[NETLINK]: Support dynamic number of multicast groups per netlink family

Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net>

[NETLINK]: Support dynamic number of multicast groups per netlink family
Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: David S. Miller <davem@davemloft.net>
Patrick McHardy · David S. Miller
1 parent ab33a1711c
Showing 1 changed file with 51 additions and 18 deletions Side-by-side Diff
net/netlink/af_netlink.c
@@ -60,21 +60,24 @@
 #include <net/scm.h>
  
 #define Nprintk(a...)
+#define NLGRPSZ(x)	(ALIGN(x, sizeof(unsigned long) * 8) / 8)
  
 struct netlink_sock {
 	/* struct sock has to be the first member of netlink_sock */
 	struct sock		sk;
 	u32			pid;
-	unsigned int		groups;
 	u32			dst_pid;
 	u32			dst_group;
+	u32			flags;
+	u32			subscriptions;
+	u32			ngroups;
+	unsigned long		*groups;
 	unsigned long		state;
 	wait_queue_head_t	wait;
 	struct netlink_callback	*cb;
 	spinlock_t		cb_lock;
 	void			(*data_ready)(struct sock *sk, int bytes);
 	struct module		*module;
-	u32			flags;
 };
  
 #define NETLINK_KERNEL_SOCKET	0x1
@@ -101,6 +104,7 @@
 	struct nl_pid_hash hash;
 	struct hlist_head mc_list;
 	unsigned int nl_nonroot;
+	unsigned int groups;
 	struct module *module;
 	int registered;
 };
@@ -138,6 +142,7 @@
 	BUG_TRAP(!atomic_read(&sk->sk_rmem_alloc));
 	BUG_TRAP(!atomic_read(&sk->sk_wmem_alloc));
 	BUG_TRAP(!nlk_sk(sk)->cb);
+	BUG_TRAP(!nlk_sk(sk)->groups);
 }
  
 /* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on SMP.
@@ -333,7 +338,7 @@
 	netlink_table_grab();
 	if (sk_del_node_init(sk))
 		nl_table[sk->sk_protocol].hash.entries--;
-	if (nlk_sk(sk)->groups)
+	if (nlk_sk(sk)->subscriptions)
 		__sk_del_bind_node(sk);
 	netlink_table_ungrab();
 }
@@ -369,6 +374,8 @@
 static int netlink_create(struct socket *sock, int protocol)
 {
 	struct module *module = NULL;
+	struct netlink_sock *nlk;
+	unsigned int groups;
 	int err = 0;
  
 	sock->state = SS_UNCONNECTED;
  
  
  
  
@@ -392,15 +399,23 @@
 		module = nl_table[protocol].module;
 	else
 		err = -EPROTONOSUPPORT;
+	groups = nl_table[protocol].groups;
 	netlink_unlock_table();
  
-	if (err)
-		goto out;
+	if (err || (err = __netlink_create(sock, protocol) < 0))
+		goto out_module;
  
-	if ((err = __netlink_create(sock, protocol) < 0))
+	nlk = nlk_sk(sock->sk);
+
+	nlk->groups = kmalloc(NLGRPSZ(groups), GFP_KERNEL);
+	if (nlk->groups == NULL) {
+		err = -ENOMEM;
 		goto out_module;
+	}
+	memset(nlk->groups, 0, NLGRPSZ(groups));
+	nlk->ngroups = groups;
  
-	nlk_sk(sock->sk)->module = module;
+	nlk->module = module;
 out:
 	return err;
  
@@ -437,7 +452,7 @@
  
 	skb_queue_purge(&sk->sk_write_queue);
  
-	if (nlk->pid && !nlk->groups) {
+	if (nlk->pid && !nlk->subscriptions) {
 		struct netlink_notify n = {
 						.protocol = sk->sk_protocol,
 						.pid = nlk->pid,
@@ -455,6 +470,9 @@
 		netlink_table_ungrab();
 	}
  
