Ceph » CephFS

This patch adds a small gtest that shows that the handling of removexattr is wrong:

diff --git a/src/test/libcephfs/test.cc b/src/test/libcephfs/test.cc
index e74e99740979..f14a4605e137 100644
--- a/src/test/libcephfs/test.cc
+++ b/src/test/libcephfs/test.cc
@@ -540,6 +540,9 @@ TEST(LibCephFS, Xattrs) {
   int fd = ceph_open(cmount, test_xattr_file, O_CREAT, 0666);
   ASSERT_GT(fd, 0);

+  // test removing non-existent xattr
+  ASSERT_EQ(-ENODATA, ceph_removexattr(cmount, test_xattr_file, "user.nosuchxattr"));
+
   char i = 'a';
   char xattrk[128];
   char xattrv[128];

libcephfs always does a synchronous RMXATTR call to the MDS for a removexattr. In the kclient, this is mostly papered over by the fact that we usually handle the removal locally when we have Xx caps, so we only rarely do a synchronous call to the MDS there.