ceph: only dirty ITER_IOVEC pages for direct read
commit 85784f9395987a422fa04263e7c0fb13da11eb5c upstream.
If a page is already locked, attempting to dirty it leads to a deadlock
in lock_page(). This is what currently happens to ITER_BVEC pages when
a dio-enabled loop device is backed by ceph:
$ losetup --direct-io /dev/loop0 /mnt/cephfs/img
$ xfs_io -c 'pread 0 4k' /dev/loop0
Follow other file systems and only dirty ITER_IOVEC pages.
Cc: stable@kernel.org
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Reviewed-by: Ilya Dryomov <idryomov@gmail.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
diff --git a/fs/ceph/file.c b/fs/ceph/file.c
index ca3f630..e7ddb23 100644
--- a/fs/ceph/file.c
+++ b/fs/ceph/file.c
@@ -598,7 +598,8 @@
struct ceph_aio_request {
struct kiocb *iocb;
size_t total_len;
- int write;
+ bool write;
+ bool should_dirty;
int error;
struct list_head osd_reqs;
unsigned num_reqs;
@@ -708,7 +709,7 @@
}
}
- ceph_put_page_vector(osd_data->pages, num_pages, !aio_req->write);
+ ceph_put_page_vector(osd_data->pages, num_pages, aio_req->should_dirty);
ceph_osdc_put_request(req);
if (rc < 0)
@@ -890,6 +891,7 @@
size_t count = iov_iter_count(iter);
loff_t pos = iocb->ki_pos;
bool write = iov_iter_rw(iter) == WRITE;
+ bool should_dirty = !write && iter_is_iovec(iter);
if (write && ceph_snap(file_inode(file)) != CEPH_NOSNAP)
return -EROFS;
@@ -954,6 +956,7 @@
if (aio_req) {
aio_req->iocb = iocb;
aio_req->write = write;
+ aio_req->should_dirty = should_dirty;
INIT_LIST_HEAD(&aio_req->osd_reqs);
if (write) {
aio_req->mtime = mtime;
@@ -1012,7 +1015,7 @@
len = ret;
}
- ceph_put_page_vector(pages, num_pages, !write);
+ ceph_put_page_vector(pages, num_pages, should_dirty);
ceph_osdc_put_request(req);
if (ret < 0)