Skip to contentSkip to conversation

network namespaces for generic netlink

PATCH In review Johannes Berg 1 participant 5 messages Jul 7, 2009

Subsystems (8)
lore
Share & tools
lore

Johannes Berg

author maintainer
#0 lore raw

cover network namespaces for generic netlink

This is a repost of my earlier patches, now with fixed semantics and locking. The first patch is required to make RCU-iteration of the network namespace list possible, which the second patch requires.

The third patch is preparation for nl80211 netns work, I'm happy to give that to John instead to go via the wireless tree with the patch that will make use of the now exported function, but it is possible that other generic netlink users will also require it once they being upgrading their family to be netns aware.

johannes

author maintainer
#1 lore raw

PATCH 1/3 net: make namespace iteration possible under RCU

All we need to take care of is using proper RCU list add/del primitives and inserting a synchronize_rcu() at one place to make sure the exit notifiers are run after everybody has stopped iterating the list.

---
 include/net/net_namespace.h |    3 +++
 net/core/net_namespace.c    |   14 +++++++++++---
 2 files changed, 14 insertions(+), 3 deletions(-)
include/net/net_namespace.h +3 -0
@@ -211,6 +211,9 @@ static inline struct net *read_pnet(stru
 #define for_each_net(VAR)				\
 	list_for_each_entry(VAR, &net_namespace_list, list)
 
+#define for_each_net_rcu(VAR)				\
+	list_for_each_entry_rcu(VAR, &net_namespace_list, list)
+
 #ifdef CONFIG_NET_NS
 #define __net_init
 #define __net_exit
net/core/net_namespace.c +11 -3
@@ -6,6 +6,7 @@
 #include <linux/delay.h>
 #include <linux/sched.h>
 #include <linux/idr.h>
+#include <linux/rculist.h>
 #include <net/net_namespace.h>
 #include <net/netns/generic.h>
 
@@ -127,7 +128,7 @@ static struct net *net_create(void)
 	rv = setup_net(net);
 	if (rv == 0) {
 		rtnl_lock();
-		list_add_tail(&net->list, &net_namespace_list);
+		list_add_tail_rcu(&net->list, &net_namespace_list);
 		rtnl_unlock();
 	}
 	mutex_unlock(&net_mutex);
@@ -156,9 +157,16 @@ static void cleanup_net(struct work_stru
 
 	/* Don't let anyone else find us. */
 	rtnl_lock();
-	list_del(&net->list);
+	list_del_rcu(&net->list);
 	rtnl_unlock();
 
+	/*
+	 * Another CPU might be rcu-iterating the list, wait for it.
+	 * This needs to be before calling the exit() notifiers, so
+	 * the rcu_barrier() below isn't sufficient alone.
+	 */
+	synchronize_rcu();
+
 	/* Run all of the network namespace exit methods */
 	list_for_each_entry_reverse(ops, &pernet_list, list) {
 		if (ops->exit)
@@ -219,7 +227,7 @@ static int __init net_ns_init(void)
 		panic("Could not setup the initial network namespace");
 
 	rtnl_lock();
-	list_add_tail(&init_net.list, &net_namespace_list);
+	list_add_tail_rcu(&init_net.list, &net_namespace_list);
 	rtnl_unlock();
 
 	mutex_unlock(&net_mutex);
author maintainer
#2 lore raw

PATCH 2/3 genetlink: make netns aware

In reply to Johannes Berg

This makes generic netlink network namespace aware. No generic netlink families except for the controller family are made namespace aware, they need to be checked one by one and then set the family->netnsok member to true.

A new function genlmsg_multicast_netns() is introduced to allow sending a multicast message in a given namespace, for example when it applies to an object that lives in that namespace, a new function genlmsg_multicast_allns() to send a message to all network namespaces (for objects that do not have an associated netns).

The function genlmsg_multicast() is changed to multicast the message in just init_net, which is currently correct for all generic netlink families since they only work in init_net right now. Some will later want to work in all net namespaces because they do not care about the netns at all -- those will have to be converted to use one of the new functions genlmsg_multicast_allns() or genlmsg_multicast_netns() whenever they are made netns aware in some way.

