Skip to contentSkip to conversation

Ease nsid allocation

PATCH In review Nicolas Dichtel 4 participants 16 messages Sep 30, 2019 – Oct 10, 2019

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

cover Ease nsid allocation

The goal of the series is to ease nsid allocation from userland. The first patch is a preparation work and the second enables to receive the new nsid in the answer to RTM_NEWNSID.

 net/core/net_namespace.c | 118 ++++++++++++++++++++++++++++-------------------
 1 file changed, 71 insertions(+), 47 deletions(-)

Comments are welcomed,
Nicolas

#1 lore raw

PATCH 1/2 netns: move rtnl_net_get_size() and rtnl_net_fill()

There is no functional change in this patch, it only prepares the next one where rtnl_net_newid() will use rtnl_net_get_size() and rtnl_net_fill().

---
 net/core/net_namespace.c | 92 ++++++++++++++++++++--------------------
 1 file changed, 46 insertions(+), 46 deletions(-)
net/core/net_namespace.c +46 -46
index a0e0d298c991..8f5fa5d5becd 100644
@@ -716,6 +716,52 @@ static const struct nla_policy rtnl_net_policy[NETNSA_MAX + 1] = {
 	[NETNSA_TARGET_NSID]	= { .type = NLA_S32 },
 };
 
+struct net_fill_args {
+	u32 portid;
+	u32 seq;
+	int flags;
+	int cmd;
+	int nsid;
+	bool add_ref;
+	int ref_nsid;
+};
+
+static int rtnl_net_get_size(void)
+{
+	return NLMSG_ALIGN(sizeof(struct rtgenmsg))
+	       + nla_total_size(sizeof(s32)) /* NETNSA_NSID */
+	       + nla_total_size(sizeof(s32)) /* NETNSA_CURRENT_NSID */
+	       ;
+}
+
+static int rtnl_net_fill(struct sk_buff *skb, struct net_fill_args *args)
+{
+	struct nlmsghdr *nlh;
+	struct rtgenmsg *rth;
+
+	nlh = nlmsg_put(skb, args->portid, args->seq, args->cmd, sizeof(*rth),
+			args->flags);
+	if (!nlh)
+		return -EMSGSIZE;
+
+	rth = nlmsg_data(nlh);
+	rth->rtgen_family = AF_UNSPEC;
+
+	if (nla_put_s32(skb, NETNSA_NSID, args->nsid))
+		goto nla_put_failure;
+
+	if (args->add_ref &&
+	    nla_put_s32(skb, NETNSA_CURRENT_NSID, args->ref_nsid))
+		goto nla_put_failure;
+
+	nlmsg_end(skb, nlh);
+	return 0;
+
+nla_put_failure:
+	nlmsg_cancel(skb, nlh);
+	return -EMSGSIZE;
+}
+
 static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh,
 			  struct netlink_ext_ack *extack)
 {
@@ -776,52 +822,6 @@ static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh,
 	return err;
 }
 
-static int rtnl_net_get_size(void)
-{
-	return NLMSG_ALIGN(sizeof(struct rtgenmsg))
-	       + nla_total_size(sizeof(s32)) /* NETNSA_NSID */
-	       + nla_total_size(sizeof(s32)) /* NETNSA_CURRENT_NSID */
-	       ;
-}
-
-struct net_fill_args {
-	u32 portid;
-	u32 seq;
-	int flags;
-	int cmd;
-	int nsid;
-	bool add_ref;
-	int ref_nsid;
-};
-
-static int rtnl_net_fill(struct sk_buff *skb, struct net_fill_args *args)
-{
-	struct nlmsghdr *nlh;
-	struct rtgenmsg *rth;
-
-	nlh = nlmsg_put(skb, args->portid, args->seq, args->cmd, sizeof(*rth),
-			args->flags);
-	if (!nlh)
-		return -EMSGSIZE;
-
-	rth = nlmsg_data(nlh);
-	rth->rtgen_family = AF_UNSPEC;
-
-	if (nla_put_s32(skb, NETNSA_NSID, args->nsid))
-		goto nla_put_failure;
-
-	if (args->add_ref &&
-	    nla_put_s32(skb, NETNSA_CURRENT_NSID, args->ref_nsid))
-		goto nla_put_failure;
-
-	nlmsg_end(skb, nlh);
-	return 0;
-
-nla_put_failure:
-	nlmsg_cancel(skb, nlh);
-	return -EMSGSIZE;
-}
-
 static int rtnl_net_valid_getid_req(struct sk_buff *skb,
 				    const struct nlmsghdr *nlh,
 				    struct nlattr **tb,
Signature · 1 line
2.23.0
#2 lore raw

