct hugetlb_cgroup *h_cg, int idx, bool reserved)
+{
+	if (reserved)
+		return &h_cg->reserved_hugepage[idx];
+	return &h_cg->hugepage[idx];
+}
+
 static inline
 struct hugetlb_cgroup *hugetlb_cgroup_from_css(struct cgroup_subsys_state *s)
 {
@@ -295,28 +308,42 @@ void hugetlb_cgroup_uncharge_cgroup(int idx, unsigned long nr_pages,

 enum {
 	RES_USAGE,
+	RES_RESERVATION_USAGE,
 	RES_LIMIT,
+	RES_RESERVATION_LIMIT,
 	RES_MAX_USAGE,
+	RES_RESERVATION_MAX_USAGE,
 	RES_FAILCNT,
+	RES_RESERVATION_FAILCNT,
 };

 static u64 hugetlb_cgroup_read_u64(struct cgroup_subsys_state *css,
 				   struct cftype *cft)
 {
 	struct page_counter *counter;
+	struct page_counter *reserved_counter;
 	struct hugetlb_cgroup *h_cg = hugetlb_cgroup_from_css(css);

 	counter = &h_cg->hugepage[MEMFILE_IDX(cft->private)];
+	reserved_counter = &h_cg->reserved_hugepage[MEMFILE_IDX(cft->private)];

 	switch (MEMFILE_ATTR(cft->private)) {
 	case RES_USAGE:
 		return (u64)page_counter_read(counter) * PAGE_SIZE;
+	case RES_RESERVATION_USAGE:
+		return (u64)page_counter_read(reserved_counter) * PAGE_SIZE;
 	case RES_LIMIT:
 		return (u64)counter->max * PAGE_SIZE;
+	case RES_RESERVATION_LIMIT:
+		return (u64)reserved_counter->max * PAGE_SIZE;
 	case RES_MAX_USAGE:
 		return (u64)counter->watermark * PAGE_SIZE;
+	case RES_RESERVATION_MAX_USAGE:
+		return (u64)reserved_counter->watermark * PAGE_SIZE;
 	case RES_FAILCNT:
 		return counter->failcnt;
+	case RES_RESERVATION_FAILCNT:
+		return reserved_counter->failcnt;
 	default:
 		BUG();
 	}
@@ -338,10 +365,16 @@ static int hugetlb_cgroup_read_u64_max(struct seq_file *seq, void *v)
 			   1 << huge_page_order(&hstates[idx]));

 	switch (MEMFILE_ATTR(cft->private)) {
+	case RES_RESERVATION_USAGE:
+		counter = &h_cg->reserved_hugepage[idx];
+		/* Fall through. */
 	case RES_USAGE:
 		val = (u64)page_counter_read(counter);
 		seq_printf(seq, "%llu\n", val * PAGE_SIZE);
 		break;
+	case RES_RESERVATION_LIMIT:
+		counter = &h_cg->reserved_hugepage[idx];
+		/* Fall through. */
 	case RES_LIMIT:
 		val = (u64)counter->max;
 		if (val == limit)
@@ -365,6 +398,7 @@ static ssize_t hugetlb_cgroup_write(struct kernfs_open_file *of,
 	int ret, idx;
 	unsigned long nr_pages;
 	struct hugetlb_cgroup *h_cg = hugetlb_cgroup_from_css(of_css(of));
+	bool reserved = false;

 	if (hugetlb_cgroup_is_root(h_cg)) /* Can't set limit on root */
 		return -EINVAL;
@@ -378,9 +412,14 @@ static ssize_t hugetlb_cgroup_write(struct kernfs_open_file *of,
 	nr_pages = round_down(nr_pages, 1 << huge_page_order(&hstates[idx]));

 	switch (MEMFILE_ATTR(of_cft(of)->private)) {
+	case RES_RESERVATION_LIMIT:
+		reserved = true;
+		/* Fall through. */
 	case RES_LIMIT:
 		mutex_lock(&hugetlb_limit_mutex);
-		ret = page_counter_set_max(&h_cg->hugepage[idx], nr_pages);
+		ret = page_counter_set_max(hugetlb_cgroup_get_counter(h_cg, idx,
+								      reserved),
+					   nr_pages);
 		mutex_unlock(&hugetlb_limit_mutex);
 		break;
 	default:
@@ -406,18 +445,26 @@ static ssize_t hugetlb_cgroup_reset(struct kernfs_open_file *of,
 				    char *buf, size_t nbytes, loff_t off)
 {
 	int ret = 0;
-	struct page_counter *counter;
+	struct page_counter *counter, *reserved_counter;
 	struct hugetlb_cgroup *h_cg = hugetlb_cgroup_from_css(of_css(of));

 	counter = &h_cg->hugepage[MEMFILE_IDX(of_cft(of)->private)];
+	reserved_counter =
+		&h_cg->reserved_hugepage[MEMFILE_IDX(of_cft(of)->private)];

 	switch (MEMFILE_ATTR(of_cft(of)->private)) {
 	case RES_MAX_USAGE:
 		page_counter_reset_watermark(counter);
 		break;
+	case RES_RESERVATION_MAX_USAGE:
+		page_counter_reset_watermark(reserved_counter);
+		break;
 	case RES_FAILCNT:
 		counter->failcnt = 0;
 		break;
+	case RES_RESERVATION_FAILCNT:
+		reserved_counter->failcnt = 0;
+		break;
 	default:
 		ret = -EINVAL;
 		break;
@@ -472,7 +519,7 @@ static void __init __hugetlb_cgroup_file_dfl_init(int idx)
 	struct hstate *h = &hstates[idx];

 	/* format the size */
-	mem_fmt(buf, 32, huge_page_size(h));
+	mem_fmt(buf, sizeof(buf), huge_page_size(h));

 	/* Add the limit file */
 	cft = &h->cgroup_files_dfl[0];
@@ -482,15 +529,30 @@ static void __init __hugetlb_cgroup_file_dfl_init(int idx)
 	cft->write = hugetlb_cgroup_write_dfl;
 	cft->flags = CFTYPE_NOT_ON_ROOT;

-	/* Add the current usage file */
+	/* Add the reservation limit file */
 	cft = &h->cgroup_files_dfl[1];
+	snprintf(cft->name, MAX_CFTYPE_NAME, "%s.reservation_max", buf);
+	cft->private = MEMFILE_PRIVATE(idx, RES_RESERVATION_LIMIT);
+	cft->seq_show = hugetlb_cgroup_read_u64_max;
+	cft->write = hugetlb_cgroup_write_dfl;
+	cft->flags = CFTYPE_NOT_ON_ROOT;
+
+	/* Add the current usage file */
+	cft = &h->cgroup_files_dfl[2];
 	snprintf(cft->name, MAX_CFTYPE_NAME, "%s.current", buf);
 	cft->private = MEMFILE_PRIVATE(idx, RES_USAGE);
 	cft->seq_show = hugetlb_cgroup_read_u64_max;
 	cft->flags = CFTYPE_NOT_ON_ROOT;

+	/* Add the current reservation usage file */
+	cft = &h->cgroup_files_dfl[3];
+	snprintf(cft->name, MAX_CFTYPE_NAME, "%s.reservation_current", buf);
+	cft->private = MEMFILE_PRIVATE(idx, RES_RESERVATION_USAGE);
+	cft->seq_show = hugetlb_cgroup_read_u64_max;
+	cft->flags = CFTYPE_NOT_ON_ROOT;
+
 	/* Add the events file */
-	cft = &h->cgroup_files_dfl[2];
+	cft = &h->cgroup_files_dfl[4];
 	snprintf(cft->name, MAX_CFTYPE_NAME, "%s.events", buf);
 	cft->private = MEMFILE_PRIVATE(idx, 0);
 	cft->seq_show = hugetlb_events_show;
@@ -498,7 +560,7 @@ static void __init __hugetlb_cgroup_file_dfl_init(int idx)
 	cft->flags = CFTYPE_NOT_ON_ROOT;

 	/* Add the events.local file */
-	cft = &h->cgroup_files_dfl[3];
+	cft = &h->cgroup_files_dfl[5];
 	snprintf(cft->name, MAX_CFTYPE_NAME, "%s.events.local", buf);
 	cft->private = MEMFILE_PRIVATE(idx, 0);
 	cft->seq_show = hugetlb_events_local_show;
@@ -507,7 +569,7 @@ static void __init __hugetlb_cgroup_file_dfl_init(int idx)
 	cft->flags = CFTYPE_NOT_ON_ROOT;

 	/* NULL terminate the last cft */
-	cft = &h->cgroup_files_dfl[4];
+	cft = &h->cgroup_files_dfl[6];
 	memset(cft, 0, sizeof(*cft));

 	WARN_ON(cgroup_add_dfl_cftypes(&hugetlb_cgrp_subsys,
@@ -530,28 +592,58 @@ static void __init __hugetlb_cgroup_file_legacy_init(int idx)
 	cft->read_u64 = hugetlb_cgroup_read_u64;
 	cft->write = hugetlb_cgroup_write_legacy;

-	/* Add the usage file */
+	/* Add the reservation limit file */
 	cft = &h->cgroup_files_legacy[1];
+	snprintf(cft->name, MAX_CFTYPE_NAME, "%s.reservation_limit_in_bytes",
+		 buf);
+	cft->private = MEMFILE_PRIVATE(idx, RES_RESERVATION_LIMIT);
+	cft->read_u64 = hugetlb_cgroup_read_u64;
+	cft->write = hugetlb_cgroup_write_legacy;
+
+	/* Add the usage file */
+	cft = &h->cgroup_files_legacy[2];
 	snprintf(cft->name, MAX_CFTYPE_NAME, "%s.usage_in_bytes", buf);
 	cft->private = MEMFILE_PRIVATE(idx, RES_USAGE);
 	cft->read_u64 = hugetlb_cgroup_read_u64;

+	/* Add the reservation usage file */
+	cft = &h->cgroup_files_legacy[3];
+	snprintf(cft->name, MAX_CFTYPE_NAME, "%s.reservation_usage_in_bytes",
+		 buf);
+	cft->private = MEMFILE_PRIVATE(idx, RES_RESERVATION_USAGE);
+	cft->read_u64 = hugetlb_cgroup_read_u64;
+
 	/* Add the MAX usage file */
-	cft = &h->cgroup_files_legacy[2];
+	cft = &h->cgroup_files_legacy[4];
 	snprintf(cft->name, MAX_CFTYPE_NAME, "%s.max_usage_in_bytes", buf);
 	cft->private = MEMFILE_PRIVATE(idx, RES_MAX_USAGE);
 	cft->write = hugetlb_cgroup_reset;
 	cft->read_u64 = hugetlb_cgroup_read_u64;

+	/* Add the MAX reservation usage file */
+	cft = &h->cgroup_files_legacy[5];
+	snprintf(cft->name, MAX_CFTYPE_NAME,
+		 "%s.reservation_max_usage_in_bytes", buf);
+	cft->private = MEMFILE_PRIVATE(idx, RES_RESERVATION_MAX_USAGE);
+	cft->write = hugetlb_cgroup_reset;
+	cft->read_u64 = hugetlb_cgroup_read_u64;
+
 	/* Add the failcntfile */
-	cft = &h->cgroup_files_legacy[3];
+	cft = &h->cgroup_files_legacy[6];
 	snprintf(cft->name, MAX_CFTYPE_NAME, "%s.failcnt", buf);
-	cft->private  = MEMFILE_PRIVATE(idx, RES_FAILCNT);
+	cft->private = MEMFILE_PRIVATE(idx, RES_FAILCNT);
+	cft->write = hugetlb_cgroup_reset;
+	cft->read_u64 = hugetlb_cgroup_read_u64;
+
+	/* Add the reservation failcntfile */
+	cft = &h->cgroup_files_legacy[7];
+	snprintf(cft->name, MAX_CFTYPE_NAME, "%s.reservation_failcnt", buf);
+	cft->private = MEMFILE_PRIVATE(idx, RES_RESERVATION_FAILCNT);
 	cft->write = hugetlb_cgroup_reset;
 	cft->read_u64 = hugetlb_cgroup_read_u64;

 	/* NULL terminate the last cft */
-	cft = &h->cgroup_files_legacy[4];
+	cft = &h->cgroup_files_legacy[8];
 	memset(cft, 0, sizeof(*cft));

 	WARN_ON(cgroup_add_legacy_cftypes(&hugetlb_cgrp_subsys,
--
2.24.1.735.g03f4e72817-goog

^ permalink raw reply related	[flat|nested] 27+ messages in thread

* [PATCH v9 2/8] hugetlb_cgroup: add interface for charge/uncharge hugetlb reservations
  2019-12-17 23:16 [PATCH v9 1/8] hugetlb_cgroup: Add hugetlb_cgroup reservation counter Mina Almasry
@ 2019-12-17 23:16 ` Mina Almasry
  2020-01-13 22:14   ` Mike Kravetz
  2019-12-17 23:16 ` [PATCH v9 3/8] hugetlb_cgroup: add reservation accounting for private mappings Mina Almasry
                   ` (8 subsequent siblings)
  9 siblings, 1 reply; 27+ messages in thread
From: Mina Almasry @ 2019-12-17 23:16 UTC (permalink / raw)
  To: mike.kravetz
  Cc: shuah, almasrymina, rientjes, shakeelb, gthelen, akpm,
	linux-kernel, linux-mm, linux-kselftest, cgroups, aneesh.kumar,
	mkoutny

Augments hugetlb_cgroup_charge_cgroup to be able to charge hugetlb
usage or hugetlb reservation counter.

Adds a new interface to uncharge a hugetlb_cgroup counter via
hugetlb_cgroup_uncharge_counter.

Integrates the counter with hugetlb_cgroup, via hugetlb_cgroup_init,
hugetlb_cgroup_have_usage, and hugetlb_cgroup_css_offline.

Signed-off-by: Mina Almasry <almasrymina@google.com>

---

Changes in V9:
- Fixed HUGETLB_CGROUP_MIN_ORDER.
- Minor variable name update.
- Moved some init/cleanup code from later patches in the series to this patch.
- Updated reparenting of reservation accounting.

---
 include/linux/hugetlb_cgroup.h | 68 ++++++++++++++---------
 mm/hugetlb.c                   | 19 ++++---
 mm/hugetlb_cgroup.c            | 98 +++++++++++++++++++++++++---------
 3 files changed, 127 insertions(+), 58 deletions(-)

diff --git a/include/linux/hugetlb_cgroup.h b/include/linux/hugetlb_cgroup.h
index 063962f6dfc6a..eab8a70d5bcb5 100644
--- a/include/linux/hugetlb_cgroup.h
+++ b/include/linux/hugetlb_cgroup.h
@@ -20,29 +20,37 @@
 struct hugetlb_cgroup;
 /*
  * Minimum page order trackable by hugetlb cgroup.
- * At least 3 pages are necessary for all the tracking information.
+ * At least 4 pages are necessary for all the tracking information.
  */
 #define HUGETLB_CGROUP_MIN_ORDER	2

 #ifdef CONFIG_CGROUP_HUGETLB

-static inline struct hugetlb_cgroup *hugetlb_cgroup_from_page(struct page *page)
+static inline struct hugetlb_cgroup *hugetlb_cgroup_from_page(struct page *page,
+							      bool reserved)
 {
 	VM_BUG_ON_PAGE(!PageHuge(page), page);

 	if (compound_order(page) < HUGETLB_CGROUP_MIN_ORDER)
 		return NULL;
-	return (struct hugetlb_cgroup *)page[2].private;
+	if (reserved)
+		return (struct hugetlb_cgroup *)page[3].private;
+	else
+		return (struct hugetlb_cgroup *)page[2].private;
 }

-static inline
-int set_hugetlb_cgroup(struct page *page, struct hugetlb_cgroup *h_cg)
+static inline int set_hugetlb_cgroup(struct page *page,
+				     struct hugetlb_cgroup *h_cg,
+				     bool reservation)
 {
 	VM_BUG_ON_PAGE(!PageHuge(page), page);

 	if (compound_order(page) < HUGETLB_CGROUP_MIN_ORDER)
 		return -1;
-	page[2].private	= (unsigned long)h_cg;
+	if (reservation)
+		page[3].private = (unsigned long)h_cg;
+	else
+		page[2].private = (unsigned long)h_cg;
 	return 0;
 }

@@ -52,26 +60,34 @@ static inline bool hugetlb_cgroup_disabled(void)
 }

 extern int hugetlb_cgroup_charge_cgroup(int idx, unsigned long nr_pages,
-					struct hugetlb_cgroup **ptr);
+					struct hugetlb_cgroup **ptr,
+					bool reserved);
 extern void hugetlb_cgroup_commit_charge(int idx, unsigned long nr_pages,
 					 struct hugetlb_cgroup *h_cg,
-					 struct page *page);
+					 struct page *page, bool reserved);
 extern void hugetlb_cgroup_uncharge_page(int idx, unsigned long nr_pages,
-					 struct page *page);
+					 struct page *page, bool reserved);
+
 extern void hugetlb_cgroup_uncharge_cgroup(int idx, unsigned long nr_pages,
-					   struct hugetlb_cgroup *h_cg);
+					   struct hugetlb_cgroup *h_cg,
+					   bool reserved);
+extern void hugetlb_cgroup_uncharge_counter(struct page_counter *p,
+					    unsigned long nr_pages,
+					    struct cgroup_subsys_state *css);
+
 extern void hugetlb_cgroup_file_init(void) __init;
 extern void hugetlb_cgroup_migrate(struct page *oldhpage,
 				   struct page *newhpage);

 #else
-static inline struct hugetlb_cgroup *hugetlb_cgroup_from_page(struct page *page)
+static inline struct hugetlb_cgroup *hugetlb_cgroup_from_page(struct page *page,
+							      bool reserved)
 {
 	return NULL;
 }

-static inline
-int set_hugetlb_cgroup(struct page *page, struct hugetlb_cgroup *h_cg)
+static inline int set_hugetlb_cgroup(struct page *page,
+				     struct hugetlb_cgroup *h_cg, bool reserved)
 {
 	return 0;
 }
@@ -81,28 +97,30 @@ static inline bool hugetlb_cgroup_disabled(void)
 	return true;
 }

-static inline int
-hugetlb_cgroup_charge_cgroup(int idx, unsigned long nr_pages,
-			     struct hugetlb_cgroup **ptr)
+static inline int hugetlb_cgroup_charge_cgroup(int idx, unsigned long nr_pages,
+					       struct hugetlb_cgroup **ptr,
+					       bool reserved)
 {
 	return 0;
 }

-static inline void
-hugetlb_cgroup_commit_charge(int idx, unsigned long nr_pages,
-			     struct hugetlb_cgroup *h_cg,
-			     struct page *page)
+static inline void hugetlb_cgroup_commit_charge(int idx, unsigned long nr_pages,
+						struct hugetlb_cgroup *h_cg,
+						struct page *page,
+						bool reserved)
 {
 }

-static inline void
-hugetlb_cgroup_uncharge_page(int idx, unsigned long nr_pages, struct page *page)
+static inline void hugetlb_cgroup_uncharge_page(int idx, unsigned long nr_pages,
+						struct page *page,
+						bool reserved)
 {
 }

-static inline void
-hugetlb_cgroup_uncharge_cgroup(int idx, unsigned long nr_pages,
-			       struct hugetlb_cgroup *h_cg)
+static inline void hugetlb_cgroup_uncharge_cgroup(int idx,
+						  unsigned long nr_pages,
+						  struct hugetlb_cgroup *h_cg,
+						  bool reserved)
 {
 }

diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index ac65bb5e38ac2..e6e8240f1718c 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -1067,7 +1067,8 @@ static void update_and_free_page(struct hstate *h, struct page *page)
 				1 << PG_active | 1 << PG_private |
 				1 << PG_writeback);
 	}
-	VM_BUG_ON_PAGE(hugetlb_cgroup_from_page(page), page);
+	VM_BUG_ON_PAGE(hugetlb_cgroup_from_page(page, false), page);
+	VM_BUG_ON_PAGE(hugetlb_cgroup_from_page(page, true), page);
 	set_compound_page_dtor(page, NULL_COMPOUND_DTOR);
 	set_page_refcounted(page);
 	if (hstate_is_gigantic(h)) {
@@ -1177,8 +1178,8 @@ void free_huge_page(struct page *page)

 	spin_lock(&hugetlb_lock);
 	clear_page_huge_active(page);
-	hugetlb_cgroup_uncharge_page(hstate_index(h),
-				     pages_per_huge_page(h), page);
+	hugetlb_cgroup_uncharge_page(hstate_index(h), pages_per_huge_page(h),
+				     page, false);
 	if (restore_reserve)
 		h->resv_huge_pages++;

@@ -1204,7 +1205,8 @@ static void prep_new_huge_page(struct hstate *h, struct page *page, int nid)
 	INIT_LIST_HEAD(&page->lru);
 	set_compound_page_dtor(page, HUGETLB_PAGE_DTOR);
 	spin_lock(&hugetlb_lock);
-	set_hugetlb_cgroup(page, NULL);
+	set_hugetlb_cgroup(page, NULL, false);
+	set_hugetlb_cgroup(page, NULL, true);
 	h->nr_huge_pages++;
 	h->nr_huge_pages_node[nid]++;
 	spin_unlock(&hugetlb_lock);
@@ -1990,7 +1992,8 @@ struct page *alloc_huge_page(struct vm_area_struct *vma,
 			gbl_chg = 1;
 	}

-	ret = hugetlb_cgroup_charge_cgroup(idx, pages_per_huge_page(h), &h_cg);
+	ret = hugetlb_cgroup_charge_cgroup(idx, pages_per_huge_page(h), &h_cg,
+					   false);
 	if (ret)
 		goto out_subpool_put;

@@ -2014,7 +2017,8 @@ struct page *alloc_huge_page(struct vm_area_struct *vma,
 		list_move(&page->lru, &h->hugepage_activelist);
 		/* Fall through */
 	}
-	hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h), h_cg, page);
+	hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h), h_cg, page,
+				     false);
 	spin_unlock(&hugetlb_lock);

 	set_page_private(page, (unsigned long)spool);
@@ -2038,7 +2042,8 @@ struct page *alloc_huge_page(struct vm_area_struct *vma,
 	return page;

 out_uncharge_cgroup:
-	hugetlb_cgroup_uncharge_cgroup(idx, pages_per_huge_page(h), h_cg);
+	hugetlb_cgroup_uncharge_cgroup(idx, pages_per_huge_page(h), h_cg,
+				       false);
 out_subpool_put:
 	if (map_chg || avoid_reserve)
 		hugepage_subpool_put_pages(spool, 1);
diff --git a/mm/hugetlb_cgroup.c b/mm/hugetlb_cgroup.c
index 35415af9ed26f..b03270b0d5833 100644
--- a/mm/hugetlb_cgroup.c
+++ b/mm/hugetlb_cgroup.c
@@ -96,8 +96,12 @@ static inline bool hugetlb_cgroup_have_usage(struct hugetlb_cgroup *h_cg)
 	int idx;

 	for (idx = 0; idx < hugetlb_max_hstate; idx++) {
-		if (page_counter_read(&h_cg->hugepage[idx]))
+		if (page_counter_read(
+			    hugetlb_cgroup_get_counter(h_cg, idx, true)) ||
+		    page_counter_read(
+			    hugetlb_cgroup_get_counter(h_cg, idx, false))) {
 			return true;
+		}
 	}
 	return false;
 }
@@ -108,18 +112,32 @@ static void hugetlb_cgroup_init(struct hugetlb_cgroup *h_cgroup,
 	int idx;

 	for (idx = 0; idx < HUGE_MAX_HSTATE; idx++) {
-		struct page_counter *counter = &h_cgroup->hugepage[idx];
-		struct page_counter *parent = NULL;
+		struct page_counter *fault_parent = NULL;
+		struct page_counter *reserved_parent = NULL;
 		unsigned long limit;
 		int ret;

-		if (parent_h_cgroup)
-			parent = &parent_h_cgroup->hugepage[idx];
-		page_counter_init(counter, parent);
+		if (parent_h_cgroup) {
+			fault_parent = hugetlb_cgroup_get_counter(
+				parent_h_cgroup, idx, false);
+			reserved_parent = hugetlb_cgroup_get_counter(
+				parent_h_cgroup, idx, true);
+		}
+		page_counter_init(hugetlb_cgroup_get_counter(h_cgroup, idx,
+							     false),
+				  fault_parent);
+		page_counter_init(hugetlb_cgroup_get_counter(h_cgroup, idx,
+							     true),
+				  reserved_parent);

 		limit = round_down(PAGE_COUNTER_MAX,
 				   1 << huge_page_order(&hstates[idx]));
-		ret = page_counter_set_max(counter, limit);
+
+		ret = page_counter_set_max(
+			hugetlb_cgroup_get_counter(h_cgroup, idx, false),
+			limit);
+		ret = page_counter_set_max(
+			hugetlb_cgroup_get_counter(h_cgroup, idx, true), limit);
 		VM_BUG_ON(ret);
 	}
 }
@@ -149,7 +167,6 @@ static void hugetlb_cgroup_css_free(struct cgroup_subsys_state *css)
 	kfree(h_cgroup);
 }

-
 /*
  * Should be called with hugetlb_lock held.
  * Since we are holding hugetlb_lock, pages cannot get moved from
@@ -165,7 +182,7 @@ static void hugetlb_cgroup_move_parent(int idx, struct hugetlb_cgroup *h_cg,
 	struct hugetlb_cgroup *page_hcg;
 	struct hugetlb_cgroup *parent = parent_hugetlb_cgroup(h_cg);

-	page_hcg = hugetlb_cgroup_from_page(page);
+	page_hcg = hugetlb_cgroup_from_page(page, false);
 	/*
 	 * We can have pages in active list without any cgroup
 	 * ie, hugepage with less than 3 pages. We can safely
@@ -184,7 +201,7 @@ static void hugetlb_cgroup_move_parent(int idx, struct hugetlb_cgroup *h_cg,
 	/* Take the pages off the local counter */
 	page_counter_cancel(counter, nr_pages);

-	set_hugetlb_cgroup(page, parent);
+	set_hugetlb_cgroup(page, parent, false);
 out:
 	return;
 }
@@ -227,7 +244,7 @@ static inline void hugetlb_event(struct hugetlb_cgroup *hugetlb, int idx,
 }

 int hugetlb_cgroup_charge_cgroup(int idx, unsigned long nr_pages,
-				 struct hugetlb_cgroup **ptr)
+				 struct hugetlb_cgroup **ptr, bool reserved)
 {
 	int ret = 0;
 	struct page_counter *counter;
@@ -250,13 +267,20 @@ int hugetlb_cgroup_charge_cgroup(int idx, unsigned long nr_pages,
 	}
 	rcu_read_unlock();

-	if (!page_counter_try_charge(&h_cg->hugepage[idx], nr_pages,
-				     &counter)) {
+	if (!page_counter_try_charge(hugetlb_cgroup_get_counter(h_cg, idx,
+								reserved),
+				     nr_pages, &counter)) {
 		ret = -ENOMEM;
 		hugetlb_event(hugetlb_cgroup_from_counter(counter, idx), idx,
 			      HUGETLB_MAX);
+		css_put(&h_cg->css);
+		goto done;
 	}
-	css_put(&h_cg->css);
+	/* Reservations take a reference to the css because they do not get
+	 * reparented.
+	 */
+	if (!reserved)
+		css_put(&h_cg->css);
 done:
 	*ptr = h_cg;
 	return ret;
@@ -265,12 +289,12 @@ int hugetlb_cgroup_charge_cgroup(int idx, unsigned long nr_pages,
 /* 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)
+				  struct page *page, bool reserved)
 {
 	if (hugetlb_cgroup_disabled() || !h_cg)
 		return;

-	set_hugetlb_cgroup(page, h_cg);
+	set_hugetlb_cgroup(page, h_cg, reserved);
 	return;
 }

@@ -278,23 +302,29 @@ void hugetlb_cgroup_commit_charge(int idx, unsigned long nr_pages,
  * Should be called with hugetlb_lock held
  */
 void hugetlb_cgroup_uncharge_page(int idx, unsigned long nr_pages,
-				  struct page *page)
+				  struct page *page, bool reserved)
 {
 	struct hugetlb_cgroup *h_cg;

 	if (hugetlb_cgroup_disabled())
 		return;
 	lockdep_assert_held(&hugetlb_lock);
-	h_cg = hugetlb_cgroup_from_page(page);
+	h_cg = hugetlb_cgroup_from_page(page, reserved);
 	if (unlikely(!h_cg))
 		return;
-	set_hugetlb_cgroup(page, NULL);
-	page_counter_uncharge(&h_cg->hugepage[idx], nr_pages);
+	set_hugetlb_cgroup(page, NULL, reserved);
+
+	page_counter_uncharge(hugetlb_cgroup_get_counter(h_cg, idx, reserved),
+			      nr_pages);
+
+	if (reserved)
+		css_put(&h_cg->css);
+
 	return;
 }

 void hugetlb_cgroup_uncharge_cgroup(int idx, unsigned long nr_pages,
-				    struct hugetlb_cgroup *h_cg)
+				    struct hugetlb_cgroup *h_cg, bool reserved)
 {
 	if (hugetlb_cgroup_disabled() || !h_cg)
 		return;
@@ -302,8 +332,22 @@ void hugetlb_cgroup_uncharge_cgroup(int idx, unsigned long nr_pages,
 	if (huge_page_order(&hstates[idx]) < HUGETLB_CGROUP_MIN_ORDER)
 		return;

-	page_counter_uncharge(&h_cg->hugepage[idx], nr_pages);
-	return;
+	page_counter_uncharge(hugetlb_cgroup_get_counter(h_cg, idx, reserved),
+			      nr_pages);
+
+	if (reserved)
+		css_put(&h_cg->css);
+}
+
+void hugetlb_cgroup_uncharge_counter(struct page_counter *p,
+				     unsigned long nr_pages,
+				     struct cgroup_subsys_state *css)
+{
+	if (hugetlb_cgroup_disabled() || !p || !css)
+		return;
+
+	page_counter_uncharge(p, nr_pages);
+	css_put(css);
 }

 enum {
@@ -678,6 +722,7 @@ void __init hugetlb_cgroup_file_init(void)
 void hugetlb_cgroup_migrate(struct page *oldhpage, struct page *newhpage)
 {
 	struct hugetlb_cgroup *h_cg;
+	struct hugetlb_cgroup *h_cg_reservation;
 	struct hstate *h = page_hstate(oldhpage);

 	if (hugetlb_cgroup_disabled())
@@ -685,11 +730,12 @@ void hugetlb_cgroup_migrate(struct page *oldhpage, struct page *newhpage)

 	VM_BUG_ON_PAGE(!PageHuge(oldhpage), oldhpage);
 	spin_lock(&hugetlb_lock);
-	h_cg = hugetlb_cgroup_from_page(oldhpage);
-	set_hugetlb_cgroup(oldhpage, NULL);
+	h_cg = hugetlb_cgroup_from_page(oldhpage, false);
+	h_cg_reservation = hugetlb_cgroup_from_page(oldhpage, true);
+	set_hugetlb_cgroup(oldhpage, NULL, false);

 	/* move the h_cg details to new cgroup */
-	set_hugetlb_cgroup(newhpage, h_cg);
+	set_hugetlb_cgroup(newhpage, h_cg_reservation, true);
 	list_move(&newhpage->lru, &h->hugepage_activelist);
 	spin_unlock(&hugetlb_lock);
 	return;
--
2.24.1.735.g03f4e72817-goog

^ permalink raw reply related	[flat|nested] 27+ messages in thread

* [PATCH v9 3/8] hugetlb_cgroup: add reservation accounting for private mappings
  2019-12-17 23:16 [PATCH v9 1/8] hugetlb_cgroup: Add hugetlb_cgroup reservation counter Mina Almasry
  2019-12-17 23:16 ` [PATCH v9 2/8] hugetlb_cgroup: add interface for charge/uncharge hugetlb reservations Mina Almasry
@ 2019-12-17 23:16 ` Mina Almasry
  2020-01-14  0:55   ` Mike Kravetz
  2019-12-17 23:16 ` [PATCH v9 4/8] hugetlb: disable region_add file_region coalescing Mina Almasry
                   ` (7 subsequent siblings)
  9 siblings, 1 reply; 27+ messages in thread
From: Mina Almasry @ 2019-12-17 23:16 UTC (permalink / raw)
  To: mike.kravetz
  Cc: shuah, almasrymina, rientjes, shakeelb, gthelen, akpm,
	linux-kernel, linux-mm, linux-kselftest, cgroups, aneesh.kumar,
	mkoutny, Hillf Danton

Normally the pointer to the cgroup to uncharge hangs off the struct
page, and gets queried when it's time to free the page. With
hugetlb_cgroup reservations, this is not possible. Because it's possible
for a page to be reserved by one task and actually faulted in by another
task.

