Skip to contentSkip to conversation

mm-convert-mms-rss-stats-into-percpu_counter.patch removed from -mm tree

merged mm-stable Andrew Morton 1 participant 1 message Dec 1, 2022

lore
Share & tools
lore

Andrew Morton

author tree notice Applied
#0 lore raw

The quilt patch titled
Subject: mm: convert mm's rss stats into percpu_counter
has been removed from the -mm tree. Its filename was
mm-convert-mms-rss-stats-into-percpu_counter.patch

This patch was dropped because it was merged into the mm-stable branch of git://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm

------------------------------------------------------
From: Shakeel Butt <[email protected]>
Subject: mm: convert mm's rss stats into percpu_counter
Date: Mon, 24 Oct 2022 05:28:41 +0000

Currently mm_struct maintains rss_stats which are updated on page fault and the unmapping codepaths. For page fault codepath the updates are cached per thread with the batch of TASK_RSS_EVENTS_THRESH which is 64. The reason for caching is performance for multithreaded applications otherwise the rss_stats updates may become hotspot for such applications.

However this optimization comes with the cost of error margin in the rss stats. The rss_stats for applications with large number of threads can be very skewed. At worst the error margin is (nr_threads * 64) and we have a lot of applications with 100s of threads, so the error margin can be very high. Internally we had to reduce TASK_RSS_EVENTS_THRESH to 32.

Recently we started seeing the unbounded errors for rss_stats for specific applications which use TCP rx0cp. It seems like vm_insert_pages() codepath does not sync rss_stats at all.

This patch converts the rss_stats into percpu_counter to convert the error margin from (nr_threads * 64) to approximately (nr_cpus ^ 2). However this conversion enable us to get the accurate stats for situations where accuracy is more important than the cpu cost.

This patch does not make such tradeoffs - we can just use percpu_counter_add_local() for the updates and percpu_counter_sum() (or percpu_counter_sync() + percpu_counter_read) for the readers. At the moment the readers are either procfs interface, oom_killer and memory reclaim which I think are not performance critical and should be ok with slow read. However I think we can make that change in a separate patch.

---

 include/linux/mm.h             |   26 +++--------
 include/linux/mm_types.h       |    7 --
 include/linux/mm_types_task.h  |   13 -----
 include/linux/percpu_counter.h |    1 
 include/linux/sched.h          |    3 -
 include/trace/events/kmem.h    |    8 +--
 kernel/fork.c                  |   16 ++++++
 mm/memory.c                    |   73 ++++---------------------------
 8 files changed, 40 insertions(+), 107 deletions(-)
include/linux/mm.h renamed from include/linux/mm.h~mm-convert-mms-rss-stats-into-percpu_counter +8 -18
@@ -2052,40 +2052,30 @@ static inline bool get_user_page_fast_on
  */
 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
 {
-	long val = atomic_long_read(&mm->rss_stat.count[member]);
-
-#ifdef SPLIT_RSS_COUNTING
-	/*
-	 * counter is updated in asynchronous manner and may go to minus.
-	 * But it's never be expected number for users.
-	 */
-	if (val < 0)
-		val = 0;
-#endif
-	return (unsigned long)val;
+	return percpu_counter_read_positive(&mm->rss_stat[member]);
 }
 
-void mm_trace_rss_stat(struct mm_struct *mm, int member, long count);
+void mm_trace_rss_stat(struct mm_struct *mm, int member);
 
 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
 {
-	long count = atomic_long_add_return(value, &mm->rss_stat.count[member]);
+	percpu_counter_add(&mm->rss_stat[member], value);
 
-	mm_trace_rss_stat(mm, member, count);
+	mm_trace_rss_stat(mm, member);
 }
 
 static inline void inc_mm_counter(struct mm_struct *mm, int member)
 {
-	long count = atomic_long_inc_return(&mm->rss_stat.count[member]);
+	percpu_counter_inc(&mm->rss_stat[member]);
 
-	mm_trace_rss_stat(mm, member, count);
+	mm_trace_rss_stat(mm, member);
 }
 
 static inline void dec_mm_counter(struct mm_struct *mm, int member)
 {
-	long count = atomic_long_dec_return(&mm->rss_stat.count[member]);
+	percpu_counter_dec(&mm->rss_stat[member]);
 
-	mm_trace_rss_stat(mm, member, count);
+	mm_trace_rss_stat(mm, member);
 }
 
 /* Optimized variant when page is already known not to be PageAnon */
include/linux/mm_types.h renamed from include/linux/mm_types.h~mm-convert-mms-rss-stats-into-percpu_counter +2 -5
@@ -18,6 +18,7 @@
 #include <linux/page-flags-layout.h>
 #include <linux/workqueue.h>
 #include <linux/seqlock.h>
+#include <linux/percpu_counter.h>
 
 #include <asm/mmu.h>
 
@@ -626,11 +627,7 @@ struct mm_struct {
 
 		unsigned long saved_auxv[AT_VECTOR_SIZE]; /* for /proc/PID/auxv */
 
-		/*
-		 * Special counters, in some configurations protected by the
-		 * page_table_lock, in other configurations by being atomic.
-		 */
-		struct mm_rss_stat rss_stat;
+		struct percpu_counter rss_stat[NR_MM_COUNTERS];
 
 		struct linux_binfmt *binfmt;
 
include/linux/mm_types_task.h renamed from include/linux/mm_types_task.h~mm-convert-mms-rss-stats-into-percpu_counter +0 -13
@@ -36,19 +36,6 @@ enum {
 	NR_MM_COUNTERS
 };
 
