From: Christian Brauner <christian.brauner@xxxxxxxxxx> When transfering ownership information to the client the k*ids are translated into raw *ids before they are sent over the wire. The function currently erroneously translates the k*ids according to the mount's idmapping. Instead, reporting the owning *ids to userspace the underlying k*ids need to be mapped up in the caller's user namespace. This is how stat() works. The caller in this instance is ksmbd itself and ksmbd always runs in the initial user namespace. Translate according to that taking any potential idmapped mounts into account. Switch to from_k*id_munged() which ensures that the overflow*id is returned instead of the (*id_t)-1 when the k*id can't be translated. Cc: Steve French <stfrench@xxxxxxxxxxxxx> Cc: Christoph Hellwig <hch@xxxxxxxxxxxxx> Cc: Namjae Jeon <namjae.jeon@xxxxxxxxxxx> Cc: Hyunchul Lee <hyc.lee@xxxxxxxxx> Cc: Sergey Senozhatsky <senozhatsky@xxxxxxxxxxxx> Cc: linux-cifs@xxxxxxxxxxxxxxx Signed-off-by: Christian Brauner <christian.brauner@xxxxxxxxxx> --- fs/ksmbd/oplock.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) diff --git a/fs/ksmbd/oplock.c b/fs/ksmbd/oplock.c index 6ace6c2f22dc..16b6236d1bd2 100644 --- a/fs/ksmbd/oplock.c +++ b/fs/ksmbd/oplock.c @@ -1614,9 +1614,11 @@ void create_posix_rsp_buf(char *cc, struct ksmbd_file *fp) buf->nlink = cpu_to_le32(inode->i_nlink); buf->reparse_tag = cpu_to_le32(fp->volatile_id); buf->mode = cpu_to_le32(inode->i_mode); - id_to_sid(from_kuid(user_ns, inode->i_uid), + id_to_sid(from_kuid_munged(&init_user_ns, + i_uid_into_mnt(user_ns, inode)), SIDNFS_USER, (struct smb_sid *)&buf->SidBuffer[0]); - id_to_sid(from_kgid(user_ns, inode->i_gid), + id_to_sid(from_kgid_munged(&init_user_ns, + i_gid_into_mnt(user_ns, inode)), SIDNFS_GROUP, (struct smb_sid *)&buf->SidBuffer[20]); } -- 2.30.2