The best place to put the hugetlb_cgroup pointer to uncharge for
reservations is in the resv_map. But, because the resv_map has different
semantics for private and shared mappings, the code patch to
charge/uncharge shared and private mappings is different. This patch
implements charging and uncharging for private mappings.

For private mappings, the counter to uncharge is in
resv_map->reservation_counter. On initializing the resv_map this is set
to NULL. On reservation of a region in private mapping, the tasks
hugetlb_cgroup is charged and the hugetlb_cgroup is placed is
resv_map->reservation_counter.

On hugetlb_vm_op_close, we uncharge resv_map->reservation_counter.

Signed-off-by: Mina Almasry <almasrymina@google.com>
Acked-by: Hillf Danton <hdanton@sina.com>

---

Changes in V9:
- Updated for reparenting of hugetlb reservation accounting.

---
 include/linux/hugetlb.h        |  9 +++++++
 include/linux/hugetlb_cgroup.h | 27 +++++++++++++++++++
 mm/hugetlb.c                   | 47 +++++++++++++++++++++++++++++++++-
 mm/hugetlb_cgroup.c            | 28 --------------------
 4 files changed, 82 insertions(+), 29 deletions(-)

diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
index dea6143aa0685..e6ab499ba2086 100644
--- a/include/linux/hugetlb.h
+++ b/include/linux/hugetlb.h
@@ -46,6 +46,15 @@ struct resv_map {
 	long adds_in_progress;
 	struct list_head region_cache;
 	long region_cache_count;
+#ifdef CONFIG_CGROUP_HUGETLB
+	/*
+	 * On private mappings, the counter to uncharge reservations is stored
+	 * here. If these fields are 0, then the mapping is shared.
+	 */
+	struct page_counter *reservation_counter;
+	unsigned long pages_per_hpage;
+	struct cgroup_subsys_state *css;
+#endif
 };
 extern struct resv_map *resv_map_alloc(void);
 void resv_map_release(struct kref *ref);
diff --git a/include/linux/hugetlb_cgroup.h b/include/linux/hugetlb_cgroup.h
index eab8a70d5bcb5..8c320accefe87 100644
--- a/include/linux/hugetlb_cgroup.h
+++ b/include/linux/hugetlb_cgroup.h
@@ -25,6 +25,33 @@ struct hugetlb_cgroup;
 #define HUGETLB_CGROUP_MIN_ORDER	2

 #ifdef CONFIG_CGROUP_HUGETLB
+enum hugetlb_memory_event {
+	HUGETLB_MAX,
+	HUGETLB_NR_MEMORY_EVENTS,
+};
+
+struct hugetlb_cgroup {
+	struct cgroup_subsys_state css;
+
+	/*
+	 * the counter to account for hugepages from hugetlb.
+	 */
+	struct page_counter hugepage[HUGE_MAX_HSTATE];
+
+	/*
+	 * the counter to account for hugepage reservations from hugetlb.
+	 */
+	struct page_counter reserved_hugepage[HUGE_MAX_HSTATE];
+
+	atomic_long_t events[HUGE_MAX_HSTATE][HUGETLB_NR_MEMORY_EVENTS];
+	atomic_long_t events_local[HUGE_MAX_HSTATE][HUGETLB_NR_MEMORY_EVENTS];
+
+	/* Handle for "hugetlb.events" */
+	struct cgroup_file events_file[HUGE_MAX_HSTATE];
+
+	/* Handle for "hugetlb.events.local" */
+	struct cgroup_file events_local_file[HUGE_MAX_HSTATE];
+};

 static inline struct hugetlb_cgroup *hugetlb_cgroup_from_page(struct page *page,
 							      bool reserved)
diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index e6e8240f1718c..7782977970301 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -665,6 +665,17 @@ struct resv_map *resv_map_alloc(void)
 	INIT_LIST_HEAD(&resv_map->regions);

 	resv_map->adds_in_progress = 0;
+#ifdef CONFIG_CGROUP_HUGETLB
+	/*
+	 * Initialize these to 0. On shared mappings, 0's here indicate these
+	 * fields don't do cgroup accounting. On private mappings, these will be
+	 * re-initialized to the proper values, to indicate that hugetlb cgroup
+	 * reservations are to be un-charged from here.
+	 */
+	resv_map->reservation_counter = NULL;
+	resv_map->pages_per_hpage = 0;
+	resv_map->css = NULL;
+#endif

 	INIT_LIST_HEAD(&resv_map->region_cache);
 	list_add(&rg->link, &resv_map->region_cache);
@@ -3145,7 +3156,20 @@ static void hugetlb_vm_op_close(struct vm_area_struct *vma)

 	reserve = (end - start) - region_count(resv, start, end);

-	kref_put(&resv->refs, resv_map_release);
+#ifdef CONFIG_CGROUP_HUGETLB
+	/*
+	 * Since we check for HPAGE_RESV_OWNER above, this must a private
+	 * mapping, and these values should be none-zero, and should point to
+	 * the hugetlb_cgroup counter to uncharge for this reservation.
+	 */
+	WARN_ON(!resv->reservation_counter);
+	WARN_ON(!resv->pages_per_hpage);
+	WARN_ON(!resv->css);
+
+	hugetlb_cgroup_uncharge_counter(resv->reservation_counter,
+					(end - start) * resv->pages_per_hpage,
+					resv->css);
+#endif

 	if (reserve) {
 		/*
@@ -3155,6 +3179,8 @@ static void hugetlb_vm_op_close(struct vm_area_struct *vma)
 		gbl_reserve = hugepage_subpool_put_pages(spool, reserve);
 		hugetlb_acct_memory(h, -gbl_reserve);
 	}
+
+	kref_put(&resv->refs, resv_map_release);
 }

 static int hugetlb_vm_op_split(struct vm_area_struct *vma, unsigned long addr)
@@ -4501,6 +4527,7 @@ int hugetlb_reserve_pages(struct inode *inode,
 	struct hstate *h = hstate_inode(inode);
 	struct hugepage_subpool *spool = subpool_inode(inode);
 	struct resv_map *resv_map;
+	struct hugetlb_cgroup *h_cg;
 	long gbl_reserve;

 	/* This should never happen */
@@ -4534,12 +4561,30 @@ int hugetlb_reserve_pages(struct inode *inode,
 		chg = region_chg(resv_map, from, to);

 	} else {
+		/* Private mapping. */
 		resv_map = resv_map_alloc();
 		if (!resv_map)
 			return -ENOMEM;

 		chg = to - from;

+		if (hugetlb_cgroup_charge_cgroup(hstate_index(h),
+						 chg * pages_per_huge_page(h),
+						 &h_cg, true)) {
+			kref_put(&resv_map->refs, resv_map_release);
+			return -ENOMEM;
+		}
+
+#ifdef CONFIG_CGROUP_HUGETLB
+		/*
+		 * Since this branch handles private mappings, we attach the
+		 * counter to uncharge for this reservation off resv_map.
+		 */
+		resv_map->reservation_counter =
+			&h_cg->reserved_hugepage[hstate_index(h)];
+		resv_map->pages_per_hpage = pages_per_huge_page(h);
+#endif
+
 		set_vma_resv_map(vma, resv_map);
 		set_vma_resv_flags(vma, HPAGE_RESV_OWNER);
 	}
diff --git a/mm/hugetlb_cgroup.c b/mm/hugetlb_cgroup.c
index b03270b0d5833..e066122ab9f49 100644
--- a/mm/hugetlb_cgroup.c
+++ b/mm/hugetlb_cgroup.c
@@ -23,34 +23,6 @@
 #include <linux/hugetlb.h>
 #include <linux/hugetlb_cgroup.h>

-enum hugetlb_memory_event {
-	HUGETLB_MAX,
-	HUGETLB_NR_MEMORY_EVENTS,
-};
-
-struct hugetlb_cgroup {
-	struct cgroup_subsys_state css;
-
-	/*
-	 * the counter to account for hugepages from hugetlb.
-	 */
-	struct page_counter hugepage[HUGE_MAX_HSTATE];
-
-	/*
-	 * the counter to account for hugepage reservations from hugetlb.
-	 */
-	struct page_counter reserved_hugepage[HUGE_MAX_HSTATE];
-
-	atomic_long_t events[HUGE_MAX_HSTATE][HUGETLB_NR_MEMORY_EVENTS];
-	atomic_long_t events_local[HUGE_MAX_HSTATE][HUGETLB_NR_MEMORY_EVENTS];
-
-	/* Handle for "hugetlb.events" */
-	struct cgroup_file events_file[HUGE_MAX_HSTATE];
-
-	/* Handle for "hugetlb.events.local" */
-	struct cgroup_file events_local_file[HUGE_MAX_HSTATE];
-};
-
 #define MEMFILE_PRIVATE(x, val)	(((x) << 16) | (val))
 #define MEMFILE_IDX(val)	(((val) >> 16) & 0xffff)
 #define MEMFILE_ATTR(val)	((val) & 0xffff)
--
2.24.1.735.g03f4e72817-goog

^ permalink raw reply related	[flat|nested] 27+ messages in thread

* [PATCH v9 4/8] hugetlb: disable region_add file_region coalescing
  2019-12-17 23:16 [PATCH v9 1/8] hugetlb_cgroup: Add hugetlb_cgroup reservation counter Mina Almasry
  2019-12-17 23:16 ` [PATCH v9 2/8] hugetlb_cgroup: add interface for charge/uncharge hugetlb reservations Mina Almasry
  2019-12-17 23:16 ` [PATCH v9 3/8] hugetlb_cgroup: add reservation accounting for private mappings Mina Almasry
@ 2019-12-17 23:16 ` Mina Almasry
  2020-01-21 17:38   ` Mike Kravetz
  2019-12-17 23:16 ` [PATCH v9 5/8] hugetlb_cgroup: add accounting for shared mappings Mina Almasry
                   ` (6 subsequent siblings)
  9 siblings, 1 reply; 27+ messages in thread
From: Mina Almasry @ 2019-12-17 23:16 UTC (permalink / raw)
  To: mike.kravetz
  Cc: shuah, almasrymina, rientjes, shakeelb, gthelen, akpm,
	linux-kernel, linux-mm, linux-kselftest, cgroups, aneesh.kumar,
	mkoutny

A follow up patch in this series adds hugetlb cgroup uncharge info the
file_region entries in resv->regions. The cgroup uncharge info may
differ for different regions, so they can no longer be coalesced at
region_add time. So, disable region coalescing in region_add in this
patch.

Behavior change:

Say a resv_map exists like this [0->1], [2->3], and [5->6].

Then a region_chg/add call comes in region_chg/add(f=0, t=5).

Old code would generate resv->regions: [0->5], [5->6].
New code would generate resv->regions: [0->1], [1->2], [2->3], [3->5],
[5->6].

Special care needs to be taken to handle the resv->adds_in_progress
variable correctly. In the past, only 1 region would be added for every
region_chg and region_add call. But now, each call may add multiple
regions, so we can no longer increment adds_in_progress by 1 in region_chg,
or decrement adds_in_progress by 1 after region_add or region_abort. Instead,
region_chg calls add_reservation_in_range() to count the number of regions
needed and allocates those, and that info is passed to region_add and
region_abort to decrement adds_in_progress correctly.

We've also modified the assumption that region_add after region_chg
never fails. region_chg now pre-allocates at least 1 region for
region_add. If region_add needs more regions than region_chg has
allocated for it, then it may fail.

Signed-off-by: Mina Almasry <almasrymina@google.com>
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>

---

Changes in v9:
- Added clarifications in the comments and addressed minor issues from
code review.
Changes in v7:
- region_chg no longer allocates (t-f) / 2 file_region entries.
Changes in v6:
- Fix bug in number of region_caches allocated by region_chg

---
 mm/hugetlb.c | 327 +++++++++++++++++++++++++++++++++++----------------
 1 file changed, 223 insertions(+), 104 deletions(-)

diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index 7782977970301..3f70742e3aee4 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -244,110 +244,179 @@ struct file_region {
 	long to;
 };

+/* Helper that removes a struct file_region from the resv_map cache and returns
+ * it for use.
+ */
+static struct file_region *
+get_file_region_entry_from_cache(struct resv_map *resv, long from, long to)
+{
+	struct file_region *nrg = NULL;
+
+	VM_BUG_ON(resv->region_cache_count <= 0);
+
+	resv->region_cache_count--;
+	nrg = list_first_entry(&resv->region_cache, struct file_region, link);
+	VM_BUG_ON(!nrg);
+	list_del(&nrg->link);
+
+	nrg->from = from;
+	nrg->to = to;
+
+	return nrg;
+}
+
 /* Must be called with resv->lock held. Calling this with count_only == true
  * will count the number of pages to be added but will not modify the linked
- * list.
+ * list. If regions_needed != NULL and count_only == true, then regions_needed
+ * will indicate the number of file_regions needed in the cache to carry out to
+ * add the regions for this range.
  */
 static long add_reservation_in_range(struct resv_map *resv, long f, long t,
-				     bool count_only)
+				     long *regions_needed, bool count_only)
 {
-	long chg = 0;
+	long add = 0;
 	struct list_head *head = &resv->regions;
+	long last_accounted_offset = f;
 	struct file_region *rg = NULL, *trg = NULL, *nrg = NULL;

-	/* Locate the region we are before or in. */
-	list_for_each_entry(rg, head, link)
-		if (f <= rg->to)
-			break;
-
-	/* Round our left edge to the current segment if it encloses us. */
-	if (f > rg->from)
-		f = rg->from;
+	if (regions_needed)
+		*regions_needed = 0;

-	chg = t - f;
+	/* In this loop, we essentially handle an entry for the range
+	 * [last_accounted_offset, rg->from), at every iteration, with some
+	 * bounds checking.
+	 */
+	list_for_each_entry_safe(rg, trg, head, link) {
+		/* Skip irrelevant regions that start before our range. */
+		if (rg->from < f) {
+			/* If this region ends after the last accounted offset,
+			 * then we need to update last_accounted_offset.
+			 */
+			if (rg->to > last_accounted_offset)
+				last_accounted_offset = rg->to;
+			continue;
+		}

-	/* Check for and consume any regions we now overlap with. */
-	nrg = rg;
-	list_for_each_entry_safe(rg, trg, rg->link.prev, link) {
-		if (&rg->link == head)
-			break;
+		/* When we find a region that starts beyond our range, we've
+		 * finished.
+		 */
 		if (rg->from > t)
 			break;

-		/* We overlap with this area, if it extends further than
-		 * us then we must extend ourselves.  Account for its
-		 * existing reservation.
+		/* Add an entry for last_accounted_offset -> rg->from, and
+		 * update last_accounted_offset.
 		 */
-		if (rg->to > t) {
-			chg += rg->to - t;
-			t = rg->to;
+		if (rg->from > last_accounted_offset) {
+			add += rg->from - last_accounted_offset;
+			if (!count_only) {
+				nrg = get_file_region_entry_from_cache(
+					resv, last_accounted_offset, rg->from);
+				list_add(&nrg->link, rg->link.prev);
+			} else if (regions_needed)
+				*regions_needed += 1;
 		}
-		chg -= rg->to - rg->from;

-		if (!count_only && rg != nrg) {
-			list_del(&rg->link);
-			kfree(rg);
-		}
+		last_accounted_offset = rg->to;
 	}

-	if (!count_only) {
-		nrg->from = f;
-		nrg->to = t;
+	/* Handle the case where our range extends beyond
+	 * last_accounted_offset.
+	 */
+	if (last_accounted_offset < t) {
+		add += t - last_accounted_offset;
+		if (!count_only) {
+			nrg = get_file_region_entry_from_cache(
+				resv, last_accounted_offset, t);
+			list_add(&nrg->link, rg->link.prev);
+		} else if (regions_needed)
+			*regions_needed += 1;
 	}

-	return chg;
+	return add;
 }

 /*
  * Add the huge page range represented by [f, t) to the reserve
- * map.  Existing regions will be expanded to accommodate the specified
- * range, or a region will be taken from the cache.  Sufficient regions
- * must exist in the cache due to the previous call to region_chg with
- * the same range.
+ * map.  Regions will be taken from the cache to fill in this range.
+ * Sufficient regions should exist in the cache due to the previous
+ * call to region_chg with the same range, but in some cases the cache will not
+ * have sufficient entries due to races with other code doing region_add or
+ * region_del.  The extra needed entries will be allocated.
  *
- * Return the number of new huge pages added to the map.  This
- * number is greater than or equal to zero.
+ * regions_needed is the out value provided by a previous call to region_chg.
+ *
+ * Return the number of new huge pages added to the map.  This number is greater
+ * than or equal to zero.  If file_region entries needed to be allocated for
+ * this operation and we were not able to allocate, it ruturns -ENOMEM.
+ * region_add of regions of length 1 never allocate file_regions and cannot
+ * fail; region_chg will always allocate at least 1 entry and a region_add for
+ * 1 page will only require at most 1 entry.
  */
-static long region_add(struct resv_map *resv, long f, long t)
+static long region_add(struct resv_map *resv, long f, long t,
+		       long in_regions_needed)
 {
-	struct list_head *head = &resv->regions;
-	struct file_region *rg, *nrg;
-	long add = 0;
+	long add = 0, actual_regions_needed = 0, i = 0;
+	struct file_region *trg = NULL, *rg = NULL;
+	struct list_head allocated_regions;
+
+	INIT_LIST_HEAD(&allocated_regions);

 	spin_lock(&resv->lock);
-	/* Locate the region we are either in or before. */
-	list_for_each_entry(rg, head, link)
-		if (f <= rg->to)
-			break;
+retry:
+
+	/* Count how many regions are actually needed to execute this add. */
+	add_reservation_in_range(resv, f, t, &actual_regions_needed, true);

 	/*
-	 * If no region exists which can be expanded to include the
-	 * specified range, pull a region descriptor from the cache
-	 * and use it for this range.
+	 * Check for sufficient descriptors in the cache to accommodate
+	 * this add operation. Note that actual_regions_needed may be greater
+	 * than in_regions_needed. In this case, we need to make sure that we
+	 * allocate extra entries, such that we have enough for all the
+	 * existing adds_in_progress, plus the excess needed for this
+	 * operation.
 	 */
-	if (&rg->link == head || t < rg->from) {
-		VM_BUG_ON(resv->region_cache_count <= 0);
+	if (resv->region_cache_count <
+	    resv->adds_in_progress +
+		    (actual_regions_needed - in_regions_needed)) {
+		/* region_add operation of range 1 should never need to
+		 * allocate file_region entries.
+		 */
+		VM_BUG_ON(t - f <= 1);

-		resv->region_cache_count--;
-		nrg = list_first_entry(&resv->region_cache, struct file_region,
-					link);
-		list_del(&nrg->link);
+		/* Must drop lock to allocate a new descriptor. */
+		spin_unlock(&resv->lock);
+		for (i = 0; i < (actual_regions_needed - in_regions_needed);
+		     i++) {
+			trg = kmalloc(sizeof(*trg), GFP_KERNEL);
+			if (!trg)
+				goto out_of_memory;
+			list_add(&trg->link, &allocated_regions);
+		}
+		spin_lock(&resv->lock);

-		nrg->from = f;
-		nrg->to = t;
-		list_add(&nrg->link, rg->link.prev);
+		list_for_each_entry_safe(rg, trg, &allocated_regions, link) {
+			list_del(&rg->link);
+			list_add(&rg->link, &resv->region_cache);
+			resv->region_cache_count++;
+		}

-		add += t - f;
-		goto out_locked;
+		goto retry;
 	}

-	add = add_reservation_in_range(resv, f, t, false);
+	add = add_reservation_in_range(resv, f, t, NULL, false);
+
+	resv->adds_in_progress -= in_regions_needed;

-out_locked:
-	resv->adds_in_progress--;
 	spin_unlock(&resv->lock);
 	VM_BUG_ON(add < 0);
 	return add;
+
+out_of_memory:
+	list_for_each_entry_safe(rg, trg, &allocated_regions, link) {
+		list_del(&rg->link);
+		kfree(rg);
+	}
+	return -ENOMEM;
 }

 /*
@@ -357,49 +426,79 @@ static long region_add(struct resv_map *resv, long f, long t)
  * call to region_add that will actually modify the reserve
  * map to add the specified range [f, t).  region_chg does
  * not change the number of huge pages represented by the
- * map.  A new file_region structure is added to the cache
- * as a placeholder, so that the subsequent region_add
- * call will have all the regions it needs and will not fail.
+ * map.  A number of new file_region structures is added to the cache as a
+ * placeholder, for the subsequent region_add call to use. At least 1
+ * file_region structure is added.
+ *
+ * out_regions_needed is the number of regions added to the
+ * resv->adds_in_progress.  This value needs to be provided to a follow up call
+ * to region_add or region_abort for proper accounting.
  *
  * Returns the number of huge pages that need to be added to the existing
  * reservation map for the range [f, t).  This number is greater or equal to
  * zero.  -ENOMEM is returned if a new file_region structure or cache entry
  * is needed and can not be allocated.
  */
-static long region_chg(struct resv_map *resv, long f, long t)
+static long region_chg(struct resv_map *resv, long f, long t,
+		       long *out_regions_needed)
 {
-	long chg = 0;
+	struct file_region *trg = NULL, *rg = NULL;
+	long chg = 0, i = 0, to_allocate = 0;
+	struct list_head allocated_regions;
+
+	INIT_LIST_HEAD(&allocated_regions);

 	spin_lock(&resv->lock);
-retry_locked:
-	resv->adds_in_progress++;
+
+	/* Count how many hugepages in this range are NOT respresented. */
+	chg = add_reservation_in_range(resv, f, t, out_regions_needed, true);
+
+	if (*out_regions_needed == 0)
+		*out_regions_needed = 1;
+
+	resv->adds_in_progress += *out_regions_needed;

 	/*
 	 * Check for sufficient descriptors in the cache to accommodate
 	 * the number of in progress add operations.
 	 */
-	if (resv->adds_in_progress > resv->region_cache_count) {
-		struct file_region *trg;
-
-		VM_BUG_ON(resv->adds_in_progress - resv->region_cache_count > 1);
-		/* Must drop lock to allocate a new descriptor. */
-		resv->adds_in_progress--;
+	while (resv->region_cache_count < resv->adds_in_progress) {
+		to_allocate = resv->adds_in_progress - resv->region_cache_count;
+
+		/* Must drop lock to allocate a new descriptor. Note that even
+		 * though we drop the lock here, we do not make another call to
+		 * add_reservation_in_range after re-acquiring the lock.
+		 * Essentially this branch makes sure that we have enough
+		 * descriptors in the cache as suggested by the first call to
+		 * add_reservation_in_range. If more regions turn out to be
+		 * required, region_add will deal with it.
+		 */
 		spin_unlock(&resv->lock);
-
-		trg = kmalloc(sizeof(*trg), GFP_KERNEL);
-		if (!trg)
-			return -ENOMEM;
+		for (i = 0; i < to_allocate; i++) {
+			trg = kmalloc(sizeof(*trg), GFP_KERNEL);
+			if (!trg)
+				goto out_of_memory;
+			list_add(&trg->link, &allocated_regions);
+		}

 		spin_lock(&resv->lock);
-		list_add(&trg->link, &resv->region_cache);
-		resv->region_cache_count++;
-		goto retry_locked;
-	}

-	chg = add_reservation_in_range(resv, f, t, true);
+		list_for_each_entry_safe(rg, trg, &allocated_regions, link) {
+			list_del(&rg->link);
+			list_add(&rg->link, &resv->region_cache);
+			resv->region_cache_count++;
+		}
+	}

 	spin_unlock(&resv->lock);
 	return chg;
+
+out_of_memory:
+	list_for_each_entry_safe(rg, trg, &allocated_regions, link) {
+		list_del(&rg->link);
+		kfree(rg);
+	}
+	return -ENOMEM;
 }

 /*
@@ -407,17 +506,20 @@ static long region_chg(struct resv_map *resv, long f, long t)
  * of the resv_map keeps track of the operations in progress between
  * calls to region_chg and region_add.  Operations are sometimes
  * aborted after the call to region_chg.  In such cases, region_abort
- * is called to decrement the adds_in_progress counter.
+ * is called to decrement the adds_in_progress counter. regions_needed
+ * is the value returned by the region_chg call, it is used to decrement
+ * the adds_in_progress counter.
  *
  * NOTE: The range arguments [f, t) are not needed or used in this
  * routine.  They are kept to make reading the calling code easier as
  * arguments will match the associated region_chg call.
  */
-static void region_abort(struct resv_map *resv, long f, long t)
+static void region_abort(struct resv_map *resv, long f, long t,
+			 long regions_needed)
 {
 	spin_lock(&resv->lock);
 	VM_BUG_ON(!resv->region_cache_count);
-	resv->adds_in_progress--;
+	resv->adds_in_progress -= regions_needed;
 	spin_unlock(&resv->lock);
 }