+	kfree(nlk->groups);
+	nlk->groups = NULL;
+
 	sock_put(sk);
 	return 0;
 }
@@ -503,6 +521,18 @@
 	       capable(CAP_NET_ADMIN);
 } 
  
+static void
+netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions)
+{
+	struct netlink_sock *nlk = nlk_sk(sk);
+
+	if (nlk->subscriptions && !subscriptions)
+		__sk_del_bind_node(sk);
+	else if (!nlk->subscriptions && subscriptions)
+		sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list);
+	nlk->subscriptions = subscriptions;
+}
+
 static int netlink_bind(struct socket *sock, struct sockaddr *addr, int addr_len)
 {
 	struct sock *sk = sock->sk;
  
@@ -528,15 +558,14 @@
 			return err;
 	}
  
-	if (!nladdr->nl_groups && !nlk->groups)
+	if (!nladdr->nl_groups && !(u32)nlk->groups[0])
 		return 0;
  
 	netlink_table_grab();
-	if (nlk->groups && !nladdr->nl_groups)
-		__sk_del_bind_node(sk);
-	else if (!nlk->groups && nladdr->nl_groups)
-		sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list);
-	nlk->groups = nladdr->nl_groups;
+	netlink_update_subscriptions(sk, nlk->subscriptions +
+	                                 hweight32(nladdr->nl_groups) -
+	                                 hweight32(nlk->groups[0]));
+	nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) | nladdr->nl_groups; 
 	netlink_table_ungrab();
  
 	return 0;
@@ -590,7 +619,7 @@
 		nladdr->nl_groups = netlink_group_mask(nlk->dst_group);
 	} else {
 		nladdr->nl_pid = nlk->pid;
-		nladdr->nl_groups = nlk->groups; 
+		nladdr->nl_groups = nlk->groups[0];
 	}
 	return 0;
 }
@@ -791,7 +820,8 @@
 	if (p->exclude_sk == sk)
 		goto out;
  
-	if (nlk->pid == p->pid || !(nlk->groups & netlink_group_mask(p->group)))
+	if (nlk->pid == p->pid || p->group - 1 >= nlk->ngroups ||
+	    !test_bit(p->group - 1, nlk->groups))
 		goto out;
  
 	if (p->failure) {
@@ -887,7 +917,8 @@
 	if (sk == p->exclude_sk)
 		goto out;
  
-	if (nlk->pid == p->pid || !(nlk->groups & netlink_group_mask(p->group)))
+	if (nlk->pid == p->pid || p->group - 1 >= nlk->ngroups ||
+	    !test_bit(p->group - 1, nlk->groups))
 		goto out;
  
 	sk->sk_err = p->code;
@@ -1112,6 +1143,7 @@
 	nlk->flags |= NETLINK_KERNEL_SOCKET;
  