PATCH 2/2 netns/rtnl: return the new nsid to the user

In reply to Nicolas Dichtel

When the user asks for a new nsid, he can let the kernel choose it (by providing -1 in NETNSA_NSID). In this case, it's useful to answer to the netlink message with the chosen nsid.

---
 net/core/net_namespace.c | 26 +++++++++++++++++++++++++-
 1 file changed, 25 insertions(+), 1 deletion(-)
net/core/net_namespace.c +25 -1
index 8f5fa5d5becd..266d095296f3 100644
@@ -810,8 +810,32 @@ static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh,
 	err = alloc_netid(net, peer, nsid);
 	spin_unlock_bh(&net->nsid_lock);
 	if (err >= 0) {
+		struct net_fill_args fillargs = {
+			.portid = NETLINK_CB(skb).portid,
+			.seq = nlh->nlmsg_seq,
+			.cmd = RTM_NEWNSID,
+			.nsid = err,
+		};
+		struct sk_buff *msg;
+
+		/* The id has been allocated, thus first notify listeners */
 		rtnl_net_notifyid(net, RTM_NEWNSID, err);
-		err = 0;
+
+		/* Then, try to send the new nsid to the sender */
+		msg = nlmsg_new(rtnl_net_get_size(), GFP_KERNEL);
+		if (!msg) {
+			err = -ENOMEM;
+			NL_SET_ERR_MSG(extack, "Unable to alloc reply msg");
+			goto out;
+		}
+
+		err = rtnl_net_fill(msg, &fillargs);
+		if (err < 0) {
+			kfree_skb(msg);
+			goto out;
+		}
+
+		err = rtnl_unicast(msg, net, NETLINK_CB(skb).portid);
 	} else if (err == -ENOSPC && nsid >= 0) {
 		err = -EEXIST;
 		NL_SET_BAD_ATTR(extack, tb[NETNSA_NSID]);
Signature · 1 line
2.23.0
maintainer
#3 lore raw
In reply to Nicolas Dichtel

From: Nicolas Dichtel <[email protected]>
Date: Mon, 30 Sep 2019 18:02:12 +0200

Quoted 3 lines

The goal of the series is to ease nsid allocation from userland. The first patch is a preparation work and the second enables to receive the new nsid in the answer to RTM_NEWNSID.

The new reply message could break existing apps.

If an app only performs netnsid operations, and fills up the receive queue because it isn't reading these new replies (it had no reason to, they didn't exist previously), operations will start failing that would not fail previously because the receive queue is full.

Given this, I don't see how we can make the change.

#4 lore raw

Le 02/10/2019 à 03:20, David Miller a écrit :

Quoted 7 lines

From: Nicolas Dichtel <[email protected]>
Date: Mon, 30 Sep 2019 18:02:12 +0200

Quoted 3 lines

The goal of the series is to ease nsid allocation from userland. The first patch is a preparation work and the second enables to receive the new nsid in the answer to RTM_NEWNSID.

Quoted 6 lines

The new reply message could break existing apps.

If an app only performs netnsid operations, and fills up the receive queue because it isn't reading these new replies (it had no reason to, they didn't exist previously), operations will start failing that would not fail previously because the receive queue is full.

Yes I see the problem. I was wondering if this was acceptable because the nl ack is sent at the end. But nl ack are optional :/

Quoted 3 lines

Given this, I don't see how we can make the change.

Is a new flag attribute ok to turn on this reply?

Thank you,
Nicolas

maintainer
#5 lore raw

From: Nicolas Dichtel <[email protected]>
Date: Wed, 2 Oct 2019 10:46:03 +0200

Quoted 1 line

Is a new flag attribute ok to turn on this reply?

