[Cluster-devel] [PATCH -v2] ext4: don't BUG if kernel subsystems dirty pages without asking ext4 first

[Theodore Ts'o]

[un]pin_user_pages_remote is dirtying pages without properly warning
the file system in advance (or faulting in the file data if the page
is not yet in the page cache).  This was noted by Jan Kara in 2018[1]
and more recently has resulted in bug reports by Syzbot in various
Android kernels[2].

This is technically a bug in the mm/gup.c codepath, but arguably ext4
is fragile in that a buggy get_user_pages() implementation causes ext4
to crash, where as other file systems are not crashing (although in
some cases the user data will be lost since gup code is not properly
informing the file system to potentially allocate blocks or reserve
space when writing into a sparse portion of file).  I suspect in real
life it is rare that people are using RDMA into file-backed memory,
which is why no one has complained to ext4 developers except fuzzing
programs.

So instead of crashing with a BUG, issue a warning (since there may be
potential data loss) and just mark the page as clean to avoid
unprivileged denial of service attacks until the problem can be
properly fixed.  More discussion and background can be found in the
thread starting at [2].

[1] https://www.spinics.net/lists/linux-mm/msg142700.html
[2] https://lore.kernel.org/r/Yg0m6IjcNmfaSokM@google.com

Reported-by: syzbot+d59332e2db681cf18f0318a06e994ebbb529a8db at syzkaller.appspotmail.com
Reported-by: Lee Jones <lee.jones at linaro.org>
Signed-off-by: Theodore Ts'o <tytso at mit.edu>
---
 fs/ext4/inode.c | 27 ++++++++++++++++++++++++++-
 1 file changed, 26 insertions(+), 1 deletion(-)

diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 01c9e4f743ba..f8fefbf67306 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -1993,6 +1993,15 @@ static int ext4_writepage(struct page *page,
 	else
 		len = PAGE_SIZE;
 
+	/* Should never happen but for buggy gup code */
+	if (!page_has_buffers(page)) {
+		ext4_warning_inode(inode,
+		   "page %lu does not have buffers attached", page->index);
+		ClearPageDirty(page);
+		unlock_page(page);
+		return 0;
+	}
+
 	page_bufs = page_buffers(page);
 	/*
 	 * We cannot do block allocation or other extent handling in this
@@ -2588,12 +2597,28 @@ static int mpage_prepare_extent_to_map(struct mpage_da_data *mpd)
 			     (mpd->wbc->sync_mode == WB_SYNC_NONE)) ||
 			    unlikely(page->mapping != mapping)) {
 				unlock_page(page);
-				continue;
+				goto out;
 			}
 
 			wait_on_page_writeback(page);
 			BUG_ON(PageWriteback(page));
 
+			/*
+			 * Should never happen but for buggy code in
+			 * other subsystemsa that call
+			 * set_page_dirty() without properly warning
+			 * the file system first.  See [1] for more
+			 * information.
+			 *
+			 * [1] https://www.spinics.net/lists/linux-mm/msg142700.html
+			 */
+			if (!page_has_buffers(page)) {
+				ext4_warning_inode(mpd->inode, "page %lu does not have buffers attached", page->index);
+				ClearPageDirty(page);
+				unlock_page(page);
+				continue;
+			}
+
 			if (mpd->map.m_len == 0)
 				mpd->first_page = page->index;
 			mpd->next_page = page->index + 1;
-- 
2.31.0