[PATCH iproute2-next 0/2] ipmaddr: read link-layer multicast addresses via netlink

[PATCH iproute2-next 0/2] ipmaddr: read link-layer multicast addresses via netlink

[Yuyang Huang]

"ip maddr show" gets the IPv4 and IPv6 multicast addresses from
RTM_GETMULTICAST dumps, but it still parses /proc/net/dev_mcast for the
link-layer addresses. The kernel now also supports RTM_GETMULTICAST
dumps with ifa_family AF_PACKET, which report the device multicast list
with every column of /proc/net/dev_mcast. Patch 2 uses this dump. The
output does not change. Older kernels reject the AF_PACKET dump with
EOPNOTSUPP, and "ip maddr" then reads /proc/net/dev_mcast as before.

Patch 1 adds IFA_F_GLOBAL to include/uapi/linux/if_addr.h. It is only
there so that the series builds, and it should not be merged: the uapi
headers are synced into iproute2-next regularly, and the next sync
brings in IFA_F_GLOBAL.

The kernel side is the net-next series "rtnetlink: dump link-layer
multicast addresses":
https://patch.msgid.link/20261012011707.3378-1-sigefriedhyy@gmail.com

Testing:
- net-next kernel in vng: "ip maddr show" output is byte-identical to
  the current code for plain, -0/-4/-6, -o, -j, dev and -n <netns>, and
  /proc/net/dev_mcast is not read. This includes a multi-part dump with
  2000 entries on one device.
- 7.1 kernel without AF_PACKET support: falls back to
  /proc/net/dev_mcast, with the same output.

Yuyang Huang (2):
  uapi: update if_addr.h
  ipmaddr: read link-layer multicast addresses via netlink

 include/uapi/linux/if_addr.h |  1 +
 ip/ipmaddr.c                 | 10 ++++++++--
 2 files changed, 9 insertions(+), 2 deletions(-)

-- 
2.56.0.455.g6c5a503b38-goog

[PATCH iproute2-next 1/2] uapi: update if_addr.h

[Yuyang Huang]

Add IFA_F_GLOBAL, added in net-next by the series "rtnetlink: dump
link-layer multicast addresses". The next patch uses it.

Signed-off-by: Yuyang Huang <yuyanghuang@google.com>
---
 include/uapi/linux/if_addr.h | 1 +
 1 file changed, 1 insertion(+)

diff --git a/include/uapi/linux/if_addr.h b/include/uapi/linux/if_addr.h
index 5a28b958..aa30ef50 100644
--- a/include/uapi/linux/if_addr.h
+++ b/include/uapi/linux/if_addr.h
@@ -57,6 +57,7 @@ enum {
 #define IFA_F_NOPREFIXROUTE	0x200
 #define IFA_F_MCAUTOJOIN	0x400
 #define IFA_F_STABLE_PRIVACY	0x800
+#define IFA_F_GLOBAL		0x1000
 
 struct ifa_cacheinfo {
 	__u32	ifa_prefered;
-- 
2.56.0.455.g6c5a503b38-goog

[PATCH iproute2-next 2/2] ipmaddr: read link-layer multicast addresses via netlink

[Yuyang Huang]

"ip maddr show" gets the IPv4 and IPv6 entries from RTM_GETMULTICAST
dumps but still parses /proc/net/dev_mcast for the link-layer ones.
The kernel now also dumps the device multicast list (dev->mc) for
RTM_GETMULTICAST requests with ifa_family AF_PACKET, so use it for
those too.

The dump carries every column of /proc/net/dev_mcast: IFA_MULTICAST is
the address, IFA_MC_USERS the reference count and IFA_F_GLOBAL in
IFA_FLAGS the "static" column, set for entries added explicitly, e.g.
by "ip maddr add". The output does not change.

Older kernels reject the AF_PACKET dump with EOPNOTSUPP, in which case
/proc/net/dev_mcast is parsed as before.

Signed-off-by: Yuyang Huang <yuyanghuang@google.com>
---
 ip/ipmaddr.c | 10 ++++++++--
 1 file changed, 8 insertions(+), 2 deletions(-)

diff --git a/ip/ipmaddr.c b/ip/ipmaddr.c
index 95b703d2..2df520a3 100644
--- a/ip/ipmaddr.c
+++ b/ip/ipmaddr.c
@@ -261,6 +261,10 @@ static int accept_maddr(struct nlmsghdr *n, void *arg)
 	ma->addr.bitlen = ma->addr.bytelen << 3;
 	memcpy(ma->addr.data, RTA_DATA(tb[IFA_MULTICAST]), ma->addr.bytelen);
 	ma->users = rta_getattr_u32(tb[IFA_MC_USERS]);
+	/* IFA_F_GLOBAL is the "static" column of /proc/net/dev_mcast */
+	if (ifm->ifa_family == AF_PACKET && tb[IFA_FLAGS] &&
+	    (rta_getattr_u32(tb[IFA_FLAGS]) & IFA_F_GLOBAL))
+		ma->features = "static";
 	maddr_ins(&ctx->list, ma);
 
 	return 0;
@@ -388,8 +392,10 @@ static int multiaddr_list(int argc, char **argv)
 		}
 	}
 
-	if (!filter.family || filter.family == AF_PACKET)
-		read_dev_mcast(&list);
+	if (!filter.family || filter.family == AF_PACKET) {
+		if (read_maddr_netlink(AF_PACKET, &list) < 0)
+			read_dev_mcast(&list);
+	}
 	if (!filter.family || filter.family == AF_INET) {
 		if (read_maddr_netlink(AF_INET, &list) < 0)
 			read_igmp(&list);
-- 
2.56.0.455.g6c5a503b38-goog