From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S934331Ab3AOW5A (ORCPT ); Tue, 15 Jan 2013 17:57:00 -0500 Received: from mail.kernel.org ([198.145.19.201]:48290 "EHLO mail.kernel.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S934301Ab3AOW46 (ORCPT ); Tue, 15 Jan 2013 17:56:58 -0500 From: Greg Kroah-Hartman To: linux-kernel@vger.kernel.org, stable@vger.kernel.org Cc: Greg Kroah-Hartman , alan@lxorguk.ukuu.org.uk, David Rientjes , Andrea Arcangeli , Johannes Weiner , KAMEZAWA Hiroyuki , Michal Hocko , Andrew Morton , Linus Torvalds Subject: [ 40/71] thp, memcg: split hugepage for memcg oom on cow Date: Tue, 15 Jan 2013 14:54:37 -0800 Message-Id: <20130115225343.315337142@linuxfoundation.org> X-Mailer: git-send-email 1.8.1.336.g94702dd In-Reply-To: <20130115225340.367496998@linuxfoundation.org> References: <20130115225340.367496998@linuxfoundation.org> User-Agent: quilt/0.60-1 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org 3.0-stable review patch. If anyone has any objections, please let me know. ------------------ From: David Rientjes commit 1f1d06c34f7675026326cd9f39ff91e4555cf355 upstream. On COW, a new hugepage is allocated and charged to the memcg. If the system is oom or the charge to the memcg fails, however, the fault handler will return VM_FAULT_OOM which results in an oom kill. Instead, it's possible to fallback to splitting the hugepage so that the COW results only in an order-0 page being allocated and charged to the memcg which has a higher liklihood to succeed. This is expensive because the hugepage must be split in the page fault handler, but it is much better than unnecessarily oom killing a process. Signed-off-by: David Rientjes Cc: Andrea Arcangeli Cc: Johannes Weiner Acked-by: KAMEZAWA Hiroyuki Cc: Michal Hocko Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds Signed-off-by: Greg Kroah-Hartman --- mm/huge_memory.c | 3 +++ mm/memory.c | 18 +++++++++++++++--- 2 files changed, 18 insertions(+), 3 deletions(-) --- a/mm/huge_memory.c +++ b/mm/huge_memory.c @@ -920,6 +920,8 @@ int do_huge_pmd_wp_page(struct mm_struct count_vm_event(THP_FAULT_FALLBACK); ret = do_huge_pmd_wp_page_fallback(mm, vma, address, pmd, orig_pmd, page, haddr); + if (ret & VM_FAULT_OOM) + split_huge_page(page); put_page(page); goto out; } @@ -927,6 +929,7 @@ int do_huge_pmd_wp_page(struct mm_struct if (unlikely(mem_cgroup_newpage_charge(new_page, mm, GFP_KERNEL))) { put_page(new_page); + split_huge_page(page); put_page(page); ret |= VM_FAULT_OOM; goto out; --- a/mm/memory.c +++ b/mm/memory.c @@ -3470,6 +3470,7 @@ int handle_mm_fault(struct mm_struct *mm if (unlikely(is_vm_hugetlb_page(vma))) return hugetlb_fault(mm, vma, address, flags); +retry: pgd = pgd_offset(mm, address); pud = pud_alloc(mm, pgd, address); if (!pud) @@ -3483,13 +3484,24 @@ int handle_mm_fault(struct mm_struct *mm pmd, flags); } else { pmd_t orig_pmd = *pmd; + int ret; + barrier(); if (pmd_trans_huge(orig_pmd)) { if (flags & FAULT_FLAG_WRITE && !pmd_write(orig_pmd) && - !pmd_trans_splitting(orig_pmd)) - return do_huge_pmd_wp_page(mm, vma, address, - pmd, orig_pmd); + !pmd_trans_splitting(orig_pmd)) { + ret = do_huge_pmd_wp_page(mm, vma, address, pmd, + orig_pmd); + /* + * If COW results in an oom, the huge pmd will + * have been split, so retry the fault on the + * pte for a smaller charge. + */ + if (unlikely(ret & VM_FAULT_OOM)) + goto retry; + return ret; + } return 0; } }