[PATCH 5.15 v3] mm/filemap: fix UAF in find_lock_entries

[PATCH 5.15 v3] mm/filemap: fix UAF in find_lock_entries

[Liu Shixin]

Release refcount after xas_set to fix UAF which may cause panic like this:

 page:ffffea000491fa40 refcount:1 mapcount:0 mapping:0000000000000000 index:0x1 pfn:0x1247e9
 head:ffffea000491fa00 order:3 compound_mapcount:0 compound_pincount:0
 memcg:ffff888104f91091
 flags: 0x2fffff80010200(slab|head|node=0|zone=2|lastcpupid=0x1fffff)
...
page dumped because: VM_BUG_ON_PAGE(PageTail(page))
 ------------[ cut here ]------------
 kernel BUG at include/linux/page-flags.h:632!
 invalid opcode: 0000 [#1] SMP DEBUG_PAGEALLOC KASAN
 CPU: 1 PID: 7642 Comm: sh Not tainted 5.15.51-dirty #26
...
 Call Trace:
  <TASK>
  __invalidate_mapping_pages+0xe7/0x540
  drop_pagecache_sb+0x159/0x320
  iterate_supers+0x120/0x240
  drop_caches_sysctl_handler+0xaa/0xe0
  proc_sys_call_handler+0x2b4/0x480
  new_sync_write+0x3d6/0x5c0
  vfs_write+0x446/0x7a0
  ksys_write+0x105/0x210
  do_syscall_64+0x35/0x80
  entry_SYSCALL_64_after_hwframe+0x44/0xae
 RIP: 0033:0x7f52b5733130
...

This problem has been fixed on mainline by patch 6b24ca4a1a8d ("mm: Use
multi-index entries in the page cache") since it deletes the related code.

Fixes: 5c211ba29deb ("mm: add and use find_lock_entries")
Signed-off-by: Liu Shixin <liushixin2@huawei.com>
---
 mm/filemap.c | 12 +++++++-----
 1 file changed, 7 insertions(+), 5 deletions(-)

diff --git a/mm/filemap.c b/mm/filemap.c
index 00e391e75880..c3e1fc0520dc 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -2090,7 +2090,11 @@ unsigned find_lock_entries(struct address_space *mapping, pgoff_t start,
 
 	rcu_read_lock();
 	while ((page = find_get_entry(&xas, end, XA_PRESENT))) {
+		unsigned long next_idx = xas.xa_index;
+
 		if (!xa_is_value(page)) {
+			if (PageTransHuge(page))
+				next_idx = page->index + thp_nr_pages(page);
 			if (page->index < start)
 				goto put;
 			if (page->index + thp_nr_pages(page) - 1 > end)
@@ -2111,13 +2115,11 @@ unsigned find_lock_entries(struct address_space *mapping, pgoff_t start,
 put:
 		put_page(page);
 next:
-		if (!xa_is_value(page) && PageTransHuge(page)) {
-			unsigned int nr_pages = thp_nr_pages(page);
-
+		if (next_idx != xas.xa_index) {
 			/* Final THP may cross MAX_LFS_FILESIZE on 32-bit */
-			xas_set(&xas, page->index + nr_pages);
-			if (xas.xa_index < nr_pages)
+			if (next_idx < xas.xa_index)
 				break;
+			xas_set(&xas, next_idx);
 		}
 	}
 	rcu_read_unlock();
-- 
2.25.1

Re: [PATCH 5.15 v3] mm/filemap: fix UAF in find_lock_entries

[Matthew Wilcox]

On Wed, Jul 06, 2022 at 03:45:27PM +0800, Liu Shixin wrote:
>  	rcu_read_lock();
>  	while ((page = find_get_entry(&xas, end, XA_PRESENT))) {
> +		unsigned long next_idx = xas.xa_index;

It's confusing to have next_idx not be the actual next index.
That was why I made it 'xas.xa_index + 1'.  I know it's somewhat
used as an indicator that we don't need to call xas_set(), and so
it doesn't really matter, but let's say what we mean.

Re: [PATCH 5.15 v3] mm/filemap: fix UAF in find_lock_entries

[Liu Shixin]

On 2022/7/6 22:08, Matthew Wilcox wrote:
> On Wed, Jul 06, 2022 at 03:45:27PM +0800, Liu Shixin wrote:
>>  	rcu_read_lock();
>>  	while ((page = find_get_entry(&xas, end, XA_PRESENT))) {
>> +		unsigned long next_idx = xas.xa_index;
> It's confusing to have next_idx not be the actual next index.
> That was why I made it 'xas.xa_index + 1'.  I know it's somewhat
> used as an indicator that we don't need to call xas_set(), and so
> it doesn't really matter, but let's say what we mean.
I'll modify it and resend again, thanks.
>
>
> .
>