@@ -1834,6 +1936,7 @@ static long __vma_reservation_common(struct hstate *h,
 	struct resv_map *resv;
 	pgoff_t idx;
 	long ret;
+	long dummy_out_regions_needed;

 	resv = vma_resv_map(vma);
 	if (!resv)
@@ -1842,20 +1945,29 @@ static long __vma_reservation_common(struct hstate *h,
 	idx = vma_hugecache_offset(h, vma, addr);
 	switch (mode) {
 	case VMA_NEEDS_RESV:
-		ret = region_chg(resv, idx, idx + 1);
+		ret = region_chg(resv, idx, idx + 1, &dummy_out_regions_needed);
+		/* We assume that vma_reservation_* routines always operate on
+		 * 1 page, and that adding to resv map a 1 page entry can only
+		 * ever require 1 region.
+		 */
+		VM_BUG_ON(dummy_out_regions_needed != 1);
 		break;
 	case VMA_COMMIT_RESV:
-		ret = region_add(resv, idx, idx + 1);
+		ret = region_add(resv, idx, idx + 1, 1);
+		/* region_add calls of range 1 should never fail. */
+		VM_BUG_ON(ret < 0);
 		break;
 	case VMA_END_RESV:
-		region_abort(resv, idx, idx + 1);
+		region_abort(resv, idx, idx + 1, 1);
 		ret = 0;
 		break;
 	case VMA_ADD_RESV:
-		if (vma->vm_flags & VM_MAYSHARE)
-			ret = region_add(resv, idx, idx + 1);
-		else {
-			region_abort(resv, idx, idx + 1);
+		if (vma->vm_flags & VM_MAYSHARE) {
+			ret = region_add(resv, idx, idx + 1, 1);
+			/* region_add calls of range 1 should never fail. */
+			VM_BUG_ON(ret < 0);
+		} else {
+			region_abort(resv, idx, idx + 1, 1);
 			ret = region_del(resv, idx, idx + 1);
 		}
 		break;
@@ -4523,12 +4635,12 @@ int hugetlb_reserve_pages(struct inode *inode,
 					struct vm_area_struct *vma,
 					vm_flags_t vm_flags)
 {
-	long ret, chg;
+	long ret, chg, add = -1;
 	struct hstate *h = hstate_inode(inode);
 	struct hugepage_subpool *spool = subpool_inode(inode);
 	struct resv_map *resv_map;
 	struct hugetlb_cgroup *h_cg;
-	long gbl_reserve;
+	long gbl_reserve, regions_needed = 0;

 	/* This should never happen */
 	if (from > to) {
@@ -4558,7 +4670,7 @@ int hugetlb_reserve_pages(struct inode *inode,
 		 */
 		resv_map = inode_resv_map(inode);

-		chg = region_chg(resv_map, from, to);
+		chg = region_chg(resv_map, from, to, &regions_needed);

 	} else {
 		/* Private mapping. */
@@ -4628,9 +4740,14 @@ int hugetlb_reserve_pages(struct inode *inode,
 	 * else has to be done for private mappings here
 	 */
 	if (!vma || vma->vm_flags & VM_MAYSHARE) {
-		long add = region_add(resv_map, from, to);
-
-		if (unlikely(chg > add)) {
+		add = region_add(resv_map, from, to, regions_needed);
+
+		if (unlikely(add < 0)) {
+			hugetlb_acct_memory(h, -gbl_reserve);
+			/* put back original number of pages, chg */
+			(void)hugepage_subpool_put_pages(spool, chg);
+			goto out_err;
+		} else if (unlikely(chg > add)) {
 			/*
 			 * pages in this range were added to the reserve
 			 * map between region_chg and region_add.  This
@@ -4648,9 +4765,11 @@ int hugetlb_reserve_pages(struct inode *inode,
 	return 0;
 out_err:
 	if (!vma || vma->vm_flags & VM_MAYSHARE)
-		/* Don't call region_abort if region_chg failed */
-		if (chg >= 0)
-			region_abort(resv_map, from, to);
+		/* Only call region_abort if the region_chg succeeded but the
+		 * region_add failed or didn't run.
+		 */
+		if (chg >= 0 && add < 0)
+			region_abort(resv_map, from, to, regions_needed);
 	if (vma && is_vma_resv_set(vma, HPAGE_RESV_OWNER))
 		kref_put(&resv_map->refs, resv_map_release);
 	return ret;
--
2.24.1.735.g03f4e72817-goog

^ permalink raw reply related	[flat|nested] 27+ messages in thread

* [PATCH v9 5/8] hugetlb_cgroup: add accounting for shared mappings
  2019-12-17 23:16 [PATCH v9 1/8] hugetlb_cgroup: Add hugetlb_cgroup reservation counter Mina Almasry
                   ` (2 preceding siblings ...)
  2019-12-17 23:16 ` [PATCH v9 4/8] hugetlb: disable region_add file_region coalescing Mina Almasry
@ 2019-12-17 23:16 ` Mina Almasry
  2019-12-17 23:16 ` [PATCH v9 6/8] hugetlb_cgroup: support noreserve mappings Mina Almasry
                   ` (5 subsequent siblings)
  9 siblings, 0 replies; 27+ messages in thread
From: Mina Almasry @ 2019-12-17 23:16 UTC (permalink / raw)
  To: mike.kravetz
  Cc: shuah, almasrymina, rientjes, shakeelb, gthelen, akpm,
	linux-kernel, linux-mm, linux-kselftest, cgroups, aneesh.kumar,
	mkoutny

For shared mappings, the pointer to the hugetlb_cgroup to uncharge lives
in the resv_map entries, in file_region->reservation_counter.

After a call to region_chg, we charge the approprate hugetlb_cgroup, and if
successful, we pass on the hugetlb_cgroup info to a follow up region_add call.
When a file_region entry is added to the resv_map via region_add, we put the
pointer to that cgroup in file_region->reservation_counter. If charging doesn't
succeed, we report the error to the caller, so that the kernel fails the
reservation.

On region_del, which is when the hugetlb memory is unreserved, we also uncharge
the file_region->reservation_counter.

Signed-off-by: Mina Almasry <almasrymina@google.com>

---

Changes in V9:
- Updated for hugetlb reservation repareting.

---
 mm/hugetlb.c | 161 +++++++++++++++++++++++++++++++++++++++++----------
 1 file changed, 129 insertions(+), 32 deletions(-)

diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index 3f70742e3aee4..39eb2fa53a420 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -242,6 +242,16 @@ struct file_region {
 	struct list_head link;
 	long from;
 	long to;
+#ifdef CONFIG_CGROUP_HUGETLB
+	/*
+	 * On shared mappings, each reserved region appears as a struct
+	 * file_region in resv_map. These fields hold the info needed to
+	 * uncharge each reservation.
+	 */
+	struct page_counter *reservation_counter;
+	unsigned long pages_per_hpage;
+	struct cgroup_subsys_state *css;
+#endif
 };

 /* Helper that removes a struct file_region from the resv_map cache and returns
@@ -265,6 +275,25 @@ get_file_region_entry_from_cache(struct resv_map *resv, long from, long to)
 	return nrg;
 }

+/* Helper that records hugetlb_cgroup uncharge info. */
+static void record_hugetlb_cgroup_uncharge_info(struct hugetlb_cgroup *h_cg,
+						struct file_region *nrg,
+						struct hstate *h)
+{
+#ifdef CONFIG_CGROUP_HUGETLB
+	if (h_cg) {
+		nrg->reservation_counter =
+			&h_cg->reserved_hugepage[hstate_index(h)];
+		nrg->pages_per_hpage = pages_per_huge_page(h);
+		nrg->css = &h_cg->css;
+	} else {
+		nrg->reservation_counter = NULL;
+		nrg->pages_per_hpage = 0;
+		nrg->css = NULL;
+	}
+#endif
+}
+
 /* Must be called with resv->lock held. Calling this with count_only == true
  * will count the number of pages to be added but will not modify the linked
  * list. If regions_needed != NULL and count_only == true, then regions_needed
@@ -272,7 +301,9 @@ get_file_region_entry_from_cache(struct resv_map *resv, long from, long to)
  * add the regions for this range.
  */
 static long add_reservation_in_range(struct resv_map *resv, long f, long t,
-				     long *regions_needed, bool count_only)
+				     struct hugetlb_cgroup *h_cg,
+				     struct hstate *h, long *regions_needed,
+				     bool count_only)
 {
 	long add = 0;
 	struct list_head *head = &resv->regions;
@@ -311,6 +342,8 @@ static long add_reservation_in_range(struct resv_map *resv, long f, long t,
 			if (!count_only) {
 				nrg = get_file_region_entry_from_cache(
 					resv, last_accounted_offset, rg->from);
+				record_hugetlb_cgroup_uncharge_info(h_cg, nrg,
+								    h);
 				list_add(&nrg->link, rg->link.prev);
 			} else if (regions_needed)
 				*regions_needed += 1;
@@ -327,11 +360,13 @@ static long add_reservation_in_range(struct resv_map *resv, long f, long t,
 		if (!count_only) {
 			nrg = get_file_region_entry_from_cache(
 				resv, last_accounted_offset, t);
+			record_hugetlb_cgroup_uncharge_info(h_cg, nrg, h);
 			list_add(&nrg->link, rg->link.prev);
 		} else if (regions_needed)
 			*regions_needed += 1;
 	}

+	VM_BUG_ON(add < 0);
 	return add;
 }

@@ -352,7 +387,8 @@ static long add_reservation_in_range(struct resv_map *resv, long f, long t,
  * fail; region_chg will always allocate at least 1 entry and a region_add for
  * 1 page will only require at most 1 entry.
  */
-static long region_add(struct resv_map *resv, long f, long t,
+static long region_add(struct hstate *h, struct hugetlb_cgroup *h_cg,
+		       struct resv_map *resv, long f, long t,
 		       long in_regions_needed)
 {
 	long add = 0, actual_regions_needed = 0, i = 0;
@@ -365,7 +401,8 @@ static long region_add(struct resv_map *resv, long f, long t,
 retry:

 	/* Count how many regions are actually needed to execute this add. */
-	add_reservation_in_range(resv, f, t, &actual_regions_needed, true);
+	add_reservation_in_range(resv, f, t, NULL, NULL, &actual_regions_needed,
+				 true);

 	/*
 	 * Check for sufficient descriptors in the cache to accommodate
@@ -403,7 +440,7 @@ static long region_add(struct resv_map *resv, long f, long t,
 		goto retry;
 	}

-	add = add_reservation_in_range(resv, f, t, NULL, false);
+	add = add_reservation_in_range(resv, f, t, h_cg, h, NULL, false);

 	resv->adds_in_progress -= in_regions_needed;

@@ -451,7 +488,8 @@ static long region_chg(struct resv_map *resv, long f, long t,
 	spin_lock(&resv->lock);

 	/* Count how many hugepages in this range are NOT respresented. */
-	chg = add_reservation_in_range(resv, f, t, out_regions_needed, true);
+	chg = add_reservation_in_range(resv, f, t, NULL, NULL,
+				       out_regions_needed, true);

 	if (*out_regions_needed == 0)
 		*out_regions_needed = 1;
@@ -523,6 +561,25 @@ static void region_abort(struct resv_map *resv, long f, long t,
 	spin_unlock(&resv->lock);
 }

+static void uncharge_cgroup_if_shared_mapping(struct resv_map *resv,
+					      struct file_region *rg,
+					      unsigned long nr_pages)
+{
+#ifdef CONFIG_CGROUP_HUGETLB
+	/*
+	 * If resv->reservation_counter is NULL, then this is shared
+	 * reservation, and the reserved memory is tracked in the file_struct
+	 * entries inside of resv_map. So we need to uncharge the memory here.
+	 */
+	if (rg->reservation_counter && rg->pages_per_hpage && nr_pages > 0 &&
+	    !resv->reservation_counter) {
+		hugetlb_cgroup_uncharge_counter(rg->reservation_counter,
+						nr_pages * rg->pages_per_hpage,
+						rg->css);
+	}
+#endif
+}
+
 /*
  * Delete the specified range [f, t) from the reserve map.  If the
  * t parameter is LONG_MAX, this indicates that ALL regions after f
@@ -587,11 +644,20 @@ static long region_del(struct resv_map *resv, long f, long t)
 			/* New entry for end of split region */
 			nrg->from = t;
 			nrg->to = rg->to;
+
+			nrg->reservation_counter = rg->reservation_counter;
+			nrg->pages_per_hpage = rg->pages_per_hpage;
+			nrg->css = rg->css;
+			css_get(rg->css);
+
 			INIT_LIST_HEAD(&nrg->link);

 			/* Original entry is trimmed */
 			rg->to = f;

+			uncharge_cgroup_if_shared_mapping(resv, rg,
+							  nrg->to - nrg->from);
+
 			list_add(&nrg->link, &rg->link);
 			nrg = NULL;
 			break;
@@ -599,6 +665,8 @@ static long region_del(struct resv_map *resv, long f, long t)

 		if (f <= rg->from && t >= rg->to) { /* Remove entire region */
 			del += rg->to - rg->from;
+			uncharge_cgroup_if_shared_mapping(resv, rg,
+							  rg->to - rg->from);
 			list_del(&rg->link);
 			kfree(rg);
 			continue;
@@ -607,14 +675,20 @@ static long region_del(struct resv_map *resv, long f, long t)
 		if (f <= rg->from) {	/* Trim beginning of region */
 			del += t - rg->from;
 			rg->from = t;
+
+			uncharge_cgroup_if_shared_mapping(resv, rg,
+							  t - rg->from);
 		} else {		/* Trim end of region */
 			del += rg->to - f;
 			rg->to = f;
+
+			uncharge_cgroup_if_shared_mapping(resv, rg, rg->to - f);
 		}
 	}

 	spin_unlock(&resv->lock);
 	kfree(nrg);
+
 	return del;
 }

@@ -1953,7 +2027,7 @@ static long __vma_reservation_common(struct hstate *h,
 		VM_BUG_ON(dummy_out_regions_needed != 1);
 		break;
 	case VMA_COMMIT_RESV:
-		ret = region_add(resv, idx, idx + 1, 1);
+		ret = region_add(NULL, NULL, resv, idx, idx + 1, 1);
 		/* region_add calls of range 1 should never fail. */
 		VM_BUG_ON(ret < 0);
 		break;
@@ -1963,7 +2037,7 @@ static long __vma_reservation_common(struct hstate *h,
 		break;
 	case VMA_ADD_RESV:
 		if (vma->vm_flags & VM_MAYSHARE) {
-			ret = region_add(resv, idx, idx + 1, 1);
+			ret = region_add(NULL, NULL, resv, idx, idx + 1, 1);
 			/* region_add calls of range 1 should never fail. */
 			VM_BUG_ON(ret < 0);
 		} else {
@@ -4639,7 +4713,7 @@ int hugetlb_reserve_pages(struct inode *inode,
 	struct hstate *h = hstate_inode(inode);
 	struct hugepage_subpool *spool = subpool_inode(inode);
 	struct resv_map *resv_map;
-	struct hugetlb_cgroup *h_cg;
+	struct hugetlb_cgroup *h_cg = NULL;
 	long gbl_reserve, regions_needed = 0;

 	/* This should never happen */
@@ -4680,23 +4754,6 @@ int hugetlb_reserve_pages(struct inode *inode,

 		chg = to - from;

-		if (hugetlb_cgroup_charge_cgroup(hstate_index(h),
-						 chg * pages_per_huge_page(h),
-						 &h_cg, true)) {
-			kref_put(&resv_map->refs, resv_map_release);
-			return -ENOMEM;
-		}
-
-#ifdef CONFIG_CGROUP_HUGETLB
-		/*
-		 * Since this branch handles private mappings, we attach the
-		 * counter to uncharge for this reservation off resv_map.
-		 */
-		resv_map->reservation_counter =
-			&h_cg->reserved_hugepage[hstate_index(h)];
-		resv_map->pages_per_hpage = pages_per_huge_page(h);
-#endif
-
 		set_vma_resv_map(vma, resv_map);
 		set_vma_resv_flags(vma, HPAGE_RESV_OWNER);
 	}
@@ -4706,6 +4763,26 @@ int hugetlb_reserve_pages(struct inode *inode,
 		goto out_err;
 	}

+	ret = hugetlb_cgroup_charge_cgroup(
+		hstate_index(h), chg * pages_per_huge_page(h), &h_cg, true);
+
+	if (ret < 0) {
+		ret = -ENOMEM;
+		goto out_err;
+	}
+
+#ifdef CONFIG_CGROUP_HUGETLB
+	if (vma && !(vma->vm_flags & VM_MAYSHARE)) {
+		/* For private mappings, the hugetlb_cgroup uncharge info hangs
+		 * of the resv_map.
+		 */
+		resv_map->reservation_counter =
+			&h_cg->reserved_hugepage[hstate_index(h)];
+		resv_map->pages_per_hpage = pages_per_huge_page(h);
+		resv_map->css = &h_cg->css;
+	}
+#endif
+
 	/*
 	 * There must be enough pages in the subpool for the mapping. If
 	 * the subpool has a minimum size, there may be some global
@@ -4714,7 +4791,7 @@ int hugetlb_reserve_pages(struct inode *inode,
 	gbl_reserve = hugepage_subpool_get_pages(spool, chg);
 	if (gbl_reserve < 0) {
 		ret = -ENOSPC;
-		goto out_err;
+		goto out_uncharge_cgroup;
 	}

 	/*
@@ -4723,9 +4800,7 @@ int hugetlb_reserve_pages(struct inode *inode,
 	 */
 	ret = hugetlb_acct_memory(h, gbl_reserve);
 	if (ret < 0) {
-		/* put back original number of pages, chg */
-		(void)hugepage_subpool_put_pages(spool, chg);
-		goto out_err;
+		goto out_put_pages;
 	}

 	/*
@@ -4740,7 +4815,7 @@ int hugetlb_reserve_pages(struct inode *inode,
 	 * else has to be done for private mappings here
 	 */
 	if (!vma || vma->vm_flags & VM_MAYSHARE) {
-		add = region_add(resv_map, from, to, regions_needed);
+		add = region_add(h, h_cg, resv_map, from, to, regions_needed);

 		if (unlikely(add < 0)) {
 			hugetlb_acct_memory(h, -gbl_reserve);
@@ -4757,12 +4832,34 @@ int hugetlb_reserve_pages(struct inode *inode,
 			 */
 			long rsv_adjust;

-			rsv_adjust = hugepage_subpool_put_pages(spool,
-								chg - add);
+			hugetlb_cgroup_uncharge_cgroup(
+				hstate_index(h),
+				(chg - add) * pages_per_huge_page(h), h_cg,
+				true);
+
+			rsv_adjust =
+				hugepage_subpool_put_pages(spool, chg - add);
 			hugetlb_acct_memory(h, -rsv_adjust);
 		}
+	} else {
+#ifdef CONFIG_CGROUP_HUGETLB
+		/*
+		 * Since this branch handles private mappings, we attach the
+		 * counter to uncharge for this reservation off resv_map.
+		 */
+		resv_map->reservation_counter =
+			&h_cg->reserved_hugepage[hstate_index(h)];
+		resv_map->pages_per_hpage = pages_per_huge_page(h);
+		resv_map->css = &h_cg->css;
+#endif
 	}
 	return 0;
+out_put_pages:
+	/* put back original number of pages, chg */
+	(void)hugepage_subpool_put_pages(spool, chg);
+out_uncharge_cgroup:
+	hugetlb_cgroup_uncharge_cgroup(
+		hstate_index(h), chg * pages_per_huge_page(h), h_cg, true);
 out_err:
 	if (!vma || vma->vm_flags & VM_MAYSHARE)
 		/* Only call region_abort if the region_chg succeeded but the
--
2.24.1.735.g03f4e72817-goog

^ permalink raw reply related	[flat|nested] 27+ messages in thread

* [PATCH v9 6/8] hugetlb_cgroup: support noreserve mappings
  2019-12-17 23:16 [PATCH v9 1/8] hugetlb_cgroup: Add hugetlb_cgroup reservation counter Mina Almasry
                   ` (3 preceding siblings ...)
  2019-12-17 23:16 ` [PATCH v9 5/8] hugetlb_cgroup: add accounting for shared mappings Mina Almasry
@ 2019-12-17 23:16 ` Mina Almasry
  2020-01-14  0:48   ` David Rientjes
  2019-12-17 23:16 ` [PATCH v9 7/8] hugetlb_cgroup: Add hugetlb_cgroup reservation tests Mina Almasry
                   ` (4 subsequent siblings)
  9 siblings, 1 reply; 27+ messages in thread
From: Mina Almasry @ 2019-12-17 23:16 UTC (permalink / raw)
  To: mike.kravetz
  Cc: shuah, almasrymina, rientjes, shakeelb, gthelen, akpm,
	linux-kernel, linux-mm, linux-kselftest, cgroups, aneesh.kumar,
	mkoutny

Support MAP_NORESERVE accounting as part of the new counter.

For each hugepage allocation, at allocation time we check if there is
a reservation for this allocation or not. If there is a reservation for
this allocation, then this allocation was charged at reservation time,
and we don't re-account it. If there is no reserevation for this
allocation, we charge the appropriate hugetlb_cgroup.

The hugetlb_cgroup to uncharge for this allocation is stored in
page[3].private. We use new APIs added in an earlier patch to set this
pointer.

Signed-off-by: Mina Almasry <almasrymina@google.com>

---
 mm/hugetlb.c | 26 +++++++++++++++++++++++++-
 1 file changed, 25 insertions(+), 1 deletion(-)

diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index 39eb2fa53a420..3e94f5c2d7cd4 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -1367,6 +1367,9 @@ void free_huge_page(struct page *page)
 	clear_page_huge_active(page);
 	hugetlb_cgroup_uncharge_page(hstate_index(h), pages_per_huge_page(h),
 				     page, false);
+	hugetlb_cgroup_uncharge_page(hstate_index(h), pages_per_huge_page(h),
+				     page, true);
+
 	if (restore_reserve)
 		h->resv_huge_pages++;

@@ -2189,10 +2192,19 @@ struct page *alloc_huge_page(struct vm_area_struct *vma,
 			gbl_chg = 1;
 	}

+	/* If this allocation is not consuming a reservation, charge it now.
+	 */
+	if (map_chg || avoid_reserve || !vma_resv_map(vma)) {
+		ret = hugetlb_cgroup_charge_cgroup(idx, pages_per_huge_page(h),
+						   &h_cg, true);
+		if (ret)
+			goto out_subpool_put;
+	}
+
 	ret = hugetlb_cgroup_charge_cgroup(idx, pages_per_huge_page(h), &h_cg,
 					   false);
 	if (ret)
-		goto out_subpool_put;
+		goto out_uncharge_cgroup_reservation;

 	spin_lock(&hugetlb_lock);
 	/*
@@ -2216,6 +2228,14 @@ struct page *alloc_huge_page(struct vm_area_struct *vma,
 	}
 	hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h), h_cg, page,
 				     false);
+	/* If allocation is not consuming a reservation, also store the
+	 * hugetlb_cgroup pointer on the page.
+	 */
+	if (!vma_resv_map(vma) || map_chg || avoid_reserve) {
+		hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h), h_cg,
+					     page, true);
+	}
+
 	spin_unlock(&hugetlb_lock);

 	set_page_private(page, (unsigned long)spool);
@@ -2241,6 +2261,10 @@ struct page *alloc_huge_page(struct vm_area_struct *vma,
 out_uncharge_cgroup:
 	hugetlb_cgroup_uncharge_cgroup(idx, pages_per_huge_page(h), h_cg,
 				       false);
+out_uncharge_cgroup_reservation:
+	if (map_chg || avoid_reserve || !vma_resv_map(vma))
+		hugetlb_cgroup_uncharge_cgroup(idx, pages_per_huge_page(h),
+					       h_cg, true);
 out_subpool_put:
 	if (map_chg || avoid_reserve)
 		hugepage_subpool_put_pages(spool, 1);
--
2.24.1.735.g03f4e72817-goog

^ permalink raw reply related	[flat|nested] 27+ messages in thread

* [PATCH v9 7/8] hugetlb_cgroup: Add hugetlb_cgroup reservation tests
  2019-12-17 23:16 [PATCH v9 1/8] hugetlb_cgroup: Add hugetlb_cgroup reservation counter Mina Almasry
                   ` (4 preceding siblings ...)
  2019-12-17 23:16 ` [PATCH v9 6/8] hugetlb_cgroup: support noreserve mappings Mina Almasry
@ 2019-12-17 23:16 ` Mina Almasry
  2019-12-17 23:16 ` [PATCH v9 8/8] hugetlb_cgroup: Add hugetlb_cgroup reservation docs Mina Almasry
                   ` (3 subsequent siblings)
  9 siblings, 0 replies; 27+ messages in thread
From: Mina Almasry @ 2019-12-17 23:16 UTC (permalink / raw)
  To: mike.kravetz
  Cc: shuah, almasrymina, rientjes, shakeelb, gthelen, akpm,
	linux-kernel, linux-mm, linux-kselftest, cgroups, aneesh.kumar,
	mkoutny

The tests use both shared and private mapped hugetlb memory, and
monitors the hugetlb usage counter as well as the hugetlb reservation
counter. They test different configurations such as hugetlb memory usage
via hugetlbfs, or MAP_HUGETLB, or shmget/shmat, and with and without
MAP_POPULATE.

Also add test for hugetlb reservation reparenting, since this is
a subtle issue.

Signed-off-by: Mina Almasry <almasrymina@google.com>

---

Changes in v9:
- Added tests for hugetlb reparenting.
- Make tests explicitly support cgroup v1 and v2 via script argument.
Changes in v6:
- Updates tests for cgroups-v2 and NORESERVE allocations.

