From nobody Sat Feb 7 17:41:05 2026 Received: from mail-dy1-f176.google.com (mail-dy1-f176.google.com [74.125.82.176]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id CC94837F100 for ; Fri, 23 Jan 2026 01:46:35 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=74.125.82.176 ARC-Seal: i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1769132811; cv=none; b=KL5l80X50hbevbHOaDruPZSggPmM9bFQEautvV86faTzzqlPcNgvPyJLedB83RBR5QXkZwpP5sg128txRWgKhZyAuW7utX7UDYef0scIFaevgx9XIJ2ddX4Toz2geiTNsKnLJcYl/AtlJ96w/FRm2tx9lp1NyIlR9Q442M7PyJA= ARC-Message-Signature: i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1769132811; c=relaxed/simple; bh=xT8DNNVs3K7gOLtjevQypsshrPtLyrgA8Nb+h2IueCI=; h=From:To:Cc:Subject:Date:Message-ID:MIME-Version; b=NQrEkjwv1TYACfN7hCwj9kFtWlxhPMKk7dYKQTfJfI+PDlE00WbyNpcSs2n94GTN37QNoXtwk1r1oZxWhCgvAR9Jc+XUFHpT/Hy7rGmB2sJ4uJWLDSCwXg64BtNSPbtov6/xy08vzj8CdlJIZj9UBXjiSihXClMLDFp4iJ1NWtg= ARC-Authentication-Results: i=1; smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=gmail.com; spf=pass smtp.mailfrom=gmail.com; dkim=pass (2048-bit key) header.d=gmail.com header.i=@gmail.com header.b=WDDY1GIS; arc=none smtp.client-ip=74.125.82.176 Authentication-Results: smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=gmail.com Authentication-Results: smtp.subspace.kernel.org; spf=pass smtp.mailfrom=gmail.com Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=gmail.com header.i=@gmail.com header.b="WDDY1GIS" Received: by mail-dy1-f176.google.com with SMTP id 5a478bee46e88-2b71515d8adso1643966eec.1 for ; Thu, 22 Jan 2026 17:46:35 -0800 (PST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20230601; t=1769132794; x=1769737594; darn=vger.kernel.org; h=content-transfer-encoding:mime-version:message-id:date:subject:cc :to:from:from:to:cc:subject:date:message-id:reply-to; bh=typTWoY1NbQ7lW/9lPbT4WdyEQtUQii14Uxv5ycxQxY=; b=WDDY1GIShxV/2ToQk1v5Z1qkEhhvkQ09sdt241fYf31dllLH9tNxIWNMnTSPpmSFNs RHUTn7wW+trr8kOLRxnasCVSz7mDyv07laLd7isbVgk4ZtgsylQCH2YjPvl00uQnZ+jB cNxZztHLTmqt6z9Pu88b/e/TSabFammKWuihM8X4inGWHdbKFxiVgwO5ERBjPgeZoEuG k1bloep9+ltt/1sIQouJ8ne99cvLy0nM/ZhLsLbfK0FE1p14I96h85TrkhDjPiYkOTTf 10k9TKSJ1s8oAh92dsRRnTdJFM2Sx0nPiBlEn8fT8ksWYOICFQ+NEEmuwPoOY4vjp1VC rw8g== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20230601; t=1769132794; x=1769737594; h=content-transfer-encoding:mime-version:message-id:date:subject:cc :to:from:x-gm-gg:x-gm-message-state:from:to:cc:subject:date :message-id:reply-to; bh=typTWoY1NbQ7lW/9lPbT4WdyEQtUQii14Uxv5ycxQxY=; b=IYcL5iQ0xa7M3mKuEmwXtC0Q7xWWlESyYkZ5Jz0pbCVQ0LxwOw5aYvnsIxz/ucEdfG fTtkqW6Pe6WTzdI4k7KD4FnHO+ON1eLcmp75cbZPDxM3PXUdLywDLWItNKQsQ2r6Tb/a 