Skip to contentSkip to conversation

net_sched: make traffic control network namespace aware

PATCH Applied Tom Goff 3 participants 5 messages Mar 20, 2010 – Apr 11, 2010

Subsystems (3)
lore
Share & tools
lore

Tom Goff

author
#0 lore raw

PATCH 1/1 net_sched: make traffic control network namespace aware

Mostly minor changes to add a net argument to various functions and remove initial network namespace checks.

Make /proc/net/psched per network namespace.

---
 net/sched/act_api.c |   45 ++++++++++----------
 net/sched/cls_api.c |   30 ++++++--------
 net/sched/sch_api.c |  112 ++++++++++++++++++++++++++++++++------------------
 3 files changed, 107 insertions(+), 80 deletions(-)
net/sched/act_api.c +22 -23
index 64f5e32..7a558da 100644
@@ -667,7 +667,8 @@ nlmsg_failure:
 }
 
 static int
-act_get_notify(u32 pid, struct nlmsghdr *n, struct tc_action *a, int event)
+act_get_notify(struct net *net, u32 pid, struct nlmsghdr *n,
+	       struct tc_action *a, int event)
 {
 	struct sk_buff *skb;
 
@@ -679,7 +680,7 @@ act_get_notify(u32 pid, struct nlmsghdr *n, struct tc_action *a, int event)
 		return -EINVAL;
 	}
 
-	return rtnl_unicast(skb, &init_net, pid);
+	return rtnl_unicast(skb, net, pid);
 }
 
 static struct tc_action *
@@ -749,7 +750,8 @@ static struct tc_action *create_a(int i)
 	return act;
 }
 
-static int tca_action_flush(struct nlattr *nla, struct nlmsghdr *n, u32 pid)
+static int tca_action_flush(struct net *net, struct nlattr *nla,
+			    struct nlmsghdr *n, u32 pid)
 {
 	struct sk_buff *skb;
 	unsigned char *b;
@@ -808,7 +810,7 @@ static int tca_action_flush(struct nlattr *nla, struct nlmsghdr *n, u32 pid)
 	nlh->nlmsg_flags |= NLM_F_ROOT;
 	module_put(a->ops->owner);
 	kfree(a);
-	err = rtnetlink_send(skb, &init_net, pid, RTNLGRP_TC, n->nlmsg_flags&NLM_F_ECHO);
+	err = rtnetlink_send(skb, net, pid, RTNLGRP_TC, n->nlmsg_flags&NLM_F_ECHO);
 	if (err > 0)
 		return 0;
 
@@ -825,7 +827,8 @@ noflush_out:
 }
 
 static int
-tca_action_gd(struct nlattr *nla, struct nlmsghdr *n, u32 pid, int event)
+tca_action_gd(struct net *net, struct nlattr *nla, struct nlmsghdr *n,
+	      u32 pid, int event)
 {
 	int i, ret;
 	struct nlattr *tb[TCA_ACT_MAX_PRIO+1];
@@ -837,7 +840,7 @@ tca_action_gd(struct nlattr *nla, struct nlmsghdr *n, u32 pid, int event)
 
 	if (event == RTM_DELACTION && n->nlmsg_flags&NLM_F_ROOT) {
 		if (tb[1] != NULL)
-			return tca_action_flush(tb[1], n, pid);
+			return tca_action_flush(net, tb[1], n, pid);
 		else
 			return -EINVAL;
 	}
@@ -858,7 +861,7 @@ tca_action_gd(struct nlattr *nla, struct nlmsghdr *n, u32 pid, int event)
 	}
 
 	if (event == RTM_GETACTION)
-		ret = act_get_notify(pid, n, head, event);
+		ret = act_get_notify(net, pid, n, head, event);
 	else { /* delete */
 		struct sk_buff *skb;
 
@@ -877,7 +880,7 @@ tca_action_gd(struct nlattr *nla, struct nlmsghdr *n, u32 pid, int event)
 
