On Mon, 2022-03-14 at 10:28 +0800, xiubli@xxxxxxxxxx wrote: > From: Xiubo Li <xiubli@xxxxxxxxxx> > > This is a new approach to improve the readdir and based the previous > discussion in another thread: > > https://patchwork.kernel.org/project/ceph-devel/list/?series=621901 > > Just start a new thread for this. > > As Jeff suggested, this patch series will dentrypt the dentry name > during parsing the readdir data in handle_reply(). And then in both > ceph_readdir_prepopulate() and ceph_readdir() we will use the > dencrypted name directly. > > NOTE: we will base64_dencode and dencrypt the names in-place instead > of allocating tmp buffers. For base64_dencode it's safe because the > dencoded string buffer will always be shorter. > > > V2: > - Fix the WARN issue reported by Luis, thanks. > > > Xiubo Li (4): > ceph: pass the request to parse_reply_info_readdir() > ceph: add ceph_encode_encrypted_dname() helper > ceph: dencrypt the dentry names early and once for readdir > ceph: clean up the ceph_readdir() code > > fs/ceph/crypto.c | 25 ++++++++--- > fs/ceph/crypto.h | 2 + > fs/ceph/dir.c | 64 +++++++++------------------ > fs/ceph/inode.c | 37 ++-------------- > fs/ceph/mds_client.c | 101 ++++++++++++++++++++++++++++++++++++------- > fs/ceph/mds_client.h | 4 +- > 6 files changed, 133 insertions(+), 100 deletions(-) > This looks good, Xiubo. I did some testing with these earlier and they seemed to work great. I've gone ahead and merged these into the wip-fscrypt branch. It may be best to eventually squash these down, but it's probably fine to leave them on top as well. Thanks! -- Jeff Layton <jlayton@xxxxxxxxxx>