---
 tools/testing/selftests/vm/.gitignore         |   1 +
 tools/testing/selftests/vm/Makefile           |   1 +
 .../selftests/vm/charge_reserved_hugetlb.sh   | 557 ++++++++++++++++++
 .../selftests/vm/hugetlb_reparenting_test.sh  | 235 ++++++++
 .../selftests/vm/write_hugetlb_memory.sh      |  23 +
 .../testing/selftests/vm/write_to_hugetlbfs.c | 261 ++++++++
 6 files changed, 1078 insertions(+)
 create mode 100755 tools/testing/selftests/vm/charge_reserved_hugetlb.sh
 create mode 100755 tools/testing/selftests/vm/hugetlb_reparenting_test.sh
 create mode 100644 tools/testing/selftests/vm/write_hugetlb_memory.sh
 create mode 100644 tools/testing/selftests/vm/write_to_hugetlbfs.c

diff --git a/tools/testing/selftests/vm/.gitignore b/tools/testing/selftests/vm/.gitignore
index 31b3c98b6d34d..d3bed9407773c 100644
--- a/tools/testing/selftests/vm/.gitignore
+++ b/tools/testing/selftests/vm/.gitignore
@@ -14,3 +14,4 @@ virtual_address_range
 gup_benchmark
 va_128TBswitch
 map_fixed_noreplace
+write_to_hugetlbfs
diff --git a/tools/testing/selftests/vm/Makefile b/tools/testing/selftests/vm/Makefile
index 7f9a8a8c31da9..662bb95e84c5d 100644
--- a/tools/testing/selftests/vm/Makefile
+++ b/tools/testing/selftests/vm/Makefile
@@ -22,6 +22,7 @@ TEST_GEN_FILES += userfaultfd
 ifneq (,$(filter $(ARCH),arm64 ia64 mips64 parisc64 ppc64 riscv64 s390x sh64 sparc64 x86_64))
 TEST_GEN_FILES += va_128TBswitch
 TEST_GEN_FILES += virtual_address_range
+TEST_GEN_FILES += write_to_hugetlbfs
 endif

 TEST_PROGS := run_vmtests
diff --git a/tools/testing/selftests/vm/charge_reserved_hugetlb.sh b/tools/testing/selftests/vm/charge_reserved_hugetlb.sh
new file mode 100755
index 0000000000000..cdb4b5fe09d01
--- /dev/null
+++ b/tools/testing/selftests/vm/charge_reserved_hugetlb.sh
@@ -0,0 +1,557 @@
+#!/bin/sh
+# SPDX-License-Identifier: GPL-2.0
+
+set -e
+
+if [[ $(id -u) -ne 0 ]]; then
+  echo "This test must be run as root. Skipping..."
+  exit 0
+fi
+
+fault_limit_file=limit_in_bytes
+reservation_limit_file=reservation_limit_in_bytes
+fault_usage_file=usage_in_bytes
+reservation_usage_file=reservation_usage_in_bytes
+
+if [[ "$1" == "-cgroup-v2" ]]; then
+  cgroup2=1
+  fault_limit_file=max
+  reservation_limit_file=reservation_max
+  fault_usage_file=current
+  reservation_usage_file=reservation_current
+fi
+
+cgroup_path=/dev/cgroup/memory
+if [[ ! -e $cgroup_path ]]; then
+  mkdir -p $cgroup_path
+  if [[ $cgroup2 ]]; then
+    mount -t cgroup2 none $cgroup_path
+  else
+    mount -t cgroup memory,hugetlb $cgroup_path
+  fi
+fi
+
+if [[ $cgroup2 ]]; then
+  echo "+hugetlb" >/dev/cgroup/memory/cgroup.subtree_control
+fi
+
+cleanup() {
+  if [[ $cgroup2 ]]; then
+    echo $$ >$cgroup_path/cgroup.procs
+  else
+    echo $$ >$cgroup_path/tasks
+  fi
+
+  if [[ -e /mnt/huge ]]; then
+    rm -rf /mnt/huge/*
+    umount /mnt/huge || echo error
+    rmdir /mnt/huge
+  fi
+  if [[ -e $cgroup_path/hugetlb_cgroup_test ]]; then
+    rmdir $cgroup_path/hugetlb_cgroup_test
+  fi
+  if [[ -e $cgroup_path/hugetlb_cgroup_test1 ]]; then
+    rmdir $cgroup_path/hugetlb_cgroup_test1
+  fi
+  if [[ -e $cgroup_path/hugetlb_cgroup_test2 ]]; then
+    rmdir $cgroup_path/hugetlb_cgroup_test2
+  fi
+  echo 0 >/proc/sys/vm/nr_hugepages
+  echo CLEANUP DONE
+}
+
+function expect_equal() {
+  local expected="$1"
+  local actual="$2"
+  local error="$3"
+
+  if [[ "$expected" != "$actual" ]]; then
+    echo "expected ($expected) != actual ($actual): $3"
+    cleanup
+    exit 1
+  fi
+}
+
+function setup_cgroup() {
+  local name="$1"
+  local cgroup_limit="$2"
+  local reservation_limit="$3"
+
+  mkdir $cgroup_path/$name
+
+  echo writing cgroup limit: "$cgroup_limit"
+  echo "$cgroup_limit" >$cgroup_path/$name/hugetlb.2MB.$fault_limit_file
+
+  echo writing reseravation limit: "$reservation_limit"
+  echo "$reservation_limit" > \
+    $cgroup_path/$name/hugetlb.2MB.$reservation_limit_file
+
+  if [ -e "$cgroup_path/$name/cpuset.cpus" ]; then
+    echo 0 >$cgroup_path/$name/cpuset.cpus
+  fi
+  if [ -e "$cgroup_path/$name/cpuset.mems" ]; then
+    echo 0 >$cgroup_path/$name/cpuset.mems
+  fi
+}
+
+function wait_for_hugetlb_memory_to_get_depleted() {
+  local cgroup="$1"
+  local path="/dev/cgroup/memory/$cgroup/hugetlb.2MB.$reservation_usage_file"
+  # Wait for hugetlbfs memory to get depleted.
+  while [ $(cat $path) != 0 ]; do
+    echo Waiting for hugetlb memory to get depleted.
+    cat $path
+    sleep 0.5
+  done
+}
+
+function wait_for_hugetlb_memory_to_get_reserved() {
+  local cgroup="$1"
+  local size="$2"
+
+  local path="/dev/cgroup/memory/$cgroup/hugetlb.2MB.$reservation_usage_file"
+  # Wait for hugetlbfs memory to get written.
+  while [ $(cat $path) != $size ]; do
+    echo Waiting for hugetlb memory reservation to reach size $size.
+    cat $path
+    sleep 0.5
+  done
+}
+
+function wait_for_hugetlb_memory_to_get_written() {
+  local cgroup="$1"
+  local size="$2"
+
+  local path="/dev/cgroup/memory/$cgroup/hugetlb.2MB.$fault_usage_file"
+  # Wait for hugetlbfs memory to get written.
+  while [ $(cat $path) != $size ]; do
+    echo Waiting for hugetlb memory to reach size $size.
+    cat $path
+    sleep 0.5
+  done
+}
+
+function write_hugetlbfs_and_get_usage() {
+  local cgroup="$1"
+  local size="$2"
+  local populate="$3"
+  local write="$4"
+  local path="$5"
+  local method="$6"
+  local private="$7"
+  local expect_failure="$8"
+  local reserve="$9"
+
+  # Function return values.
+  reservation_failed=0
+  oom_killed=0
+  hugetlb_difference=0
+  reserved_difference=0
+
+  local hugetlb_usage=$cgroup_path/$cgroup/hugetlb.2MB.$fault_usage_file
+  local reserved_usage=$cgroup_path/$cgroup/hugetlb.2MB.$reservation_usage_file
+
+  local hugetlb_before=$(cat $hugetlb_usage)
+  local reserved_before=$(cat $reserved_usage)
+
+  echo
+  echo Starting:
+  echo hugetlb_usage="$hugetlb_before"
+  echo reserved_usage="$reserved_before"
+  echo expect_failure is "$expect_failure"
+
+  set +e
+  if [[ "$method" == "1" ]] || [[ "$method" == 2 ]] ||
+    [[ "$private" == "-r" ]] && [[ "$expect_failure" != 1 ]]; then
+
+    bash write_hugetlb_memory.sh "$size" "$populate" "$write" \
+      "$cgroup" "$path" "$method" "$private" "-l" "$reserve" &
+
+    local write_result=$?
+
+    if [[ "$reserve" != "-n" ]]; then
+      wait_for_hugetlb_memory_to_get_reserved "$cgroup" "$size"
+    elif [[ "$populate" == "-o" ]] || [[ "$write" == "-w" ]]; then
+      wait_for_hugetlb_memory_to_get_written "$cgroup" "$size"
+    else
+      # This case doesn't produce visible effects, but we still have
+      # to wait for the async process to start and execute...
+      sleep 0.5
+    fi
+
+    echo write_result is $write_result
+  else
+    bash write_hugetlb_memory.sh "$size" "$populate" "$write" \
+      "$cgroup" "$path" "$method" "$private" "$reserve"
+    local write_result=$?
+
+    if [[ "$reserve" != "-n" ]]; then
+      wait_for_hugetlb_memory_to_get_reserved "$cgroup" "$size"
+    fi
+  fi
+  set -e
+
+  if [[ "$write_result" == 1 ]]; then
+    reservation_failed=1
+  fi
+
+  # On linus/master, the above process gets SIGBUS'd on oomkill, with
+  # return code 135. On earlier kernels, it gets actual oomkill, with return
+  # code 137, so just check for both conditions in case we're testing
+  # against an earlier kernel.
+  if [[ "$write_result" == 135 ]] || [[ "$write_result" == 137 ]]; then
+    oom_killed=1
+  fi
+
+  local hugetlb_after=$(cat $hugetlb_usage)
+  local reserved_after=$(cat $reserved_usage)
+
+  echo After write:
+  echo hugetlb_usage="$hugetlb_after"
+  echo reserved_usage="$reserved_after"
+
+  hugetlb_difference=$(($hugetlb_after - $hugetlb_before))
+  reserved_difference=$(($reserved_after - $reserved_before))
+}
+
+function cleanup_hugetlb_memory() {
+  set +e
+  local cgroup="$1"
+  if [[ "$(pgrep write_to_hugetlbfs)" != "" ]]; then
+    echo kiling write_to_hugetlbfs
+    killall -2 write_to_hugetlbfs
+    wait_for_hugetlb_memory_to_get_depleted $cgroup
+  fi
+  set -e
+
+  if [[ -e /mnt/huge ]]; then
+    rm -rf /mnt/huge/*
+    umount /mnt/huge
+    rmdir /mnt/huge
+  fi
+}
+
+function run_test() {
+  local size="$1"
+  local populate="$2"
+  local write="$3"
+  local cgroup_limit="$4"
+  local reservation_limit="$5"
+  local nr_hugepages="$6"
+  local method="$7"
+  local private="$8"
+  local expect_failure="$9"
+  local reserve="${10}"
+
+  # Function return values.
+  hugetlb_difference=0
+  reserved_difference=0
+  reservation_failed=0
+  oom_killed=0
+
+  echo nr hugepages = "$nr_hugepages"
+  echo "$nr_hugepages" >/proc/sys/vm/nr_hugepages
+
+  setup_cgroup "hugetlb_cgroup_test" "$cgroup_limit" "$reservation_limit"
+
+  mkdir -p /mnt/huge
+  mount -t hugetlbfs \
+    -o pagesize=2M,size=256M none /mnt/huge
+
+  write_hugetlbfs_and_get_usage "hugetlb_cgroup_test" "$size" "$populate" \
+    "$write" "/mnt/huge/test" "$method" "$private" "$expect_failure" \
+    "$reserve"
+
+  cleanup_hugetlb_memory "hugetlb_cgroup_test"
+
+  local final_hugetlb=$(cat $cgroup_path/hugetlb_cgroup_test/hugetlb.2MB.$fault_usage_file)
+  local final_reservation=$(cat $cgroup_path/hugetlb_cgroup_test/hugetlb.2MB.$reservation_usage_file)
+
+  echo $hugetlb_difference
+  echo $reserved_difference
+  expect_equal "0" "$final_hugetlb" "final hugetlb is not zero"
+  expect_equal "0" "$final_reservation" "final reservation is not zero"
+}
+
+function run_multiple_cgroup_test() {
+  local size1="$1"
+  local populate1="$2"
+  local write1="$3"
+  local cgroup_limit1="$4"
+  local reservation_limit1="$5"
+
+  local size2="$6"
+  local populate2="$7"
+  local write2="$8"
+  local cgroup_limit2="$9"
+  local reservation_limit2="${10}"
+
+  local nr_hugepages="${11}"
+  local method="${12}"
+  local private="${13}"
+  local expect_failure="${14}"
+  local reserve="${15}"
+
+  # Function return values.
+  hugetlb_difference1=0
+  reserved_difference1=0
+  reservation_failed1=0
+  oom_killed1=0
+
+  hugetlb_difference2=0
+  reserved_difference2=0
+  reservation_failed2=0
+  oom_killed2=0
+
+  echo nr hugepages = "$nr_hugepages"
+  echo "$nr_hugepages" >/proc/sys/vm/nr_hugepages
+
+  setup_cgroup "hugetlb_cgroup_test1" "$cgroup_limit1" "$reservation_limit1"
+  setup_cgroup "hugetlb_cgroup_test2" "$cgroup_limit2" "$reservation_limit2"
+
+  mkdir -p /mnt/huge
+  mount -t hugetlbfs \
+    -o pagesize=2M,size=256M none /mnt/huge
+
+  write_hugetlbfs_and_get_usage "hugetlb_cgroup_test1" "$size1" \
+    "$populate1" "$write1" "/mnt/huge/test1" "$method" "$private" \
+    "$expect_failure" "$reserve"
+
+  hugetlb_difference1=$hugetlb_difference
+  reserved_difference1=$reserved_difference
+  reservation_failed1=$reservation_failed
+  oom_killed1=$oom_killed
+
+  local cgroup1_hugetlb_usage=$cgroup_path/hugetlb_cgroup_test1/hugetlb.2MB.$fault_usage_file
+  local cgroup1_reservation_usage=$cgroup_path/hugetlb_cgroup_test1/hugetlb.2MB.$reservation_usage_file
+  local cgroup2_hugetlb_usage=$cgroup_path/hugetlb_cgroup_test2/hugetlb.2MB.$fault_usage_file
+  local cgroup2_reservation_usage=$cgroup_path/hugetlb_cgroup_test2/hugetlb.2MB.$reservation_usage_file
+
+  local usage_before_second_write=$(cat $cgroup1_hugetlb_usage)
+  local reservation_usage_before_second_write=$(cat \
+    $cgroup1_reservation_usage)
+
+  write_hugetlbfs_and_get_usage "hugetlb_cgroup_test2" "$size2" \
+    "$populate2" "$write2" "/mnt/huge/test2" "$method" "$private" \
+    "$expect_failure" "$reserve"
+
+  hugetlb_difference2=$hugetlb_difference
+  reserved_difference2=$reserved_difference
+  reservation_failed2=$reservation_failed
+  oom_killed2=$oom_killed
+
+  expect_equal "$usage_before_second_write" \
+    "$(cat $cgroup1_hugetlb_usage)" "Usage changed."
+  expect_equal "$reservation_usage_before_second_write" \
+    "$(cat $cgroup1_reservation_usage)" "Reservation usage changed."
+
+  cleanup_hugetlb_memory
+
+  local final_hugetlb=$(cat $cgroup1_hugetlb_usage)
+  local final_reservation=$(cat $cgroup1_reservation_usage)
+
+  expect_equal "0" "$final_hugetlb" \
+    "hugetlbt_cgroup_test1 final hugetlb is not zero"
+  expect_equal "0" "$final_reservation" \
+    "hugetlbt_cgroup_test1 final reservation is not zero"
+
+  local final_hugetlb=$(cat $cgroup2_hugetlb_usage)
+  local final_reservation=$(cat $cgroup2_reservation_usage)
+
+  expect_equal "0" "$final_hugetlb" \
+    "hugetlb_cgroup_test2 final hugetlb is not zero"
+  expect_equal "0" "$final_reservation" \
+    "hugetlb_cgroup_test2 final reservation is not zero"
+}
+
+cleanup
+
+for populate in "" "-o"; do
+  for method in 0 1 2; do
+    for private in "" "-r"; do
+      for reserve in "" "-n"; do
+
+        # Skip mmap(MAP_HUGETLB | MAP_SHARED). Doesn't seem to be supported.
+        if [[ "$method" == 1 ]] && [[ "$private" == "" ]]; then
+          continue
+        fi
+
+        # Skip populated shmem tests. Doesn't seem to be supported.
+        if [[ "$method" == 2"" ]] && [[ "$populate" == "-o" ]]; then
+          continue
+        fi
+
+        if [[ "$method" == 2"" ]] && [[ "$reserve" == "-n" ]]; then
+          continue
+        fi
+
+        cleanup
+        echo
+        echo
+        echo
+        echo Test normal case.
+        echo private=$private, populate=$populate, method=$method, reserve=$reserve
+        run_test $((10 * 1024 * 1024)) "$populate" "" $((20 * 1024 * 1024)) \
+          $((20 * 1024 * 1024)) 10 "$method" "$private" "0" "$reserve"
+
+        echo Memory charged to hugtlb=$hugetlb_difference
+        echo Memory charged to reservation=$reserved_difference
+
+        if [[ "$populate" == "-o" ]]; then
+          expect_equal "$((10 * 1024 * 1024))" "$hugetlb_difference" \
+            "Reserved memory charged to hugetlb cgroup."
+        else
+          expect_equal "0" "$hugetlb_difference" \
+            "Reserved memory charged to hugetlb cgroup."
+        fi
+
+        if [[ "$reserve" != "-n" ]] || [[ "$populate" == "-o" ]]; then
+          expect_equal "$((10 * 1024 * 1024))" "$reserved_difference" \
+            "Reserved memory not charged to reservation usage."
+        else
+          expect_equal "0" "$reserved_difference" \
+            "Reserved memory not charged to reservation usage."
+        fi
+
+        echo 'PASS'
+
+        cleanup
+        echo
+        echo
+        echo
+        echo Test normal case with write.
+        echo private=$private, populate=$populate, method=$method, reserve=$reserve
+        run_test $((10 * 1024 * 1024)) "$populate" '-w' $((20 * 1024 * 1024)) \
+          $((20 * 1024 * 1024)) 10 "$method" "$private" "0" "$reserve"
+
+        echo Memory charged to hugtlb=$hugetlb_difference
+        echo Memory charged to reservation=$reserved_difference
+
+        expect_equal "$((10 * 1024 * 1024))" "$hugetlb_difference" \
+          "Reserved memory charged to hugetlb cgroup."
+
+        expect_equal "$((10 * 1024 * 1024))" "$reserved_difference" \
+          "Reserved memory not charged to reservation usage."
+
+        echo 'PASS'
+
+        cleanup
+        continue
+        echo
+        echo
+        echo
+        echo Test more than reservation case.
+        echo private=$private, populate=$populate, method=$method, reserve=$reserve
+
+        if [ "$reserve" != "-n" ]; then
+          run_test "$((10 * 1024 * 1024))" "$populate" '' "$((20 * 1024 * 1024))" \
+            "$((5 * 1024 * 1024))" "10" "$method" "$private" "1" "$reserve"
+
+          expect_equal "1" "$reservation_failed" "Reservation succeeded."
+        fi
+
+        echo 'PASS'
+
+        cleanup
+
+        echo
+        echo
+        echo
+        echo Test more than cgroup limit case.
+        echo private=$private, populate=$populate, method=$method, reserve=$reserve
+
+        # Not sure if shm memory can be cleaned up when the process gets sigbus'd.
+        if [[ "$method" != 2 ]]; then
+          run_test $((10 * 1024 * 1024)) "$populate" "-w" $((5 * 1024 * 1024)) \
+            $((20 * 1024 * 1024)) 10 "$method" "$private" "1" "$reserve"
+
+          expect_equal "1" "$oom_killed" "Not oom killed."
+        fi
+        echo 'PASS'
+
+        cleanup
+
+        echo
+        echo
+        echo
+        echo Test normal case, multiple cgroups.
+        echo private=$private, populate=$populate, method=$method, reserve=$reserve
+        run_multiple_cgroup_test "$((6 * 1024 * 1024))" "$populate" "" \
+          "$((20 * 1024 * 1024))" "$((20 * 1024 * 1024))" "$((10 * 1024 * 1024))" \
+          "$populate" "" "$((20 * 1024 * 1024))" "$((20 * 1024 * 1024))" "10" \
+          "$method" "$private" "0" "$reserve"
+
+        echo Memory charged to hugtlb1=$hugetlb_difference1
+        echo Memory charged to reservation1=$reserved_difference1
+        echo Memory charged to hugtlb2=$hugetlb_difference2
+        echo Memory charged to reservation2=$reserved_difference2
+
+        if [[ "$reserve" != "-n" ]] || [[ "$populate" == "-o" ]]; then
+          expect_equal "$((6 * 1024 * 1024))" "$reserved_difference1" \
+            "Incorrect reservations charged to cgroup 1."
+
+          expect_equal "$((10 * 1024 * 1024))" "$reserved_difference2" \
+            "Incorrect reservation charged to cgroup 2."
+
+        else
+          expect_equal "0" "$reserved_difference1" \
+            "Incorrect reservations charged to cgroup 1."
+
+          expect_equal "0" "$reserved_difference2" \
+            "Incorrect reservation charged to cgroup 2."
+        fi
+
+        if [[ "$populate" == "-o" ]]; then
+          expect_equal "$((6 * 1024 * 1024))" "$hugetlb_difference1" \
+            "Incorrect hugetlb charged to cgroup 1."
+
+          expect_equal "$((10 * 1024 * 1024))" "$hugetlb_difference2" \
+            "Incorrect hugetlb charged to cgroup 2."
+
+        else
+          expect_equal "0" "$hugetlb_difference1" \
+            "Incorrect hugetlb charged to cgroup 1."
+
+          expect_equal "0" "$hugetlb_difference2" \
+            "Incorrect hugetlb charged to cgroup 2."
+        fi
+        echo 'PASS'
+
+        cleanup
+        echo
+        echo
+        echo
+        echo Test normal case with write, multiple cgroups.
+        echo private=$private, populate=$populate, method=$method, reserve=$reserve
+        run_multiple_cgroup_test "$((6 * 1024 * 1024))" "$populate" "-w" \
+          "$((20 * 1024 * 1024))" "$((20 * 1024 * 1024))" "$((10 * 1024 * 1024))" \
+          "$populate" "-w" "$((20 * 1024 * 1024))" "$((20 * 1024 * 1024))" "10" \
+          "$method" "$private" "0" "$reserve"
+
+        echo Memory charged to hugtlb1=$hugetlb_difference1
+        echo Memory charged to reservation1=$reserved_difference1
+        echo Memory charged to hugtlb2=$hugetlb_difference2
+        echo Memory charged to reservation2=$reserved_difference2
+
+        expect_equal "$((6 * 1024 * 1024))" "$hugetlb_difference1" \
+          "Incorrect hugetlb charged to cgroup 1."
+
+        expect_equal "$((6 * 1024 * 1024))" "$reserved_difference1" \
+          "Incorrect reservation charged to cgroup 1."
+
+        expect_equal "$((10 * 1024 * 1024))" "$hugetlb_difference2" \
+          "Incorrect hugetlb charged to cgroup 2."
+
+        expect_equal "$((10 * 1024 * 1024))" "$reserved_difference2" \
+          "Incorrected reservation charged to cgroup 2."
+        echo 'PASS'
+
+        cleanup
+
+      done # reserve
+    done   # private
+  done     # populate
+done       # method
+
+umount $cgroup_path
+rmdir $cgroup_path
diff --git a/tools/testing/selftests/vm/hugetlb_reparenting_test.sh b/tools/testing/selftests/vm/hugetlb_reparenting_test.sh
new file mode 100755
index 0000000000000..2be672c2b311a
--- /dev/null
+++ b/tools/testing/selftests/vm/hugetlb_reparenting_test.sh
@@ -0,0 +1,235 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0
+
+set -e
+
+if [[ $(id -u) -ne 0 ]]; then
+  echo "This test must be run as root. Skipping..."
+  exit 0
+fi
+
+usage_file=usage_in_bytes
+
+if [[ "$1" == "-cgroup-v2" ]]; then
+  cgroup2=1
+  usage_file=current
+fi
+
+CGROUP_ROOT='/dev/cgroup/memory'
+MNT='/mnt/huge/'
+
+if [[ ! -e $CGROUP_ROOT ]]; then
+  mkdir -p $CGROUP_ROOT
+  if [[ $cgroup2 ]]; then
+    mount -t cgroup2 none $CGROUP_ROOT
+    sleep 1
+    echo "+hugetlb +memory" >$CGROUP_ROOT/cgroup.subtree_control
+  else
+    mount -t cgroup memory,hugetlb $CGROUP_ROOT
+  fi
+fi
+
+function cleanup() {
+  echo cleanup
+  set +e
+  rm -rf "$MNT"/* 2>/dev/null
+  umount "$MNT" 2>/dev/null
+  rmdir "$MNT" 2>/dev/null
+  rmdir "$CGROUP_ROOT"/a/b 2>/dev/null
+  rmdir "$CGROUP_ROOT"/a 2>/dev/null
+  rmdir "$CGROUP_ROOT"/test1 2>/dev/null
+  echo 0 >/proc/sys/vm/nr_hugepages
+  set -e
+}
+
+function assert_state() {
+  local expected_a="$1"
+  local expected_a_hugetlb="$2"
+  local expected_b=""
+  local expected_b_hugetlb=""
+
+  if [ ! -z ${3:-} ] && [ ! -z ${4:-} ]; then
+    expected_b="$3"
+    expected_b_hugetlb="$4"
+  fi
+  local tolerance=$((5 * 1024 * 1024))
+
+  local actual_a
+  actual_a="$(cat "$CGROUP_ROOT"/a/memory.$usage_file)"
+  if [[ $actual_a -lt $(($expected_a - $tolerance)) ]] ||
+    [[ $actual_a -gt $(($expected_a + $tolerance)) ]]; then
+    echo actual a = $((${actual_a%% *} / 1024 / 1024)) MB
+    echo expected a = $((${expected_a%% *} / 1024 / 1024)) MB
+    echo fail
+
+    cleanup
+    exit 1
+  fi
+
+  local actual_a_hugetlb
+  actual_a_hugetlb="$(cat "$CGROUP_ROOT"/a/hugetlb.2MB.$usage_file)"
+  if [[ $actual_a_hugetlb -lt $(($expected_a_hugetlb - $tolerance)) ]] ||
+    [[ $actual_a_hugetlb -gt $(($expected_a_hugetlb + $tolerance)) ]]; then
+    echo actual a hugetlb = $((${actual_a_hugetlb%% *} / 1024 / 1024)) MB
+    echo expected a hugetlb = $((${expected_a_hugetlb%% *} / 1024 / 1024)) MB
+    echo fail
+
+    cleanup
+    exit 1
+  fi
+
+  if [[ -z "$expected_b" || -z "$expected_b_hugetlb" ]]; then
+    return
+  fi
+
+  local actual_b
+  actual_b="$(cat "$CGROUP_ROOT"/a/b/memory.$usage_file)"
+  if [[ $actual_b -lt $(($expected_b - $tolerance)) ]] ||
+    [[ $actual_b -gt $(($expected_b + $tolerance)) ]]; then
+    echo actual b = $((${actual_b%% *} / 1024 / 1024)) MB
+    echo expected b = $((${expected_b%% *} / 1024 / 1024)) MB
+    echo fail
+
+    cleanup
+    exit 1
+  fi
+
+  local actual_b_hugetlb
+  actual_b_hugetlb="$(cat "$CGROUP_ROOT"/a/b/hugetlb.2MB.$usage_file)"
+  if [[ $actual_b_hugetlb -lt $(($expected_b_hugetlb - $tolerance)) ]] ||
+    [[ $actual_b_hugetlb -gt $(($expected_b_hugetlb + $tolerance)) ]]; then
+    echo actual b hugetlb = $((${actual_b_hugetlb%% *} / 1024 / 1024)) MB
+    echo expected b hugetlb = $((${expected_b_hugetlb%% *} / 1024 / 1024)) MB
+    echo fail
+
+    cleanup
+    exit 1
+  fi
+}
+
+function setup() {
+  echo 100 >/proc/sys/vm/nr_hugepages
+  mkdir "$CGROUP_ROOT"/a
+  sleep 1
+  if [[ $cgroup2 ]]; then
+    echo "+hugetlb +memory" >$CGROUP_ROOT/a/cgroup.subtree_control
+  else
+    echo 0 >$CGROUP_ROOT/a/cpuset.mems
+    echo 0 >$CGROUP_ROOT/a/cpuset.cpus
+  fi
+
+  mkdir "$CGROUP_ROOT"/a/b
+
+  if [[ ! $cgroup2 ]]; then
+    echo 0 >$CGROUP_ROOT/a/b/cpuset.mems
+    echo 0 >$CGROUP_ROOT/a/b/cpuset.cpus
+  fi
+
+  mkdir -p "$MNT"
+  mount -t hugetlbfs none "$MNT"
+}
+
+write_hugetlbfs() {
+  local cgroup="$1"
+  local path="$2"
+  local size="$3"
+
+  if [[ $cgroup2 ]]; then
+    echo $$ >$CGROUP_ROOT/$cgroup/cgroup.procs
+  else
+    echo 0 >$CGROUP_ROOT/$cgroup/cpuset.mems
+    echo 0 >$CGROUP_ROOT/$cgroup/cpuset.cpus
+    echo $$ >"$CGROUP_ROOT/$cgroup/tasks"
+  fi
+  ./write_to_hugetlbfs -p "$path" -s "$size" -m 0 -o
+  if [[ $cgroup2 ]]; then
+    echo $$ >$CGROUP_ROOT/cgroup.procs
+  else
+    echo $$ >"$CGROUP_ROOT/tasks"
+  fi
+  echo
+}
+
+set -e
+
+size=$((2 * 1024 * 1024 * 25)) # 50MB = 25 * 2MB hugepages.
+
+cleanup
+
+echo
+echo
+echo Test charge, rmdir, uncharge
+setup
+echo mkdir
+mkdir $CGROUP_ROOT/test1
+
+echo write
+write_hugetlbfs test1 "$MNT"/test $size
+
+echo rmdir
+rmdir $CGROUP_ROOT/test1
+mkdir $CGROUP_ROOT/test1
+
+echo uncharge
+rm -rf /mnt/huge/*
+
+cleanup
+
+echo done
+echo
+echo
+if [[ ! $cgroup2 ]]; then
+  echo "Test parent and child hugetlb usage"
+  setup
+
+  echo write
+  write_hugetlbfs a "$MNT"/test $size
+
+  echo Assert memory charged correctly for parent use.
+  assert_state 0 $size 0 0
+
+  write_hugetlbfs a/b "$MNT"/test2 $size
+
+  echo Assert memory charged correctly for child use.
+  assert_state 0 $(($size * 2)) 0 $size
+
+  rmdir "$CGROUP_ROOT"/a/b
+  sleep 5
+  echo Assert memory reparent correctly.
+  assert_state 0 $(($size * 2))
+
+  rm -rf "$MNT"/*
+  umount "$MNT"
+  echo Assert memory uncharged correctly.
+  assert_state 0 0
+
+  cleanup
+fi
+
+echo
+echo
+echo "Test child only hugetlb usage"
+echo setup
+setup
+
+echo write
+write_hugetlbfs a/b "$MNT"/test2 $size
+
+echo Assert memory charged correctly for child only use.
+assert_state 0 $(($size)) 0 $size
+
+rmdir "$CGROUP_ROOT"/a/b
+echo Assert memory reparent correctly.
+assert_state 0 $size
+
+rm -rf "$MNT"/*
+umount "$MNT"
+echo Assert memory uncharged correctly.
+assert_state 0 0
+
+cleanup
+
+echo ALL PASS
+
+umount $CGROUP_ROOT
+rm -rf $CGROUP_ROOT
diff --git a/tools/testing/selftests/vm/write_hugetlb_memory.sh b/tools/testing/selftests/vm/write_hugetlb_memory.sh
new file mode 100644
index 0000000000000..d3d0d108924d4
--- /dev/null
+++ b/tools/testing/selftests/vm/write_hugetlb_memory.sh
@@ -0,0 +1,23 @@
+#!/bin/sh
+# SPDX-License-Identifier: GPL-2.0
+
+set -e
+
+size=$1
+populate=$2
+write=$3
+cgroup=$4
+path=$5
+method=$6
+private=$7
+want_sleep=$8
+reserve=$9
+
+echo "Putting task in cgroup '$cgroup'"
+echo $$ > /dev/cgroup/memory/"$cgroup"/cgroup.procs
+
+echo "Method is $method"
+
+set +e
+./write_to_hugetlbfs -p "$path" -s "$size" "$write" "$populate" -m "$method" \
+      "$private" "$want_sleep" "$reserve"
diff --git a/tools/testing/selftests/vm/write_to_hugetlbfs.c b/tools/testing/selftests/vm/write_to_hugetlbfs.c
new file mode 100644
index 0000000000000..85811c3384a10
--- /dev/null
+++ b/tools/testing/selftests/vm/write_to_hugetlbfs.c
@@ -0,0 +1,261 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * This program reserves and uses hugetlb memory, supporting a bunch of
+ * scenarios needed by the charged_reserved_hugetlb.sh test.
+ */
+
+#include <err.h>
+#include <errno.h>
+#include <signal.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+#include <fcntl.h>
+#include <sys/types.h>
+#include <sys/shm.h>
+#include <sys/stat.h>
+#include <sys/mman.h>
+
+/* Global definitions. */
+enum method {
+	HUGETLBFS,
+	MMAP_MAP_HUGETLB,
+	SHM,
+	MAX_METHOD
+};
+
+
+/* Global variables. */
+static const char *self;
+static char *shmaddr;
+static int shmid;
+
+/*
+ * Show usage and exit.
+ */
+static void exit_usage(void)
+{
+	printf("Usage: %s -p <path to hugetlbfs file> -s <size to map> "
+	       "[-m <0=hugetlbfs | 1=mmap(MAP_HUGETLB)>] [-l] [-r] "
+	       "[-o] [-w] [-n]\n",
+	       self);
+	exit(EXIT_FAILURE);
+}
+
+void sig_handler(int signo)
+{
+	printf("Received %d.\n", signo);
+	if (signo == SIGINT) {
+		printf("Deleting the memory\n");
+		if (shmdt((const void *)shmaddr) != 0) {
+			perror("Detach failure");
+			shmctl(shmid, IPC_RMID, NULL);
+			exit(4);
+		}
+
+		shmctl(shmid, IPC_RMID, NULL);
+		printf("Done deleting the memory\n");
+	}
+	exit(2);
+}
+
+int main(int argc, char **argv)
+{
+	int fd = 0;
+	int key = 0;
+	int *ptr = NULL;
+	int c = 0;
+	int size = 0;
+	char path[256] = "";
+	enum method method = MAX_METHOD;
+	int want_sleep = 0, private = 0;
+	int populate = 0;
+	int write = 0;
+	int reserve = 1;
+
+	unsigned long i;
+
+	if (signal(SIGINT, sig_handler) == SIG_ERR)
+		err(1, "\ncan't catch SIGINT\n");
+
+	/* Parse command-line arguments. */
+	setvbuf(stdout, NULL, _IONBF, 0);
+	self = argv[0];
+
+	while ((c = getopt(argc, argv, "s:p:m:owlrn")) != -1) {
+		switch (c) {
+		case 's':
+			size = atoi(optarg);
+			break;
+		case 'p':
+			strncpy(path, optarg, sizeof(path));
+			break;
+		case 'm':
+			if (atoi(optarg) >= MAX_METHOD) {
+				errno = EINVAL;
+				perror("Invalid -m.");
+				exit_usage();
+			}
+			method = atoi(optarg);
+			break;
+		case 'o':
+			populate = 1;
+			break;
+		case 'w':
+			write = 1;
+			break;
+		case 'l':
+			want_sleep = 1;
+			break;
+		case 'r':
+		    private
+			= 1;
+			break;
+		case 'n':
+			reserve = 0;
+			break;
+		default:
+			errno = EINVAL;
+			perror("Invalid arg");
+			exit_usage();
+		}
+	}
+
+	if (strncmp(path, "", sizeof(path)) != 0) {
+		printf("Writing to this path: %s\n", path);
+	} else {
+		errno = EINVAL;
+		perror("path not found");
+		exit_usage();
+	}
+
+	if (size != 0) {
+		printf("Writing this size: %d\n", size);
+	} else {
+		errno = EINVAL;
+		perror("size not found");
+		exit_usage();
+	}
+
+	if (!populate)
+		printf("Not populating.\n");
+	else
+		printf("Populating.\n");
+
+	if (!write)
+		printf("Not writing to memory.\n");
+
+	if (method == MAX_METHOD) {
+		errno = EINVAL;
+		perror("-m Invalid");
+		exit_usage();
+	} else
+		printf("Using method=%d\n", method);
+
+	if (!private)
+		printf("Shared mapping.\n");
+	else
+		printf("Private mapping.\n");
+
+	if (!reserve)
+		printf("NO_RESERVE mapping.\n");
+	else
+		printf("RESERVE mapping.\n");
+
+	switch (method) {
+	case HUGETLBFS:
+		printf("Allocating using HUGETLBFS.\n");
+		fd = open(path, O_CREAT | O_RDWR, 0777);
+		if (fd == -1)
+			err(1, "Failed to open file.");
+
+		ptr = mmap(NULL, size, PROT_READ | PROT_WRITE,
+			   (private ? MAP_PRIVATE : MAP_SHARED) |
+				   (populate ? MAP_POPULATE : 0) |
+				   (reserve ? 0 : MAP_NORESERVE),
+			   fd, 0);
+
+		if (ptr == MAP_FAILED) {
+			close(fd);
+			err(1, "Error mapping the file");
+		}
+		break;
+	case MMAP_MAP_HUGETLB:
+		printf("Allocating using MAP_HUGETLB.\n");
+		ptr = mmap(NULL, size, PROT_READ | PROT_WRITE,
+			   (private ? (MAP_PRIVATE | MAP_ANONYMOUS) :
+				      MAP_SHARED) |
+				   MAP_HUGETLB | (populate ? MAP_POPULATE : 0) |
+				   (reserve ? 0 : MAP_NORESERVE),
+			   -1, 0);
+
+		if (ptr == MAP_FAILED)
+			err(1, "mmap");
+
+		printf("Returned address is %p\n", ptr);
+		break;
+	case SHM:
+		printf("Allocating using SHM.\n");
+		shmid = shmget(key, size,
+			       SHM_HUGETLB | IPC_CREAT | SHM_R | SHM_W);
+		if (shmid < 0) {
+			shmid = shmget(++key, size,
+				       SHM_HUGETLB | IPC_CREAT | SHM_R | SHM_W);
+			if (shmid < 0)
+				err(1, "shmget");
+		}
+		printf("shmid: 0x%x, shmget key:%d\n", shmid, key);
+
+		shmaddr = shmat(shmid, NULL, 0);
+		if (shmaddr == (char *)-1) {
+			perror("Shared memory attach failure");
+			shmctl(shmid, IPC_RMID, NULL);
+			exit(2);
+		}
+		printf("shmaddr: %p\n", shmaddr);
+
+		break;
+	default:
+		errno = EINVAL;
+		err(1, "Invalid method.");
+	}
+
+	if (write) {
+		printf("Writing to memory.\n");
+		if (method != SHM) {
+			memset(ptr, 1, size);
+		} else {
+			printf("Starting the writes:\n");
+			for (i = 0; i < size; i++) {
+				shmaddr[i] = (char)(i);
+				if (!(i % (1024 * 1024)))
+					printf(".");
+			}
+			printf("\n");
+
+			printf("Starting the Check...");
+			for (i = 0; i < size; i++)
+				if (shmaddr[i] != (char)i) {
+					printf("\nIndex %lu mismatched\n", i);
+					exit(3);
+				}
+			printf("Done.\n");
+		}
+	}
+
+	if (want_sleep) {
+		/* Signal to caller that we're done. */
+		printf("DONE\n");
+
+		/* Hold memory until external kill signal is delivered. */
+		while (1)
+			sleep(100);
+	}
+
+	switch (method == HUGETLBFS) {
+		close(fd);
+	}
+
+	return 0;
+}
--
2.24.1.735.g03f4e72817-goog