 	netlink_table_grab();
+	nl_table[unit].groups = 32;
 	nl_table[unit].module = module;
 	nl_table[unit].registered = 1;
 	netlink_table_ungrab();
@@ -1358,7 +1390,8 @@
 			   s,
 			   s->sk_protocol,
 			   nlk->pid,
-			   nlk->groups,
+			   nlk->flags & NETLINK_KERNEL_SOCKET ?
+				0 : (unsigned int)nlk->groups[0],
 			   atomic_read(&s->sk_rmem_alloc),
 			   atomic_read(&s->sk_wmem_alloc),
 			   nlk->cb,
...	...	@@ -60,21 +60,24 @@
60	60	#include <net/scm.h>
61	61
62	62	#define Nprintk(a...)
	63	+#define NLGRPSZ(x) (ALIGN(x, sizeof(unsigned long) * 8) / 8)
63	64
64	65	struct netlink_sock {
65	66	/* struct sock has to be the first member of netlink_sock */
66	67	struct sock sk;
67	68	u32 pid;
68		- unsigned int groups;
69	69	u32 dst_pid;
70	70	u32 dst_group;
	71	+ u32 flags;
	72	+ u32 subscriptions;
	73	+ u32 ngroups;
	74	+ unsigned long *groups;
71	75	unsigned long state;
72	76	wait_queue_head_t wait;
73	77	struct netlink_callback *cb;
74	78	spinlock_t cb_lock;
75	79	void (data_ready)(struct sock sk, int bytes);
76	80	struct module *module;
77		- u32 flags;
78	81	};
79	82
80	83	#define NETLINK_KERNEL_SOCKET 0x1
...	...	@@ -101,6 +104,7 @@
101	104	struct nl_pid_hash hash;
102	105	struct hlist_head mc_list;
103	106	unsigned int nl_nonroot;
	107	+ unsigned int groups;
104	108	struct module *module;
105	109	int registered;
106	110	};
...	...	@@ -138,6 +142,7 @@
138	142	BUG_TRAP(!atomic_read(&sk->sk_rmem_alloc));
139	143	BUG_TRAP(!atomic_read(&sk->sk_wmem_alloc));
140	144	BUG_TRAP(!nlk_sk(sk)->cb);
	145	+ BUG_TRAP(!nlk_sk(sk)->groups);
141	146	}
142	147
143	148	/* This lock without WQ_FLAG_EXCLUSIVE is good on UP and it is _very_ bad on SMP.
...	...	@@ -333,7 +338,7 @@
333	338	netlink_table_grab();
334	339	if (sk_del_node_init(sk))
335	340	nl_table[sk->sk_protocol].hash.entries--;
336		- if (nlk_sk(sk)->groups)
	341	+ if (nlk_sk(sk)->subscriptions)
337	342	__sk_del_bind_node(sk);
338	343	netlink_table_ungrab();
339	344	}
...	...	@@ -369,6 +374,8 @@
369	374	static int netlink_create(struct socket *sock, int protocol)
370	375	{
371	376	struct module *module = NULL;
	377	+ struct netlink_sock *nlk;
	378	+ unsigned int groups;
372	379	int err = 0;
373	380
374	381	sock->state = SS_UNCONNECTED;
375	382
376	383
377	384
378	385
...	...	@@ -392,15 +399,23 @@
392	399	module = nl_table[protocol].module;
393	400	else
394	401	err = -EPROTONOSUPPORT;
	402	+ groups = nl_table[protocol].groups;
395	403	netlink_unlock_table();
396	404
397		- if (err)
398		- goto out;
	405	+ if (err \|\| (err = __netlink_create(sock, protocol) < 0))
	406	+ goto out_module;
399	407
400		- if ((err = __netlink_create(sock, protocol) < 0))
	408	+ nlk = nlk_sk(sock->sk);
	409	+
	410	+ nlk->groups = kmalloc(NLGRPSZ(groups), GFP_KERNEL);
	411	+ if (nlk->groups == NULL) {
	412	+ err = -ENOMEM;
401	413	goto out_module;
	414	+ }
	415	+ memset(nlk->groups, 0, NLGRPSZ(groups));
	416	+ nlk->ngroups = groups;
402	417
403		- nlk_sk(sock->sk)->module = module;
	418	+ nlk->module = module;
404	419	out:
405	420	return err;
406	421
...	...	@@ -437,7 +452,7 @@
437	452
438	453	skb_queue_purge(&sk->sk_write_queue);
439	454
440		- if (nlk->pid && !nlk->groups) {
	455	+ if (nlk->pid && !nlk->subscriptions) {