5nvEQErjVvAXpjlErFB0F3Htvu9m1j61LqsejFsl4hQhU/QXYdYz8njuKaRSWy1nt2Zv sYdA1qNhWV09iYnbMHl7xLXbMMnz1IC2plQmzkwPaqie6siZ38gmhsig66y6Jasjjsu3 vwbw1p0sHH5665ulXkQGNWHuKQO+CmxQShKmeAAKxlFLEXNn1d72JzpS5/QZ9o472PBN CUsA== X-Forwarded-Encrypted: i=1; AJvYcCUH4wXm/QHq3dr/w0rEpYzcUm9OBJ83FOoKpAHIt/n2lnS4FwzuQn/myQc5kHyfELa6fPL0CvqOHtPVKuo=@vger.kernel.org X-Gm-Message-State: AOJu0YyF9L7oa6PKczVGpV3Zx6AIqnc5SFGTs0orSVRwIH5xYyhj0KES ym+k92e1HyIilideFTfP+/eqELsZQ0KmV8iHZdN4o4fK8FQJHMfrAkEh X-Gm-Gg: AZuq6aI8ZBaA4BKPleZsAzEG7j4EGm7v7D/1Qi0i1aMVii1QMfWI2YUg8dTFzLHNjb7 RjREAHSKds0SCNN7RBME/9yT+SHq/BM5vhCU3CxHxbYgfFhj3p/bjmngF0daxDjm0QOXqFOPA/H /bQZq49k+Y4jUtyN4KUVL4uuf4w9rAcFLF38M3HvKomvRHmKb79PQ05emW2peEE8GbwzqNADtAo lUPIhBjSwB4XlA2e6W9lbkNayqwv5UATufO0UJW4sZn+m14FSWjYg4fQrl2MP50DIBjYjkoMewG YC818Gl8RxXjQTzSsGwQ4UNdrHmw/pw8t36LoR45YAadIbuVO+HJBo+ATyA66VSuFH1Lu+WdEC3 +vD5DudUC0iEHvQ+bpeKAWx/O4STkLLgYWSfaVURFklusEuVA8uHn5S4d2A67FBEBoQQDo/0/sV qZVaY= X-Received: by 2002:a05:7301:fa0e:b0:2b6:c617:f79d with SMTP id 5a478bee46e88-2b7399978f2mr560822eec.13.1769132794331; Thu, 22 Jan 2026 17:46:34 -0800 (PST) Received: from debian ([74.48.213.230]) by smtp.gmail.com with ESMTPSA id 5a478bee46e88-2b73a6919dfsm1420424eec.4.2026.01.22.17.46.31 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Thu, 22 Jan 2026 17:46:34 -0800 (PST) From: Qiliang Yuan To: Ingo Molnar , Peter Zijlstra , Juri Lelli , Vincent Guittot Cc: Dietmar Eggemann , Steven Rostedt , Ben Segall , Mel Gorman , Valentin Schneider , linux-kernel@vger.kernel.org, Qiliang Yuan , Qiliang Yuan Subject: [PATCH v2] sched/numa: Optimize NUMA placement algorithm complexity from O(Nodes) to O(Active_Nodes) Date: Thu, 22 Jan 2026 20:46:26 -0500 Message-ID: <20260123014626.197138-1-realwujing@gmail.com> X-Mailer: git-send-email 2.51.0 Precedence: bulk X-Mailing-List: linux-kernel@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset="utf-8" On systems with a large number of NUMA nodes, periodic scanning of all onli= ne nodes for fault decay and task placement becomes a bottleneck. This patch introduces 'numa_faults_nodes_mask' in task_struct to track nodes where the task has actually incurred faults. By replacing for_each_online_n= ode() with for_each_node_mask(), we reduce the search space and decay overhead, especially for tasks whose memory footprint is localized to a few nodes. Signed-off-by: Qiliang Yuan Signed-off-by: Qiliang Yuan --- include/linux/sched.h | 1 + kernel/sched/fair.c | 18 +++++++++++++++--- 2 files changed, 16 insertions(+), 3 deletions(-) diff --git a/include/linux/sched.h b/include/linux/sched.h index d395f2810fac..2c426e10c9d5 100644 --- a/include/linux/sched.h +++ b/include/linux/sched.h @@ -1397,6 +1397,7 @@ struct