After this patch families can easily decide whether or not they should be available in all net namespaces. Many genl families us it for objects not related to networking and should therefore be available in all namespaces, but that will have to be done on a per family basis.

Note that this doesn't touch on the checkpoint/restart problem where network namespaces could be used, genl families and multicast groups are numbered globally and I see no easy way of changing that, especially since it must be possible to multicast to all network namespaces for those families that do not care about netns.

---
 fs/dlm/netlink.c               |    2 
 include/net/genetlink.h        |   66 +++++++++++++-
 include/net/net_namespace.h    |    2 
 kernel/taskstats.c             |   10 +-
 net/irda/irnetlink.c           |    2 
 net/netfilter/ipvs/ip_vs_ctl.c |    2 
 net/netlink/genetlink.c        |  186 ++++++++++++++++++++++++++++++++---------
 net/tipc/netlink.c             |    2 
 net/wireless/nl80211.c         |   14 +--
 9 files changed, 223 insertions(+), 63 deletions(-)
include/net/genetlink.h +59 -7
@@ -3,6 +3,7 @@
 
 #include <linux/genetlink.h>
 #include <net/netlink.h>
+#include <net/net_namespace.h>
 
 /**
  * struct genl_multicast_group - generic netlink multicast group
@@ -27,6 +28,8 @@ struct genl_multicast_group
  * @name: name of family
  * @version: protocol version
  * @maxattr: maximum number of attributes supported
+ * @netnsok: set to true if the family can handle network
+ *	namespaces and should be presented in all of them
  * @attrbuf: buffer to store parsed attributes
  * @ops_list: list of all assigned operations
  * @family_list: family list
@@ -39,6 +42,7 @@ struct genl_family
 	char			name[GENL_NAMSIZ];
 	unsigned int		version;
 	unsigned int		maxattr;
+	bool			netnsok;
 	struct nlattr **	attrbuf;	/* private */
 	struct list_head	ops_list;	/* private */
 	struct list_head	family_list;	/* private */
@@ -62,8 +66,32 @@ struct genl_info
 	struct genlmsghdr *	genlhdr;
 	void *			userhdr;
 	struct nlattr **	attrs;
+#ifdef CONFIG_NET_NS
+	struct net *		_net;
+#endif
 };
 
+#ifdef CONFIG_NET_NS
+static inline struct net *genl_info_net(struct genl_info *info)
+{
+	return info->_net;
+}
+
+static inline void genl_info_net_set(struct genl_info *info, struct net *net)
+{
+	info->_net = net;
+}
+#else
+static inline struct net *genl_info_net(struct genl_info *info)
+{
+	return &init_net;
+}
+
+static inline void genl_info_net_set(struct genl_info *info, struct net *net)
+{
+}
+#endif
+
 /**
  * struct genl_ops - generic netlink operations
  * @cmd: command identifier
@@ -98,8 +126,6 @@ extern int genl_register_mc_group(struct
 extern void genl_unregister_mc_group(struct genl_family *family,
 				     struct genl_multicast_group *grp);
 
-extern struct sock *genl_sock;
-
 /**
  * genlmsg_put - Add generic netlink header to netlink message
  * @skb: socket buffer holding the message
@@ -170,7 +196,21 @@ static inline void genlmsg_cancel(struct
 }
 
 /**
- * genlmsg_multicast - multicast a netlink message
+ * genlmsg_multicast_netns - multicast a netlink message to a specific netns
+ * @net: the net namespace
+ * @skb: netlink message as socket buffer
+ * @pid: own netlink pid to avoid sending to yourself
+ * @group: multicast group id
+ * @flags: allocation flags
+ */
+static inline int genlmsg_multicast_netns(struct net *net, struct sk_buff *skb,
+					  u32 pid, unsigned int group, gfp_t flags)
+{
+	return nlmsg_multicast(net->genl_sock, skb, pid, group, flags);
+}
+
+/**
+ * genlmsg_multicast - multicast a netlink message to the default netns
  * @skb: netlink message as socket buffer
  * @pid: own netlink pid to avoid sending to yourself
  * @group: multicast group id
@@ -179,17 +219,29 @@ static inline void genlmsg_cancel(struct
 static inline int genlmsg_multicast(struct sk_buff *skb, u32 pid,
 				    unsigned int group, gfp_t flags)
 {
-	return nlmsg_multicast(genl_sock, skb, pid, group, flags);
+	return genlmsg_multicast_netns(&init_net, skb, pid, group, flags);
 }
 