^ permalink raw reply related	[flat|nested] 27+ messages in thread

* [PATCH v9 8/8] hugetlb_cgroup: Add hugetlb_cgroup reservation docs
  2019-12-17 23:16 [PATCH v9 1/8] hugetlb_cgroup: Add hugetlb_cgroup reservation counter Mina Almasry
                   ` (5 preceding siblings ...)
  2019-12-17 23:16 ` [PATCH v9 7/8] hugetlb_cgroup: Add hugetlb_cgroup reservation tests Mina Almasry
@ 2019-12-17 23:16 ` Mina Almasry
  2020-01-14  0:54   ` David Rientjes
  2019-12-19  1:12 ` [PATCH v9 1/8] hugetlb_cgroup: Add hugetlb_cgroup reservation counter Andrew Morton
                   ` (2 subsequent siblings)
  9 siblings, 1 reply; 27+ messages in thread
From: Mina Almasry @ 2019-12-17 23:16 UTC (permalink / raw)
  To: mike.kravetz
  Cc: shuah, almasrymina, rientjes, shakeelb, gthelen, akpm,
	linux-kernel, linux-mm, linux-kselftest, cgroups, aneesh.kumar,
	mkoutny, Hillf Danton

Add docs for how to use hugetlb_cgroup reservations, and their behavior.

Signed-off-by: Mina Almasry <almasrymina@google.com>
Acked-by: Hillf Danton <hdanton@sina.com>

---

Changes in v6:
- Updated docs to reflect the new design based on a new counter that
tracks both reservations and faults.

---
 .../admin-guide/cgroup-v1/hugetlb.rst         | 64 +++++++++++++++----
 1 file changed, 53 insertions(+), 11 deletions(-)

diff --git a/Documentation/admin-guide/cgroup-v1/hugetlb.rst b/Documentation/admin-guide/cgroup-v1/hugetlb.rst
index a3902aa253a96..efb94e4db9d5a 100644
--- a/Documentation/admin-guide/cgroup-v1/hugetlb.rst
+++ b/Documentation/admin-guide/cgroup-v1/hugetlb.rst
@@ -2,13 +2,6 @@
 HugeTLB Controller
 ==================

-The HugeTLB controller allows to limit the HugeTLB usage per control group and
-enforces the controller limit during page fault. Since HugeTLB doesn't
-support page reclaim, enforcing the limit at page fault time implies that,
-the application will get SIGBUS signal if it tries to access HugeTLB pages
-beyond its limit. This requires the application to know beforehand how much
-HugeTLB pages it would require for its use.
-
 HugeTLB controller can be created by first mounting the cgroup filesystem.

 # mount -t cgroup -o hugetlb none /sys/fs/cgroup
@@ -28,10 +21,14 @@ process (bash) into it.

 Brief summary of control files::

- hugetlb.<hugepagesize>.limit_in_bytes     # set/show limit of "hugepagesize" hugetlb usage
- hugetlb.<hugepagesize>.max_usage_in_bytes # show max "hugepagesize" hugetlb  usage recorded
- hugetlb.<hugepagesize>.usage_in_bytes     # show current usage for "hugepagesize" hugetlb
- hugetlb.<hugepagesize>.failcnt		   # show the number of allocation failure due to HugeTLB limit
+ hugetlb.<hugepagesize>.reservation_limit_in_bytes     # set/show limit of "hugepagesize" hugetlb reservations
+ hugetlb.<hugepagesize>.reservation_max_usage_in_bytes # show max "hugepagesize" hugetlb reservations and no-reserve faults.
+ hugetlb.<hugepagesize>.reservation_usage_in_bytes     # show current reservations and no-reserve faults for "hugepagesize" hugetlb
+ hugetlb.<hugepagesize>.reservation_failcnt            # show the number of allocation failure due to HugeTLB reservation limit
+ hugetlb.<hugepagesize>.limit_in_bytes                 # set/show limit of "hugepagesize" hugetlb faults
+ hugetlb.<hugepagesize>.max_usage_in_bytes             # show max "hugepagesize" hugetlb  usage recorded
+ hugetlb.<hugepagesize>.usage_in_bytes                 # show current usage for "hugepagesize" hugetlb
+ hugetlb.<hugepagesize>.failcnt                        # show the number of allocation failure due to HugeTLB usage limit

 For a system supporting three hugepage sizes (64k, 32M and 1G), the control
 files include::
@@ -40,11 +37,56 @@ files include::
   hugetlb.1GB.max_usage_in_bytes
   hugetlb.1GB.usage_in_bytes
   hugetlb.1GB.failcnt
+  hugetlb.1GB.reservation_limit_in_bytes
+  hugetlb.1GB.reservation_max_usage_in_bytes
+  hugetlb.1GB.reservation_usage_in_bytes
+  hugetlb.1GB.reservation_failcnt
   hugetlb.64KB.limit_in_bytes
   hugetlb.64KB.max_usage_in_bytes
   hugetlb.64KB.usage_in_bytes
   hugetlb.64KB.failcnt
+  hugetlb.64KB.reservation_limit_in_bytes
+  hugetlb.64KB.reservation_max_usage_in_bytes
+  hugetlb.64KB.reservation_usage_in_bytes
+  hugetlb.64KB.reservation_failcnt
   hugetlb.32MB.limit_in_bytes
   hugetlb.32MB.max_usage_in_bytes
   hugetlb.32MB.usage_in_bytes
   hugetlb.32MB.failcnt
+  hugetlb.32MB.reservation_limit_in_bytes
+  hugetlb.32MB.reservation_max_usage_in_bytes
+  hugetlb.32MB.reservation_usage_in_bytes
+  hugetlb.32MB.reservation_failcnt
+
+
+1. Reservation limits
+
+The HugeTLB controller allows to limit the HugeTLB reservations per control
+group and enforces the controller limit at reservation time and at the fault of
+hugetlb memory for which no reservation exists. Reservation limits
+are superior to Page fault limits (see section 2), since Reservation limits are
+enforced at reservation time (on mmap or shget), and never causes the
+application to get SIGBUS signal if the memory was reserved before hand. For
+MAP_NORESERVE allocations, the reservation limit behaves the same as the fault
+limit, enforcing memory usage at fault time and causing the application to
+receive a SIGBUS if it's crossing its limit.
+
+2. Page fault limits
+
+The HugeTLB controller allows to limit the HugeTLB usage (page fault) per
+control group and enforces the controller limit during page fault. Since HugeTLB
+doesn't support page reclaim, enforcing the limit at page fault time implies
+that, the application will get SIGBUS signal if it tries to access HugeTLB
+pages beyond its limit. This requires the application to know beforehand how
+much HugeTLB pages it would require for its use.
+
+
+3. Caveats with shared memory
+
+For shared hugetlb memory, both hugetlb reservation and page faults are charged
+to the first task that causes the memory to be reserved or faulted, and all
+subsequent uses of this reserved or faulted memory is done without charging.
+
+Shared hugetlb memory is only uncharged when it is unreserved or deallocated.
+This is usually when the hugetlbfs file is deleted, and not when the task that
+caused the reservation or fault has exited.
--
2.24.1.735.g03f4e72817-goog

^ permalink raw reply related	[flat|nested] 27+ messages in thread

* Re: [PATCH v9 1/8] hugetlb_cgroup: Add hugetlb_cgroup reservation counter
  2019-12-17 23:16 [PATCH v9 1/8] hugetlb_cgroup: Add hugetlb_cgroup reservation counter Mina Almasry
                   ` (6 preceding siblings ...)
  2019-12-17 23:16 ` [PATCH v9 8/8] hugetlb_cgroup: Add hugetlb_cgroup reservation docs Mina Almasry
@ 2019-12-19  1:12 ` Andrew Morton
  2019-12-19  1:37   ` Mike Kravetz
  2020-01-13 18:43 ` Mike Kravetz
  2020-01-14  0:45 ` David Rientjes
  9 siblings, 1 reply; 27+ messages in thread
From: Andrew Morton @ 2019-12-19  1:12 UTC (permalink / raw)
  To: Mina Almasry
  Cc: mike.kravetz, shuah, rientjes, shakeelb, gthelen, linux-kernel,
	linux-mm, linux-kselftest, cgroups, aneesh.kumar, mkoutny,
	Hillf Danton, Giuseppe Scrivano, Tejun Heo

On Tue, 17 Dec 2019 15:16:08 -0800 Mina Almasry <almasrymina@google.com> wrote:

> These counters will track hugetlb reservations rather than hugetlb
> memory faulted in. This patch only adds the counter, following patches
> add the charging and uncharging of the counter.
> 
> This is patch 1 of an 8 patch series.
> 
> Problem:
> Currently tasks attempting to allocate more hugetlb memory than is available get
> a failure at mmap/shmget time. This is thanks to Hugetlbfs Reservations [1].
> However, if a task attempts to allocate hugetlb memory only more than its
> hugetlb_cgroup limit allows, the kernel will allow the mmap/shmget call,
> but will SIGBUS the task when it attempts to fault the memory in.
> 
> We have developers interested in using hugetlb_cgroups, and they have expressed
> dissatisfaction regarding this behavior. We'd like to improve this
> behavior such that tasks violating the hugetlb_cgroup limits get an error on
> mmap/shmget time, rather than getting SIGBUS'd when they try to fault
> the excess memory in.
> 
> The underlying problem is that today's hugetlb_cgroup accounting happens
> at hugetlb memory *fault* time, rather than at *reservation* time.
> Thus, enforcing the hugetlb_cgroup limit only happens at fault time, and
> the offending task gets SIGBUS'd.
> 
> Proposed Solution:
> A new page counter named hugetlb.xMB.reservation_[limit|usage]_in_bytes. This
> counter has slightly different semantics than
> hugetlb.xMB.[limit|usage]_in_bytes:
> 
> - While usage_in_bytes tracks all *faulted* hugetlb memory,
> reservation_usage_in_bytes tracks all *reserved* hugetlb memory and
> hugetlb memory faulted in without a prior reservation.
> 
> - If a task attempts to reserve more memory than limit_in_bytes allows,
> the kernel will allow it to do so. But if a task attempts to reserve
> more memory than reservation_limit_in_bytes, the kernel will fail this
> reservation.
> 
> This proposal is implemented in this patch series, with tests to verify
> functionality and show the usage. We also added cgroup-v2 support to
> hugetlb_cgroup so that the new use cases can be extended to v2.

This would make
http://lkml.kernel.org/r/20191216193831.540953-1-gscrivan@redhat.com
obsolete?


> Alternatives considered:
> 1. A new cgroup, instead of only a new page_counter attached to
>    the existing hugetlb_cgroup. Adding a new cgroup seemed like a lot of code
>    duplication with hugetlb_cgroup. Keeping hugetlb related page counters under
>    hugetlb_cgroup seemed cleaner as well.
> 
> 2. Instead of adding a new counter, we considered adding a sysctl that modifies
>    the behavior of hugetlb.xMB.[limit|usage]_in_bytes, to do accounting at
>    reservation time rather than fault time. Adding a new page_counter seems
>    better as userspace could, if it wants, choose to enforce different cgroups
>    differently: one via limit_in_bytes, and another via
>    reservation_limit_in_bytes. This could be very useful if you're
>    transitioning how hugetlb memory is partitioned on your system one
>    cgroup at a time, for example. Also, someone may find usage for both
>    limit_in_bytes and reservation_limit_in_bytes concurrently, and this
>    approach gives them the option to do so.
> 


^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH v9 1/8] hugetlb_cgroup: Add hugetlb_cgroup reservation counter
  2019-12-19  1:12 ` [PATCH v9 1/8] hugetlb_cgroup: Add hugetlb_cgroup reservation counter Andrew Morton
@ 2019-12-19  1:37   ` Mike Kravetz
  2019-12-19  1:59     ` Mina Almasry
  0 siblings, 1 reply; 27+ messages in thread
From: Mike Kravetz @ 2019-12-19  1:37 UTC (permalink / raw)
  To: Andrew Morton, Mina Almasry
  Cc: shuah, rientjes, shakeelb, gthelen, linux-kernel, linux-mm,
	linux-kselftest, cgroups, aneesh.kumar, mkoutny, Hillf Danton,
	Giuseppe Scrivano, Tejun Heo

On 12/18/19 5:12 PM, Andrew Morton wrote:
> On Tue, 17 Dec 2019 15:16:08 -0800 Mina Almasry <almasrymina@google.com> wrote:
<snip>
>> This proposal is implemented in this patch series, with tests to verify
>> functionality and show the usage. We also added cgroup-v2 support to
>> hugetlb_cgroup so that the new use cases can be extended to v2.
> 
> This would make
> http://lkml.kernel.org/r/20191216193831.540953-1-gscrivan@redhat.com
> obsolete?

I haven't started looking at this series yet.  However, since Mina was
involved in the discussion of that patch (hugetlb controller for cgroups v2)
my assumption is that this patch would simply build on that v2 support?
Seems like the above patch would be a prereq for this series.

Mina, are those assumptions correct and perhaps this is an old/obsolete
comment?  Does this series apply 'on top' of the above patch?  That patch
is already in Andrew's tree.
-- 
Mike Kravetz

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH v9 1/8] hugetlb_cgroup: Add hugetlb_cgroup reservation counter
  2019-12-19  1:37   ` Mike Kravetz
@ 2019-12-19  1:59     ` Mina Almasry
  0 siblings, 0 replies; 27+ messages in thread
From: Mina Almasry @ 2019-12-19  1:59 UTC (permalink / raw)
  To: Mike Kravetz
  Cc: Andrew Morton, shuah, David Rientjes, Shakeel Butt, Greg Thelen,
	open list, linux-mm, linux-kselftest, cgroups, Aneesh Kumar,
	Michal Koutný,
	Hillf Danton, Giuseppe Scrivano, Tejun Heo

On Wed, Dec 18, 2019 at 5:37 PM Mike Kravetz <mike.kravetz@oracle.com> wrote:
>
> On 12/18/19 5:12 PM, Andrew Morton wrote:
> > On Tue, 17 Dec 2019 15:16:08 -0800 Mina Almasry <almasrymina@google.com> wrote:
> <snip>
> >> This proposal is implemented in this patch series, with tests to verify
> >> functionality and show the usage. We also added cgroup-v2 support to
> >> hugetlb_cgroup so that the new use cases can be extended to v2.
> >
> > This would make
> > http://lkml.kernel.org/r/20191216193831.540953-1-gscrivan@redhat.com
> > obsolete?
>
> I haven't started looking at this series yet.  However, since Mina was
> involved in the discussion of that patch (hugetlb controller for cgroups v2)
> my assumption is that this patch would simply build on that v2 support?
> Seems like the above patch would be a prereq for this series.
>
> Mina, are those assumptions correct and perhaps this is an old/obsolete
> comment?  Does this series apply 'on top' of the above patch?  That patch
> is already in Andrew's tree.

Gah, I just forgot to update this commit message. This patch series is
indeed on top of Giuseppe's v2 support and doesn't conflict or
duplicate that functionality. Sorry for the confusion. I'll fix the
commit message in the next iteration.

> --
> Mike Kravetz

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH v9 1/8] hugetlb_cgroup: Add hugetlb_cgroup reservation counter
  2019-12-17 23:16 [PATCH v9 1/8] hugetlb_cgroup: Add hugetlb_cgroup reservation counter Mina Almasry
                   ` (7 preceding siblings ...)
  2019-12-19  1:12 ` [PATCH v9 1/8] hugetlb_cgroup: Add hugetlb_cgroup reservation counter Andrew Morton
@ 2020-01-13 18:43 ` Mike Kravetz
  2020-01-13 21:03   ` Mina Almasry
  2020-01-14  0:45 ` David Rientjes
  9 siblings, 1 reply; 27+ messages in thread
From: Mike Kravetz @ 2020-01-13 18:43 UTC (permalink / raw)
  To: Mina Almasry
  Cc: shuah, rientjes, shakeelb, gthelen, akpm, linux-kernel, linux-mm,
	linux-kselftest, cgroups, aneesh.kumar, mkoutny, Hillf Danton

On 12/17/19 3:16 PM, Mina Almasry wrote:
> These counters will track hugetlb reservations rather than hugetlb
> memory faulted in. This patch only adds the counter, following patches
> add the charging and uncharging of the counter.
> 
> This is patch 1 of an 8 patch series.
> 
> Problem:
> Currently tasks attempting to allocate more hugetlb memory than is available get
> a failure at mmap/shmget time. This is thanks to Hugetlbfs Reservations [1].
> However, if a task attempts to allocate hugetlb memory only more than its
> hugetlb_cgroup limit allows, the kernel will allow the mmap/shmget call,
> but will SIGBUS the task when it attempts to fault the memory in.
> 
> We have developers interested in using hugetlb_cgroups, and they have expressed
> dissatisfaction regarding this behavior. We'd like to improve this
> behavior such that tasks violating the hugetlb_cgroup limits get an error on
> mmap/shmget time, rather than getting SIGBUS'd when they try to fault
> the excess memory in.
> 
> The underlying problem is that today's hugetlb_cgroup accounting happens
> at hugetlb memory *fault* time, rather than at *reservation* time.
> Thus, enforcing the hugetlb_cgroup limit only happens at fault time, and
> the offending task gets SIGBUS'd.
> 
> Proposed Solution:
> A new page counter named hugetlb.xMB.reservation_[limit|usage]_in_bytes. This
> counter has slightly different semantics than
> hugetlb.xMB.[limit|usage]_in_bytes:
> 
> - While usage_in_bytes tracks all *faulted* hugetlb memory,
> reservation_usage_in_bytes tracks all *reserved* hugetlb memory and
> hugetlb memory faulted in without a prior reservation.

To me, this implies that 'faults without reservations' could cause
reservation usage to exceed reservation limit?  Or, does the faulting
process get a SIGBUS because of the reservation limit even though it
is not using reservations?

We shall see in subsequent patches.

> 
> - If a task attempts to reserve more memory than limit_in_bytes allows,
> the kernel will allow it to do so. But if a task attempts to reserve
> more memory than reservation_limit_in_bytes, the kernel will fail this
> reservation.
> 
> This proposal is implemented in this patch series, with tests to verify
> functionality and show the usage. We also added cgroup-v2 support to
> hugetlb_cgroup so that the new use cases can be extended to v2.

As previously discussed, cgroup-v2 support for hugetlb_cgroup will exist
before this patch series.

> 
> Alternatives considered:
> 1. A new cgroup, instead of only a new page_counter attached to
>    the existing hugetlb_cgroup. Adding a new cgroup seemed like a lot of code
>    duplication with hugetlb_cgroup. Keeping hugetlb related page counters under
>    hugetlb_cgroup seemed cleaner as well.
> 
> 2. Instead of adding a new counter, we considered adding a sysctl that modifies
>    the behavior of hugetlb.xMB.[limit|usage]_in_bytes, to do accounting at
>    reservation time rather than fault time. Adding a new page_counter seems
>    better as userspace could, if it wants, choose to enforce different cgroups
>    differently: one via limit_in_bytes, and another via
>    reservation_limit_in_bytes. This could be very useful if you're
>    transitioning how hugetlb memory is partitioned on your system one
>    cgroup at a time, for example. Also, someone may find usage for both
>    limit_in_bytes and reservation_limit_in_bytes concurrently, and this
>    approach gives them the option to do so.
> 
> Testing:
> - Added tests passing.
> - Used libhugetlbfs for regression testing.
> 
> [1]: https://www.kernel.org/doc/html/latest/vm/hugetlbfs_reserv.html
> 
> Signed-off-by: Mina Almasry <almasrymina@google.com>
> Acked-by: Hillf Danton <hdanton@sina.com>

I think the ACK by Hillf happened some time back.  You may want to check
to see if it still applies.