task_struct { */ unsigned long *numa_faults; unsigned long total_numa_faults; + nodemask_t numa_faults_nodes_mask; =20 /* * numa_faults_locality tracks if faults recorded during the last diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c index e71302282671..44cf35c43684 100644 --- a/kernel/sched/fair.c +++ b/kernel/sched/fair.c @@ -2538,8 +2538,9 @@ static int task_numa_migrate(struct task_struct *p) */ ng =3D deref_curr_numa_group(p); if (env.best_cpu =3D=3D -1 || (ng && ng->active_nodes > 1)) { - for_each_node_state(nid, N_CPU) { - if (nid =3D=3D env.src_nid || nid =3D=3D p->numa_preferred_nid) + for_each_node_mask(nid, p->numa_faults_nodes_mask) { + if (nid =3D=3D env.src_nid || nid =3D=3D p->numa_preferred_nid || + !node_state(nid, N_CPU)) continue; =20 dist =3D node_distance(env.src_nid, env.dst_nid); @@ -2892,11 +2893,12 @@ static void task_numa_placement(struct task_struct = *p) } =20 /* Find the node with the highest number of faults */ - for_each_online_node(nid) { + for_each_node_mask(nid, p->numa_faults_nodes_mask) { /* Keep track of the offsets in numa_faults array */ int mem_idx, membuf_idx, cpu_idx, cpubuf_idx; unsigned long faults =3D 0, group_faults =3D 0; int priv; + bool node_has_faults =3D false; =20 for (priv =3D 0; priv < NR_NUMA_HINT_FAULT_TYPES; priv++) { long diff, f_diff, f_weight; @@ -2928,6 +2930,10 @@ static void task_numa_placement(struct task_struct *= p) p->numa_faults[cpu_idx] +=3D f_diff; faults +=3D p->numa_faults[mem_idx]; p->total_numa_faults +=3D diff; + + if (p->numa_faults[mem_idx] || p->numa_faults[cpu_idx]) + node_has_faults =3D true; + if (ng) { /* * safe because we can only change our own group @@ -2952,6 +2958,9 @@ static void task_numa_placement(struct task_struct *p) max_faults =3D group_faults; max_nid =3D nid; } + + if (!node_has_faults) + node_clear(nid, p->numa_faults_nodes_mask); } =20 /* Cannot migrate task to CPU-less node */ @@ -3209,6 +3218,8 @@ void task_numa_fault(int last_cpupid, int mem_node, i= nt pages, int flags) =20 p->numa_faults[task_faults_idx(NUMA_MEMBUF, mem_node, priv)] +=3D pages; p->numa_faults[task_faults_idx(NUMA_CPUBUF, cpu_node, priv)] +=3D pages; + node_set(mem_node, p->numa_faults_nodes_mask); + node_set(cpu_node, p->numa_faults_nodes_mask); p->numa_faults_locality[local] +=3D pages; } =20 @@ -3545,6 +3556,7 @@ void init_numa_balancing(u64 clone_flags, struct task= _struct *p) /* Protect against double add, see task_tick_numa and task_numa_work */ p->numa_work.next =3D &p->numa_work; p->numa_faults =3D NULL; + nodes_clear(p->numa_faults_nodes_mask); p->numa_pages_migrated =3D 0; p->total_numa_faults =3D 0; RCU_INIT_POINTER(p->numa_group, NULL); --=20 2.51.0