linux-kernel - [PATCH] mm/mmap: do the check on expand instead of on insert

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [day] [month] [year] [list]

Message-Id: <20200917224722.54428-1-richard.weiyang@linux.alibaba.com>
Date:   Fri, 18 Sep 2020 06:47:22 +0800
From:   Wei Yang <richard.weiyang@...ux.alibaba.com>
To:     akpm@...ux-foundation.org
Cc:     linux-mm@...ck.org, linux-kernel@...r.kernel.org,
        Wei Yang <richard.weiyang@...ux.alibaba.com>
Subject: [PATCH] mm/mmap: do the check on expand instead of on insert

Function __vma_adjust() checks on *insert* to decide whether it is
necessary to remove_next/adjust_next. While it is more reasonable to do
the check on *expand* instead of on *insert*, since this is only necessary
when *expand* is non-NULL.

There are several users for __vma_adjust, let's classify them based on
the value of *insert*/*expand*:

   caller                  |    insert/expand
   ------------------------+-----------------
   vma_merge               |    NULL/non-NULL
   __split_vma             |    non-NULL/NULL
   shift_arg_pages/mremap  |    NULL/NULL

To make this change, we need to make sure those non-NULL *insert* cases
wouldn't go into this if branch except vma_merge. There are two cases we
need to take care of: shift_arg_pages and mremap. Let's look at it one
by one.

For mremap, it is for sure we won't go into this if branch since
vma_adjust tries to expand the vma and the vma is expandable(the end
wouldn't interact with vma and next).

For shift_arg_pages, this case is a little tricky. Actually, for this
case, vma->vm_next should be NULL. Otherwise we would go into the branch
of "end < vma->vm_end" since we are shifting left. This means we would
expand the vma->vm_next by accident. Luckily, in current code, we won't
fall into this situation because shift_arg_pages only shift the stack
which is the highest one in virtual space.

To make the code more easy to understand(only vma_merge has a non-NULL
expand), and to make it handle the corner case(shift_arg_pages)
properly, let's do the check on *expand* instead of *insert*.

Signed-off-by: Wei Yang <richard.weiyang@...ux.alibaba.com>
---
 mm/mmap.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mm/mmap.c b/mm/mmap.c
index 829897646a9c..ca31b405fbfa 100644
--- a/mm/mmap.c
+++ b/mm/mmap.c
@@ -706,7 +706,7 @@ int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
 	long adjust_next = 0;
 	int remove_next = 0;

-	if (next && !insert) {
+	if (next && expand) {
 		struct vm_area_struct *exporter = NULL, *importer = NULL;

 		if (end >= next->vm_end) {
-- 
2.20.1 (Apple Git-117)