 		/* now do the delete */
 		tcf_action_destroy(head, 0);
-		ret = rtnetlink_send(skb, &init_net, pid, RTNLGRP_TC,
+		ret = rtnetlink_send(skb, net, pid, RTNLGRP_TC,
 				     n->nlmsg_flags&NLM_F_ECHO);
 		if (ret > 0)
 			return 0;
@@ -888,8 +891,8 @@ err:
 	return ret;
 }
 
-static int tcf_add_notify(struct tc_action *a, u32 pid, u32 seq, int event,
-			  u16 flags)
+static int tcf_add_notify(struct net *net, struct tc_action *a,
+			  u32 pid, u32 seq, int event, u16 flags)
 {
 	struct tcamsg *t;
 	struct nlmsghdr *nlh;
@@ -922,7 +925,7 @@ static int tcf_add_notify(struct tc_action *a, u32 pid, u32 seq, int event,
 	nlh->nlmsg_len = skb_tail_pointer(skb) - b;
 	NETLINK_CB(skb).dst_group = RTNLGRP_TC;
 
-	err = rtnetlink_send(skb, &init_net, pid, RTNLGRP_TC, flags&NLM_F_ECHO);
+	err = rtnetlink_send(skb, net, pid, RTNLGRP_TC, flags&NLM_F_ECHO);
 	if (err > 0)
 		err = 0;
 	return err;
@@ -935,7 +938,8 @@ nlmsg_failure:
 
 
 static int
-tcf_action_add(struct nlattr *nla, struct nlmsghdr *n, u32 pid, int ovr)
+tcf_action_add(struct net *net, struct nlattr *nla, struct nlmsghdr *n,
+	       u32 pid, int ovr)
 {
 	int ret = 0;
 	struct tc_action *act;
@@ -953,7 +957,7 @@ tcf_action_add(struct nlattr *nla, struct nlmsghdr *n, u32 pid, int ovr)
 	/* dump then free all the actions after update; inserted policy
 	 * stays intact
 	 * */
-	ret = tcf_add_notify(act, pid, seq, RTM_NEWACTION, n->nlmsg_flags);
+	ret = tcf_add_notify(net, act, pid, seq, RTM_NEWACTION, n->nlmsg_flags);
 	for (a = act; a; a = act) {
 		act = a->next;
 		kfree(a);
@@ -969,9 +973,6 @@ static int tc_ctl_action(struct sk_buff *skb, struct nlmsghdr *n, void *arg)
 	u32 pid = skb ? NETLINK_CB(skb).pid : 0;
 	int ret = 0, ovr = 0;
 
-	if (!net_eq(net, &init_net))
-		return -EINVAL;
-
 	ret = nlmsg_parse(n, sizeof(struct tcamsg), tca, TCA_ACT_MAX, NULL);
 	if (ret < 0)
 		return ret;
@@ -994,15 +995,17 @@ static int tc_ctl_action(struct sk_buff *skb, struct nlmsghdr *n, void *arg)
 		if (n->nlmsg_flags&NLM_F_REPLACE)
 			ovr = 1;
 replay:
-		ret = tcf_action_add(tca[TCA_ACT_TAB], n, pid, ovr);
+		ret = tcf_action_add(net, tca[TCA_ACT_TAB], n, pid, ovr);
 		if (ret == -EAGAIN)
 			goto replay;
 		break;
 	case RTM_DELACTION:
-		ret = tca_action_gd(tca[TCA_ACT_TAB], n, pid, RTM_DELACTION);
+		ret = tca_action_gd(net, tca[TCA_ACT_TAB], n,
+				    pid, RTM_DELACTION);
 		break;
 	case RTM_GETACTION:
-		ret = tca_action_gd(tca[TCA_ACT_TAB], n, pid, RTM_GETACTION);
+		ret = tca_action_gd(net, tca[TCA_ACT_TAB], n,
+				    pid, RTM_GETACTION);
 		break;
 	default:
 		BUG();
@@ -1042,7 +1045,6 @@ find_dump_kind(const struct nlmsghdr *n)
 static int
 tc_dump_action(struct sk_buff *skb, struct netlink_callback *cb)
 {
-	struct net *net = sock_net(skb->sk);
 	struct nlmsghdr *nlh;
 	unsigned char *b = skb_tail_pointer(skb);
 	struct nlattr *nest;
@@ -1052,9 +1054,6 @@ tc_dump_action(struct sk_buff *skb, struct netlink_callback *cb)
 	struct tcamsg *t = (struct tcamsg *) NLMSG_DATA(cb->nlh);
 	struct nlattr *kind = find_dump_kind(cb->nlh);
 
