From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753439Ab1LUM1K (ORCPT ); Wed, 21 Dec 2011 07:27:10 -0500 Received: from merlin.infradead.org ([205.233.59.134]:35558 "EHLO merlin.infradead.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751245Ab1LUM1J convert rfc822-to-8bit (ORCPT ); Wed, 21 Dec 2011 07:27:09 -0500 Message-ID: <1324470416.10752.1.camel@twins> Subject: [PATCH] futex: Fix uninterruptble loop due to gate_area From: Peter Zijlstra To: Thomas Gleixner , Sasha Levin , Andrew Morton Cc: linux-kernel , "linux-mm@kvack.org" Date: Wed, 21 Dec 2011 13:26:56 +0100 Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: 8BIT X-Mailer: Evolution 3.2.1- Mime-Version: 1.0 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Subject: futex: Fix uninterruptble loop due to gate_area From: Peter Zijlstra Date: Fri Dec 02 14:12:06 CET 2011 It was found (by Sasha) that if you use a futex located in the gate area we get stuck in an uninterruptible infinite loop, much like the ZERO_PAGE issue. While looking at this problem, I realized you'll get into similar trouble when hitting any install_special_pages() mapping. The solution chosen was not to modify special_mapping_fault() to install a non-zero page->mapping because that might lead to issues when freeing these pages. Instead do a find_vma() when we find we're again in the !mapping branch. Cc: stable@kernel.org Reported-by: Sasha Levin Signed-off-by: Peter Zijlstra --- include/linux/mm.h | 1 + kernel/futex.c | 40 +++++++++++++++++++++++++++++++++++----- mm/mmap.c | 5 +++++ 3 files changed, 41 insertions(+), 5 deletions(-) Index: linux-2.6/include/linux/mm.h =================================================================== --- linux-2.6.orig/include/linux/mm.h +++ linux-2.6/include/linux/mm.h @@ -1394,6 +1394,7 @@ extern int may_expand_vm(struct mm_struc extern int install_special_mapping(struct mm_struct *mm, unsigned long addr, unsigned long len, unsigned long flags, struct page **pages); +extern bool is_special_mapping(struct vm_area_struct *vma); extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); Index: linux-2.6/kernel/futex.c =================================================================== --- linux-2.6.orig/kernel/futex.c +++ linux-2.6/kernel/futex.c @@ -59,6 +59,7 @@ #include #include #include +#include #include @@ -236,7 +237,7 @@ get_futex_key(u32 __user *uaddr, int fsh unsigned long address = (unsigned long)uaddr; struct mm_struct *mm = current->mm; struct page *page, *page_head; - int err, ro = 0; + int err, ro = 0, no_mapping_tries = 0; /* * The futex address must be "naturally" aligned. @@ -317,13 +318,42 @@ get_futex_key(u32 __user *uaddr, int fsh if (!page_head->mapping) { unlock_page(page_head); put_page(page_head); + /* - * ZERO_PAGE pages don't have a mapping. Avoid a busy loop - * trying to find one. RW mapping would have COW'd (and thus - * have a mapping) so this page is RO and won't ever change. - */ + * ZERO_PAGE pages don't have a mapping. Avoid a busy loop + * trying to find one. RW mapping would have COW'd (and thus + * have a mapping) so this page is RO and won't ever change. + */ if ((page_head == ZERO_PAGE(address))) return -EFAULT; + + /* + * Similar problem for the gate area. + */ + if (in_gate_area(mm, address)) + return -EFAULT; + + /* + * There is a special class of pages that will have no mapping + * and yet is perfectly valid and not going anywhere. These + * are the pages from install_special_mapping(). Since looking + * up the vma is expensive, don't do so on the first go round. + */ + if (no_mapping_tries) { + struct vm_area_struct *vma; + + err = 0; + down_read(&mm->mmap_sem); + vma = find_vma(mm, address); + if (vma && is_special_mapping(vma)) + err = -EFAULT; + up_read(&mm->mmap_sem); + + if (err) + return err; + } + + ++no_mapping_tries; goto again; } Index: linux-2.6/mm/mmap.c =================================================================== --- linux-2.6.orig/mm/mmap.c +++ linux-2.6/mm/mmap.c @@ -2479,6 +2479,11 @@ int install_special_mapping(struct mm_st return ret; } +bool is_special_mapping(struct vm_area_struct *vma) +{ + return vma->vm_ops == &special_mapping_vmops; +} + static DEFINE_MUTEX(mm_all_locks_mutex); static void vm_lock_anon_vma(struct mm_struct *mm, struct anon_vma *anon_vma)