That might work, yes.

#6 lore raw
In reply to Nicolas Dichtel
Quoted 14 lines · Nicolas Dichtel

Le 02/10/2019 à 03:20, David Miller a écrit :

Quoted 7 lines

From: Nicolas Dichtel <[email protected]>
Date: Mon, 30 Sep 2019 18:02:12 +0200

Quoted 3 lines

The goal of the series is to ease nsid allocation from userland. The first patch is a preparation work and the second enables to receive the new nsid in the answer to RTM_NEWNSID.

Quoted 6 lines

The new reply message could break existing apps.

If an app only performs netnsid operations, and fills up the receive queue because it isn't reading these new replies (it had no reason to, they didn't exist previously), operations will start failing that would not fail previously because the receive queue is full.

Quoted 8 lines

Yes I see the problem. I was wondering if this was acceptable because the nl ack is sent at the end. But nl ack are optional :/

Quoted 3 lines

Given this, I don't see how we can make the change.

Is a new flag attribute ok to turn on this reply?

Why not using the existing NLM_F_ECHO mechanism?

IIUC, if rtnl_net_notifyid() did pass the proper nlmsghdr and portid to rtnl_notify(), the later would automatically notify the caller with updated information if the original request had the NLM_F_ECHO flag.

#7 lore raw

Le 03/10/2019 à 18:19, Guillaume Nault a écrit :
[snip]

Quoted 6 lines

Why not using the existing NLM_F_ECHO mechanism?

IIUC, if rtnl_net_notifyid() did pass the proper nlmsghdr and portid to rtnl_notify(), the later would automatically notify the caller with updated information if the original request had the NLM_F_ECHO flag.

Good point. Note that with library like libnl, the auto sequence number check will fail (seq number is 0 instead of the previous + 1) and thus must be bypassed.

#8 lore raw
Quoted 11 lines · Nicolas Dichtel

Le 03/10/2019 à 18:19, Guillaume Nault a écrit :
[snip]

Quoted 6 lines

Why not using the existing NLM_F_ECHO mechanism?

IIUC, if rtnl_net_notifyid() did pass the proper nlmsghdr and portid to rtnl_notify(), the later would automatically notify the caller with updated information if the original request had the NLM_F_ECHO flag.

Good point. Note that with library like libnl, the auto sequence number check will fail (seq number is 0 instead of the previous + 1) and thus must be bypassed.

Well, it's up to the caller of rtnl_notify() to build a netlink message with proper sequence number. Currently, rtnl_net_notifyid() always sets it to 0. That's probably why the libnl test fails.

#9 lore raw

PATCH 1/1 netns: fix NLM_F_ECHO mechanism for RTM_NEWNSID

In reply to Guillaume Nault

The flag NLM_F_ECHO aims to reply to the user the message notified to all listeners.
It was not the case with the command RTM_NEWNSID, let's fix this.

Fixes 0c7aecd4bde4 ("netns: add rtnl cmd to add and get peer netns ids")
---
 net/core/net_namespace.c | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)
net/core/net_namespace.c +9 -6
index a0e0d298c991..f496ce0e8da8 100644
@@ -245,7 +245,8 @@ static int __peernet2id(struct net *net, struct net *peer)
 	return __peernet2id_alloc(net, peer, &no);
 }
 