-	if (!net_eq(net, &init_net))
-		return 0;
-
 	if (kind == NULL) {
 		printk("tc_dump_action: action bad kind\n");
 		return 0;
net/sched/cls_api.c +13 -17
index 3725d8f..4a795d9 100644
@@ -98,8 +98,9 @@ out:
 }
 EXPORT_SYMBOL(unregister_tcf_proto_ops);
 
-static int tfilter_notify(struct sk_buff *oskb, struct nlmsghdr *n,
-			  struct tcf_proto *tp, unsigned long fh, int event);
+static int tfilter_notify(struct net *net, struct sk_buff *oskb,
+			  struct nlmsghdr *n, struct tcf_proto *tp,
+			  unsigned long fh, int event);
 
 
 /* Select new prio value from the range, managed by kernel. */
@@ -137,9 +138,6 @@ static int tc_ctl_tfilter(struct sk_buff *skb, struct nlmsghdr *n, void *arg)
 	int err;
 	int tp_created = 0;
 
-	if (!net_eq(net, &init_net))
-		return -EINVAL;
-
 replay:
 	t = NLMSG_DATA(n);
 	protocol = TC_H_MIN(t->tcm_info);
@@ -158,7 +156,7 @@ replay:
 	/* Find head of filter chain. */
 
 	/* Find link */
-	dev = __dev_get_by_index(&init_net, t->tcm_ifindex);
+	dev = __dev_get_by_index(net, t->tcm_ifindex);
 	if (dev == NULL)
 		return -ENODEV;
 
@@ -282,7 +280,7 @@ replay:
 			*back = tp->next;
 			spin_unlock_bh(root_lock);
 
-			tfilter_notify(skb, n, tp, fh, RTM_DELTFILTER);
+			tfilter_notify(net, skb, n, tp, fh, RTM_DELTFILTER);
 			tcf_destroy(tp);
 			err = 0;
 			goto errout;
@@ -305,10 +303,10 @@ replay:
 		case RTM_DELTFILTER:
 			err = tp->ops->delete(tp, fh);
 			if (err == 0)
-				tfilter_notify(skb, n, tp, fh, RTM_DELTFILTER);
+				tfilter_notify(net, skb, n, tp, fh, RTM_DELTFILTER);
 			goto errout;
 		case RTM_GETTFILTER:
-			err = tfilter_notify(skb, n, tp, fh, RTM_NEWTFILTER);
+			err = tfilter_notify(net, skb, n, tp, fh, RTM_NEWTFILTER);
 			goto errout;
 		default:
 			err = -EINVAL;
@@ -324,7 +322,7 @@ replay:
 			*back = tp;
 			spin_unlock_bh(root_lock);
 		}
-		tfilter_notify(skb, n, tp, fh, RTM_NEWTFILTER);
+		tfilter_notify(net, skb, n, tp, fh, RTM_NEWTFILTER);
 	} else {
 		if (tp_created)
 			tcf_destroy(tp);
@@ -370,8 +368,9 @@ nla_put_failure:
 	return -1;
 }
 