441	456	struct netlink_notify n = {
442	457	.protocol = sk->sk_protocol,
443	458	.pid = nlk->pid,
...	...	@@ -455,6 +470,9 @@
455	470	netlink_table_ungrab();
456	471	}
457	472
	473	+ kfree(nlk->groups);
	474	+ nlk->groups = NULL;
	475	+
458	476	sock_put(sk);
459	477	return 0;
460	478	}
...	...	@@ -503,6 +521,18 @@
503	521	capable(CAP_NET_ADMIN);
504	522	}
505	523
	524	+static void
	525	+netlink_update_subscriptions(struct sock *sk, unsigned int subscriptions)
	526	+{
	527	+ struct netlink_sock *nlk = nlk_sk(sk);
	528	+
	529	+ if (nlk->subscriptions && !subscriptions)
	530	+ __sk_del_bind_node(sk);
	531	+ else if (!nlk->subscriptions && subscriptions)
	532	+ sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list);
	533	+ nlk->subscriptions = subscriptions;
	534	+}
	535	+
506	536	static int netlink_bind(struct socket sock, struct sockaddr addr, int addr_len)
507	537	{
508	538	struct sock *sk = sock->sk;
509	539
...	...	@@ -528,15 +558,14 @@
528	558	return err;
529	559	}
530	560
531		- if (!nladdr->nl_groups && !nlk->groups)
	561	+ if (!nladdr->nl_groups && !(u32)nlk->groups[0])
532	562	return 0;
533	563
534	564	netlink_table_grab();
535		- if (nlk->groups && !nladdr->nl_groups)
536		- __sk_del_bind_node(sk);
537		- else if (!nlk->groups && nladdr->nl_groups)
538		- sk_add_bind_node(sk, &nl_table[sk->sk_protocol].mc_list);
539		- nlk->groups = nladdr->nl_groups;
	565	+ netlink_update_subscriptions(sk, nlk->subscriptions +
	566	+ hweight32(nladdr->nl_groups) -
	567	+ hweight32(nlk->groups[0]));
	568	+ nlk->groups[0] = (nlk->groups[0] & ~0xffffffffUL) \| nladdr->nl_groups;
540	569	netlink_table_ungrab();
541	570
542	571	return 0;
...	...	@@ -590,7 +619,7 @@
590	619	nladdr->nl_groups = netlink_group_mask(nlk->dst_group);
591	620	} else {
592	621	nladdr->nl_pid = nlk->pid;
593		- nladdr->nl_groups = nlk->groups;
	622	+ nladdr->nl_groups = nlk->groups[0];
594	623	}
595	624	return 0;
596	625	}
...	...	@@ -791,7 +820,8 @@
791	820	if (p->exclude_sk == sk)
792	821	goto out;
793	822
794		- if (nlk->pid == p->pid \|\| !(nlk->groups & netlink_group_mask(p->group)))
	823	+ if (nlk->pid == p->pid \|\| p->group - 1 >= nlk->ngroups \|\|
	824	+ !test_bit(p->group - 1, nlk->groups))
795	825	goto out;
796	826
797	827	if (p->failure) {
...	...	@@ -887,7 +917,8 @@
887	917	if (sk == p->exclude_sk)
888	918	goto out;
889	919
890		- if (nlk->pid == p->pid \|\| !(nlk->groups & netlink_group_mask(p->group)))
	920	+ if (nlk->pid == p->pid \|\| p->group - 1 >= nlk->ngroups \|\|
	921	+ !test_bit(p->group - 1, nlk->groups))
891	922	goto out;
892	923
893	924	sk->sk_err = p->code;
...	...	@@ -1112,6 +1143,7 @@
1112	1143	nlk->flags \|= NETLINK_KERNEL_SOCKET;
1113	1144
1114	1145	netlink_table_grab();
	1146	+ nl_table[unit].groups = 32;
1115	1147	nl_table[unit].module = module;
1116	1148	nl_table[unit].registered = 1;
1117	1149	netlink_table_ungrab();
...	...	@@ -1358,7 +1390,8 @@
1358	1390	s,
1359	1391	s->sk_protocol,
1360	1392	nlk->pid,
1361		- nlk->groups,
	1393	+ nlk->flags & NETLINK_KERNEL_SOCKET ?
	1394	+ 0 : (unsigned int)nlk->groups[0],
1362	1395	atomic_read(&s->sk_rmem_alloc),
1363	1396	atomic_read(&s->sk_wmem_alloc),
1364	1397	nlk->cb,