[SRU][Trusty][PATCH 1/2] fix unbalanced page refcounting in bio_map_user_iov

Kleber Sacilotto de Souza kleber.souza at canonical.com
Fri Dec 8 13:37:55 UTC 2017


From: Vitaly Mayatskikh <v.mayatskih at gmail.com>

bio_map_user_iov and bio_unmap_user do unbalanced pages refcounting if
IO vector has small consecutive buffers belonging to the same page.
bio_add_pc_page merges them into one, but the page reference is never
dropped.

Cc: stable at vger.kernel.org
Signed-off-by: Vitaly Mayatskikh <v.mayatskih at gmail.com>
Signed-off-by: Al Viro <viro at zeniv.linux.org.uk>

CVE-2017-12190
(backported from commit 95d78c28b5a85bacbc29b8dba7c04babb9b0d467)
[klebers:
 - Fixed file path from block/bio.c to fs/bio.c.
 - Adjusted for context.]
Signed-off-by: Kleber Sacilotto de Souza <kleber.souza at canonical.com>
---
 fs/bio.c | 8 ++++++++
 1 file changed, 8 insertions(+)

diff --git a/fs/bio.c b/fs/bio.c
index 99367f46ec86..3311aa0f3dd9 100644
--- a/fs/bio.c
+++ b/fs/bio.c
@@ -1285,6 +1285,7 @@ static struct bio *__bio_map_user_iov(struct request_queue *q,
 		offset = uaddr & ~PAGE_MASK;
 		for (j = cur_page; j < page_limit; j++) {
 			unsigned int bytes = PAGE_SIZE - offset;
+			unsigned short prev_bi_vcnt = bio->bi_vcnt;
 
 			if (len <= 0)
 				break;
@@ -1299,6 +1300,13 @@ static struct bio *__bio_map_user_iov(struct request_queue *q,
 					    bytes)
 				break;
 
+			/*
+			 * check if vector was merged with previous
+			 * drop page reference if needed
+			 */
+			if (bio->bi_vcnt == prev_bi_vcnt)
+				put_page(pages[j]);
+
 			len -= bytes;
 			offset = 0;
 		}
-- 
2.14.1





More information about the kernel-team mailing list