> 
> ---
>  include/linux/hugetlb.h |   4 +-
>  mm/hugetlb_cgroup.c     | 116 +++++++++++++++++++++++++++++++++++-----
>  2 files changed, 106 insertions(+), 14 deletions(-)

Only one minor nit in the code.

You made this cleanup,
@@ -472,7 +519,7 @@ static void __init __hugetlb_cgroup_file_dfl_init(int idx)
        struct hstate *h = &hstates[idx];

        /* format the size */
-       mem_fmt(buf, 32, huge_page_size(h));
+       mem_fmt(buf, sizeof(buf), huge_page_size(h));

        /* Add the limit file */
        cft = &h->cgroup_files_dfl[0];

But did not make the same cleanup in __hugetlb_cgroup_file_legacy_init()
-- 
Mike Kravetz

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH v9 1/8] hugetlb_cgroup: Add hugetlb_cgroup reservation counter
  2020-01-13 18:43 ` Mike Kravetz
@ 2020-01-13 21:03   ` Mina Almasry
  2020-01-13 22:05     ` Mike Kravetz
  0 siblings, 1 reply; 27+ messages in thread
From: Mina Almasry @ 2020-01-13 21:03 UTC (permalink / raw)
  To: Mike Kravetz
  Cc: shuah, David Rientjes, Shakeel Butt, Greg Thelen, Andrew Morton,
	open list, linux-mm, linux-kselftest, cgroups, Aneesh Kumar,
	Michal Koutný,
	Hillf Danton

On Mon, Jan 13, 2020 at 10:44 AM Mike Kravetz <mike.kravetz@oracle.com> wrote:
>
> On 12/17/19 3:16 PM, Mina Almasry wrote:
> > These counters will track hugetlb reservations rather than hugetlb
> > memory faulted in. This patch only adds the counter, following patches
> > add the charging and uncharging of the counter.
> >
> > This is patch 1 of an 8 patch series.
> >
> > Problem:
> > Currently tasks attempting to allocate more hugetlb memory than is available get
> > a failure at mmap/shmget time. This is thanks to Hugetlbfs Reservations [1].
> > However, if a task attempts to allocate hugetlb memory only more than its
> > hugetlb_cgroup limit allows, the kernel will allow the mmap/shmget call,
> > but will SIGBUS the task when it attempts to fault the memory in.
> >
> > We have developers interested in using hugetlb_cgroups, and they have expressed
> > dissatisfaction regarding this behavior. We'd like to improve this
> > behavior such that tasks violating the hugetlb_cgroup limits get an error on
> > mmap/shmget time, rather than getting SIGBUS'd when they try to fault
> > the excess memory in.
> >
> > The underlying problem is that today's hugetlb_cgroup accounting happens
> > at hugetlb memory *fault* time, rather than at *reservation* time.
> > Thus, enforcing the hugetlb_cgroup limit only happens at fault time, and
> > the offending task gets SIGBUS'd.
> >
> > Proposed Solution:
> > A new page counter named hugetlb.xMB.reservation_[limit|usage]_in_bytes. This
> > counter has slightly different semantics than
> > hugetlb.xMB.[limit|usage]_in_bytes:
> >
> > - While usage_in_bytes tracks all *faulted* hugetlb memory,
> > reservation_usage_in_bytes tracks all *reserved* hugetlb memory and
> > hugetlb memory faulted in without a prior reservation.
>
> To me, this implies that 'faults without reservations' could cause
> reservation usage to exceed reservation limit?  Or, does the faulting
> process get a SIGBUS because of the reservation limit even though it
> is not using reservations?
>
> We shall see in subsequent patches.
>

The design we went with based on previous discussions is as follows:
hugetlb pages faulted without a prior reservation get accounted at
fault time, rather than reservation time, and if the fault causes the
counter to cross the limit, the charge fails, hence the fault fails,
hence the process gets sigbus'd.

This means that one counter I'm adding here can cover both use cases:
if the userspace uses MAP_NORESERVE, then their memory is accounted at
fault time and they may get sigbus'd. If the userspace does *not* use
MAP_NORESERVE, then their memory is charged at reservation (mmap) time
and should be completely immune to runtime sigbus, unless they do
something really weird like reserve a bunch of memory, punch a hole in
the reserved memory, lower their limit, then fault in memory from the
hole.

> >
> > - If a task attempts to reserve more memory than limit_in_bytes allows,
> > the kernel will allow it to do so. But if a task attempts to reserve
> > more memory than reservation_limit_in_bytes, the kernel will fail this
> > reservation.
> >
> > This proposal is implemented in this patch series, with tests to verify
> > functionality and show the usage. We also added cgroup-v2 support to
> > hugetlb_cgroup so that the new use cases can be extended to v2.
>
> As previously discussed, cgroup-v2 support for hugetlb_cgroup will exist
> before this patch series.
>

Yes, this will be fixed in the next iteration.

> >
> > Alternatives considered:
> > 1. A new cgroup, instead of only a new page_counter attached to
> >    the existing hugetlb_cgroup. Adding a new cgroup seemed like a lot of code
> >    duplication with hugetlb_cgroup. Keeping hugetlb related page counters under
> >    hugetlb_cgroup seemed cleaner as well.
> >
> > 2. Instead of adding a new counter, we considered adding a sysctl that modifies
> >    the behavior of hugetlb.xMB.[limit|usage]_in_bytes, to do accounting at
> >    reservation time rather than fault time. Adding a new page_counter seems
> >    better as userspace could, if it wants, choose to enforce different cgroups
> >    differently: one via limit_in_bytes, and another via
> >    reservation_limit_in_bytes. This could be very useful if you're
> >    transitioning how hugetlb memory is partitioned on your system one
> >    cgroup at a time, for example. Also, someone may find usage for both
> >    limit_in_bytes and reservation_limit_in_bytes concurrently, and this
> >    approach gives them the option to do so.
> >
> > Testing:
> > - Added tests passing.
> > - Used libhugetlbfs for regression testing.
> >
> > [1]: https://www.kernel.org/doc/html/latest/vm/hugetlbfs_reserv.html
> >
> > Signed-off-by: Mina Almasry <almasrymina@google.com>
> > Acked-by: Hillf Danton <hdanton@sina.com>
>
> I think the ACK by Hillf happened some time back.  You may want to check
> to see if it still applies.
>

Yes, will remove the ack in the next iteration. I'll re-add it if Hilf
reviews again.

> >
> > ---
> >  include/linux/hugetlb.h |   4 +-
> >  mm/hugetlb_cgroup.c     | 116 +++++++++++++++++++++++++++++++++++-----
> >  2 files changed, 106 insertions(+), 14 deletions(-)
>
> Only one minor nit in the code.
>
> You made this cleanup,
> @@ -472,7 +519,7 @@ static void __init __hugetlb_cgroup_file_dfl_init(int idx)
>         struct hstate *h = &hstates[idx];
>
>         /* format the size */
> -       mem_fmt(buf, 32, huge_page_size(h));
> +       mem_fmt(buf, sizeof(buf), huge_page_size(h));
>
>         /* Add the limit file */
>         cft = &h->cgroup_files_dfl[0];
>
> But did not make the same cleanup in __hugetlb_cgroup_file_legacy_init()

Will be fixed in the next iteration.

> --
> Mike Kravetz

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH v9 1/8] hugetlb_cgroup: Add hugetlb_cgroup reservation counter
  2020-01-13 21:03   ` Mina Almasry
@ 2020-01-13 22:05     ` Mike Kravetz
  2020-01-13 22:21       ` Mina Almasry
  0 siblings, 1 reply; 27+ messages in thread
From: Mike Kravetz @ 2020-01-13 22:05 UTC (permalink / raw)
  To: Mina Almasry
  Cc: shuah, David Rientjes, Shakeel Butt, Greg Thelen, Andrew Morton,
	open list, linux-mm, linux-kselftest, cgroups, Aneesh Kumar,
	Michal Koutný,
	Hillf Danton

On 1/13/20 1:03 PM, Mina Almasry wrote:
> On Mon, Jan 13, 2020 at 10:44 AM Mike Kravetz <mike.kravetz@oracle.com> wrote:
>>
>> On 12/17/19 3:16 PM, Mina Almasry wrote:
>>> - While usage_in_bytes tracks all *faulted* hugetlb memory,
>>> reservation_usage_in_bytes tracks all *reserved* hugetlb memory and
>>> hugetlb memory faulted in without a prior reservation.
>>
>> To me, this implies that 'faults without reservations' could cause
>> reservation usage to exceed reservation limit?  Or, does the faulting
>> process get a SIGBUS because of the reservation limit even though it
>> is not using reservations?
>>
>> We shall see in subsequent patches.
>>
> 
> The design we went with based on previous discussions is as follows:
> hugetlb pages faulted without a prior reservation get accounted at
> fault time, rather than reservation time, and if the fault causes the
> counter to cross the limit, the charge fails, hence the fault fails,
> hence the process gets sigbus'd.

Ok, sorry I did not recall the design discussion.

> This means that one counter I'm adding here can cover both use cases:
> if the userspace uses MAP_NORESERVE, then their memory is accounted at
> fault time and they may get sigbus'd.

Let's make sure this is clearly documented.  Someone could be surprised
if their application not using reserves gets a SIGBUS because there is a
reserve limit.
-- 
Mike Kravetz

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH v9 2/8] hugetlb_cgroup: add interface for charge/uncharge hugetlb reservations
  2019-12-17 23:16 ` [PATCH v9 2/8] hugetlb_cgroup: add interface for charge/uncharge hugetlb reservations Mina Almasry
@ 2020-01-13 22:14   ` Mike Kravetz
  2020-01-14  0:45     ` David Rientjes
  2020-01-14 22:55     ` Mina Almasry
  0 siblings, 2 replies; 27+ messages in thread
From: Mike Kravetz @ 2020-01-13 22:14 UTC (permalink / raw)
  To: Mina Almasry
  Cc: shuah, rientjes, shakeelb, gthelen, akpm, linux-kernel, linux-mm,
	linux-kselftest, cgroups, aneesh.kumar, mkoutny

On 12/17/19 3:16 PM, Mina Almasry wrote:
> Augments hugetlb_cgroup_charge_cgroup to be able to charge hugetlb
> usage or hugetlb reservation counter.
> 
> Adds a new interface to uncharge a hugetlb_cgroup counter via
> hugetlb_cgroup_uncharge_counter.
> 
> Integrates the counter with hugetlb_cgroup, via hugetlb_cgroup_init,
> hugetlb_cgroup_have_usage, and hugetlb_cgroup_css_offline.
> 
> Signed-off-by: Mina Almasry <almasrymina@google.com>
> 
> ---
> 
> Changes in V9:
> - Fixed HUGETLB_CGROUP_MIN_ORDER.
> - Minor variable name update.
> - Moved some init/cleanup code from later patches in the series to this patch.
> - Updated reparenting of reservation accounting.
> 
> ---
>  include/linux/hugetlb_cgroup.h | 68 ++++++++++++++---------
>  mm/hugetlb.c                   | 19 ++++---
>  mm/hugetlb_cgroup.c            | 98 +++++++++++++++++++++++++---------
>  3 files changed, 127 insertions(+), 58 deletions(-)
> 
> diff --git a/include/linux/hugetlb_cgroup.h b/include/linux/hugetlb_cgroup.h
> index 063962f6dfc6a..eab8a70d5bcb5 100644
> --- a/include/linux/hugetlb_cgroup.h
> +++ b/include/linux/hugetlb_cgroup.h
> @@ -20,29 +20,37 @@
>  struct hugetlb_cgroup;
>  /*
>   * Minimum page order trackable by hugetlb cgroup.
> - * At least 3 pages are necessary for all the tracking information.
> + * At least 4 pages are necessary for all the tracking information.
>   */
>  #define HUGETLB_CGROUP_MIN_ORDER	2
> 
>  #ifdef CONFIG_CGROUP_HUGETLB
> 
> -static inline struct hugetlb_cgroup *hugetlb_cgroup_from_page(struct page *page)
> +static inline struct hugetlb_cgroup *hugetlb_cgroup_from_page(struct page *page,
> +							      bool reserved)
>  {
>  	VM_BUG_ON_PAGE(!PageHuge(page), page);
> 
>  	if (compound_order(page) < HUGETLB_CGROUP_MIN_ORDER)
>  		return NULL;
> -	return (struct hugetlb_cgroup *)page[2].private;
> +	if (reserved)
> +		return (struct hugetlb_cgroup *)page[3].private;
> +	else
> +		return (struct hugetlb_cgroup *)page[2].private;
>  }
> 
> -static inline
> -int set_hugetlb_cgroup(struct page *page, struct hugetlb_cgroup *h_cg)
> +static inline int set_hugetlb_cgroup(struct page *page,
> +				     struct hugetlb_cgroup *h_cg,
> +				     bool reservation)
>  {
>  	VM_BUG_ON_PAGE(!PageHuge(page), page);
> 
>  	if (compound_order(page) < HUGETLB_CGROUP_MIN_ORDER)
>  		return -1;
> -	page[2].private	= (unsigned long)h_cg;
> +	if (reservation)
> +		page[3].private = (unsigned long)h_cg;
> +	else
> +		page[2].private = (unsigned long)h_cg;
>  	return 0;
>  }
> 
> @@ -52,26 +60,34 @@ static inline bool hugetlb_cgroup_disabled(void)
>  }
> 
>  extern int hugetlb_cgroup_charge_cgroup(int idx, unsigned long nr_pages,
> -					struct hugetlb_cgroup **ptr);
> +					struct hugetlb_cgroup **ptr,
> +					bool reserved);
>  extern void hugetlb_cgroup_commit_charge(int idx, unsigned long nr_pages,
>  					 struct hugetlb_cgroup *h_cg,
> -					 struct page *page);
> +					 struct page *page, bool reserved);
>  extern void hugetlb_cgroup_uncharge_page(int idx, unsigned long nr_pages,
> -					 struct page *page);
> +					 struct page *page, bool reserved);
> +
>  extern void hugetlb_cgroup_uncharge_cgroup(int idx, unsigned long nr_pages,
> -					   struct hugetlb_cgroup *h_cg);
> +					   struct hugetlb_cgroup *h_cg,
> +					   bool reserved);
> +extern void hugetlb_cgroup_uncharge_counter(struct page_counter *p,
> +					    unsigned long nr_pages,
> +					    struct cgroup_subsys_state *css);
> +

Do we need a corresponding stub for the !CONFIG_CGROUP_HUGETLB case?

>  extern void hugetlb_cgroup_file_init(void) __init;
>  extern void hugetlb_cgroup_migrate(struct page *oldhpage,
>  				   struct page *newhpage);
> 
>  #else
> -static inline struct hugetlb_cgroup *hugetlb_cgroup_from_page(struct page *page)
> +static inline struct hugetlb_cgroup *hugetlb_cgroup_from_page(struct page *page,
> +							      bool reserved)
>  {
>  	return NULL;
>  }
> 
> -static inline
> -int set_hugetlb_cgroup(struct page *page, struct hugetlb_cgroup *h_cg)
> +static inline int set_hugetlb_cgroup(struct page *page,
> +				     struct hugetlb_cgroup *h_cg, bool reserved)
>  {
>  	return 0;
>  }
> @@ -81,28 +97,30 @@ static inline bool hugetlb_cgroup_disabled(void)
>  	return true;
>  }
> 
> -static inline int
> -hugetlb_cgroup_charge_cgroup(int idx, unsigned long nr_pages,
> -			     struct hugetlb_cgroup **ptr)
> +static inline int hugetlb_cgroup_charge_cgroup(int idx, unsigned long nr_pages,
> +					       struct hugetlb_cgroup **ptr,
> +					       bool reserved)
>  {
>  	return 0;
>  }
> 
> -static inline void
> -hugetlb_cgroup_commit_charge(int idx, unsigned long nr_pages,
> -			     struct hugetlb_cgroup *h_cg,
> -			     struct page *page)
> +static inline void hugetlb_cgroup_commit_charge(int idx, unsigned long nr_pages,
> +						struct hugetlb_cgroup *h_cg,
> +						struct page *page,
> +						bool reserved)
>  {
>  }
> 
> -static inline void
> -hugetlb_cgroup_uncharge_page(int idx, unsigned long nr_pages, struct page *page)
> +static inline void hugetlb_cgroup_uncharge_page(int idx, unsigned long nr_pages,
> +						struct page *page,
> +						bool reserved)
>  {
>  }
> 
> -static inline void
> -hugetlb_cgroup_uncharge_cgroup(int idx, unsigned long nr_pages,
> -			       struct hugetlb_cgroup *h_cg)
> +static inline void hugetlb_cgroup_uncharge_cgroup(int idx,
> +						  unsigned long nr_pages,
> +						  struct hugetlb_cgroup *h_cg,
> +						  bool reserved)
>  {
>  }
> 
> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> index ac65bb5e38ac2..e6e8240f1718c 100644
> --- a/mm/hugetlb.c
> +++ b/mm/hugetlb.c
> @@ -1067,7 +1067,8 @@ static void update_and_free_page(struct hstate *h, struct page *page)
>  				1 << PG_active | 1 << PG_private |
>  				1 << PG_writeback);
>  	}
> -	VM_BUG_ON_PAGE(hugetlb_cgroup_from_page(page), page);
> +	VM_BUG_ON_PAGE(hugetlb_cgroup_from_page(page, false), page);
> +	VM_BUG_ON_PAGE(hugetlb_cgroup_from_page(page, true), page);
>  	set_compound_page_dtor(page, NULL_COMPOUND_DTOR);
>  	set_page_refcounted(page);
>  	if (hstate_is_gigantic(h)) {
> @@ -1177,8 +1178,8 @@ void free_huge_page(struct page *page)
> 
>  	spin_lock(&hugetlb_lock);
>  	clear_page_huge_active(page);
> -	hugetlb_cgroup_uncharge_page(hstate_index(h),
> -				     pages_per_huge_page(h), page);
> +	hugetlb_cgroup_uncharge_page(hstate_index(h), pages_per_huge_page(h),
> +				     page, false);

hugetlb code has been restructured a bit and this now resides in the
function __free_huge_page().

>  	if (restore_reserve)
>  		h->resv_huge_pages++;
> 
> @@ -1204,7 +1205,8 @@ static void prep_new_huge_page(struct hstate *h, struct page *page, int nid)
>  	INIT_LIST_HEAD(&page->lru);
>  	set_compound_page_dtor(page, HUGETLB_PAGE_DTOR);
>  	spin_lock(&hugetlb_lock);
> -	set_hugetlb_cgroup(page, NULL);
> +	set_hugetlb_cgroup(page, NULL, false);
> +	set_hugetlb_cgroup(page, NULL, true);
>  	h->nr_huge_pages++;
>  	h->nr_huge_pages_node[nid]++;
>  	spin_unlock(&hugetlb_lock);
> @@ -1990,7 +1992,8 @@ struct page *alloc_huge_page(struct vm_area_struct *vma,
>  			gbl_chg = 1;
>  	}
> 
> -	ret = hugetlb_cgroup_charge_cgroup(idx, pages_per_huge_page(h), &h_cg);
> +	ret = hugetlb_cgroup_charge_cgroup(idx, pages_per_huge_page(h), &h_cg,
> +					   false);
>  	if (ret)
>  		goto out_subpool_put;
> 
> @@ -2014,7 +2017,8 @@ struct page *alloc_huge_page(struct vm_area_struct *vma,
>  		list_move(&page->lru, &h->hugepage_activelist);
>  		/* Fall through */
>  	}
> -	hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h), h_cg, page);
> +	hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h), h_cg, page,
> +				     false);
>  	spin_unlock(&hugetlb_lock);
> 
>  	set_page_private(page, (unsigned long)spool);
> @@ -2038,7 +2042,8 @@ struct page *alloc_huge_page(struct vm_area_struct *vma,
>  	return page;
> 
>  out_uncharge_cgroup:
> -	hugetlb_cgroup_uncharge_cgroup(idx, pages_per_huge_page(h), h_cg);
> +	hugetlb_cgroup_uncharge_cgroup(idx, pages_per_huge_page(h), h_cg,
> +				       false);
>  out_subpool_put:
>  	if (map_chg || avoid_reserve)
>  		hugepage_subpool_put_pages(spool, 1);
> diff --git a/mm/hugetlb_cgroup.c b/mm/hugetlb_cgroup.c
> index 35415af9ed26f..b03270b0d5833 100644
> --- a/mm/hugetlb_cgroup.c
> +++ b/mm/hugetlb_cgroup.c
> @@ -96,8 +96,12 @@ static inline bool hugetlb_cgroup_have_usage(struct hugetlb_cgroup *h_cg)
>  	int idx;
> 
>  	for (idx = 0; idx < hugetlb_max_hstate; idx++) {
> -		if (page_counter_read(&h_cg->hugepage[idx]))
> +		if (page_counter_read(
> +			    hugetlb_cgroup_get_counter(h_cg, idx, true)) ||
> +		    page_counter_read(
> +			    hugetlb_cgroup_get_counter(h_cg, idx, false))) {
>  			return true;
> +		}
>  	}
>  	return false;
>  }
> @@ -108,18 +112,32 @@ static void hugetlb_cgroup_init(struct hugetlb_cgroup *h_cgroup,
>  	int idx;
> 
>  	for (idx = 0; idx < HUGE_MAX_HSTATE; idx++) {
> -		struct page_counter *counter = &h_cgroup->hugepage[idx];
> -		struct page_counter *parent = NULL;
> +		struct page_counter *fault_parent = NULL;
> +		struct page_counter *reserved_parent = NULL;
>  		unsigned long limit;
>  		int ret;
> 
> -		if (parent_h_cgroup)
> -			parent = &parent_h_cgroup->hugepage[idx];
> -		page_counter_init(counter, parent);
> +		if (parent_h_cgroup) {
> +			fault_parent = hugetlb_cgroup_get_counter(
> +				parent_h_cgroup, idx, false);
> +			reserved_parent = hugetlb_cgroup_get_counter(
> +				parent_h_cgroup, idx, true);
> +		}
> +		page_counter_init(hugetlb_cgroup_get_counter(h_cgroup, idx,
> +							     false),
> +				  fault_parent);
> +		page_counter_init(hugetlb_cgroup_get_counter(h_cgroup, idx,
> +							     true),
> +				  reserved_parent);
> 
>  		limit = round_down(PAGE_COUNTER_MAX,
>  				   1 << huge_page_order(&hstates[idx]));
> -		ret = page_counter_set_max(counter, limit);
> +
> +		ret = page_counter_set_max(
> +			hugetlb_cgroup_get_counter(h_cgroup, idx, false),
> +			limit);
> +		ret = page_counter_set_max(
> +			hugetlb_cgroup_get_counter(h_cgroup, idx, true), limit);
>  		VM_BUG_ON(ret);

The second page_counter_set_max() call overwrites ret before the check in
VM_BUG_ON().

>  	}
>  }
> @@ -149,7 +167,6 @@ static void hugetlb_cgroup_css_free(struct cgroup_subsys_state *css)
>  	kfree(h_cgroup);
>  }
> 
> -
>  /*
>   * Should be called with hugetlb_lock held.
>   * Since we are holding hugetlb_lock, pages cannot get moved from
> @@ -165,7 +182,7 @@ static void hugetlb_cgroup_move_parent(int idx, struct hugetlb_cgroup *h_cg,
>  	struct hugetlb_cgroup *page_hcg;
>  	struct hugetlb_cgroup *parent = parent_hugetlb_cgroup(h_cg);
> 
> -	page_hcg = hugetlb_cgroup_from_page(page);
> +	page_hcg = hugetlb_cgroup_from_page(page, false);
>  	/*
>  	 * We can have pages in active list without any cgroup
>  	 * ie, hugepage with less than 3 pages. We can safely
> @@ -184,7 +201,7 @@ static void hugetlb_cgroup_move_parent(int idx, struct hugetlb_cgroup *h_cg,
>  	/* Take the pages off the local counter */
>  	page_counter_cancel(counter, nr_pages);
> 
> -	set_hugetlb_cgroup(page, parent);
> +	set_hugetlb_cgroup(page, parent, false);
>  out:
>  	return;
>  }
> @@ -227,7 +244,7 @@ static inline void hugetlb_event(struct hugetlb_cgroup *hugetlb, int idx,
>  }
> 
>  int hugetlb_cgroup_charge_cgroup(int idx, unsigned long nr_pages,
> -				 struct hugetlb_cgroup **ptr)
> +				 struct hugetlb_cgroup **ptr, bool reserved)
>  {
>  	int ret = 0;
>  	struct page_counter *counter;
> @@ -250,13 +267,20 @@ int hugetlb_cgroup_charge_cgroup(int idx, unsigned long nr_pages,
>  	}
>  	rcu_read_unlock();
> 
> -	if (!page_counter_try_charge(&h_cg->hugepage[idx], nr_pages,
> -				     &counter)) {
> +	if (!page_counter_try_charge(hugetlb_cgroup_get_counter(h_cg, idx,
> +								reserved),
> +				     nr_pages, &counter)) {
>  		ret = -ENOMEM;
>  		hugetlb_event(hugetlb_cgroup_from_counter(counter, idx), idx,
>  			      HUGETLB_MAX);
> +		css_put(&h_cg->css);
> +		goto done;
>  	}
> -	css_put(&h_cg->css);
> +	/* Reservations take a reference to the css because they do not get
> +	 * reparented.

I'm hoping someone with more cgroup knowledge can comment on this and any
consequences of not reparenting reservations.  We previously talked about
why reparenting would be very difficult/expensive.  I understand why you are
nopt doing it.  Just do not fully understand what needs to be done from the
cgroup side.

