From nobody Sat Jul 25 03:20:10 2026 Received: from mail-pg1-f178.google.com (mail-pg1-f178.google.com [209.85.215.178]) (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 B06662765DF for ; Mon, 20 Jul 2026 03:19:33 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=209.85.215.178 ARC-Seal: i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1784517575; cv=none; b=dCqgbh42IiCLjCDkHlDlS9Q2BR7pcTb/stcNVMwsyXdW2ToP740L7JkC3N8H056NtGAdB6qNEeDOM7hoJtn/F5bTROL02QXzZL8JfOqfDmSvyjxbeJlyh+qqvtAzo5WYxFo0bRwMnszldF8ZNmSa3hVHsHz/Z7ixK2Dp3p3vCA4= ARC-Message-Signature: i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1784517575; c=relaxed/simple; bh=+nd0ZPd9dH2GOmF1MkjC1JoKQOltjiZpZRbJFkdi3wc=; h=From:To:Cc:Subject:Date:Message-ID:MIME-Version; b=MTulO7CyFqk0qknDDu2K3WhpbOhZebAXGLYIc8/R8ygqqWVgennXwC9VG4p55xos/+2znRogGSnXGxtl0+KDpa47puSUV0s0TVym8Fm8L11NZ1q8VFsY+rOrTqLQeG38eC8N0/K1EwyNCR56CJpb9N5WBfBoeqJFijLavf6fCiM= 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=DyVdqpga; arc=none smtp.client-ip=209.85.215.178 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="DyVdqpga" Received: by mail-pg1-f178.google.com with SMTP id 41be03b00d2f7-c95d0a54ea5so7064106a12.0 for ; Sun, 19 Jul 2026 20:19:33 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20251104; t=1784517573; x=1785122373; 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:content-type; bh=v2IFBQgGLhV/2VZq8Hcmtgn4n+KDnEkgWAqmJMMgNfY=; b=DyVdqpga1SEIVGLctFqvDFUWFH6pi4bGlddumhec47dMu8mWAYXFDC0ZGg8GCrNOPk 5ps+rYkCUQ4lC1PLAZma5F+oPDVgsJCExDpNX+wWxEmB0z8I3x0nI7NcJtimglK2H9DV z6tfi7XrI0N8TCa1YDkrjlaYlQrPoSh5vGxKSljG3ExAycShNKuQq9qbTAff1w3RWZSZ eEiiKj3uzZB+FINKDkuWQBpydxCAly4bLI7skGQQChLyxr0/D4F8cxWRZS03zBDFiLPt TWSfqT0mzaYII/67dbxGG0qjrPdf0GvQZF52QAp6q9gqNUq3xzgFOMq9xTFupkYnby4f lT1w== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20251104; t=1784517573; x=1785122373; 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:content-type; bh=v2IFBQgGLhV/2VZq8Hcmtgn4n+KDnEkgWAqmJMMgNfY=; b=M6lUv9ciWOqtyjLtlMBYedAsjEGXqA0VG7tuTBqbZcfY4dkzlhkksjCFqbv3DTCEG3 VPHKyhyCGxbZpNjpe9/jBncnKFGXPDyIg8TQJP4y+CnF9JmT5aaGQgBCD2/ByJTgloM2 T13PkNsqKa0XE5n+losZf3JMTkvPamJqbhCYzpyRbN2iiAUQ7nRVmoauCF4AfNqkALMo RXCNbm9RxTXknBQXi2obQcKMVb4dnADrR0bh/rNKMzhBqg5qNntLXAzMdlpiFZtFrFJP L97u9wx2FxIgrPX3J5eI7xN9TGnYALwlZFKIGFJUw7Mq+6xpxZlucconAiykVgh72B4j GenQ== X-Gm-Message-State: AOJu0YyKJVVPEabo4V5PBwp3R21zSmTSvl0QBLmnNitDUtGhe4qxlMR+ tPn+0hojC5ePdHM1Z0wDBHIudokL9t1f8IDtqyW0xnMpkblFlxDaiLPB X-Gm-Gg: AfdE7cmwCg8BxxHEC27kCI9XgT5TQz4xoSjjPqHzwNSYGypi0LmvmAof7auIpgYhoM1 8oMQ9NZWyC3dvZKVZfSSGYufRHt/irQOJJA4xmmoucrXqCRBvgFjuoyciTIVVPq5MJ2G5/51VfD 8LHDzqq6gqM2SDZYv+6t9wM9BYqrqi4/0fx1h9cZadikLpe1MqdhNQVmQBcdVRGJfcbj8Z7MGUO HQIOtjhaf1Sn7QyjNgTPLmr6N90QrD/SnJUU1DcnbrYn4XAEKoEGwWhi2W464Oro3JEVNvdN4N9 