]> git.ipfire.org Git - thirdparty/linux.git/commitdiff
nilfs2: fix potential bug in end_buffer_async_write
authorRyusuke Konishi <konishi.ryusuke@gmail.com>
Sat, 3 Feb 2024 16:16:45 +0000 (01:16 +0900)
committerAndrew Morton <akpm@linux-foundation.org>
Thu, 8 Feb 2024 05:20:37 +0000 (21:20 -0800)
According to a syzbot report, end_buffer_async_write(), which handles the
completion of block device writes, may detect abnormal condition of the
buffer async_write flag and cause a BUG_ON failure when using nilfs2.

Nilfs2 itself does not use end_buffer_async_write().  But, the async_write
flag is now used as a marker by commit 7f42ec394156 ("nilfs2: fix issue
with race condition of competition between segments for dirty blocks") as
a means of resolving double list insertion of dirty blocks in
nilfs_lookup_dirty_data_buffers() and nilfs_lookup_node_buffers() and the
resulting crash.

This modification is safe as long as it is used for file data and b-tree
node blocks where the page caches are independent.  However, it was
irrelevant and redundant to also introduce async_write for segment summary
and super root blocks that share buffers with the backing device.  This
led to the possibility that the BUG_ON check in end_buffer_async_write
would fail as described above, if independent writebacks of the backing
device occurred in parallel.

The use of async_write for segment summary buffers has already been
removed in a previous change.

Fix this issue by removing the manipulation of the async_write flag for
the remaining super root block buffer.

Link: https://lkml.kernel.org/r/20240203161645.4992-1-konishi.ryusuke@gmail.com
Fixes: 7f42ec394156 ("nilfs2: fix issue with race condition of competition between segments for dirty blocks")
Signed-off-by: Ryusuke Konishi <konishi.ryusuke@gmail.com>
Reported-by: syzbot+5c04210f7c7f897c1e7f@syzkaller.appspotmail.com
Closes: https://lkml.kernel.org/r/00000000000019a97c05fd42f8c8@google.com
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
fs/nilfs2/segment.c

index 2590a0860eab022ba68a18b9db8fe181faab5069..2bfb08052d399972dee9fd49583b77b95104ac83 100644 (file)
@@ -1703,7 +1703,6 @@ static void nilfs_segctor_prepare_write(struct nilfs_sc_info *sci)
 
                list_for_each_entry(bh, &segbuf->sb_payload_buffers,
                                    b_assoc_buffers) {
-                       set_buffer_async_write(bh);
                        if (bh == segbuf->sb_super_root) {
                                if (bh->b_folio != bd_folio) {
                                        folio_lock(bd_folio);
@@ -1714,6 +1713,7 @@ static void nilfs_segctor_prepare_write(struct nilfs_sc_info *sci)
                                }
                                break;
                        }
+                       set_buffer_async_write(bh);
                        if (bh->b_folio != fs_folio) {
                                nilfs_begin_folio_io(fs_folio);
                                fs_folio = bh->b_folio;
@@ -1800,7 +1800,6 @@ static void nilfs_abort_logs(struct list_head *logs, int err)
 
                list_for_each_entry(bh, &segbuf->sb_payload_buffers,
                                    b_assoc_buffers) {
-                       clear_buffer_async_write(bh);
                        if (bh == segbuf->sb_super_root) {
                                clear_buffer_uptodate(bh);
                                if (bh->b_folio != bd_folio) {
@@ -1809,6 +1808,7 @@ static void nilfs_abort_logs(struct list_head *logs, int err)
                                }
                                break;
                        }
+                       clear_buffer_async_write(bh);
                        if (bh->b_folio != fs_folio) {
                                nilfs_end_folio_io(fs_folio, err);
                                fs_folio = bh->b_folio;
@@ -1896,8 +1896,9 @@ static void nilfs_segctor_complete_write(struct nilfs_sc_info *sci)
                                 BIT(BH_Delay) | BIT(BH_NILFS_Volatile) |
                                 BIT(BH_NILFS_Redirected));
 
-                       set_mask_bits(&bh->b_state, clear_bits, set_bits);
                        if (bh == segbuf->sb_super_root) {
+                               set_buffer_uptodate(bh);
+                               clear_buffer_dirty(bh);
                                if (bh->b_folio != bd_folio) {
                                        folio_end_writeback(bd_folio);
                                        bd_folio = bh->b_folio;
@@ -1905,6 +1906,7 @@ static void nilfs_segctor_complete_write(struct nilfs_sc_info *sci)
                                update_sr = true;
                                break;
                        }
+                       set_mask_bits(&bh->b_state, clear_bits, set_bits);
                        if (bh->b_folio != fs_folio) {
                                nilfs_end_folio_io(fs_folio, 0);
                                fs_folio = bh->b_folio;