fuse: atomic_o_trunc should truncate pagecache
authorMiklos Szeredi <mszeredi@redhat.com>
Thu, 8 Feb 2018 14:17:38 +0000 (15:17 +0100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Tue, 3 Jul 2018 09:24:49 +0000 (11:24 +0200)
commit df0e91d488276086bc07da2e389986cae0048c37 upstream.

Fuse has an "atomic_o_trunc" mode, where userspace filesystem uses the
O_TRUNC flag in the OPEN request to truncate the file atomically with the
open.

In this mode there's no need to send a SETATTR request to userspace after
the open, so fuse_do_setattr() checks this mode and returns.  But this
misses the important step of truncating the pagecache.

Add the missing parts of truncation to the ATTR_OPEN branch.

Reported-by: Chad Austin <chadaustin@fb.com>
Fixes: 6ff958edbf39 ("fuse: add atomic open+truncate support")
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
fs/fuse/dir.c

index 2496738..7a980b4 100644 (file)
@@ -1629,8 +1629,19 @@ int fuse_do_setattr(struct dentry *dentry, struct iattr *attr,
                return err;
 
        if (attr->ia_valid & ATTR_OPEN) {
-               if (fc->atomic_o_trunc)
+               /* This is coming from open(..., ... | O_TRUNC); */
+               WARN_ON(!(attr->ia_valid & ATTR_SIZE));
+               WARN_ON(attr->ia_size != 0);
+               if (fc->atomic_o_trunc) {
+                       /*
+                        * No need to send request to userspace, since actual
+                        * truncation has already been done by OPEN.  But still
+                        * need to truncate page cache.
+                        */
+                       i_size_write(inode, 0);
+                       truncate_pagecache(inode, 0);
                        return 0;
+               }
                file = NULL;
        }