aVXRr04obhaQQxLOoT0rELMuKMdZl2DXv/RlNF+PE5o2L2G01+Hine/fxaZrxZ6jfNtCAYTs0dr 45BLIc7GHidS9YNAMczo2NkTqqQLQB7CdaqOZGRRZFqcZ/zjfG5D+a+OXiil/6xqNHspSN2Pe3Q Zwz3GHFm5BJnFBw+uC X-Received: by 2002:a05:6a21:114c:b0:3c3:8f3d:832f with SMTP id adf61e73a8af0-3c3ad9620f6mr12983309637.49.1784517572647; Sun, 19 Jul 2026 20:19:32 -0700 (PDT) Received: from localhost.localdomain ([221.227.131.254]) by smtp.gmail.com with ESMTPSA id 41be03b00d2f7-cb517aada63sm3827465a12.3.2026.07.19.20.19.29 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Sun, 19 Jul 2026 20:19:32 -0700 (PDT) From: Vernon Yang To: akpm@linux-foundation.org, david@kernel.org, ljs@kernel.org, liam@infradead.org, vbabka@kernel.org, jannh@google.com, baohua@kernel.org Cc: linux-kernel@vger.kernel.org, linux-mm@kvack.org, Vernon Yang Subject: [RFC PATCH] mm: madvise: use per_vma lock for MADV_COLD/PAGEOUT Date: Mon, 20 Jul 2026 11:19:01 +0800 Message-ID: <20260720031901.1724737-1-vernon2gm@gmail.com> X-Mailer: git-send-email 2.53.0 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" From: Vernon Yang MADV_COLD and MADV_PAGEOUT hold the mmap_lock read lock while walking the target range. When a concurrent mmap_lock write lock is held, MADV_COLD/PAGEOUT must wait and execute serially, even for unrelated VMAs. MADV_FREE/DONTNEED have already achieved this goal using the per_vma read lock. With only simple modifications, MADV_COLD/PAGEOUT can also use the per_vma read lock to replace the mmap_lock read lock. Start 16 threads running mmap/unmap on separate VMAs, and benchmark the time taken by 10,000 madvise(MADV_COLD/PAGEOUT) calls on another VMA[1]. before after MADV_COLD 8.09 s 0.07 s MADV_PAGEOUT 11.68 s 0.40 s [1] https://github.com/vernon2gh/app_and_module/blob/main/madvise/app.c Signed-off-by: Vernon Yang --- mm/madvise.c | 52 ++++++++++++++++++++++++++++------------------------ 1 file changed, 28 insertions(+), 24 deletions(-) diff --git a/mm/madvise.c b/mm/madvise.c index cd9bb077072c..795a8a6fb237 100644 --- a/mm/madvise.c +++ b/mm/madvise.c @@ -566,10 +566,19 @@ static int madvise_cold_or_pageout_pte_range(pmd_t *p= md, return 0; } =20 -static const struct mm_walk_ops cold_walk_ops =3D { - .pmd_entry =3D madvise_cold_or_pageout_pte_range, - .walk_lock =3D PGWALK_RDLOCK, -}; +static inline enum page_walk_lock get_walk_lock(enum madvise_lock_mode mod= e) +{ + switch (mode) { + case MADVISE_VMA_READ_LOCK: + return PGWALK_VMA_RDLOCK_VERIFY; + case MADVISE_MMAP_READ_LOCK: + return PGWALK_RDLOCK; + default: + /* Other modes don't require fixing up the walk_lock */ + WARN_ON_ONCE(1); + return PGWALK_RDLOCK; + } +} =20 static void madvise_cold_page_range(struct mmu_gather *tlb, struct madvise_behavior *madv_behavior) @@ -577,13 +586,17 @@ static void madvise_cold_page_range(struct mmu_gather= *tlb, { struct