Copy dq_dqb into rquota structure more carefully

[nfs-utils.git] / utils / idmapd / idmapd.c

diff --git a/utils/idmapd/idmapd.c b/utils/idmapd/idmapd.c
index b1da786455c40a8bc1c2a691bfa3c1109056ea43..80819d4d53e34037d06f705bf34cdc0f317adee7 100644 (file)
--- a/utils/idmapd/idmapd.c
+++ b/utils/idmapd/idmapd.c
@@ -714,9 +714,7 @@ idtonameres(struct idmap_msg *im)
                                        sizeof(im->im_name));
                break;
        }
-       /* XXX Hack? would rather return failure instead of writing nobody
-        * as above, but kernel seems not to deal well with that as of
-        * 2.6.8-rc3. */
+       /* XXX Hack? */
        im->im_status = IDMAP_STATUS_SUCCESS;
 }
 
@@ -725,6 +723,14 @@ nametoidres(struct idmap_msg *im)
 {
        int ret = 0;
 
+       /* XXX: nobody fallbacks shouldn't always happen:
+        *      server id -> name should be OK
+        *      client name -> id should be OK
+        * but not otherwise */
+       /* XXX: move nobody stuff to library calls
+        * (nfs4_get_nobody_user(domain), nfs4_get_nobody_group(domain)) */
+       /* XXX: should make this call higher up in the call chain (so we'd
+        * have a chance on looking up server/whatever. */
        switch (im->im_type) {
        case IDMAP_TYPE_USER:
                ret = nfs4_name_to_uid(im->im_name, &im->im_id);
@@ -737,9 +743,7 @@ nametoidres(struct idmap_msg *im)
                        im->im_id = nobodygid;
                break;
        }
-       /* XXX Hack? would rather return failure instead of writing nobody
-        * as above, but kernel seems not to deal well with that as of
-        * 2.6.8-rc3. */
+       /* XXX? */
        im->im_status = IDMAP_STATUS_SUCCESS;
 }