[PATCH v2 1/1] route/addr: only sent IFA_FLAGS when needed to workaround picky older kernels

[Thomas Haller]

Older kernels don't accept receiving unknown netlink attributes.
See net/core/rtnetlink.c, rtnetlink_rcv_msg(). This was fixed by kernel
commit 661d2967b3f1b34eeaa7e212e7b9bbe8ee072b59.

As a workaround, only set the additional attributes, when the user
provided flags that makes this necessary and useful.

https://github.com/thom311/libnl/issues/56
https://bugzilla.redhat.com/show_bug.cgi?id=1063885

Based-on-patch-by: Pavel Kankovsky <peak at argo.troja.mff.cuni.cz>
Signed-off-by: Thomas Haller <thaller at redhat.com>
---

 Obsoletes previous patch.

 changes: reword commit message and code comments.

 lib/route/addr.c | 14 +++++++++++++-
 1 file changed, 13 insertions(+), 1 deletion(-)

diff --git a/lib/route/addr.c b/lib/route/addr.c
index 97905f0..e6e91d2 100644
--- a/lib/route/addr.c
+++ b/lib/route/addr.c
@@ -598,7 +598,19 @@ static int build_addr_msg(struct rtnl_addr *tmpl, int cmd, int flags,
 		NLA_PUT(msg, IFA_CACHEINFO, sizeof(ca), &ca);
 	}
 
-	NLA_PUT_U32(msg, IFA_FLAGS, tmpl->a_flags);
+	if (tmpl->a_flags & ~0xFF) {
+		/* only set the IFA_FLAGS attribute, if they actually contain additional
+		 * flags that are not already set to am.ifa_flags.
+		 *
+		 * Older kernels refuse RTM_NEWADDR and RTM_NEWROUTE messages with EINVAL
+		 * if they contain unknown netlink attributes. See net/core/rtnetlink.c, which
+		 * was fixed by kernel commit 661d2967b3f1b34eeaa7e212e7b9bbe8ee072b59.
+		 *
+		 * With this workaround, libnl will function correctly with older kernels,
+		 * unless there is a new libnl user that wants to set these flags. In this
+		 * case it's up to the user to workaround this issue. */
+		NLA_PUT_U32(msg, IFA_FLAGS, tmpl->a_flags);
+	}
 
 	*result = msg;
 	return 0;
-- 
1.9.0