Several of the following modifications deal with this.
--
Mike kravetz

> +	 */
> +	if (!reserved)
> +		css_put(&h_cg->css);
>  done:
>  	*ptr = h_cg;
>  	return ret;
> @@ -265,12 +289,12 @@ int hugetlb_cgroup_charge_cgroup(int idx, unsigned long nr_pages,
>  /* 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)
> +				  struct page *page, bool reserved)
>  {
>  	if (hugetlb_cgroup_disabled() || !h_cg)
>  		return;
> 
> -	set_hugetlb_cgroup(page, h_cg);
> +	set_hugetlb_cgroup(page, h_cg, reserved);
>  	return;
>  }
> 
> @@ -278,23 +302,29 @@ void hugetlb_cgroup_commit_charge(int idx, unsigned long nr_pages,
>   * Should be called with hugetlb_lock held
>   */
>  void hugetlb_cgroup_uncharge_page(int idx, unsigned long nr_pages,
> -				  struct page *page)
> +				  struct page *page, bool reserved)
>  {
>  	struct hugetlb_cgroup *h_cg;
> 
>  	if (hugetlb_cgroup_disabled())
>  		return;
>  	lockdep_assert_held(&hugetlb_lock);
> -	h_cg = hugetlb_cgroup_from_page(page);
> +	h_cg = hugetlb_cgroup_from_page(page, reserved);
>  	if (unlikely(!h_cg))
>  		return;
> -	set_hugetlb_cgroup(page, NULL);
> -	page_counter_uncharge(&h_cg->hugepage[idx], nr_pages);
> +	set_hugetlb_cgroup(page, NULL, reserved);
> +
> +	page_counter_uncharge(hugetlb_cgroup_get_counter(h_cg, idx, reserved),
> +			      nr_pages);
> +
> +	if (reserved)
> +		css_put(&h_cg->css);
> +
>  	return;
>  }
> 
>  void hugetlb_cgroup_uncharge_cgroup(int idx, unsigned long nr_pages,
> -				    struct hugetlb_cgroup *h_cg)
> +				    struct hugetlb_cgroup *h_cg, bool reserved)
>  {
>  	if (hugetlb_cgroup_disabled() || !h_cg)
>  		return;
> @@ -302,8 +332,22 @@ void hugetlb_cgroup_uncharge_cgroup(int idx, unsigned long nr_pages,
>  	if (huge_page_order(&hstates[idx]) < HUGETLB_CGROUP_MIN_ORDER)
>  		return;
> 
> -	page_counter_uncharge(&h_cg->hugepage[idx], nr_pages);
> -	return;
> +	page_counter_uncharge(hugetlb_cgroup_get_counter(h_cg, idx, reserved),
> +			      nr_pages);
> +
> +	if (reserved)
> +		css_put(&h_cg->css);
> +}
> +
> +void hugetlb_cgroup_uncharge_counter(struct page_counter *p,
> +				     unsigned long nr_pages,
> +				     struct cgroup_subsys_state *css)
> +{
> +	if (hugetlb_cgroup_disabled() || !p || !css)
> +		return;
> +
> +	page_counter_uncharge(p, nr_pages);
> +	css_put(css);
>  }
> 
>  enum {
> @@ -678,6 +722,7 @@ void __init hugetlb_cgroup_file_init(void)
>  void hugetlb_cgroup_migrate(struct page *oldhpage, struct page *newhpage)
>  {
>  	struct hugetlb_cgroup *h_cg;
> +	struct hugetlb_cgroup *h_cg_reservation;
>  	struct hstate *h = page_hstate(oldhpage);
> 
>  	if (hugetlb_cgroup_disabled())
> @@ -685,11 +730,12 @@ void hugetlb_cgroup_migrate(struct page *oldhpage, struct page *newhpage)
> 
>  	VM_BUG_ON_PAGE(!PageHuge(oldhpage), oldhpage);
>  	spin_lock(&hugetlb_lock);
> -	h_cg = hugetlb_cgroup_from_page(oldhpage);
> -	set_hugetlb_cgroup(oldhpage, NULL);
> +	h_cg = hugetlb_cgroup_from_page(oldhpage, false);
> +	h_cg_reservation = hugetlb_cgroup_from_page(oldhpage, true);
> +	set_hugetlb_cgroup(oldhpage, NULL, false);
> 
>  	/* move the h_cg details to new cgroup */
> -	set_hugetlb_cgroup(newhpage, h_cg);
> +	set_hugetlb_cgroup(newhpage, h_cg_reservation, true);
>  	list_move(&newhpage->lru, &h->hugepage_activelist);
>  	spin_unlock(&hugetlb_lock);
>  	return;
> --
> 2.24.1.735.g03f4e72817-goog
> 

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH v9 1/8] hugetlb_cgroup: Add hugetlb_cgroup reservation counter
  2020-01-13 22:05     ` Mike Kravetz
@ 2020-01-13 22:21       ` Mina Almasry
  0 siblings, 0 replies; 27+ messages in thread
From: Mina Almasry @ 2020-01-13 22:21 UTC (permalink / raw)
  To: Mike Kravetz
  Cc: shuah, David Rientjes, Shakeel Butt, Greg Thelen, Andrew Morton,
	open list, linux-mm, linux-kselftest, cgroups, Aneesh Kumar,
	Michal Koutný,
	Hillf Danton

> >> On 12/17/19 3:16 PM, Mina Almasry wrote:
> >
> > The design we went with based on previous discussions is as follows:
> > hugetlb pages faulted without a prior reservation get accounted at
> > fault time, rather than reservation time, and if the fault causes the
> > counter to cross the limit, the charge fails, hence the fault fails,
> > hence the process gets sigbus'd.
>
> Ok, sorry I did not recall the design discussion.
>

No worries! It has indeed been a while since that discussion.

> > This means that one counter I'm adding here can cover both use cases:
> > if the userspace uses MAP_NORESERVE, then their memory is accounted at
> > fault time and they may get sigbus'd.
>
> Let's make sure this is clearly documented.  Someone could be surprised
> if their application not using reserves gets a SIGBUS because there is a
> reserve limit.

I have some stuff on that already in the docs patch, but I'll beef
that section up to ensure there is no confusion.

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH v9 2/8] hugetlb_cgroup: add interface for charge/uncharge hugetlb reservations
  2020-01-13 22:14   ` Mike Kravetz
@ 2020-01-14  0:45     ` David Rientjes
  2020-01-14 22:55     ` Mina Almasry
  1 sibling, 0 replies; 27+ messages in thread
From: David Rientjes @ 2020-01-14  0:45 UTC (permalink / raw)
  To: Mike Kravetz
  Cc: Mina Almasry, shuah, shakeelb, gthelen, akpm, linux-kernel,
	linux-mm, linux-kselftest, cgroups, aneesh.kumar, mkoutny

On Mon, 13 Jan 2020, Mike Kravetz wrote:

> > diff --git a/mm/hugetlb_cgroup.c b/mm/hugetlb_cgroup.c
> > index 35415af9ed26f..b03270b0d5833 100644
> > --- a/mm/hugetlb_cgroup.c
> > +++ b/mm/hugetlb_cgroup.c
> > @@ -96,8 +96,12 @@ static inline bool hugetlb_cgroup_have_usage(struct hugetlb_cgroup *h_cg)
> >  	int idx;
> > 
> >  	for (idx = 0; idx < hugetlb_max_hstate; idx++) {
> > -		if (page_counter_read(&h_cg->hugepage[idx]))
> > +		if (page_counter_read(
> > +			    hugetlb_cgroup_get_counter(h_cg, idx, true)) ||
> > +		    page_counter_read(
> > +			    hugetlb_cgroup_get_counter(h_cg, idx, false))) {
> >  			return true;
> > +		}
> >  	}
> >  	return false;
> >  }
> > @@ -108,18 +112,32 @@ static void hugetlb_cgroup_init(struct hugetlb_cgroup *h_cgroup,
> >  	int idx;
> > 
> >  	for (idx = 0; idx < HUGE_MAX_HSTATE; idx++) {
> > -		struct page_counter *counter = &h_cgroup->hugepage[idx];
> > -		struct page_counter *parent = NULL;
> > +		struct page_counter *fault_parent = NULL;
> > +		struct page_counter *reserved_parent = NULL;
> >  		unsigned long limit;
> >  		int ret;
> > 
> > -		if (parent_h_cgroup)
> > -			parent = &parent_h_cgroup->hugepage[idx];
> > -		page_counter_init(counter, parent);
> > +		if (parent_h_cgroup) {
> > +			fault_parent = hugetlb_cgroup_get_counter(
> > +				parent_h_cgroup, idx, false);
> > +			reserved_parent = hugetlb_cgroup_get_counter(
> > +				parent_h_cgroup, idx, true);
> > +		}
> > +		page_counter_init(hugetlb_cgroup_get_counter(h_cgroup, idx,
> > +							     false),
> > +				  fault_parent);
> > +		page_counter_init(hugetlb_cgroup_get_counter(h_cgroup, idx,
> > +							     true),
> > +				  reserved_parent);
> > 
> >  		limit = round_down(PAGE_COUNTER_MAX,
> >  				   1 << huge_page_order(&hstates[idx]));
> > -		ret = page_counter_set_max(counter, limit);
> > +
> > +		ret = page_counter_set_max(
> > +			hugetlb_cgroup_get_counter(h_cgroup, idx, false),
> > +			limit);
> > +		ret = page_counter_set_max(
> > +			hugetlb_cgroup_get_counter(h_cgroup, idx, true), limit);
> >  		VM_BUG_ON(ret);
> 
> The second page_counter_set_max() call overwrites ret before the check in
> VM_BUG_ON().
> 
> >  	}
> >  }
> > @@ -149,7 +167,6 @@ static void hugetlb_cgroup_css_free(struct cgroup_subsys_state *css)
> >  	kfree(h_cgroup);
> >  }
> > 
> > -
> >  /*
> >   * Should be called with hugetlb_lock held.
> >   * Since we are holding hugetlb_lock, pages cannot get moved from
> > @@ -165,7 +182,7 @@ static void hugetlb_cgroup_move_parent(int idx, struct hugetlb_cgroup *h_cg,
> >  	struct hugetlb_cgroup *page_hcg;
> >  	struct hugetlb_cgroup *parent = parent_hugetlb_cgroup(h_cg);
> > 
> > -	page_hcg = hugetlb_cgroup_from_page(page);
> > +	page_hcg = hugetlb_cgroup_from_page(page, false);
> >  	/*
> >  	 * We can have pages in active list without any cgroup
> >  	 * ie, hugepage with less than 3 pages. We can safely
> > @@ -184,7 +201,7 @@ static void hugetlb_cgroup_move_parent(int idx, struct hugetlb_cgroup *h_cg,
> >  	/* Take the pages off the local counter */
> >  	page_counter_cancel(counter, nr_pages);
> > 
> > -	set_hugetlb_cgroup(page, parent);
> > +	set_hugetlb_cgroup(page, parent, false);
> >  out:
> >  	return;
> >  }
> > @@ -227,7 +244,7 @@ static inline void hugetlb_event(struct hugetlb_cgroup *hugetlb, int idx,
> >  }
> > 
> >  int hugetlb_cgroup_charge_cgroup(int idx, unsigned long nr_pages,
> > -				 struct hugetlb_cgroup **ptr)
> > +				 struct hugetlb_cgroup **ptr, bool reserved)
> >  {
> >  	int ret = 0;
> >  	struct page_counter *counter;
> > @@ -250,13 +267,20 @@ int hugetlb_cgroup_charge_cgroup(int idx, unsigned long nr_pages,
> >  	}
> >  	rcu_read_unlock();
> > 
> > -	if (!page_counter_try_charge(&h_cg->hugepage[idx], nr_pages,
> > -				     &counter)) {
> > +	if (!page_counter_try_charge(hugetlb_cgroup_get_counter(h_cg, idx,
> > +								reserved),
> > +				     nr_pages, &counter)) {
> >  		ret = -ENOMEM;
> >  		hugetlb_event(hugetlb_cgroup_from_counter(counter, idx), idx,
> >  			      HUGETLB_MAX);
> > +		css_put(&h_cg->css);
> > +		goto done;
> >  	}
> > -	css_put(&h_cg->css);
> > +	/* Reservations take a reference to the css because they do not get
> > +	 * reparented.
> 
> I'm hoping someone with more cgroup knowledge can comment on this and any
> consequences of not reparenting reservations.  We previously talked about
> why reparenting would be very difficult/expensive.  I understand why you are
> nopt doing it.  Just do not fully understand what needs to be done from the
> cgroup side.
> 

I don't see any description of how hugetlb_cgroup currently acts wrt 
reparenting in the last patch in the series and how this is the same or 
different for reservations.  I think the discussion that is referenced 
here is probably lost in some previous posting of the series.  I think 
it's particularly useful information that the end user will need to know 
about for its handling so it would benefit from some documentation in the 
last patch.

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH v9 1/8] hugetlb_cgroup: Add hugetlb_cgroup reservation counter
  2019-12-17 23:16 [PATCH v9 1/8] hugetlb_cgroup: Add hugetlb_cgroup reservation counter Mina Almasry
                   ` (8 preceding siblings ...)
  2020-01-13 18:43 ` Mike Kravetz
@ 2020-01-14  0:45 ` David Rientjes
  9 siblings, 0 replies; 27+ messages in thread
From: David Rientjes @ 2020-01-14  0:45 UTC (permalink / raw)
  To: Mina Almasry
  Cc: mike.kravetz, shuah, shakeelb, gthelen, akpm, linux-kernel,
	linux-mm, linux-kselftest, cgroups, aneesh.kumar, mkoutny,
	Hillf Danton

On Tue, 17 Dec 2019, Mina Almasry wrote:

> These counters will track hugetlb reservations rather than hugetlb
> memory faulted in. This patch only adds the counter, following patches
> add the charging and uncharging of the counter.
> 
> This is patch 1 of an 8 patch series.
> 
> Problem:
> Currently tasks attempting to allocate more hugetlb memory than is available get
> a failure at mmap/shmget time. This is thanks to Hugetlbfs Reservations [1].
> However, if a task attempts to allocate hugetlb memory only more than its
> hugetlb_cgroup limit allows, the kernel will allow the mmap/shmget call,
> but will SIGBUS the task when it attempts to fault the memory in.
> 

I think it's subtle, but the use of the word "allocate" instead of using 
"reserve" might be confusing here.  Might want to reword it.

> We have developers interested in using hugetlb_cgroups, and they have expressed
> dissatisfaction regarding this behavior. We'd like to improve this
> behavior such that tasks violating the hugetlb_cgroup limits get an error on
> mmap/shmget time, rather than getting SIGBUS'd when they try to fault
> the excess memory in.
> 

I'm not sure the developers are interested in being restricted by 
hugetlb_cgroups :)  I think users get constrained by hugetlb_cgroup so the 
developers are interested in the failure more: do we want to SIGBUS at 
fault and not be allowed an opportunity to influence that (today) or do we 
want to fallback to non-hugetlbfs memory and just keep going (tomorrow, 
after your patchset).

> The underlying problem is that today's hugetlb_cgroup accounting happens
> at hugetlb memory *fault* time, rather than at *reservation* time.
> Thus, enforcing the hugetlb_cgroup limit only happens at fault time, and
> the offending task gets SIGBUS'd.
> 
> Proposed Solution:
> A new page counter named hugetlb.xMB.reservation_[limit|usage]_in_bytes. This
> counter has slightly different semantics than
> hugetlb.xMB.[limit|usage]_in_bytes:
> 

Is there a max_usage_in_bytes equivalent?  It's a page_counter so I assume 
it's easy to support.

I'll defer the naming to Mike here, "rsvd" seems to be the hugetlb way of 
saying "reserved".

> - While usage_in_bytes tracks all *faulted* hugetlb memory,
> reservation_usage_in_bytes tracks all *reserved* hugetlb memory and
> hugetlb memory faulted in without a prior reservation.
> 
> - If a task attempts to reserve more memory than limit_in_bytes allows,
> the kernel will allow it to do so. But if a task attempts to reserve
> more memory than reservation_limit_in_bytes, the kernel will fail this
> reservation.
> 
> This proposal is implemented in this patch series, with tests to verify
> functionality and show the usage. We also added cgroup-v2 support to
> hugetlb_cgroup so that the new use cases can be extended to v2.
> 
> Alternatives considered:
> 1. A new cgroup, instead of only a new page_counter attached to
>    the existing hugetlb_cgroup. Adding a new cgroup seemed like a lot of code
>    duplication with hugetlb_cgroup. Keeping hugetlb related page counters under
>    hugetlb_cgroup seemed cleaner as well.
> 
> 2. Instead of adding a new counter, we considered adding a sysctl that modifies
>    the behavior of hugetlb.xMB.[limit|usage]_in_bytes, to do accounting at
>    reservation time rather than fault time. Adding a new page_counter seems
>    better as userspace could, if it wants, choose to enforce different cgroups
>    differently: one via limit_in_bytes, and another via
>    reservation_limit_in_bytes. This could be very useful if you're
>    transitioning how hugetlb memory is partitioned on your system one
>    cgroup at a time, for example. Also, someone may find usage for both
>    limit_in_bytes and reservation_limit_in_bytes concurrently, and this
>    approach gives them the option to do so.
> 
> Testing:
> - Added tests passing.
> - Used libhugetlbfs for regression testing.
> 
> [1]: https://www.kernel.org/doc/html/latest/vm/hugetlbfs_reserv.html
> 
> Signed-off-by: Mina Almasry <almasrymina@google.com>
> Acked-by: Hillf Danton <hdanton@sina.com>

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH v9 6/8] hugetlb_cgroup: support noreserve mappings
  2019-12-17 23:16 ` [PATCH v9 6/8] hugetlb_cgroup: support noreserve mappings Mina Almasry
@ 2020-01-14  0:48   ` David Rientjes
  0 siblings, 0 replies; 27+ messages in thread
From: David Rientjes @ 2020-01-14  0:48 UTC (permalink / raw)
  To: Mina Almasry
  Cc: mike.kravetz, shuah, shakeelb, gthelen, akpm, linux-kernel,
	linux-mm, linux-kselftest, cgroups, aneesh.kumar, mkoutny

On Tue, 17 Dec 2019, Mina Almasry wrote:

> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> index 39eb2fa53a420..3e94f5c2d7cd4 100644
> --- a/mm/hugetlb.c
> +++ b/mm/hugetlb.c
> @@ -1367,6 +1367,9 @@ void free_huge_page(struct page *page)
>  	clear_page_huge_active(page);
>  	hugetlb_cgroup_uncharge_page(hstate_index(h), pages_per_huge_page(h),
>  				     page, false);
> +	hugetlb_cgroup_uncharge_page(hstate_index(h), pages_per_huge_page(h),
> +				     page, true);
> +
>  	if (restore_reserve)
>  		h->resv_huge_pages++;
> 
> @@ -2189,10 +2192,19 @@ struct page *alloc_huge_page(struct vm_area_struct *vma,
>  			gbl_chg = 1;
>  	}
> 
> +	/* If this allocation is not consuming a reservation, charge it now.
> +	 */
> +	if (map_chg || avoid_reserve || !vma_resv_map(vma)) {

These conditions are checked three times in this function, maybe better 
set a bool on the stack and use it throughout the function to guarantee it 
remains consistent?

bool deferred_reserve = !vma_resv_map(vma) || map_chg || avoid_reserve;

> +		ret = hugetlb_cgroup_charge_cgroup(idx, pages_per_huge_page(h),
> +						   &h_cg, true);
> +		if (ret)
> +			goto out_subpool_put;
> +	}
> +
>  	ret = hugetlb_cgroup_charge_cgroup(idx, pages_per_huge_page(h), &h_cg,
>  					   false);
>  	if (ret)
> -		goto out_subpool_put;
> +		goto out_uncharge_cgroup_reservation;
> 
>  	spin_lock(&hugetlb_lock);
>  	/*
> @@ -2216,6 +2228,14 @@ struct page *alloc_huge_page(struct vm_area_struct *vma,
>  	}
>  	hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h), h_cg, page,
>  				     false);
> +	/* If allocation is not consuming a reservation, also store the
> +	 * hugetlb_cgroup pointer on the page.
> +	 */
> +	if (!vma_resv_map(vma) || map_chg || avoid_reserve) {
> +		hugetlb_cgroup_commit_charge(idx, pages_per_huge_page(h), h_cg,
> +					     page, true);
> +	}
> +
>  	spin_unlock(&hugetlb_lock);
> 
>  	set_page_private(page, (unsigned long)spool);
> @@ -2241,6 +2261,10 @@ struct page *alloc_huge_page(struct vm_area_struct *vma,
>  out_uncharge_cgroup:
>  	hugetlb_cgroup_uncharge_cgroup(idx, pages_per_huge_page(h), h_cg,
>  				       false);
> +out_uncharge_cgroup_reservation:
> +	if (map_chg || avoid_reserve || !vma_resv_map(vma))
> +		hugetlb_cgroup_uncharge_cgroup(idx, pages_per_huge_page(h),
> +					       h_cg, true);
>  out_subpool_put:
>  	if (map_chg || avoid_reserve)
>  		hugepage_subpool_put_pages(spool, 1);
> --
> 2.24.1.735.g03f4e72817-goog
> 

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH v9 8/8] hugetlb_cgroup: Add hugetlb_cgroup reservation docs
  2019-12-17 23:16 ` [PATCH v9 8/8] hugetlb_cgroup: Add hugetlb_cgroup reservation docs Mina Almasry
@ 2020-01-14  0:54   ` David Rientjes
  0 siblings, 0 replies; 27+ messages in thread
From: David Rientjes @ 2020-01-14  0:54 UTC (permalink / raw)
  To: Mina Almasry
  Cc: mike.kravetz, shuah, shakeelb, gthelen, akpm, linux-kernel,
	linux-mm, linux-kselftest, cgroups, aneesh.kumar, mkoutny,
	Hillf Danton

On Tue, 17 Dec 2019, Mina Almasry wrote:

> diff --git a/Documentation/admin-guide/cgroup-v1/hugetlb.rst b/Documentation/admin-guide/cgroup-v1/hugetlb.rst
> index a3902aa253a96..efb94e4db9d5a 100644
> --- a/Documentation/admin-guide/cgroup-v1/hugetlb.rst
> +++ b/Documentation/admin-guide/cgroup-v1/hugetlb.rst
> @@ -2,13 +2,6 @@
>  HugeTLB Controller
>  ==================
> 
> -The HugeTLB controller allows to limit the HugeTLB usage per control group and
> -enforces the controller limit during page fault. Since HugeTLB doesn't
> -support page reclaim, enforcing the limit at page fault time implies that,
> -the application will get SIGBUS signal if it tries to access HugeTLB pages
> -beyond its limit. This requires the application to know beforehand how much
> -HugeTLB pages it would require for its use.
> -
>  HugeTLB controller can be created by first mounting the cgroup filesystem.
> 
>  # mount -t cgroup -o hugetlb none /sys/fs/cgroup
> @@ -28,10 +21,14 @@ process (bash) into it.
> 
>  Brief summary of control files::
> 
> - hugetlb.<hugepagesize>.limit_in_bytes     # set/show limit of "hugepagesize" hugetlb usage
> - hugetlb.<hugepagesize>.max_usage_in_bytes # show max "hugepagesize" hugetlb  usage recorded
> - hugetlb.<hugepagesize>.usage_in_bytes     # show current usage for "hugepagesize" hugetlb
> - hugetlb.<hugepagesize>.failcnt		   # show the number of allocation failure due to HugeTLB limit
> + hugetlb.<hugepagesize>.reservation_limit_in_bytes     # set/show limit of "hugepagesize" hugetlb reservations
> + hugetlb.<hugepagesize>.reservation_max_usage_in_bytes # show max "hugepagesize" hugetlb reservations and no-reserve faults.
> + hugetlb.<hugepagesize>.reservation_usage_in_bytes     # show current reservations and no-reserve faults for "hugepagesize" hugetlb
> + hugetlb.<hugepagesize>.reservation_failcnt            # show the number of allocation failure due to HugeTLB reservation limit
> + hugetlb.<hugepagesize>.limit_in_bytes                 # set/show limit of "hugepagesize" hugetlb faults
> + hugetlb.<hugepagesize>.max_usage_in_bytes             # show max "hugepagesize" hugetlb  usage recorded
> + hugetlb.<hugepagesize>.usage_in_bytes                 # show current usage for "hugepagesize" hugetlb
> + hugetlb.<hugepagesize>.failcnt                        # show the number of allocation failure due to HugeTLB usage limit
> 

I assume these are better named hugetlb.<hugepagesize>.reservation.* 
rather than reservation_*, or perhaps shortened to resv.*, so for example 
hugetlb.<hugepagesize>.resv.limit_in_bytes.

>  For a system supporting three hugepage sizes (64k, 32M and 1G), the control
>  files include::
> @@ -40,11 +37,56 @@ files include::
>    hugetlb.1GB.max_usage_in_bytes
>    hugetlb.1GB.usage_in_bytes
>    hugetlb.1GB.failcnt
> +  hugetlb.1GB.reservation_limit_in_bytes
> +  hugetlb.1GB.reservation_max_usage_in_bytes
> +  hugetlb.1GB.reservation_usage_in_bytes
> +  hugetlb.1GB.reservation_failcnt
>    hugetlb.64KB.limit_in_bytes
>    hugetlb.64KB.max_usage_in_bytes
>    hugetlb.64KB.usage_in_bytes
>    hugetlb.64KB.failcnt
> +  hugetlb.64KB.reservation_limit_in_bytes
> +  hugetlb.64KB.reservation_max_usage_in_bytes
> +  hugetlb.64KB.reservation_usage_in_bytes
> +  hugetlb.64KB.reservation_failcnt
>    hugetlb.32MB.limit_in_bytes
>    hugetlb.32MB.max_usage_in_bytes
>    hugetlb.32MB.usage_in_bytes
>    hugetlb.32MB.failcnt
> +  hugetlb.32MB.reservation_limit_in_bytes
> +  hugetlb.32MB.reservation_max_usage_in_bytes
> +  hugetlb.32MB.reservation_usage_in_bytes
> +  hugetlb.32MB.reservation_failcnt
> +
> +
> +1. Reservation limits

Should probably be described after the page fault limits since those are 
the canonical limits that already exist and the "reservation_.*" 
equivalents are supplementary.

> +
> +The HugeTLB controller allows to limit the HugeTLB reservations per control
> +group and enforces the controller limit at reservation time and at the fault of
> +hugetlb memory for which no reservation exists. Reservation limits
> +are superior to Page fault limits (see section 2), since Reservation limits are
> +enforced at reservation time (on mmap or shget), and never causes the
> +application to get SIGBUS signal if the memory was reserved before hand. For
> +MAP_NORESERVE allocations, the reservation limit behaves the same as the fault
> +limit, enforcing memory usage at fault time and causing the application to
> +receive a SIGBUS if it's crossing its limit.
> +

When saying that reservation limits are superior to page fault limits, it 
might be helpful to expand on the downsides of page fault limits.  The 
existing documentation calls out that the application needs to know its 
expected usage; it does not call attention to the fact that several 
different applications may be accessing an overcommited system-wide pool 
of hugetlb memory.  So it might be possible that the application 
understands its own usage but it may not understand how that is 
orchestrated with other applications on the same system accessing a shared 
pool of hugetlb pages.

But yes, I think MAP_FAILED and allow for fallback or freeing of hugetlb 
memory is far superior to SIGBUS :)

> +2. Page fault limits
> +
> +The HugeTLB controller allows to limit the HugeTLB usage (page fault) per
> +control group and enforces the controller limit during page fault. Since HugeTLB
> +doesn't support page reclaim, enforcing the limit at page fault time implies
> +that, the application will get SIGBUS signal if it tries to access HugeTLB
> +pages beyond its limit. This requires the application to know beforehand how
> +much HugeTLB pages it would require for its use.
> +
> +
> +3. Caveats with shared memory
> +
> +For shared hugetlb memory, both hugetlb reservation and page faults are charged
> +to the first task that causes the memory to be reserved or faulted, and all
> +subsequent uses of this reserved or faulted memory is done without charging.
> +
> +Shared hugetlb memory is only uncharged when it is unreserved or deallocated.
> +This is usually when the hugetlbfs file is deleted, and not when the task that
> +caused the reservation or fault has exited.

Discussion of reparenting?

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH v9 3/8] hugetlb_cgroup: add reservation accounting for private mappings
  2019-12-17 23:16 ` [PATCH v9 3/8] hugetlb_cgroup: add reservation accounting for private mappings Mina Almasry
@ 2020-01-14  0:55   ` Mike Kravetz
  2020-01-14 22:52     ` Mina Almasry
  0 siblings, 1 reply; 27+ messages in thread
From: Mike Kravetz @ 2020-01-14  0:55 UTC (permalink / raw)
  To: Mina Almasry
  Cc: shuah, rientjes, shakeelb, gthelen, akpm, linux-kernel, linux-mm,
	linux-kselftest, cgroups, aneesh.kumar, mkoutny, Hillf Danton

On 12/17/19 3:16 PM, Mina Almasry wrote:
> Normally the pointer to the cgroup to uncharge hangs off the struct
> page, and gets queried when it's time to free the page. With
> hugetlb_cgroup reservations, this is not possible. Because it's possible
> for a page to be reserved by one task and actually faulted in by another
> task.
> 
> The best place to put the hugetlb_cgroup pointer to uncharge for
> reservations is in the resv_map. But, because the resv_map has different
> semantics for private and shared mappings, the code patch to
> charge/uncharge shared and private mappings is different. This patch
> implements charging and uncharging for private mappings.
> 
> For private mappings, the counter to uncharge is in
> resv_map->reservation_counter. On initializing the resv_map this is set
> to NULL. On reservation of a region in private mapping, the tasks
> hugetlb_cgroup is charged and the hugetlb_cgroup is placed is
> resv_map->reservation_counter.
> 
> On hugetlb_vm_op_close, we uncharge resv_map->reservation_counter.
> 
> Signed-off-by: Mina Almasry <almasrymina@google.com>
> Acked-by: Hillf Danton <hdanton@sina.com>
> 
> ---
> 
> Changes in V9:
> - Updated for reparenting of hugetlb reservation accounting.
> 
> ---
>  include/linux/hugetlb.h        |  9 +++++++
>  include/linux/hugetlb_cgroup.h | 27 +++++++++++++++++++
>  mm/hugetlb.c                   | 47 +++++++++++++++++++++++++++++++++-
>  mm/hugetlb_cgroup.c            | 28 --------------------
>  4 files changed, 82 insertions(+), 29 deletions(-)
> 
> diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
> index dea6143aa0685..e6ab499ba2086 100644
> --- a/include/linux/hugetlb.h
> +++ b/include/linux/hugetlb.h
> @@ -46,6 +46,15 @@ struct resv_map {
>  	long adds_in_progress;
>  	struct list_head region_cache;
>  	long region_cache_count;
> +#ifdef CONFIG_CGROUP_HUGETLB
> +	/*
> +	 * On private mappings, the counter to uncharge reservations is stored
> +	 * here. If these fields are 0, then the mapping is shared.

