Ceph » rgw

sorry， the condition when last_added_entries is not updated should be count >= num_entries

    if (count < num_entries) {
      marker = last_added_entry = dirent.key; // double assign
      ent_list.emplace_back(std::move(dirent));
      ++count;
    } else {
      *is_truncated = true;
      goto check_updates;
    }

when object name adorned with upload_id, we might get wrong current_shard

    // test whether object name is a multipart meta name
    if(! multipart_meta_filter.filter(start.name, key)) { <=== do not stripe upload_id when multipart name doesn't have meta suffix 
      // if multipart_meta_filter fails, must be "regular" (i.e.,
      // unadorned) and the name is the key
      key = start.name;
    }

    // now convert the key (oid) to an rgw_obj_key since that will
    // separate out the namespace, name, and instance
    rgw_obj_key obj_key;
    bool parsed = rgw_obj_key::parse_raw_oid(key, &obj_key); <=== do not stripe upload_id
    if (!parsed) {
      ldout(cct, 0) <<
    "ERROR: RGWRados::cls_bucket_list_unordered received an invalid " 
    "start marker: '" << start << "'" << dendl;
      return -EINVAL;
    } else if (obj_key.name.empty()) {
      // if the name is empty that means the object name came in with
      // a namespace only, and therefore we need to start our scan at
      // the first bucket index shard
      current_shard = 0u;
    } else {
      // so now we have the key used to compute the bucket index shard
      // and can extract the specific shard from it
      current_shard = rgw_bucket_shard_index(obj_key.name, num_shards); <=== here we might get wrong current_shard
    }

fixes ： http://tracker.ceph.com/issues/49206

when object name adorned with upload_id, we might get wrong current_shard
[...]

pr: https://github.com/ceph/ceph/pull/42151