Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
From:   mateusznosek0@gmail.com
To:     linux-mm@kvack.org, linux-kernel@vger.kernel.org
Cc:     Mateusz Nosek <mateusznosek0@gmail.com>, akpm@linux-foundation.org
Subject: [RFC PATCH] mm/page_alloc.c: micro-optimization reduce oom critical section size
Date:   Mon, 14 Sep 2020 12:06:54 +0200
Message-Id: <20200914100654.21746-1-mateusznosek0@gmail.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

From: Mateusz Nosek <mateusznosek0@gmail.com>

Most operations from '__alloc_pages_may_oom' do not require oom_mutex hold.
Exception is 'out_of_memory'. The patch refactors '__alloc_pages_may_oom'
to reduce critical section size and improve overall system performance.

Signed-off-by: Mateusz Nosek <mateusznosek0@gmail.com>
---
 mm/page_alloc.c | 45 ++++++++++++++++++++++++---------------------
 1 file changed, 24 insertions(+), 21 deletions(-)

diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index b9bd75cacf02..b07f950a5825 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -3935,18 +3935,7 @@ __alloc_pages_may_oom(gfp_t gfp_mask, unsigned int order,
 		.order = order,
 	};
 	struct page *page;
-
-	*did_some_progress = 0;
-
-	/*
-	 * Acquire the oom lock.  If that fails, somebody else is
-	 * making progress for us.
-	 */
-	if (!mutex_trylock(&oom_lock)) {
-		*did_some_progress = 1;
-		schedule_timeout_uninterruptible(1);
-		return NULL;
-	}
+	bool success;
 
 	/*
 	 * Go through the zonelist yet one more time, keep very high watermark
@@ -3959,14 +3948,17 @@ __alloc_pages_may_oom(gfp_t gfp_mask, unsigned int order,
 				      ~__GFP_DIRECT_RECLAIM, order,
 				      ALLOC_WMARK_HIGH|ALLOC_CPUSET, ac);
 	if (page)
-		goto out;
+		return page;
+
+	/* Check if somebody else is making progress for us. */
+	*did_some_progress = mutex_is_locked(&oom_lock);
 
 	/* Coredumps can quickly deplete all memory reserves */
 	if (current->flags & PF_DUMPCORE)
-		goto out;
+		return NULL;
 	/* The OOM killer will not help higher order allocs */
 	if (order > PAGE_ALLOC_COSTLY_ORDER)
-		goto out;
+		return NULL;
 	/*
 	 * We have already exhausted all our reclaim opportunities without any
 	 * success so it is time to admit defeat. We will skip the OOM killer
@@ -3976,12 +3968,12 @@ __alloc_pages_may_oom(gfp_t gfp_mask, unsigned int order,
 	 * The OOM killer may not free memory on a specific node.
 	 */
 	if (gfp_mask & (__GFP_RETRY_MAYFAIL | __GFP_THISNODE))
-		goto out;
+		return NULL;
 	/* The OOM killer does not needlessly kill tasks for lowmem */
 	if (ac->highest_zoneidx < ZONE_NORMAL)
-		goto out;
+		return NULL;
 	if (pm_suspended_storage())
-		goto out;
+		return NULL;
 	/*
 	 * XXX: GFP_NOFS allocations should rather fail than rely on
 	 * other request to make a forward progress.
@@ -3992,8 +3984,20 @@ __alloc_pages_may_oom(gfp_t gfp_mask, unsigned int order,
 	 * failures more gracefully we should just bail out here.
 	 */
 
+	/*
+	 * Acquire the oom lock.  If that fails, somebody else is
+	 * making progress for us.
+	 */
+	if (!mutex_trylock(&oom_lock)) {
+		*did_some_progress = 1;
+		schedule_timeout_uninterruptible(1);
+		return NULL;
+	}
+	success = out_of_memory(&oc);
+	mutex_unlock(&oom_lock);
+
 	/* Exhausted what can be done so it's blame time */
-	if (out_of_memory(&oc) || WARN_ON_ONCE(gfp_mask & __GFP_NOFAIL)) {
+	if (success || WARN_ON_ONCE(gfp_mask & __GFP_NOFAIL)) {
 		*did_some_progress = 1;
 
 		/*
@@ -4004,8 +4008,7 @@ __alloc_pages_may_oom(gfp_t gfp_mask, unsigned int order,
 			page = __alloc_pages_cpuset_fallback(gfp_mask, order,
 					ALLOC_NO_WATERMARKS, ac);
 	}
-out:
-	mutex_unlock(&oom_lock);
+
 	return page;
 }
 
-- 
2.20.1