From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751814AbcBNFdA (ORCPT ); Sun, 14 Feb 2016 00:33:00 -0500 Received: from e31.co.us.ibm.com ([32.97.110.149]:53552 "EHLO e31.co.us.ibm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751386AbcBNFc7 (ORCPT ); Sun, 14 Feb 2016 00:32:59 -0500 X-IBM-Helo: d03dlp01.boulder.ibm.com X-IBM-MailFrom: aneesh.kumar@linux.vnet.ibm.com X-IBM-RcptTo: linux-kernel@vger.kernel.org From: "Aneesh Kumar K.V" To: Michael Ellerman , benh@kernel.crashing.org, paulus@samba.org, akpm@linux-foundation.org, Mel Gorman , "Kirill A. Shutemov" Cc: linux-mm@kvack.org, linuxppc-dev@lists.ozlabs.org, linux-kernel@vger.kernel.org Subject: Re: [V3] powerpc/mm: Fix Multi hit ERAT cause by recent THP update In-Reply-To: <20160209121606.EA46C140B97@ozlabs.org> References: <20160209121606.EA46C140B97@ozlabs.org> User-Agent: Notmuch/0.20.2 (http://notmuchmail.org) Emacs/24.5.1 (x86_64-pc-linux-gnu) Date: Sun, 14 Feb 2016 11:02:47 +0530 Message-ID: <877fi7ambk.fsf@linux.vnet.ibm.com> MIME-Version: 1.0 Content-Type: text/plain X-TM-AS-MML: disable X-Content-Scanned: Fidelis XPS MAILER x-cbid: 16021405-8236-0000-0000-00001619B677 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Michael Ellerman writes: > On Tue, 2016-09-02 at 01:20:31 UTC, "Aneesh Kumar K.V" wrote: >> With ppc64 we use the deposited pgtable_t to store the hash pte slot >> information. We should not withdraw the deposited pgtable_t without >> marking the pmd none. This ensure that low level hash fault handling >> will skip this huge pte and we will handle them at upper levels. >> >> Recent change to pmd splitting changed the above in order to handle the >> race between pmd split and exit_mmap. The race is explained below. >> >> Consider following race: >> >> CPU0 CPU1 >> shrink_page_list() >> add_to_swap() >> split_huge_page_to_list() >> __split_huge_pmd_locked() >> pmdp_huge_clear_flush_notify() >> // pmd_none() == true >> exit_mmap() >> unmap_vmas() >> zap_pmd_range() >> // no action on pmd since pmd_none() == true >> pmd_populate() >> >> As result the THP will not be freed. The leak is detected by check_mm(): >> >> BUG: Bad rss-counter state mm:ffff880058d2e580 idx:1 val:512 >> >> The above required us to not mark pmd none during a pmd split. >> >> The fix for ppc is to clear the huge pte of _PAGE_USER, so that low >> level fault handling code skip this pte. At higher level we do take ptl >> lock. That should serialze us against the pmd split. Once the lock is >> acquired we do check the pmd again using pmd_same. That should always >> return false for us and hence we should retry the access. We do the >> pmd_same check in all case after taking plt with >> THP (do_huge_pmd_wp_page, do_huge_pmd_numa_page and >> huge_pmd_set_accessed) >> >> Also make sure we wait for irq disable section in other cpus to finish >> before flipping a huge pte entry with a regular pmd entry. Code paths >> like find_linux_pte_or_hugepte depend on irq disable to get >> a stable pte_t pointer. A parallel thp split need to make sure we >> don't convert a pmd pte to a regular pmd entry without waiting for the >> irq disable section to finish. >> >> Acked-by: Kirill A. Shutemov >> Signed-off-by: Aneesh Kumar K.V > > Applied to powerpc fixes, thanks. > > https://git.kernel.org/powerpc/c/9db4cd6c21535a4846b38808f3 > Can we apply the below hunk ?. The reason for marking pmd none was to avoid clearing both _PAGE_USER and _PAGE_PRESENT on the pte. At pmd level that used to mean a hugepd pointer before. We did fix that earlier by introducing _PAGE_PTE. But then I was thinking it was harmless to mark pmd none. Now marking it one will still result in the race I explained above, eventhough the window is much smaller now. diff --git a/arch/powerpc/mm/pgtable_64.c b/arch/powerpc/mm/pgtable_64.c index c8a00da39969..03f6e72697d0 100644 --- a/arch/powerpc/mm/pgtable_64.c +++ b/arch/powerpc/mm/pgtable_64.c @@ -694,7 +694,7 @@ void set_pmd_at(struct mm_struct *mm, unsigned long addr, void pmdp_invalidate(struct vm_area_struct *vma, unsigned long address, pmd_t *pmdp) { - pmd_hugepage_update(vma->vm_mm, address, pmdp, ~0UL, 0); + pmd_hugepage_update(vma->vm_mm, address, pmdp, _PAGE_PRESENT, 0); /* * This ensures that generic code that rely on IRQ disabling * to prevent a parallel THP split work as expected.