It is trivial to trigger a WARN_ON_ONCE(1) in iomap_dio_actor() by
unprivileged users which would taint the kernel, or worse - panic if
panic_on_warn or panic_on_taint is set. Hence, just convert it to
pr_warn_ratelimited() to let users know their workloads are racing.
Thank Dave Chinner for the initial analysis of the racing reproducers.
Signed-off-by: Qian Cai <c...@lca.pw>
---
v4: use %pD4.
v3: Keep the default case and update the message.
v2: Record the path, pid and command as well.
fs/iomap/direct-io.c | 10 ++++++++++
1 file changed, 10 insertions(+)
diff --git a/fs/iomap/direct-io.c b/fs/iomap/direct-io.c
index c1aafb2ab990..9519113ebc35 100644
--- a/fs/iomap/direct-io.c
+++ b/fs/iomap/direct-io.c
@@ -388,6 +388,16 @@ iomap_dio_actor(struct inode *inode, loff_t pos, loff_t
length,
return iomap_dio_bio_actor(inode, pos, length, dio, iomap);
case IOMAP_INLINE:
return iomap_dio_inline_actor(inode, pos, length, dio, iomap);
+ case IOMAP_DELALLOC:
+ /*
+ * DIO is not serialised against mmap() access at all, and so
+ * if the page_mkwrite occurs between the writeback and the
+ * iomap_apply() call in the DIO path, then it will see the
+ * DELALLOC block that the page-mkwrite allocated.
+ */
+ pr_warn_ratelimited("Direct I/O collision with buffered writes!
File: %pD4 Comm: %.20s\n",
+ dio->iocb->ki_filp, current->comm);
+ return -EIO;
default:
WARN_ON_ONCE(1);
return -EIO;
--
2.18.4