-#if USE_SPLIT_PTE_PTLOCKS && defined(CONFIG_MMU)
-#define SPLIT_RSS_COUNTING
-/* per-thread cached information, */
-struct task_rss_stat {
-	int events;	/* for synchronization threshold */
-	int count[NR_MM_COUNTERS];
-};
-#endif /* USE_SPLIT_PTE_PTLOCKS */
-
-struct mm_rss_stat {
-	atomic_long_t count[NR_MM_COUNTERS];
-};
-
 struct page_frag {
 	struct page *page;
 #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
include/linux/percpu_counter.h renamed from include/linux/percpu_counter.h~mm-convert-mms-rss-stats-into-percpu_counter +0 -1
@@ -13,7 +13,6 @@
 #include <linux/threads.h>
 #include <linux/percpu.h>
 #include <linux/types.h>
-#include <linux/gfp.h>
 
 /* percpu_counter batch for local add or sub */
 #define PERCPU_COUNTER_LOCAL_BATCH	INT_MAX
include/linux/sched.h renamed from include/linux/sched.h~mm-convert-mms-rss-stats-into-percpu_counter +0 -3
@@ -870,9 +870,6 @@ struct task_struct {
 	struct mm_struct		*mm;
 	struct mm_struct		*active_mm;
 
-#ifdef SPLIT_RSS_COUNTING
-	struct task_rss_stat		rss_stat;
-#endif
 	int				exit_state;
 	int				exit_code;
 	int				exit_signal;
include/trace/events/kmem.h renamed from include/trace/events/kmem.h~mm-convert-mms-rss-stats-into-percpu_counter +4 -4
@@ -346,10 +346,9 @@ TRACE_MM_PAGES
 TRACE_EVENT(rss_stat,
 
 	TP_PROTO(struct mm_struct *mm,
-		int member,
-		long count),
+		int member),
 
-	TP_ARGS(mm, member, count),
+	TP_ARGS(mm, member),
 
 	TP_STRUCT__entry(
 		__field(unsigned int, mm_id)
@@ -362,7 +361,8 @@ TRACE_EVENT(rss_stat,
 		__entry->mm_id = mm_ptr_to_hash(mm);
 		__entry->curr = !!(current->mm == mm);
 		__entry->member = member;
-		__entry->size = (count << PAGE_SHIFT);
+		__entry->size = (percpu_counter_sum_positive(&mm->rss_stat[member])
+							    << PAGE_SHIFT);
 	),
 
 	TP_printk("mm_id=%u curr=%d type=%s size=%ldB",
kernel/fork.c renamed from kernel/fork.c~mm-convert-mms-rss-stats-into-percpu_counter +15 -1
@@ -753,7 +753,7 @@ static void check_mm(struct mm_struct *m
 			 "Please make sure 'struct resident_page_types[]' is updated as well");
 
 	for (i = 0; i < NR_MM_COUNTERS; i++) {
-		long x = atomic_long_read(&mm->rss_stat.count[i]);
+		long x = percpu_counter_sum(&mm->rss_stat[i]);
 
 		if (unlikely(x))
 			pr_alert("BUG: Bad rss-counter state mm:%p type:%s val:%ld\n",
@@ -779,6 +779,8 @@ static void check_mm(struct mm_struct *m
  */
 void __mmdrop(struct mm_struct *mm)
 {
+	int i;
+
 	BUG_ON(mm == &init_mm);
 	WARN_ON_ONCE(mm == current->mm);
 	WARN_ON_ONCE(mm == current->active_mm);
@@ -788,6 +790,9 @@ void __mmdrop(struct mm_struct *mm)
 	check_mm(mm);
 	put_user_ns(mm->user_ns);
 	mm_pasid_drop(mm);
+
+	for (i = 0; i < NR_MM_COUNTERS; i++)
+		percpu_counter_destroy(&mm->rss_stat[i]);
 	free_mm(mm);
 }
 EXPORT_SYMBOL_GPL(__mmdrop);
@@ -1107,6 +1112,8 @@ static void mm_init_uprobes_state(struct
 static struct mm_struct *mm_init(struct mm_struct *mm, struct task_struct *p,
 	struct user_namespace *user_ns)
 {
+	int i;
+
 	mt_init_flags(&mm->mm_mt, MM_MT_FLAGS);
 	mt_set_external_lock(&mm->mm_mt, &mm->mmap_lock);
 	atomic_set(&mm->mm_users, 1);
@@ -1148,10 +1155,17 @@ static struct mm_struct *mm_init(struct
 	if (init_new_context(p, mm))
 		goto fail_nocontext;
 
+	for (i = 0; i < NR_MM_COUNTERS; i++)
+		if (percpu_counter_init(&mm->rss_stat[i], 0, GFP_KERNEL_ACCOUNT))
+			goto fail_pcpu;
+
 	mm->user_ns = get_user_ns(user_ns);
 	lru_gen_init_mm(mm);
 	return mm;
 
+fail_pcpu:
+	while (i > 0)
+		percpu_counter_destroy(&mm->rss_stat[--i]);
 fail_nocontext:
 	mm_free_pgd(mm);
 fail_nopgd:
mm/memory.c renamed from mm/memory.c~mm-convert-mms-rss-stats-into-percpu_counter +11 -62

123 lines not on this page · raw message

_

Patches currently in -mm which might be from [email protected] are