Received-SPF: pass (google.com: domain of linux-kernel+bounces-142652-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.48.161 as permitted sender) client-ip=147.75.48.161;
From: Barry Song <21cnbao@gmail.com>
To: akpm@linux-foundation.org,
	willy@infradead.org,
	vbabka@suse.cz,
	hannes@cmpxchg.org,
	linux-mm@kvack.org
Cc: linux-kernel@vger.kernel.org,
	Barry Song <v-songbaohua@oppo.com>,
	Ryan Roberts <ryan.roberts@arm.com>
Subject: [RFC PATCH] mm/vmstat: sum up all possible CPUs instead of using vm_events_fold_cpu
Date: Sat, 13 Apr 2024 00:30:39 +1200
Message-Id: <20240412123039.442743-1-21cnbao@gmail.com>
Precedence: bulk
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit

From: Barry Song <v-songbaohua@oppo.com>

When unplugging a CPU, the current code merges its vm_events
with an online CPU. Because, during summation, it only considers
online CPUs, which is a crude workaround. By transitioning to
summing up all possible CPUs, we can eliminate the need for
vm_events_fold_cpu.

Suggested-by: Ryan Roberts <ryan.roberts@arm.com>
Signed-off-by: Barry Song <v-songbaohua@oppo.com>
---
 -v1:
 originally suggested by Ryan while he reviewed mTHP counters
 patchset[1]; I am also applying this suggestion to vm_events

 [1] https://lore.kernel.org/linux-mm/ca73cbf1-8304-4790-a721-3c3a42f9d293@arm.com/

 include/linux/vmstat.h |  5 -----
 mm/page_alloc.c        |  8 --------
 mm/vmstat.c            | 19 +------------------
 3 files changed, 1 insertion(+), 31 deletions(-)

diff --git a/include/linux/vmstat.h b/include/linux/vmstat.h
index 735eae6e272c..f7eaeb8bfa47 100644
--- a/include/linux/vmstat.h
+++ b/include/linux/vmstat.h
@@ -83,8 +83,6 @@ static inline void count_vm_events(enum vm_event_item item, long delta)
 
 extern void all_vm_events(unsigned long *);
 
-extern void vm_events_fold_cpu(int cpu);
-
 #else
 
 /* Disable counters */
@@ -103,9 +101,6 @@ static inline void __count_vm_events(enum vm_event_item item, long delta)
 static inline void all_vm_events(unsigned long *ret)
 {
 }
-static inline void vm_events_fold_cpu(int cpu)
-{
-}
 
 #endif /* CONFIG_VM_EVENT_COUNTERS */
 
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index b51becf03d1e..640d5752e400 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -5832,14 +5832,6 @@ static int page_alloc_cpu_dead(unsigned int cpu)
 	mlock_drain_remote(cpu);
 	drain_pages(cpu);
 
-	/*
-	 * Spill the event counters of the dead processor
-	 * into the current processors event counters.
-	 * This artificially elevates the count of the current
-	 * processor.
-	 */
-	vm_events_fold_cpu(cpu);
-
 	/*
 	 * Zero the differential counters of the dead processor
 	 * so that the vm statistics are consistent.
diff --git a/mm/vmstat.c b/mm/vmstat.c
index db79935e4a54..bae6ff476870 100644
--- a/mm/vmstat.c
+++ b/mm/vmstat.c
@@ -114,7 +114,7 @@ static void sum_vm_events(unsigned long *ret)
 
 	memset(ret, 0, NR_VM_EVENT_ITEMS * sizeof(unsigned long));
 
-	for_each_online_cpu(cpu) {
+	for_each_possible_cpu(cpu) {
 		struct vm_event_state *this = &per_cpu(vm_event_states, cpu);
 
 		for (i = 0; i < NR_VM_EVENT_ITEMS; i++)
@@ -135,23 +135,6 @@ void all_vm_events(unsigned long *ret)
 }
 EXPORT_SYMBOL_GPL(all_vm_events);
 
-/*
- * Fold the foreign cpu events into our own.
- *
- * This is adding to the events on one processor
- * but keeps the global counts constant.
- */
-void vm_events_fold_cpu(int cpu)
-{
-	struct vm_event_state *fold_state = &per_cpu(vm_event_states, cpu);
-	int i;
-
-	for (i = 0; i < NR_VM_EVENT_ITEMS; i++) {
-		count_vm_events(i, fold_state->event[i]);
-		fold_state->event[i] = 0;
-	}
-}
-
 #endif /* CONFIG_VM_EVENT_COUNTERS */
 
 /*
-- 
2.34.1