 /**
+ * genlmsg_multicast_allns - multicast a netlink message to all net namespaces
+ * @skb: netlink message as socket buffer
+ * @pid: own netlink pid to avoid sending to yourself
+ * @group: multicast group id
+ * @flags: allocation flags
+ *
+ * This function must hold the RTNL or rcu_read_lock().
+ */
+int genlmsg_multicast_allns(struct sk_buff *skb, u32 pid,
+			    unsigned int group, gfp_t flags);
+
+/**
  * genlmsg_unicast - unicast a netlink message
  * @skb: netlink message as socket buffer
  * @pid: netlink pid of the destination socket
  */
-static inline int genlmsg_unicast(struct sk_buff *skb, u32 pid)
+static inline int genlmsg_unicast(struct net *net, struct sk_buff *skb, u32 pid)
 {
-	return nlmsg_unicast(genl_sock, skb, pid);
+	return nlmsg_unicast(net->genl_sock, skb, pid);
 }
 
 /**
@@ -199,7 +251,7 @@ static inline int genlmsg_unicast(struct
  */
 static inline int genlmsg_reply(struct sk_buff *skb, struct genl_info *info)
 {
-	return genlmsg_unicast(skb, info->snd_pid);
+	return genlmsg_unicast(genl_info_net(info), skb, info->snd_pid);
 }
 
 /**
net/netlink/genetlink.c +146 -40

312 lines not on this page · raw message

fs/dlm/netlink.c +1 -1
@@ -63,7 +63,7 @@ static int send_data(struct sk_buff *skb
 		return rv;
 	}
 
-	return genlmsg_unicast(skb, listener_nlpid);
+	return genlmsg_unicast(&init_net, skb, listener_nlpid);
 }
 
 static int user_cmd(struct sk_buff *skb, struct genl_info *info)
kernel/taskstats.c +5 -5
@@ -108,7 +108,7 @@ static int prepare_reply(struct genl_inf
 /*
  * Send taskstats data in @skb to listener with nl_pid @pid
  */
-static int send_reply(struct sk_buff *skb, pid_t pid)
+static int send_reply(struct sk_buff *skb, struct genl_info *info)
 {
 	struct genlmsghdr *genlhdr = nlmsg_data(nlmsg_hdr(skb));
 	void *reply = genlmsg_data(genlhdr);
@@ -120,7 +120,7 @@ static int send_reply(struct sk_buff *sk
 		return rc;
 	}
 
-	return genlmsg_unicast(skb, pid);
+	return genlmsg_reply(skb, info);
 }
 
 /*
@@ -150,7 +150,7 @@ static void send_cpu_listeners(struct sk
 			if (!skb_next)
 				break;
 		}
-		rc = genlmsg_unicast(skb_cur, s->pid);
+		rc = genlmsg_unicast(&init_net, skb_cur, s->pid);
 		if (rc == -ECONNREFUSED) {
 			s->valid = 0;
 			delcount++;
@@ -418,7 +418,7 @@ static int cgroupstats_user_cmd(struct s
 		goto err;
 	}
 
-	rc = send_reply(rep_skb, info->snd_pid);
+	rc = send_reply(rep_skb, info);
 
 err:
 	fput_light(file, fput_needed);
@@ -487,7 +487,7 @@ free_return_rc:
 	} else
 		goto err;
 
-	return send_reply(rep_skb, info->snd_pid);
+	return send_reply(rep_skb, info);
 err:
 	nlmsg_free(rep_skb);
 	return rc;
net/irda/irnetlink.c +1 -1
@@ -115,7 +115,7 @@ static int irda_nl_get_mode(struct sk_bu
 
 	genlmsg_end(msg, hdr);
 
