Don't always use default mapping to "nobody".

Signed-off-by: Kevin Coffman <kwc@citi.umich.edu>

We've been ignoring all mapping errors and instead mapping to a "nobody" user
or group.

This is arguably OK for the cases where we're returning a value to the user
(so, id->name mapping on the server or name->id mapping on the client).

But it's a disaster in the other direction (id->name on the server or id->name
on the client): for example, a chown to an unknown user should *not*
automatically be translated into a succesful chown to "nobody".

This patch fixes that problem on the server side.

Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>

1 files changed, 35 insertions, 17 deletions

diff --git a/utils/idmapd/idmapd.c b/utils/idmapd/idmapd.c
index 21a1916..19bf7a6 100644
--- a/utils/idmapd/idmapd.c
+++ b/utils/idmapd/idmapd.c
@@ -626,10 +626,18 @@ nfsdcb(int fd, short which, void *data)
 		snprintf(buf1, sizeof(buf1), "%lu",
 			 time(NULL) + cache_entry_expiration);
 		addfield(&bp, &bsiz, buf1);
-		/* ID */
-		snprintf(buf1, sizeof(buf1), "%u", im.im_id);
-		addfield(&bp, &bsiz, buf1);
+		/* Note that we don't want to write the id if the mapping
+		 * failed; instead, by leaving it off, we write a negative
+		 * cache entry which will result in an error returned to
+		 * the client.  We don't want a chown or setacl referring
+		 * to an unknown user to result in giving permissions to
+		 * "nobody"! */
+		if (im.im_status == IDMAP_STATUS_SUCCESS) {
+			/* ID */
+			snprintf(buf1, sizeof(buf1), "%u", im.im_id);
+			addfield(&bp, &bsiz, buf1);
 
+		}
 		//if (bsiz == sizeof(buf)) /* XXX */
 
 		bp[-1] = '\n';
@@ -646,6 +654,8 @@ nfsdcb(int fd, short which, void *data)
 		snprintf(buf1, sizeof(buf1), "%lu",
 			 time(NULL) + cache_entry_expiration);
 		addfield(&bp, &bsiz, buf1);
+		/* Note we're ignoring the status field in this case; we'll
+		 * just map to nobody instead. */
 		/* Name */
 		addfield(&bp, &bsiz, im.im_name);
 
@@ -718,6 +728,13 @@ nfscb(int fd, short which, void *data)
 
 	imconv(ic, &im);
 
+	/* XXX: I don't like ignoring this error in the id->name case,
+	 * but we've never returned it, and I need to check that the client
+	 * can handle it gracefully before starting to return it now. */
+
+	if (im.im_status == IDMAP_STATUS_LOOKUPFAIL)
+		im.im_status = IDMAP_STATUS_SUCCESS;
+
 	if (atomicio(write, ic->ic_fd, &im, sizeof(im)) != sizeof(im))
 		idmapd_warn("nfscb: write(%s)", ic->ic_path);
 out:
@@ -837,8 +854,10 @@ idtonameres(struct idmap_msg *im)
 		}
 		break;
 	}
-	/* XXX Hack? */
-	im->im_status = IDMAP_STATUS_SUCCESS;
+	if (ret)
+		im->im_status = IDMAP_STATUS_LOOKUPFAIL;
+	else
+		im->im_status = IDMAP_STATUS_SUCCESS;
 }
 
 static void
@@ -848,30 +867,29 @@ nametoidres(struct idmap_msg *im)
 	gid_t gid;
 	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. */
+
+	im->im_status = IDMAP_STATUS_SUCCESS;
+
 	switch (im->im_type) {
 	case IDMAP_TYPE_USER:
 		ret = nfs4_name_to_uid(im->im_name, &uid);
 		im->im_id = (u_int32_t) uid;
-		if (ret)
+		if (ret) {
+			im->im_status = IDMAP_STATUS_LOOKUPFAIL;
 			im->im_id = nobodyuid;
-		break;
+		}
+		return;
 	case IDMAP_TYPE_GROUP:
 		ret = nfs4_name_to_gid(im->im_name, &gid);
 		im->im_id = (u_int32_t) gid;
-		if (ret)
+		if (ret) {
+			im->im_status = IDMAP_STATUS_LOOKUPFAIL;
 			im->im_id = nobodygid;
-		break;
+		}
+		return;
 	}
-	/* XXX? */
-	im->im_status = IDMAP_STATUS_SUCCESS;
 }
 
 static int