-static int tfilter_notify(struct sk_buff *oskb, struct nlmsghdr *n,
-			  struct tcf_proto *tp, unsigned long fh, int event)
+static int tfilter_notify(struct net *net, struct sk_buff *oskb,
+			  struct nlmsghdr *n, struct tcf_proto *tp,
+			  unsigned long fh, int event)
 {
 	struct sk_buff *skb;
 	u32 pid = oskb ? NETLINK_CB(oskb).pid : 0;
@@ -385,7 +384,7 @@ static int tfilter_notify(struct sk_buff *oskb, struct nlmsghdr *n,
 		return -EINVAL;
 	}
 
-	return rtnetlink_send(skb, &init_net, pid, RTNLGRP_TC,
+	return rtnetlink_send(skb, net, pid, RTNLGRP_TC,
 			      n->nlmsg_flags & NLM_F_ECHO);
 }
 
@@ -418,12 +417,9 @@ static int tc_dump_tfilter(struct sk_buff *skb, struct netlink_callback *cb)
 	const struct Qdisc_class_ops *cops;
 	struct tcf_dump_args arg;
 
-	if (!net_eq(net, &init_net))
-		return 0;
-
 	if (cb->nlh->nlmsg_len < NLMSG_LENGTH(sizeof(*tcm)))
 		return skb->len;
-	if ((dev = __dev_get_by_index(&init_net, tcm->tcm_ifindex)) == NULL)
+	if ((dev = __dev_get_by_index(net, tcm->tcm_ifindex)) == NULL)
 		return skb->len;
 
 	if (!tcm->tcm_parent)
net/sched/sch_api.c +72 -40

257 lines not on this page · raw message

Signature · 1 line
1.6.3.3
#1 lore raw
Quoted 4 lines · Tom Goff

Mostly minor changes to add a net argument to various functions and remove initial network namespace checks.

Make /proc/net/psched per network namespace.

Looks fine from a qdisc POV. One thing that appears to be missing though is teql master netdev registration in other than the initial namespace.

#2 lore raw
Quoted 9 lines · Patrick McHardy
Quoted 4 lines · Tom Goff

Mostly minor changes to add a net argument to various functions and remove initial network namespace checks.

Make /proc/net/psched per network namespace.

Looks fine from a qdisc POV. One thing that appears to be missing though is teql master netdev registration in other than the initial namespace.

Actually we could take this opportunity and add rtnl_link support for teql device registration. I can look into this in a couple of days.

maintainer Applied
#3 lore raw

From: Patrick McHardy <[email protected]>
Date: Mon, 22 Mar 2010 19:04:44 +0100

Quoted 7 lines
Quoted 6 lines · Patrick McHardy
Quoted 4 lines · Tom Goff

Mostly minor changes to add a net argument to various functions and remove initial network namespace checks.

Make /proc/net/psched per network namespace.

Quoted 7 lines
Quoted 3 lines

Looks fine from a qdisc POV. One thing that appears to be missing though is teql master netdev registration in other than the initial namespace.

Actually we could take this opportunity and add rtnl_link support for teql device registration. I can look into this in a couple of days.

I've applied Tom's patch to net-next-2.6, thanks everyone.

#4 lore raw
In reply to Patrick McHardy
Quoted 5 lines · Patrick McHardy
Quoted 4 lines · Patrick McHardy
Quoted 3 lines · Tom Goff

Mostly minor changes to add a net argument to various functions and remove initial network namespace checks.

Quoted 8 lines
Quoted 4 lines
Quoted 1 line

Make /proc/net/psched per network namespace.

Looks fine from a qdisc POV. One thing that appears to be missing though is teql master netdev registration in other than the initial namespace.

Actually we could take this opportunity and add rtnl_link support for teql device registration. I can look into this in a couple of days.

I tried to do that, but adding proper netns support is more complicated than I expected. sch_teql registers a qdisc for each master device using the name of the master. Currently qdisc registrations are global, so this doesn't work with network namespaces.

We could of course make them per netns, but that would require duplicating all global registrations for each namespace. I'm not convinced that its worth doing this since its only teql that needs it and it doesn't seem to be very useful to use teql in a virtual environment.