Re: [PATCH] ip-nexthop: support flush by id

Yes, I just refer to the support of protocol, not considering that id is unique. It is too heavy to get all.

2021年3月31日 下午11:36，David Ahern [off-list ref] 写道：

On 3/31/21 5:53 AM, Ido Schimmel wrote:

quoted

quoted

@@ -124,6 +125,9 @@ static int flush_nexthop(struct nlmsghdr *nlh, void *arg)

	if (tb[NHA_ID])
		id = rta_getattr_u32(tb[NHA_ID]);

+	if (filter.id && filter.id != id)
+		return 0;
+
	if (id && !delete_nexthop(id))
		filter.flushed++;

@@ -491,7 +495,10 @@ static int ipnh_list_flush(int argc, char **argv, int action)

			NEXT_ARG();
			if (get_unsigned(&id, *argv, 0))
				invarg("invalid id value", *argv);
-			return ipnh_get_id(id);
+			if (action == IPNH_FLUSH)
+				filter.id = id;
+			else
+				return ipnh_get_id(id);

I think it's quite weird to ask for a dump of all nexthops only to
delete a specific one. How about this:

+1

quoted

diff --git a/ip/ipnexthop.c b/ip/ipnexthop.c
index 0263307c49df..09a3076231aa 100644
--- a/ip/ipnexthop.c
+++ b/ip/ipnexthop.c
@@ -765,8 +765,16 @@ static int ipnh_list_flush(int argc, char **argv, int action)
                       if (!filter.master)
                               invarg("VRF does not exist\n", *argv);
               } else if (!strcmp(*argv, "id")) {
-                       NEXT_ARG();
-                       return ipnh_get_id(ipnh_parse_id(*argv));
+                       /* When 'id' is specified with 'flush' / 'list' we do
+                        * not need to perform a dump.
+                        */
+                       if (action == IPNH_LIST) {
+                               NEXT_ARG();
+                               return ipnh_get_id(ipnh_parse_id(*argv));
+                       } else {
+                               return ipnh_modify(RTM_DELNEXTHOP, 0, argc,
+                                                  argv);
+                       }

since delete just needs the id, you could refactor ipnh_modify and
create a ipnh_delete_id.