[BUG] rgw/sfs: crash when calling CompleteMultipartUpload with a non-existent uploadId

[giubacc]

Describe the bug A clear and concise description of what the bug is.

Putting an non existing uploadId in the final CompleteMultipartUpload call guarantees a crash due to the ceph_assert(mp.has_value()); in the get_meta_obj() function in multipart.cc.

To Reproduce Steps to reproduce the behavior:

Just call CompleteMultipartUpload with a non-existing UploadId.

[jecluis]

This might be a bug in rgw itself, as that assert was expected to be a sanity check.

@0xavi0 and @irq0 do you have opinions on where this should be fixed?

[0xavi0]

This needs further investigation (to confirm if it's an rgw bug or in sfs)

Code in rgw_op.cc looks like this:

  upload = s->bucket->get_multipart_upload(s->object->get_name(), upload_id);

  RGWCompressionInfo cs_info;
  bool compressed = false;
  uint64_t accounted_size = 0;

  list<rgw_obj_index_key> remove_objs; /* objects to be removed from index listing */

  meta_obj = upload->get_meta_obj();
  meta_obj->set_in_extra_data(true);
  meta_obj->set_hash_source(s->object->get_name());

Looks like it follows an optimistic approach (it doesn't check if the upload is valid or not and just uses it)
Crash happens when calling meta_obj = upload->get_meta_obj();

According to Amazon docs it should return a 404 error when upload id is not valid.

Error code: NoSuchUpload

Description: The specified multipart upload does not exist. The upload ID might be invalid, or the multipart upload might have been aborted or completed.

404 Not Found