Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
From:   Omar Sandoval <osandov@osandov.com>
To:     linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org,
        Andrew Morton <akpm@linux-foundation.org>
Cc:     Alexey Dobriyan <adobriyan@gmail.com>,
        Eric Biederman <ebiederm@xmission.com>,
        James Morse <james.morse@arm.com>,
        Bhupesh Sharma <bhsharma@redhat.com>, kernel-team@fb.com
Subject: [PATCH v4 2/9] proc/kcore: don't grab lock for memory hotplug notifier
Date:   Wed, 25 Jul 2018 16:59:13 -0700
Message-Id: <93fa8dc895f40f2b85d706779b3cd2902567ff7a.1532563124.git.osandov@fb.com>
In-Reply-To: <cover.1532563124.git.osandov@fb.com>
References: <cover.1532563124.git.osandov@fb.com>
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

From: Omar Sandoval <osandov@fb.com>

The memory hotplug notifier kcore_callback() only needs kclist_lock to
prevent races with __kcore_update_ram(), but we can easily eliminate
that race by using an atomic xchg() in __kcore_update_ram(). This is
preparation for converting kclist_lock to an rwsem.

Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Omar Sandoval <osandov@fb.com>
---
 fs/proc/kcore.c | 6 ++----
 1 file changed, 2 insertions(+), 4 deletions(-)

diff --git a/fs/proc/kcore.c b/fs/proc/kcore.c
index b0b9a76f28d6..e83f15a4f66d 100644
--- a/fs/proc/kcore.c
+++ b/fs/proc/kcore.c
@@ -118,7 +118,7 @@ static void __kcore_update_ram(struct list_head *list)
 	LIST_HEAD(garbage);
 
 	write_lock(&kclist_lock);
-	if (kcore_need_update) {
+	if (xchg(&kcore_need_update, 0)) {
 		list_for_each_entry_safe(pos, tmp, &kclist_head, list) {
 			if (pos->type == KCORE_RAM
 				|| pos->type == KCORE_VMEMMAP)
@@ -127,7 +127,6 @@ static void __kcore_update_ram(struct list_head *list)
 		list_splice_tail(list, &kclist_head);
 	} else
 		list_splice(list, &garbage);
-	kcore_need_update = 0;
 	proc_root_kcore->size = get_kcore_size(&nphdr, &size);
 	write_unlock(&kclist_lock);
 
@@ -593,9 +592,8 @@ static int __meminit kcore_callback(struct notifier_block *self,
 	switch (action) {
 	case MEM_ONLINE:
 	case MEM_OFFLINE:
-		write_lock(&kclist_lock);
 		kcore_need_update = 1;
-		write_unlock(&kclist_lock);
+		break;
 	}
 	return NOTIFY_OK;
 }
-- 
2.18.0