From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-wm0-f68.google.com ([74.125.82.68]:44352 "EHLO mail-wm0-f68.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752654AbdLGKW6 (ORCPT ); Thu, 7 Dec 2017 05:22:58 -0500 Received: by mail-wm0-f68.google.com with SMTP id t8so12090240wmc.3 for ; Thu, 07 Dec 2017 02:22:57 -0800 (PST) From: Jack Wang To: Greg Kroah-Hartman Cc: "Kirill A. Shutemov" , Andrea Arcangeli , Hillf Danton , stable@vger.kernel.org, Andrew Morton , Linus Torvalds , Jack Wang Subject: [PATCH 2/3] thp: fix MADV_DONTNEED vs. numa balancing race Date: Thu, 7 Dec 2017 11:22:51 +0100 Message-Id: <1512642172-7981-3-git-send-email-jinpu.wangl@profitbricks.com> In-Reply-To: <1512642172-7981-1-git-send-email-jinpu.wangl@profitbricks.com> References: <1512642172-7981-1-git-send-email-jinpu.wangl@profitbricks.com> Sender: stable-owner@vger.kernel.org List-ID: From: "Kirill A. Shutemov" commit ced108037c2aa542b3ed8b7afd1576064ad1362a upstream In case prot_numa, we are under down_read(mmap_sem). It's critical to not clear pmd intermittently to avoid race with MADV_DONTNEED which is also under down_read(mmap_sem): CPU0: CPU1: change_huge_pmd(prot_numa=1) pmdp_huge_get_and_clear_notify() madvise_dontneed() zap_pmd_range() pmd_trans_huge(*pmd) == 0 (without ptl) // skip the pmd set_pmd_at(); // pmd is re-established The race makes MADV_DONTNEED miss the huge pmd and don't clear it which may break userspace. Found by code analysis, never saw triggered. Link: http://lkml.kernel.org/r/20170302151034.27829-3-kirill.shutemov@linux.intel.com Signed-off-by: Kirill A. Shutemov Cc: Andrea Arcangeli Cc: Hillf Danton Cc: Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds [jwang: adjust context for 4.4] Signed-off-by: Jack Wang --- mm/huge_memory.c | 34 +++++++++++++++++++++++++++++++++- 1 file changed, 33 insertions(+), 1 deletion(-) diff --git a/mm/huge_memory.c b/mm/huge_memory.c index ea013cb..0127b78 100644 --- a/mm/huge_memory.c +++ b/mm/huge_memory.c @@ -1588,7 +1588,39 @@ int change_huge_pmd(struct vm_area_struct *vma, pmd_t *pmd, if (prot_numa && pmd_protnone(*pmd)) goto unlock; - entry = pmdp_huge_get_and_clear_notify(mm, addr, pmd); + /* + * In case prot_numa, we are under down_read(mmap_sem). It's critical + * to not clear pmd intermittently to avoid race with MADV_DONTNEED + * which is also under down_read(mmap_sem): + * + * CPU0: CPU1: + * change_huge_pmd(prot_numa=1) + * pmdp_huge_get_and_clear_notify() + * madvise_dontneed() + * zap_pmd_range() + * pmd_trans_huge(*pmd) == 0 (without ptl) + * // skip the pmd + * set_pmd_at(); + * // pmd is re-established + * + * The race makes MADV_DONTNEED miss the huge pmd and don't clear it + * which may break userspace. + * + * pmdp_invalidate() is required to make sure we don't miss + * dirty/young flags set by hardware. + */ + entry = *pmd; + pmdp_invalidate(vma, addr, pmd); + + /* + * Recover dirty/young flags. It relies on pmdp_invalidate to not + * corrupt them. + */ + if (pmd_dirty(*pmd)) + entry = pmd_mkdirty(entry); + if (pmd_young(*pmd)) + entry = pmd_mkyoung(entry); + entry = pmd_modify(entry, newprot); if (preserve_write) entry = pmd_mkwrite(entry); -- 2.7.4