ceph: handle race between vmtruncate and queuing cap snap

It's possible that we create a cap snap while there is pending
vmtruncate (truncate hasn't been processed by worker thread).
We should truncate dirty pages beyond capsnap->size in that case.

Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>

diff --git a/fs/ceph/inode.c b/fs/ceph/inode.c
index a19fafdf87f8f861658e5397ef9e084c714a6811..373dab5173cac540d27eb3b6b889569cb69415db 100644 (file)
--- a/fs/ceph/inode.c
+++ b/fs/ceph/inode.c
@@ -1833,9 +1833,20 @@ retry:
         * possibly truncate them.. so write AND block!
         */
        if (ci->i_wrbuffer_ref_head < ci->i_wrbuffer_ref) {
+               struct ceph_cap_snap *capsnap;
+               to = ci->i_truncate_size;
+               list_for_each_entry(capsnap, &ci->i_cap_snaps, ci_item) {
+                       // MDS should have revoked Frw caps
+                       WARN_ON_ONCE(capsnap->writing);
+                       if (capsnap->dirty_pages && capsnap->size > to)
+                               to = capsnap->size;
+               }
+               spin_unlock(&ci->i_ceph_lock);
                dout("__do_pending_vmtruncate %p flushing snaps first\n",
                     inode);
-               spin_unlock(&ci->i_ceph_lock);
+
+               truncate_pagecache(inode, to);
+
                filemap_write_and_wait_range(&inode->i_data, 0,
                                             inode->i_sb->s_maxbytes);
                goto retry;