-	return genlmsg_unicast(msg, info->snd_pid);
+	return genlmsg_reply(msg, info);
 
  err_out:
 	nlmsg_free(msg);
net/netfilter/ipvs/ip_vs_ctl.c +1 -1
@@ -3231,7 +3231,7 @@ static int ip_vs_genl_get_cmd(struct sk_
 	}
 
 	genlmsg_end(msg, reply);
-	ret = genlmsg_unicast(msg, info->snd_pid);
+	ret = genlmsg_reply(msg, info);
 	goto out;
 
 nla_put_failure:
net/tipc/netlink.c +1 -1
@@ -62,7 +62,7 @@ static int handle_cmd(struct sk_buff *sk
 		rep_nlh = nlmsg_hdr(rep_buf);
 		memcpy(rep_nlh, req_nlh, hdr_space);
 		rep_nlh->nlmsg_len = rep_buf->len;
-		genlmsg_unicast(rep_buf, NETLINK_CB(skb).pid);
+		genlmsg_unicast(&init_net, rep_buf, NETLINK_CB(skb).pid);
 	}
 
 	return 0;
net/wireless/nl80211.c +7 -7
@@ -413,7 +413,7 @@ static int nl80211_get_wiphy(struct sk_b
 
 	cfg80211_unlock_rdev(dev);
 
-	return genlmsg_unicast(msg, info->snd_pid);
+	return genlmsg_reply(msg, info);
 
  out_free:
 	nlmsg_free(msg);
@@ -739,7 +739,7 @@ static int nl80211_get_interface(struct 
 	dev_put(netdev);
 	cfg80211_unlock_rdev(dev);
 
-	return genlmsg_unicast(msg, info->snd_pid);
+	return genlmsg_reply(msg, info);
 
  out_free:
 	nlmsg_free(msg);
@@ -1030,7 +1030,7 @@ static int nl80211_get_key(struct sk_buf
 		goto nla_put_failure;
 
 	genlmsg_end(msg, hdr);
-	err = genlmsg_unicast(msg, info->snd_pid);
+	err = genlmsg_reply(msg, info);
 	goto out;
 
  nla_put_failure:
@@ -1618,7 +1618,7 @@ static int nl80211_get_station(struct sk
 				 dev, mac_addr, &sinfo) < 0)
 		goto out_free;
 
-	err = genlmsg_unicast(msg, info->snd_pid);
+	err = genlmsg_reply(msg, info);
 	goto out;
 
  out_free:
@@ -2087,7 +2087,7 @@ static int nl80211_get_mpath(struct sk_b
 				 dev, dst, next_hop, &pinfo) < 0)
 		goto out_free;
 
-	err = genlmsg_unicast(msg, info->snd_pid);
+	err = genlmsg_reply(msg, info);
 	goto out;
 
  out_free:
@@ -2436,7 +2436,7 @@ static int nl80211_get_mesh_params(struc
 			cur_params.dot11MeshHWMPnetDiameterTraversalTime);
 	nla_nest_end(msg, pinfoattr);
 	genlmsg_end(msg, hdr);
-	err = genlmsg_unicast(msg, info->snd_pid);
+	err = genlmsg_reply(msg, info);
 	goto out;
 
  nla_put_failure:
@@ -2624,7 +2624,7 @@ static int nl80211_get_reg(struct sk_buf
 	nla_nest_end(msg, nl_reg_rules);
 
 	genlmsg_end(msg, hdr);
-	err = genlmsg_unicast(msg, info->snd_pid);
+	err = genlmsg_reply(msg, info);
 	goto out;
 
 nla_put_failure:
include/net/net_namespace.h +2 -0
@@ -26,6 +26,7 @@ struct net_device;
 struct sock;
 struct ctl_table_header;
 struct net_generic;
+struct sock;
 
 struct net {
 	atomic_t		count;		/* To decided when the network
@@ -57,6 +58,7 @@ struct net {
 	spinlock_t		rules_mod_lock;
 
 	struct sock 		*rtnl;			/* rtnetlink socket */
+	struct sock		*genl_sock;
 
 	struct netns_core	core;
 	struct netns_mib	mib;
author maintainer
#4 lore raw

PATCH 3/3 net: move and export get_net_ns_by_pid

In reply to Johannes Berg

The function get_net_ns_by_pid(), to get a network namespace from a pid_t, will be required in cfg80211 as well. Therefore, let's move it to net_namespace.c and export it. We can't make it a static inline in the !NETNS case because it needs to verify that the given pid even exists (and return -ESRCH).