vm_area_struct *vma =3D madv_behavior->vma; struct madvise_behavior_range *range =3D &madv_behavior->range; + struct mm_walk_ops walk_ops =3D { + .pmd_entry =3D madvise_cold_or_pageout_pte_range, + }; struct madvise_walk_private walk_private =3D { .pageout =3D false, .tlb =3D tlb, }; =20 + walk_ops.walk_lock =3D get_walk_lock(madv_behavior->lock_mode); tlb_start_vma(tlb, vma); - walk_page_range_vma(vma, range->start, range->end, &cold_walk_ops, + walk_page_range_vma(vma, range->start, range->end, &walk_ops, &walk_private); tlb_end_vma(tlb, vma); } @@ -611,15 +624,20 @@ static long madvise_cold(struct madvise_behavior *mad= v_behavior) =20 static void madvise_pageout_page_range(struct mmu_gather *tlb, struct vm_area_struct *vma, - struct madvise_behavior_range *range) + struct madvise_behavior *madv_behavior) { + struct madvise_behavior_range *range =3D &madv_behavior->range; + struct mm_walk_ops walk_ops =3D { + .pmd_entry =3D madvise_cold_or_pageout_pte_range, + }; struct madvise_walk_private walk_private =3D { .pageout =3D true, .tlb =3D tlb, }; =20 + walk_ops.walk_lock =3D get_walk_lock(madv_behavior->lock_mode); tlb_start_vma(tlb, vma); - walk_page_range_vma(vma, range->start, range->end, &cold_walk_ops, + walk_page_range_vma(vma, range->start, range->end, &walk_ops, &walk_private); tlb_end_vma(tlb, vma); } @@ -644,7 +662,7 @@ static long madvise_pageout(struct madvise_behavior *ma= dv_behavior) =20 lru_add_drain(); tlb_gather_mmu(&tlb, madv_behavior->mm); - madvise_pageout_page_range(&tlb, vma, &madv_behavior->range); + madvise_pageout_page_range(&tlb, vma, madv_behavior); tlb_finish_mmu(&tlb); =20 return 0; @@ -783,20 +801,6 @@ static int madvise_free_pte_range(pmd_t *pmd, unsigned= long addr, return 0; } =20 -static inline enum page_walk_lock get_walk_lock(enum madvise_lock_mode mod= e) -{ - switch (mode) { - case MADVISE_VMA_READ_LOCK: - return PGWALK_VMA_RDLOCK_VERIFY; - case MADVISE_MMAP_READ_LOCK: - return PGWALK_RDLOCK; - default: - /* Other modes don't require fixing up the walk_lock */ - WARN_ON_ONCE(1); - return PGWALK_RDLOCK; - } -} - static int madvise_free_single_vma(struct madvise_behavior *madv_behavior) { struct mm_struct *mm =3D madv_behavior->mm; @@ -1748,8 +1752,6 @@ static enum madvise_lock_mode get_lock_mode(struct ma= dvise_behavior *madv_behavi switch (madv_behavior->behavior) { case MADV_REMOVE: case MADV_WILLNEED: - case MADV_COLD: - case MADV_PAGEOUT: case MADV_POPULATE_READ: case MADV_POPULATE_WRITE: case MADV_COLLAPSE: @@ -1759,6 +1761,8 @@ static enum madvise_lock_mode get_lock_mode(struct ma= dvise_behavior *madv_behavi case MADV_DONTNEED: case MADV_DONTNEED_LOCKED: case MADV_FREE: + case MADV_COLD: + case MADV_PAGEOUT: return MADVISE_VMA_READ_LOCK; default: return MADVISE_MMAP_WRITE_LOCK; --=20 2.53.0