pg scrub stat mismatch with special objects that have hash 'ffffffff'
Here is the steps to reproduce the pg inconsistents:
1. build a cluster with vstart
MON=1 OSD=3 MDS=0 MGR=1 ../src/vstart.sh -X -n
2. create a pool with only 1 pg
./bin/ceph osd pool create rbd 1
3. put an object with special name, which will let the object hash as 'ffffffff'
./bin/rados -p rbd put 59431f0f-60c1-49be-ab18-9bfbee59869c.1019453.1_obj0344489466 /etc/passwd --debug_objecter=20 --debug_ms=5
4. trigger the pg scrub
./bin/ceph pg scrub 1.0
5. the expect output of 'ceph -s' will include 'active+clean+inconsistent'
6. the problem maybe the objectstore::get_coll_range interface which will not include the object with hash 'ffffffff'