From: Robert Doebbelin <[email protected]> commit 7cabc61e01a0a8b663bd2b4c982aa53048218734 upstream.
There's a race in fuse_direct_IO(), whereby is_sync_kiocb() is called on an iocb that could have been freed if async io has already completed. The fix in this case is simple and obvious: cache the result before starting io. It was discovered by KASan: Kernel: ================================================================== Kernel: BUG: KASan: use after free in fuse_direct_IO+0xb1a/0xcc0 at addr ffff88036c414390 Signed-off-by: Robert Doebbelin <[email protected]> Signed-off-by: Miklos Szeredi <[email protected]> Fixes: bcba24ccdc82 ("fuse: enable asynchronous processing direct IO") Signed-off-by: Jan Kara <[email protected]> Signed-off-by: Willy Tarreau <[email protected]> --- fs/fuse/file.c | 5 +++-- 1 file changed, 3 insertions(+), 2 deletions(-) diff --git a/fs/fuse/file.c b/fs/fuse/file.c index 35f604b..7ada0f0 100644 --- a/fs/fuse/file.c +++ b/fs/fuse/file.c @@ -2398,6 +2398,7 @@ fuse_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, loff_t i_size; size_t count = iov_length(iov, nr_segs); struct fuse_io_priv *io; + bool is_sync = is_sync_kiocb(iocb); pos = offset; inode = file->f_mapping->host; @@ -2433,7 +2434,7 @@ fuse_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, * to wait on real async I/O requests, so we must submit this request * synchronously. */ - if (!is_sync_kiocb(iocb) && (offset + count > i_size) && rw == WRITE) + if (!is_sync && (offset + count > i_size) && rw == WRITE) io->async = false; if (rw == WRITE) @@ -2445,7 +2446,7 @@ fuse_direct_IO(int rw, struct kiocb *iocb, const struct iovec *iov, fuse_aio_complete(io, ret < 0 ? ret : 0, -1); /* we have a non-extending, async request, so return */ - if (!is_sync_kiocb(iocb)) + if (!is_sync) return -EIOCBQUEUED; ret = wait_on_sync_kiocb(iocb); -- 2.8.0.rc2.1.gbe9624a
