Netlink: Add option to specify netlink socket receive buffer size

Add option 'netlink rx buffer' to specify netlink socket receive buffer
size. Uses SO_RCVBUFFORCE, so it can override rmem_max limit.

Thanks to Trisha Biswas and Michal for the original patches.

doc/bird.sgml

@@ -3248,6 +3248,12 @@ channels.
 	allows to specify a limit on maximal number of nexthops in one route. By
 	default, multipath merging is disabled. If enabled, default value of the
 	limit is 16.
+
+	<tag><label id="krt-netlink-rx-buffer">netlink rx buffer <m/number/</tag> (Linux)
+	Set kernel receive buffer size (in bytes) for the netlink socket. The default
+	value is OS-dependent (from the <file>/proc/sys/net/core/rmem_default</file>
+	file), If you get some "Kernel dropped some netlink message ..." warnings,
+	you may increase this value.
 </descrip>
 
 <sect1>Attributes

sysdep/linux/krt-sys.h

@@ -69,6 +69,7 @@ static inline struct ifa * kif_get_primary_ip(struct iface *i UNUSED) { return N
 struct krt_params {
   u32 table_id;				/* Kernel table ID we sync with */
   u32 metric;				/* Kernel metric used for all routes */
+  uint netlink_rx_buffer;		/* Rx buffer size for the netlink socket */
 };
 
 struct krt_state {

sysdep/linux/netlink.Y

@@ -10,7 +10,8 @@ CF_HDR
 
 CF_DECLS
 
-CF_KEYWORDS(KERNEL, TABLE, METRIC, KRT_PREFSRC, KRT_REALM, KRT_SCOPE, KRT_MTU, KRT_WINDOW,
+CF_KEYWORDS(KERNEL, TABLE, METRIC, NETLINK, RX, BUFFER,
+	    KRT_PREFSRC, KRT_REALM, KRT_SCOPE, KRT_MTU, KRT_WINDOW,
 	    KRT_RTT, KRT_RTTVAR, KRT_SSTRESH, KRT_CWND, KRT_ADVMSS, KRT_REORDERING,
 	    KRT_HOPLIMIT, KRT_INITCWND, KRT_RTO_MIN, KRT_INITRWND, KRT_QUICKACK,
 	    KRT_LOCK_MTU, KRT_LOCK_WINDOW, KRT_LOCK_RTT, KRT_LOCK_RTTVAR,
@@ -24,6 +25,7 @@ kern_proto: kern_proto kern_sys_item ';' ;
 kern_sys_item:
    KERNEL TABLE expr { THIS_KRT->sys.table_id = $3; }
  | METRIC expr { THIS_KRT->sys.metric = $2; }
+ | NETLINK RX BUFFER expr { THIS_KRT->sys.netlink_rx_buffer = $4; }
  ;
 
 dynamic_attr: KRT_PREFSRC	{ $$ = f_new_dynamic_attr(EAF_TYPE_IP_ADDRESS, T_IP, EA_KRT_PREFSRC); } ;

sysdep/linux/netlink.c

@@ -174,6 +174,27 @@ nl_set_strict_dump(struct nl_sock *nl, int strict)
 #endif
 }
 
+static void
+nl_set_rcvbuf(int fd, uint val)
+{
+  if (setsockopt(fd, SOL_SOCKET, SO_RCVBUFFORCE, &val, sizeof(val)) < 0)
+    log(L_WARN "KRT: Cannot set netlink rx buffer size to %u: %m", val);
+}
+
+static uint
+nl_cfg_rx_buffer_size(struct config *cfg)
+{
+  uint bufsize = 0;
+
+  struct proto_config *pc;
+  WALK_LIST(pc, cfg->protos)
+    if ((pc->protocol == &proto_unix_kernel) && !pc->disabled)
+      bufsize = MAX(bufsize, ((struct krt_config *) pc)->sys.netlink_rx_buffer);
+
+  return bufsize;
+}
+
+
 static void
 nl_open(void)
 {
@@ -1976,6 +1997,8 @@ krt_do_scan(struct krt_proto *p UNUSED)	/* CONFIG_ALL_TABLES_AT_ONCE => p is NUL
 
 static sock *nl_async_sk;		/* BIRD socket for asynchronous notifications */
 static byte *nl_async_rx_buffer;	/* Receive buffer */
+static uint nl_async_bufsize;		/* Kernel rx buffer size for the netlink socket */
+static struct config *nl_last_config;	/* For tracking changes to nl_async_bufsize */
 
 static void
 nl_async_msg(struct nlmsghdr *h)
@@ -2111,6 +2134,32 @@ nl_open_async(void)
     bug("Netlink: sk_open failed");
 }
 
+static void
+nl_update_async_bufsize(void)
+{
+  /* No async socket */
+  if (!nl_async_sk)
+    return;
+
+  /* Already reconfigured */
+  if (nl_last_config == config)
+    return;
+
+  /* Update netlink buffer size */
+  uint bufsize = nl_cfg_rx_buffer_size(config);
+  if (bufsize && (bufsize != nl_async_bufsize))
+  {
+    /* Log message for reconfigurations only */
+    if (nl_last_config)
+      log(L_INFO "KRT: Changing netlink rx buffer size to %u", bufsize);
+
+    nl_set_rcvbuf(nl_async_sk->fd, bufsize);
+    nl_async_bufsize = bufsize;
+  }
+
+  nl_last_config = config;
+}
+
 
 /*
  *	Interface to the UNIX krt module
@@ -2139,6 +2188,7 @@ krt_sys_start(struct krt_proto *p)
 
   nl_open();
   nl_open_async();
+  nl_update_async_bufsize();
 
   return 1;
 }
@@ -2146,12 +2196,16 @@ krt_sys_start(struct krt_proto *p)
 void
 krt_sys_shutdown(struct krt_proto *p)
 {
+  nl_update_async_bufsize();
+
   HASH_REMOVE2(nl_table_map, RTH, krt_pool, p);
 }
 
 int
 krt_sys_reconfigure(struct krt_proto *p UNUSED, struct krt_config *n, struct krt_config *o)
 {
+  nl_update_async_bufsize();
+
   return (n->sys.table_id == o->sys.table_id) && (n->sys.metric == o->sys.metric);
 }