From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757050Ab2EVVFI (ORCPT ); Tue, 22 May 2012 17:05:08 -0400 Received: from smtp102.prem.mail.ac4.yahoo.com ([76.13.13.41]:44891 "HELO smtp102.prem.mail.ac4.yahoo.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with SMTP id S1754933Ab2EVVFG (ORCPT ); Tue, 22 May 2012 17:05:06 -0400 X-Yahoo-Newman-Property: ymail-3 X-YMail-OSG: hdZcRcAVM1nUlM_T5fTYsqn3jsC4hn2IBWtbttVJthEysb1 NJ0HTcjtnakIXiW59uwC6.4ZMYJMfOZdjKWBLTuxxgIRyF0edgzfGltQd.RZ 8R.Cwlq95.8LhLSUz0HMin5PMWeySCrFS_RJCpnz7GvBpIL1EM68KcTWurpb ZJcbtKfvj_Btri5e.oUClyZRCjr3SRq1wXRK3TFzDr4zQ_Zei6X3OKELUKoA rp0qVmoIllZQZ0pSgS7Hyqn9hm3moCRoQvHINxZ2_qwZ0G.tKAWVO7Kj8S.F vi3gpnNTTI7fXPlABCUHynQfjw80KJVX3oKgYx.vaYc.rphrWWvoDx3DEf3H BEbzrvtCAYM1BP55GISRuGiBC3f6mHnNSm5izQDmCz83MFMR8v09F_aXiUt9 N X-Yahoo-SMTP: _Dag8S.swBC1p4FJKLCXbs8NQzyse1SYSgnAbY0- Date: Tue, 22 May 2012 16:05:02 -0500 (CDT) From: Christoph Lameter X-X-Sender: cl@router.home To: Andrew Morton cc: Dave Hansen , linux-kernel@vger.kernel.org, linux-mm@kvack.org, aarcange@redhat.com, kosaki.motohiro@jp.fujitsu.com, hughd@google.com, rientjes@google.com, adobriyan@gmail.com, mel@csn.ul.ie Subject: Re: [PATCH] hugetlb: fix resv_map leak in error path In-Reply-To: <20120522134558.49255899.akpm@linux-foundation.org> Message-ID: References: <20120521202814.E01F0FE1@kernel> <20120522134558.49255899.akpm@linux-foundation.org> User-Agent: Alpine 2.00 (DEB 1167 2008-08-23) MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Tue, 22 May 2012, Andrew Morton wrote: > On Mon, 21 May 2012 13:28:14 -0700 > Dave Hansen wrote: > > > When called for anonymous (non-shared) mappings, > > hugetlb_reserve_pages() does a resv_map_alloc(). It depends on > > code in hugetlbfs's vm_ops->close() to release that allocation. > > > > However, in the mmap() failure path, we do a plain unmap_region() > > without the remove_vma() which actually calls vm_ops->close(). > > > > This is a decent fix. This leak could get reintroduced if > > new code (say, after hugetlb_reserve_pages() in > > hugetlbfs_file_mmap()) decides to return an error. But, I think > > it would have to unroll the reservation anyway. > > How far back does this bug go? The patch applies to 3.4 but gets > rejects in 3.3 and earlier. The earliest that I have seen it on was 2.6.32. I have rediffed the patch against 2.6.32 and 3.2.0. ---- >>From dave@linux.vnet.ibm.com Fri May 18 13:50:17 2012 Date: Fri, 18 May 2012 11:46:30 -0700 From: Dave Hansen To: cl@linux.com Cc: linux-kernel@vger.kernel.org, linux-mm@kvack.org, aarcange@redhat.com, kosaki.motohiro@jp.fujitsu.com, hughd@google.com, rientjes@google.com, adobriyan@gmail.com, akpm@linux-foundation.org, mel@csn.ul.ie, Dave Hansen Subject: [RFC][PATCH] hugetlb: fix resv_map leak in error path When called for anonymous (non-shared) mappings, hugetlb_reserve_pages() does a resv_map_alloc(). It depends on code in hugetlbfs's vm_ops->close() to release that allocation. However, in the mmap() failure path, we do a plain unmap_region() without the remove_vma() which actually calls vm_ops->close(). This is a decent fix. This leak could get reintroduced if new code (say, after hugetlb_reserve_pages() in hugetlbfs_file_mmap()) decides to return an error. But, I think it would have to unroll the reservation anyway. This hasn't been extensively tested. Pretty much compile and boot tested along with Christoph's test case. Comments? Signed-off-by: Dave Hansen --- mm/hugetlb.c | 28 ++++++++++++++++++++++------ 1 file changed, 22 insertions(+), 6 deletions(-) Index: linux-3.2.0/mm/hugetlb.c =================================================================== --- linux-3.2.0.orig/mm/hugetlb.c 2012-01-04 17:55:44.000000000 -0600 +++ linux-3.2.0/mm/hugetlb.c 2012-05-22 03:32:37.760054550 -0500 @@ -2068,6 +2068,15 @@ static void hugetlb_vm_op_open(struct vm kref_get(&reservations->refs); } +static void resv_map_put(struct vm_area_struct *vma) +{ + struct resv_map *reservations = vma_resv_map(vma); + + if (!reservations) + return; + kref_put(&reservations->refs, resv_map_release); +} + static void hugetlb_vm_op_close(struct vm_area_struct *vma) { struct hstate *h = hstate_vma(vma); @@ -2083,7 +2092,7 @@ static void hugetlb_vm_op_close(struct v reserve = (end - start) - region_count(&reservations->regions, start, end); - kref_put(&reservations->refs, resv_map_release); + resv_map_put(vma); if (reserve) { hugetlb_acct_memory(h, -reserve); @@ -2883,12 +2892,16 @@ int hugetlb_reserve_pages(struct inode * set_vma_resv_flags(vma, HPAGE_RESV_OWNER); } - if (chg < 0) - return chg; + if (chg < 0) { + ret = chg; + goto out_err; + } /* There must be enough filesystem quota for the mapping */ - if (hugetlb_get_quota(inode->i_mapping, chg)) - return -ENOSPC; + if (hugetlb_get_quota(inode->i_mapping, chg)) { + ret = -ENOSPC; + goto out_err; + } /* * Check enough hugepages are available for the reservation. @@ -2897,7 +2910,7 @@ int hugetlb_reserve_pages(struct inode * ret = hugetlb_acct_memory(h, chg); if (ret < 0) { hugetlb_put_quota(inode->i_mapping, chg); - return ret; + goto out_err; } /* @@ -2914,6 +2927,9 @@ int hugetlb_reserve_pages(struct inode * if (!vma || vma->vm_flags & VM_MAYSHARE) region_add(&inode->i_mapping->private_list, from, to); return 0; +out_err: + resv_map_put(vma); + return ret; } void hugetlb_unreserve_pages(struct inode *inode, long offset, long freed)