diff mbox series

[V3,2/2] block: fix page leak in case of merging to same page

Message ID 20190610090215.14412-3-ming.lei@redhat.com (mailing list archive)
State Deferred, archived
Headers show
Series block: fix page leak by merging to same page | expand

Commit Message

Ming Lei June 10, 2019, 9:02 a.m. UTC
Different iovec may use one same page, then 'pages' array filled
by iov_iter_get_pages() may get reference of the same page several
times. If some elements in 'pages' can be merged to same page in
one bvec by bio_add_page(), bio_release_pages() only drops the
page's reference once.

This way causes page leak reported by David Gibson.

This issue can be triggered since 576ed913 ("block: use bio_add_page in
bio_iov_iter_get_pages").

Fixes the issue by putting the page's ref if it is merged to same page.

Cc: David Gibson <david@gibson.dropbear.id.au>
Cc: "Darrick J. Wong" <darrick.wong@oracle.com>
Cc: linux-xfs@vger.kernel.org
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@infradead.org>
Link: https://lkml.org/lkml/2019/4/23/64
Fixes: 576ed913 ("block: use bio_add_page in bio_iov_iter_get_pages")
Reported-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
---
 block/bio.c         | 12 ++++++++++--
 include/linux/bio.h |  8 ++++++++
 2 files changed, 18 insertions(+), 2 deletions(-)
diff mbox series

Patch

diff --git a/block/bio.c b/block/bio.c
index 39e3b931dc3b..358ccb5086e6 100644
--- a/block/bio.c
+++ b/block/bio.c
@@ -652,6 +652,9 @@  static inline bool page_is_mergeable(const struct bio_vec *bv,
 			return false;
 		if (pfn_to_page(PFN_DOWN(vec_end_addr)) + 1 != page)
 			return false;
+	/* drop page ref if the page has been added and user asks to do that */
+	} else if (flags & BVEC_MERGE_PUT_SAME_PAGE) {
+		put_page(page);
 	}
 
 	WARN_ON_ONCE((flags & BVEC_MERGE_TO_SAME_PAGE) &&
@@ -924,8 +927,13 @@  static int __bio_iov_iter_get_pages(struct bio *bio, struct iov_iter *iter)
 		struct page *page = pages[i];
 
 		len = min_t(size_t, PAGE_SIZE - offset, left);
-		if (WARN_ON_ONCE(bio_add_page(bio, page, len, offset) != len))
-			return -EINVAL;
+
+		if (!__bio_try_merge_page(bio, page, len, offset,
+					BVEC_MERGE_PUT_SAME_PAGE)) {
+			if (WARN_ON_ONCE(bio_full(bio)))
+                                return -EINVAL;
+			__bio_add_page(bio, page, len, offset);
+		}
 		offset = 0;
 	}
 
diff --git a/include/linux/bio.h b/include/linux/bio.h
index ee18895431ba..0168bc5df8e4 100644
--- a/include/linux/bio.h
+++ b/include/linux/bio.h
@@ -428,6 +428,14 @@  enum bvec_merge_flags {
 	 * bvec
 	 */
 	BVEC_MERGE_TO_SAME_PAGE = BIT(0),
+
+	/*
+	 * put refcount of bio's last page if the start page to add is
+	 * same with bio's last page. If user gets refcount of every
+	 * page added to bio before calling bio_add_page, please consider
+	 * to use this flag for avoiding page leak
+	 */
+	BVEC_MERGE_PUT_SAME_PAGE = BIT(1),
 };
 
 extern int bio_add_page(struct bio *, struct page *, unsigned int,unsigned int);