Will *reservation_counter ALWAYS be non-NULL for private mappings?

More on this below.

> +	 */
> +	struct page_counter *reservation_counter;
> +	unsigned long pages_per_hpage;
> +	struct cgroup_subsys_state *css;
> +#endif
>  };
>  extern struct resv_map *resv_map_alloc(void);
>  void resv_map_release(struct kref *ref);
> diff --git a/include/linux/hugetlb_cgroup.h b/include/linux/hugetlb_cgroup.h
> index eab8a70d5bcb5..8c320accefe87 100644
> --- a/include/linux/hugetlb_cgroup.h
> +++ b/include/linux/hugetlb_cgroup.h
> @@ -25,6 +25,33 @@ struct hugetlb_cgroup;
>  #define HUGETLB_CGROUP_MIN_ORDER	2
> 
>  #ifdef CONFIG_CGROUP_HUGETLB
> +enum hugetlb_memory_event {
> +	HUGETLB_MAX,
> +	HUGETLB_NR_MEMORY_EVENTS,
> +};
> +
> +struct hugetlb_cgroup {
> +	struct cgroup_subsys_state css;
> +
> +	/*
> +	 * the counter to account for hugepages from hugetlb.
> +	 */
> +	struct page_counter hugepage[HUGE_MAX_HSTATE];
> +
> +	/*
> +	 * the counter to account for hugepage reservations from hugetlb.
> +	 */
> +	struct page_counter reserved_hugepage[HUGE_MAX_HSTATE];
> +
> +	atomic_long_t events[HUGE_MAX_HSTATE][HUGETLB_NR_MEMORY_EVENTS];
> +	atomic_long_t events_local[HUGE_MAX_HSTATE][HUGETLB_NR_MEMORY_EVENTS];
> +
> +	/* Handle for "hugetlb.events" */
> +	struct cgroup_file events_file[HUGE_MAX_HSTATE];
> +
> +	/* Handle for "hugetlb.events.local" */
> +	struct cgroup_file events_local_file[HUGE_MAX_HSTATE];
> +};
> 
>  static inline struct hugetlb_cgroup *hugetlb_cgroup_from_page(struct page *page,
>  							      bool reserved)
> diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> index e6e8240f1718c..7782977970301 100644
> --- a/mm/hugetlb.c
> +++ b/mm/hugetlb.c
> @@ -665,6 +665,17 @@ struct resv_map *resv_map_alloc(void)
>  	INIT_LIST_HEAD(&resv_map->regions);
> 
>  	resv_map->adds_in_progress = 0;
> +#ifdef CONFIG_CGROUP_HUGETLB
> +	/*
> +	 * Initialize these to 0. On shared mappings, 0's here indicate these
> +	 * fields don't do cgroup accounting. On private mappings, these will be
> +	 * re-initialized to the proper values, to indicate that hugetlb cgroup
> +	 * reservations are to be un-charged from here.
> +	 */
> +	resv_map->reservation_counter = NULL;
> +	resv_map->pages_per_hpage = 0;
> +	resv_map->css = NULL;
> +#endif
> 
>  	INIT_LIST_HEAD(&resv_map->region_cache);
>  	list_add(&rg->link, &resv_map->region_cache);
> @@ -3145,7 +3156,20 @@ static void hugetlb_vm_op_close(struct vm_area_struct *vma)
> 
>  	reserve = (end - start) - region_count(resv, start, end);
> 
> -	kref_put(&resv->refs, resv_map_release);
> +#ifdef CONFIG_CGROUP_HUGETLB
> +	/*
> +	 * Since we check for HPAGE_RESV_OWNER above, this must a private
> +	 * mapping, and these values should be none-zero, and should point to
> +	 * the hugetlb_cgroup counter to uncharge for this reservation.
> +	 */
> +	WARN_ON(!resv->reservation_counter);
> +	WARN_ON(!resv->pages_per_hpage);
> +	WARN_ON(!resv->css);

I was once again wondering if these were always non-NULL for private mappings.
It seems that reservation_counter (h_gc) would be NULL in these cases from
these early checks in hugetlb_cgroup_charge_cgroup().

int hugetlb_cgroup_charge_cgroup(int idx, unsigned long nr_pages,
                                 struct hugetlb_cgroup **ptr, bool reserved)
{
        int ret = 0;
        struct page_counter *counter;
        struct hugetlb_cgroup *h_cg = NULL;

        if (hugetlb_cgroup_disabled())
                goto done;
        /*
         * We don't charge any cgroup if the compound page have less
         * than 3 pages.
         */
        if (huge_page_order(&hstates[idx]) < HUGETLB_CGROUP_MIN_ORDER)
                goto done;
	...

It seems like the following hugetlb_cgroup_uncharge_counter() guards
against reservation_counter being NULL (for some of the same reasons).

> +
> +	hugetlb_cgroup_uncharge_counter(resv->reservation_counter,
> +					(end - start) * resv->pages_per_hpage,
> +					resv->css);
> +#endif
> 
>  	if (reserve) {
>  		/*
> @@ -3155,6 +3179,8 @@ static void hugetlb_vm_op_close(struct vm_area_struct *vma)
>  		gbl_reserve = hugepage_subpool_put_pages(spool, reserve);
>  		hugetlb_acct_memory(h, -gbl_reserve);
>  	}
> +
> +	kref_put(&resv->refs, resv_map_release);
>  }
> 
>  static int hugetlb_vm_op_split(struct vm_area_struct *vma, unsigned long addr)
> @@ -4501,6 +4527,7 @@ int hugetlb_reserve_pages(struct inode *inode,
>  	struct hstate *h = hstate_inode(inode);
>  	struct hugepage_subpool *spool = subpool_inode(inode);
>  	struct resv_map *resv_map;
> +	struct hugetlb_cgroup *h_cg;
>  	long gbl_reserve;
> 
>  	/* This should never happen */
> @@ -4534,12 +4561,30 @@ int hugetlb_reserve_pages(struct inode *inode,
>  		chg = region_chg(resv_map, from, to);
> 
>  	} else {
> +		/* Private mapping. */
>  		resv_map = resv_map_alloc();
>  		if (!resv_map)
>  			return -ENOMEM;
> 
>  		chg = to - from;
> 
> +		if (hugetlb_cgroup_charge_cgroup(hstate_index(h),
> +						 chg * pages_per_huge_page(h),
> +						 &h_cg, true)) {
> +			kref_put(&resv_map->refs, resv_map_release);
> +			return -ENOMEM;
> +		}
> +

Shouldn't this code be in the #ifdef CONFIG_CGROUP_HUGETLB block?
-- 
Mike Kravetz

> +#ifdef CONFIG_CGROUP_HUGETLB
> +		/*
> +		 * Since this branch handles private mappings, we attach the
> +		 * counter to uncharge for this reservation off resv_map.
> +		 */
> +		resv_map->reservation_counter =
> +			&h_cg->reserved_hugepage[hstate_index(h)];
> +		resv_map->pages_per_hpage = pages_per_huge_page(h);
> +#endif
> +
>  		set_vma_resv_map(vma, resv_map);
>  		set_vma_resv_flags(vma, HPAGE_RESV_OWNER);
>  	}

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH v9 3/8] hugetlb_cgroup: add reservation accounting for private mappings
  2020-01-14  0:55   ` Mike Kravetz
@ 2020-01-14 22:52     ` Mina Almasry
  2020-01-17 22:09       ` Mike Kravetz
  0 siblings, 1 reply; 27+ messages in thread
From: Mina Almasry @ 2020-01-14 22:52 UTC (permalink / raw)
  To: Mike Kravetz
  Cc: shuah, David Rientjes, Shakeel Butt, Greg Thelen, Andrew Morton,
	open list, linux-mm, linux-kselftest, cgroups, Aneesh Kumar,
	Michal Koutný,
	Hillf Danton

On Mon, Jan 13, 2020 at 4:55 PM Mike Kravetz <mike.kravetz@oracle.com> wrote:
> > diff --git a/include/linux/hugetlb.h b/include/linux/hugetlb.h
> > index dea6143aa0685..e6ab499ba2086 100644
> > --- a/include/linux/hugetlb.h
> > +++ b/include/linux/hugetlb.h
> > @@ -46,6 +46,15 @@ struct resv_map {
> >       long adds_in_progress;
> >       struct list_head region_cache;
> >       long region_cache_count;
> > +#ifdef CONFIG_CGROUP_HUGETLB
> > +     /*
> > +      * On private mappings, the counter to uncharge reservations is stored
> > +      * here. If these fields are 0, then the mapping is shared.
>
> Will *reservation_counter ALWAYS be non-NULL for private mappings?
>
> More on this below.
>
> > +      */
> > +     struct page_counter *reservation_counter;
> > +     unsigned long pages_per_hpage;
> > +     struct cgroup_subsys_state *css;
> > +#endif
> >  };
> >  extern struct resv_map *resv_map_alloc(void);
> >  void resv_map_release(struct kref *ref);
> > diff --git a/include/linux/hugetlb_cgroup.h b/include/linux/hugetlb_cgroup.h
> > index eab8a70d5bcb5..8c320accefe87 100644
> > --- a/include/linux/hugetlb_cgroup.h
> > +++ b/include/linux/hugetlb_cgroup.h
> > @@ -25,6 +25,33 @@ struct hugetlb_cgroup;
> >  #define HUGETLB_CGROUP_MIN_ORDER     2
> >
> >  #ifdef CONFIG_CGROUP_HUGETLB
> > +enum hugetlb_memory_event {
> > +     HUGETLB_MAX,
> > +     HUGETLB_NR_MEMORY_EVENTS,
> > +};
> > +
> > +struct hugetlb_cgroup {
> > +     struct cgroup_subsys_state css;
> > +
> > +     /*
> > +      * the counter to account for hugepages from hugetlb.
> > +      */
> > +     struct page_counter hugepage[HUGE_MAX_HSTATE];
> > +
> > +     /*
> > +      * the counter to account for hugepage reservations from hugetlb.
> > +      */
> > +     struct page_counter reserved_hugepage[HUGE_MAX_HSTATE];
> > +
> > +     atomic_long_t events[HUGE_MAX_HSTATE][HUGETLB_NR_MEMORY_EVENTS];
> > +     atomic_long_t events_local[HUGE_MAX_HSTATE][HUGETLB_NR_MEMORY_EVENTS];
> > +
> > +     /* Handle for "hugetlb.events" */
> > +     struct cgroup_file events_file[HUGE_MAX_HSTATE];
> > +
> > +     /* Handle for "hugetlb.events.local" */
> > +     struct cgroup_file events_local_file[HUGE_MAX_HSTATE];
> > +};
> >
> >  static inline struct hugetlb_cgroup *hugetlb_cgroup_from_page(struct page *page,
> >                                                             bool reserved)
> > diff --git a/mm/hugetlb.c b/mm/hugetlb.c
> > index e6e8240f1718c..7782977970301 100644
> > --- a/mm/hugetlb.c
> > +++ b/mm/hugetlb.c
> > @@ -665,6 +665,17 @@ struct resv_map *resv_map_alloc(void)
> >       INIT_LIST_HEAD(&resv_map->regions);
> >
> >       resv_map->adds_in_progress = 0;
> > +#ifdef CONFIG_CGROUP_HUGETLB
> > +     /*
> > +      * Initialize these to 0. On shared mappings, 0's here indicate these
> > +      * fields don't do cgroup accounting. On private mappings, these will be
> > +      * re-initialized to the proper values, to indicate that hugetlb cgroup
> > +      * reservations are to be un-charged from here.
> > +      */
> > +     resv_map->reservation_counter = NULL;
> > +     resv_map->pages_per_hpage = 0;
> > +     resv_map->css = NULL;
> > +#endif
> >
> >       INIT_LIST_HEAD(&resv_map->region_cache);
> >       list_add(&rg->link, &resv_map->region_cache);
> > @@ -3145,7 +3156,20 @@ static void hugetlb_vm_op_close(struct vm_area_struct *vma)
> >
> >       reserve = (end - start) - region_count(resv, start, end);
> >
> > -     kref_put(&resv->refs, resv_map_release);
> > +#ifdef CONFIG_CGROUP_HUGETLB
> > +     /*
> > +      * Since we check for HPAGE_RESV_OWNER above, this must a private
> > +      * mapping, and these values should be none-zero, and should point to
> > +      * the hugetlb_cgroup counter to uncharge for this reservation.
> > +      */
> > +     WARN_ON(!resv->reservation_counter);
> > +     WARN_ON(!resv->pages_per_hpage);
> > +     WARN_ON(!resv->css);
>
> I was once again wondering if these were always non-NULL for private mappings.
> It seems that reservation_counter (h_gc) would be NULL in these cases from
> these early checks in hugetlb_cgroup_charge_cgroup().
>

You are right. I'm fixing in v10 the code and comments to account for
h_cg potentially being NULL, but I'm having trouble testing. Looking
at the code, I'm a bit confused by the checks. Seems to me
hugetlb_cgroup_disabled() is the same as #ifdef CONFIG_CGROUP_HUGETLB;
I can't find a way to enable the Kconfig but have that return false
unless I hack the code. Also seems to me checking huge_page_order is
just super definsive; I skimmed the hugepage sizes allowed code and I
can't find an arch that allows you to configure hugetlb page size to <
2^HUGETLB_CGROUP_MIN_ORDER pages. So in reality these will never fire,
IIUC.

> int hugetlb_cgroup_charge_cgroup(int idx, unsigned long nr_pages,
>                                  struct hugetlb_cgroup **ptr, bool reserved)
> {
>         int ret = 0;
>         struct page_counter *counter;
>         struct hugetlb_cgroup *h_cg = NULL;
>
>         if (hugetlb_cgroup_disabled())
>                 goto done;
>         /*
>          * We don't charge any cgroup if the compound page have less
>          * than 3 pages.
>          */
>         if (huge_page_order(&hstates[idx]) < HUGETLB_CGROUP_MIN_ORDER)
>                 goto done;
>         ...
>
> It seems like the following hugetlb_cgroup_uncharge_counter() guards
> against reservation_counter being NULL (for some of the same reasons).
>
> > +
> > +     hugetlb_cgroup_uncharge_counter(resv->reservation_counter,
> > +                                     (end - start) * resv->pages_per_hpage,
> > +                                     resv->css);
> > +#endif
> >
> >       if (reserve) {
> >               /*
> > @@ -3155,6 +3179,8 @@ static void hugetlb_vm_op_close(struct vm_area_struct *vma)
> >               gbl_reserve = hugepage_subpool_put_pages(spool, reserve);
> >               hugetlb_acct_memory(h, -gbl_reserve);
> >       }
> > +
> > +     kref_put(&resv->refs, resv_map_release);
> >  }
> >
> >  static int hugetlb_vm_op_split(struct vm_area_struct *vma, unsigned long addr)
> > @@ -4501,6 +4527,7 @@ int hugetlb_reserve_pages(struct inode *inode,
> >       struct hstate *h = hstate_inode(inode);
> >       struct hugepage_subpool *spool = subpool_inode(inode);
> >       struct resv_map *resv_map;
> > +     struct hugetlb_cgroup *h_cg;
> >       long gbl_reserve;
> >
> >       /* This should never happen */
> > @@ -4534,12 +4561,30 @@ int hugetlb_reserve_pages(struct inode *inode,
> >               chg = region_chg(resv_map, from, to);
> >
> >       } else {
> > +             /* Private mapping. */
> >               resv_map = resv_map_alloc();
> >               if (!resv_map)
> >                       return -ENOMEM;
> >
> >               chg = to - from;
> >
> > +             if (hugetlb_cgroup_charge_cgroup(hstate_index(h),
> > +                                              chg * pages_per_huge_page(h),
> > +                                              &h_cg, true)) {
> > +                     kref_put(&resv_map->refs, resv_map_release);
> > +                     return -ENOMEM;
> > +             }
> > +
>
> Shouldn't this code be in the #ifdef CONFIG_CGROUP_HUGETLB block?

Not necessary AFAICT, hugetlb_cgroup_charge_cgroup stub returns 0 (no-op).

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH v9 2/8] hugetlb_cgroup: add interface for charge/uncharge hugetlb reservations
  2020-01-13 22:14   ` Mike Kravetz
  2020-01-14  0:45     ` David Rientjes
@ 2020-01-14 22:55     ` Mina Almasry
  1 sibling, 0 replies; 27+ messages in thread
From: Mina Almasry @ 2020-01-14 22:55 UTC (permalink / raw)
  To: Mike Kravetz
  Cc: shuah, David Rientjes, Shakeel Butt, Greg Thelen, Andrew Morton,
	open list, linux-mm, linux-kselftest, cgroups, Aneesh Kumar,
	Michal Koutný

On Mon, Jan 13, 2020 at 2:14 PM Mike Kravetz <mike.kravetz@oracle.com> wrote:
> > +extern void hugetlb_cgroup_uncharge_counter(struct page_counter *p,
> > +                                         unsigned long nr_pages,
> > +                                         struct cgroup_subsys_state *css);
> > +
>
> Do we need a corresponding stub for the !CONFIG_CGROUP_HUGETLB case?
>

Not necessary AFAICT because the code is only called from code guarded
by CONFIG_CGROUP_HUGETLB.

Addressing comments from you and David in v10 shortly. Thanks for the review!

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH v9 3/8] hugetlb_cgroup: add reservation accounting for private mappings
  2020-01-14 22:52     ` Mina Almasry
@ 2020-01-17 22:09       ` Mike Kravetz
  2020-01-22 21:40         ` Mina Almasry
  0 siblings, 1 reply; 27+ messages in thread
From: Mike Kravetz @ 2020-01-17 22:09 UTC (permalink / raw)
  To: Mina Almasry
  Cc: shuah, David Rientjes, Shakeel Butt, Greg Thelen, Andrew Morton,
	open list, linux-mm, linux-kselftest, cgroups, Aneesh Kumar,
	Michal Koutný,
	Hillf Danton

On 1/14/20 2:52 PM, Mina Almasry wrote:
> On Mon, Jan 13, 2020 at 4:55 PM Mike Kravetz <mike.kravetz@oracle.com> wrote:
>>> +#ifdef CONFIG_CGROUP_HUGETLB
>>> +     /*
>>> +      * Since we check for HPAGE_RESV_OWNER above, this must a private
>>> +      * mapping, and these values should be none-zero, and should point to
>>> +      * the hugetlb_cgroup counter to uncharge for this reservation.
>>> +      */
>>> +     WARN_ON(!resv->reservation_counter);
>>> +     WARN_ON(!resv->pages_per_hpage);
>>> +     WARN_ON(!resv->css);
>>
>> I was once again wondering if these were always non-NULL for private mappings.
>> It seems that reservation_counter (h_gc) would be NULL in these cases from
>> these early checks in hugetlb_cgroup_charge_cgroup().
>>
> 
> You are right. I'm fixing in v10 the code and comments to account for
> h_cg potentially being NULL, but I'm having trouble testing. Looking
> at the code, I'm a bit confused by the checks. Seems to me
> hugetlb_cgroup_disabled() is the same as #ifdef CONFIG_CGROUP_HUGETLB;
> I can't find a way to enable the Kconfig but have that return false
> unless I hack the code.

What about the boot options?

cgroup_disable=
cgroup_no_v1=

-- 
Mike Kravetz

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH v9 4/8] hugetlb: disable region_add file_region coalescing
  2019-12-17 23:16 ` [PATCH v9 4/8] hugetlb: disable region_add file_region coalescing Mina Almasry
@ 2020-01-21 17:38   ` Mike Kravetz
  2020-01-21 17:40     ` Mike Kravetz
  0 siblings, 1 reply; 27+ messages in thread
From: Mike Kravetz @ 2020-01-21 17:38 UTC (permalink / raw)
  To: Mina Almasry
  Cc: shuah, rientjes, shakeelb, gthelen, akpm, linux-kernel, linux-mm,
	linux-kselftest, cgroups, aneesh.kumar, mkoutny

On 12/17/19 3:16 PM, Mina Almasry wrote:
> A follow up patch in this series adds hugetlb cgroup uncharge info the

*nit*
A follow up patch in this series adds hugetlb cgroup uncharge info to the

> file_region entries in resv->regions. The cgroup uncharge info may
> differ for different regions, so they can no longer be coalesced at
> region_add time. So, disable region coalescing in region_add in this
> patch.
> 
> Behavior change:
> 
> Say a resv_map exists like this [0->1], [2->3], and [5->6].
> 
> Then a region_chg/add call comes in region_chg/add(f=0, t=5).
> 
> Old code would generate resv->regions: [0->5], [5->6].
> New code would generate resv->regions: [0->1], [1->2], [2->3], [3->5],
> [5->6].
> 
> Special care needs to be taken to handle the resv->adds_in_progress
> variable correctly. In the past, only 1 region would be added for every
> region_chg and region_add call. But now, each call may add multiple
> regions, so we can no longer increment adds_in_progress by 1 in region_chg,
> or decrement adds_in_progress by 1 after region_add or region_abort. Instead,
> region_chg calls add_reservation_in_range() to count the number of regions
> needed and allocates those, and that info is passed to region_add and
> region_abort to decrement adds_in_progress correctly.
> 
> We've also modified the assumption that region_add after region_chg
> never fails. region_chg now pre-allocates at least 1 region for
> region_add. If region_add needs more regions than region_chg has
> allocated for it, then it may fail.
> 
> Signed-off-by: Mina Almasry <almasrymina@google.com>
> Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>

Thanks for the updated comments.  Still,
Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>

-- 
Mike Kravetz

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH v9 4/8] hugetlb: disable region_add file_region coalescing
  2020-01-21 17:38   ` Mike Kravetz
@ 2020-01-21 17:40     ` Mike Kravetz
  0 siblings, 0 replies; 27+ messages in thread
From: Mike Kravetz @ 2020-01-21 17:40 UTC (permalink / raw)
  To: Mina Almasry
  Cc: shuah, rientjes, shakeelb, gthelen, akpm, linux-kernel, linux-mm,
	linux-kselftest, cgroups, aneesh.kumar, mkoutny

On 1/21/20 9:38 AM, Mike Kravetz wrote:
> On 12/17/19 3:16 PM, Mina Almasry wrote:
>> A follow up patch in this series adds hugetlb cgroup uncharge info the
> 
> *nit*
> A follow up patch in this series adds hugetlb cgroup uncharge info to the
> 
>> file_region entries in resv->regions. The cgroup uncharge info may
>> differ for different regions, so they can no longer be coalesced at
>> region_add time. So, disable region coalescing in region_add in this
>> patch.
>>
>> Behavior change:
>>
>> Say a resv_map exists like this [0->1], [2->3], and [5->6].
>>
>> Then a region_chg/add call comes in region_chg/add(f=0, t=5).
>>
>> Old code would generate resv->regions: [0->5], [5->6].
>> New code would generate resv->regions: [0->1], [1->2], [2->3], [3->5],
>> [5->6].
>>
>> Special care needs to be taken to handle the resv->adds_in_progress
>> variable correctly. In the past, only 1 region would be added for every
>> region_chg and region_add call. But now, each call may add multiple
>> regions, so we can no longer increment adds_in_progress by 1 in region_chg,
>> or decrement adds_in_progress by 1 after region_add or region_abort. Instead,
>> region_chg calls add_reservation_in_range() to count the number of regions
>> needed and allocates those, and that info is passed to region_add and
>> region_abort to decrement adds_in_progress correctly.
>>
>> We've also modified the assumption that region_add after region_chg
>> never fails. region_chg now pre-allocates at least 1 region for
>> region_add. If region_add needs more regions than region_chg has
>> allocated for it, then it may fail.
>>
>> Signed-off-by: Mina Almasry <almasrymina@google.com>
>> Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>
> 
> Thanks for the updated comments.  Still,
> Reviewed-by: Mike Kravetz <mike.kravetz@oracle.com>


Sorry, that comment was directed at V10 of the patch.

-- 
Mike Kravetz

^ permalink raw reply	[flat|nested] 27+ messages in thread

* Re: [PATCH v9 3/8] hugetlb_cgroup: add reservation accounting for private mappings
  2020-01-17 22:09       ` Mike Kravetz
@ 2020-01-22 21:40         ` Mina Almasry
  0 siblings, 0 replies; 27+ messages in thread
From: Mina Almasry @ 2020-01-22 21:40 UTC (permalink / raw)
  To: Mike Kravetz
  Cc: shuah, David Rientjes, Shakeel Butt, Greg Thelen, Andrew Morton,
	open list, linux-mm, linux-kselftest, cgroups, Aneesh Kumar,
	Michal Koutný,
	Hillf Danton

On Fri, Jan 17, 2020 at 2:09 PM Mike Kravetz <mike.kravetz@oracle.com> wrote:
>
> On 1/14/20 2:52 PM, Mina Almasry wrote:
> > On Mon, Jan 13, 2020 at 4:55 PM Mike Kravetz <mike.kravetz@oracle.com> wrote:
> >>> +#ifdef CONFIG_CGROUP_HUGETLB
> >>> +     /*
> >>> +      * Since we check for HPAGE_RESV_OWNER above, this must a private
> >>> +      * mapping, and these values should be none-zero, and should point to
> >>> +      * the hugetlb_cgroup counter to uncharge for this reservation.
> >>> +      */
> >>> +     WARN_ON(!resv->reservation_counter);
> >>> +     WARN_ON(!resv->pages_per_hpage);
> >>> +     WARN_ON(!resv->css);
> >>
> >> I was once again wondering if these were always non-NULL for private mappings.
> >> It seems that reservation_counter (h_gc) would be NULL in these cases from
> >> these early checks in hugetlb_cgroup_charge_cgroup().
> >>
> >
> > You are right. I'm fixing in v10 the code and comments to account for
> > h_cg potentially being NULL, but I'm having trouble testing. Looking
> > at the code, I'm a bit confused by the checks. Seems to me
> > hugetlb_cgroup_disabled() is the same as #ifdef CONFIG_CGROUP_HUGETLB;
> > I can't find a way to enable the Kconfig but have that return false
> > unless I hack the code.
>
> What about the boot options?
>
> cgroup_disable=
> cgroup_no_v1=

Thanks, cgroup_disable=hugetlb does it. I ran the the libhugetlbfs
tests with patchset v10 and it passed, so it seems the latest version
of the patch should be fine. Of course my hugetlb cgroup tests fail
outright when hugetlb cgroups are disabled so those don't say anything
useful.

^ permalink raw reply	[flat|nested] 27+ messages in thread

end of thread, other threads:[~2020-01-22 21:41 UTC | newest]

Thread overview: 27+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-12-17 23:16 [PATCH v9 1/8] hugetlb_cgroup: Add hugetlb_cgroup reservation counter Mina Almasry
2019-12-17 23:16 ` [PATCH v9 2/8] hugetlb_cgroup: add interface for charge/uncharge hugetlb reservations Mina Almasry
2020-01-13 22:14   ` Mike Kravetz
2020-01-14  0:45     ` David Rientjes
2020-01-14 22:55     ` Mina Almasry
2019-12-17 23:16 ` [PATCH v9 3/8] hugetlb_cgroup: add reservation accounting for private mappings Mina Almasry
2020-01-14  0:55   ` Mike Kravetz
2020-01-14 22:52     ` Mina Almasry
2020-01-17 22:09       ` Mike Kravetz
2020-01-22 21:40         ` Mina Almasry
2019-12-17 23:16 ` [PATCH v9 4/8] hugetlb: disable region_add file_region coalescing Mina Almasry
2020-01-21 17:38   ` Mike Kravetz
2020-01-21 17:40     ` Mike Kravetz
2019-12-17 23:16 ` [PATCH v9 5/8] hugetlb_cgroup: add accounting for shared mappings Mina Almasry
2019-12-17 23:16 ` [PATCH v9 6/8] hugetlb_cgroup: support noreserve mappings Mina Almasry
2020-01-14  0:48   ` David Rientjes
2019-12-17 23:16 ` [PATCH v9 7/8] hugetlb_cgroup: Add hugetlb_cgroup reservation tests Mina Almasry
2019-12-17 23:16 ` [PATCH v9 8/8] hugetlb_cgroup: Add hugetlb_cgroup reservation docs Mina Almasry
2020-01-14  0:54   ` David Rientjes
2019-12-19  1:12 ` [PATCH v9 1/8] hugetlb_cgroup: Add hugetlb_cgroup reservation counter Andrew Morton
2019-12-19  1:37   ` Mike Kravetz
2019-12-19  1:59     ` Mina Almasry
2020-01-13 18:43 ` Mike Kravetz
2020-01-13 21:03   ` Mina Almasry
2020-01-13 22:05     ` Mike Kravetz
2020-01-13 22:21       ` Mina Almasry
2020-01-14  0:45 ` David Rientjes

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).