fs/9p/xattr.c: catch the error of p9_client_clunk when setting xattr failed
authorpiaojun <piaojun@huawei.com>
Wed, 25 Jul 2018 03:13:16 +0000 (11:13 +0800)
committerDominique Martinet <dominique.martinet@cea.fr>
Mon, 13 Aug 2018 00:35:16 +0000 (09:35 +0900)
In my testing, v9fs_fid_xattr_set will return successfully even if the
backend ext4 filesystem has no space to store xattr key-value. That will
cause inconsistent behavior between front end and back end. The reason is
that lsetxattr will be triggered by p9_client_clunk, and unfortunately we
did not catch the error. This patch will catch the error to notify upper

p9_client_clunk (in 9p)
  p9_client_rpc(clnt, P9_TCLUNK, "d", fid->fid);
    v9fs_clunk (in qemu)
                ext4_xattr_user_set (in host ext4 filesystem)

Link: http://lkml.kernel.org/r/5B57EACC.2060900@huawei.com
Signed-off-by: Jun Piao <piaojun@huawei.com>
Cc: Eric Van Hensbergen <ericvh@gmail.com>
Cc: Ron Minnich <rminnich@sandia.gov>
Cc: Latchesar Ionkov <lucho@ionkov.net>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: stable@vger.kernel.org
Signed-off-by: Dominique Martinet <dominique.martinet@cea.fr>

index f329eee6dc93e3f75033aa4c52e459aceeeb8df0..352abc39e891a1468d3576cc199fa13b089f32f7 100644 (file)
@@ -105,7 +105,7 @@ int v9fs_fid_xattr_set(struct p9_fid *fid, const char *name,
        struct kvec kvec = {.iov_base = (void *)value, .iov_len = value_len};
        struct iov_iter from;
-       int retval;
+       int retval, err;
        iov_iter_kvec(&from, WRITE | ITER_KVEC, &kvec, 1, value_len);
@@ -126,7 +126,9 @@ int v9fs_fid_xattr_set(struct p9_fid *fid, const char *name,
                p9_client_write(fid, 0, &from, &retval);
-       p9_client_clunk(fid);
+       err = p9_client_clunk(fid);
+       if (!retval && err)
+               retval = err;
        return retval;