---
 include/net/net_namespace.h |    2 ++
 net/core/net_namespace.c    |   21 +++++++++++++++++++++
 net/core/rtnetlink.c        |   21 +--------------------
 3 files changed, 24 insertions(+), 20 deletions(-)
include/net/net_namespace.h +2 -0
@@ -111,6 +111,8 @@ static inline struct net *copy_net_ns(un
 
 extern struct list_head net_namespace_list;
 
+extern struct net *get_net_ns_by_pid(pid_t pid);
+
 #ifdef CONFIG_NET_NS
 extern void __put_net(struct net *net);
 
net/core/net_namespace.c +21 -0
@@ -7,6 +7,7 @@
 #include <linux/sched.h>
 #include <linux/idr.h>
 #include <linux/rculist.h>
+#include <linux/nsproxy.h>
 #include <net/net_namespace.h>
 #include <net/netns/generic.h>
 
@@ -201,6 +202,26 @@ struct net *copy_net_ns(unsigned long fl
 }
 #endif
 
+struct net *get_net_ns_by_pid(pid_t pid)
+{
+	struct task_struct *tsk;
+	struct net *net;
+
+	/* Lookup the network namespace */
+	net = ERR_PTR(-ESRCH);
+	rcu_read_lock();
+	tsk = find_task_by_vpid(pid);
+	if (tsk) {
+		struct nsproxy *nsproxy;
+		nsproxy = task_nsproxy(tsk);
+		if (nsproxy)
+			net = get_net(nsproxy->net_ns);
+	}
+	rcu_read_unlock();
+	return net;
+}
+EXPORT_SYMBOL_GPL(get_net_ns_by_pid);
+
 static int __init net_ns_init(void)
 {
 	struct net_generic *ng;
net/core/rtnetlink.c +1 -20
@@ -35,7 +35,6 @@
 #include <linux/security.h>
 #include <linux/mutex.h>
 #include <linux/if_addr.h>
-#include <linux/nsproxy.h>
 
 #include <asm/uaccess.h>
 #include <asm/system.h>
@@ -52,6 +51,7 @@
 #include <net/pkt_sched.h>
 #include <net/fib_rules.h>
 #include <net/rtnetlink.h>
+#include <net/net_namespace.h>
 
 struct rtnl_link
 {
@@ -725,25 +725,6 @@ static const struct nla_policy ifla_info
 	[IFLA_INFO_DATA]	= { .type = NLA_NESTED },
 };
 
-static struct net *get_net_ns_by_pid(pid_t pid)
-{
-	struct task_struct *tsk;
-	struct net *net;
-
-	/* Lookup the network namespace */
-	net = ERR_PTR(-ESRCH);
-	rcu_read_lock();
-	tsk = find_task_by_vpid(pid);
-	if (tsk) {
-		struct nsproxy *nsproxy;
-		nsproxy = task_nsproxy(tsk);
-		if (nsproxy)
-			net = get_net(nsproxy->net_ns);
-	}
-	rcu_read_unlock();
-	return net;
-}
-
 static int validate_linkmsg(struct net_device *dev, struct nlattr *tb[])
 {
 	if (dev) {
author maintainer
#3 lore raw

Re: [PATCH 2/3] genetlink: make netns aware

In reply to Johannes Berg
Quoted 2 lines · Johannes Berg

plain text document attachment (030-genl-netns.patch) This makes generic netlink network namespace aware.

No, buggy again, and again for some reason I didn't notice on first testing :(

johannes