sd-dhcp-client: --omg-optimized

Passing the protocol to socket() is redundant as it will be specified again in
bind(). Dropping the redundancy reduces the cost of bind() from ~30ms to ~0ms.
For details see [0].

networkd in a container (i.e., with next to no network latency) can now
negotiate a DHCP lease in 0.7 - 5 ms.

Thanks to Kay for help with debugging and to Daniel Borkmann for the pointer
to fix the problem.

[0]: <https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=902fefb82ef72a50c78cb4a20cc954b037a98d1c>

diff --git a/src/libsystemd-network/dhcp-network.c b/src/libsystemd-network/dhcp-network.c
index 29e915c473f6026781483ebbf9ba370c9d339f11..ecc94b9be782946a30d89bfcd96c2ce44e980b44 100644 (file)
--- a/src/libsystemd-network/dhcp-network.c
+++ b/src/libsystemd-network/dhcp-network.c
@@ -37,8 +37,7 @@ int dhcp_network_bind_raw_socket(int index, union sockaddr_union *link)
         assert(index > 0);
         assert(link);
 
-        s = socket(AF_PACKET, SOCK_DGRAM | SOCK_CLOEXEC | SOCK_NONBLOCK,
-                   htons(ETH_P_IP));
+        s = socket(AF_PACKET, SOCK_DGRAM | SOCK_CLOEXEC | SOCK_NONBLOCK, 0);
         if (s < 0)
                 return -errno;