-static void rtnl_net_notifyid(struct net *net, int cmd, int id);
+static void rtnl_net_notifyid(struct net *net, int cmd, int id, u32 portid,
+			      struct nlmsghdr *nlh);
 /* This function returns the id of a peer netns. If no id is assigned, one will
  * be allocated and returned.
  */
@@ -268,7 +269,7 @@ int peernet2id_alloc(struct net *net, struct net *peer)
 	id = __peernet2id_alloc(net, peer, &alloc);
 	spin_unlock_bh(&net->nsid_lock);
 	if (alloc && id >= 0)
-		rtnl_net_notifyid(net, RTM_NEWNSID, id);
+		rtnl_net_notifyid(net, RTM_NEWNSID, id, 0, NULL);
 	if (alive)
 		put_net(peer);
 	return id;
@@ -532,7 +533,7 @@ static void unhash_nsid(struct net *net, struct net *last)
 			idr_remove(&tmp->netns_ids, id);
 		spin_unlock_bh(&tmp->nsid_lock);
 		if (id >= 0)
-			rtnl_net_notifyid(tmp, RTM_DELNSID, id);
+			rtnl_net_notifyid(tmp, RTM_DELNSID, id, 0, NULL);
 		if (tmp == last)
 			break;
 	}
@@ -764,7 +765,8 @@ static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh,
 	err = alloc_netid(net, peer, nsid);
 	spin_unlock_bh(&net->nsid_lock);
 	if (err >= 0) {
-		rtnl_net_notifyid(net, RTM_NEWNSID, err);
+		rtnl_net_notifyid(net, RTM_NEWNSID, err, NETLINK_CB(skb).portid,
+				  nlh);
 		err = 0;
 	} else if (err == -ENOSPC && nsid >= 0) {
 		err = -EEXIST;
@@ -1051,7 +1053,8 @@ static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb)
 	return err < 0 ? err : skb->len;
 }
 
-static void rtnl_net_notifyid(struct net *net, int cmd, int id)
+static void rtnl_net_notifyid(struct net *net, int cmd, int id, u32 portid,
+			      struct nlmsghdr *nlh)
 {
 	struct net_fill_args fillargs = {
 		.cmd = cmd,
@@ -1068,7 +1071,7 @@ static void rtnl_net_notifyid(struct net *net, int cmd, int id)
 	if (err < 0)
 		goto err_out;
 
-	rtnl_notify(msg, net, 0, RTNLGRP_NSID, NULL, 0);
+	rtnl_notify(msg, net, portid, RTNLGRP_NSID, nlh, 0);
 	return;
 
 err_out:
Signature · 1 line
2.23.0
#10 lore raw

Re: [PATCH net] netns: fix NLM_F_ECHO mechanism for RTM_NEWNSID

Quoted 15 lines · Nicolas Dichtel

The flag NLM_F_ECHO aims to reply to the user the message notified to all listeners.
It was not the case with the command RTM_NEWNSID, let's fix this.

Fixes 0c7aecd4bde4 ("netns: add rtnl cmd to add and get peer netns ids")
---
 net/core/net_namespace.c | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)
net/core/net_namespace.c +0 -0
index a0e0d298c991..f496ce0e8da8 100644

net/core/net_namespace.c static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb) · line 1060

Quoted 6 lines
net/core/net_namespace.c +2 -1
-static void rtnl_net_notifyid(struct net *net, int cmd, int id)
+static void rtnl_net_notifyid(struct net *net, int cmd, int id, u32 portid,
+			      struct nlmsghdr *nlh)
 {
 	struct net_fill_args fillargs = {
 		.cmd = cmd,

We also need to set .portid and .seq otherwise rtnl_net_fill() builds a netlink message with invalid port id and sequence number (as you noted in your previous message).

#11 lore raw

Re: [PATCH net] netns: fix NLM_F_ECHO mechanism for RTM_NEWNSID

Le 09/10/2019 à 01:10, Guillaume Nault a écrit :
[snip]

Quoted 4 lines

We also need to set .portid and .seq otherwise rtnl_net_fill() builds a netlink message with invalid port id and sequence number (as you noted in your previous message).

Yes you're right. I don't know why, I had in mind that nl msg sent by the kernel should have the portid and seq number set to 0.
Will send a v2.

Thank you,
Nicolas

#12 lore raw

Re: [PATCH net] netns: fix NLM_F_ECHO mechanism for RTM_NEWNSID

Quoted 10 lines · Nicolas Dichtel

Le 09/10/2019 à 01:10, Guillaume Nault a écrit :
[snip]

Quoted 4 lines

We also need to set .portid and .seq otherwise rtnl_net_fill() builds a netlink message with invalid port id and sequence number (as you noted in your previous message).

Yes you're right. I don't know why, I had in mind that nl msg sent by the kernel should have the portid and seq number set to 0.
Will send a v2.

I guess this idea comes from the fact that portid and seq don't carry any meaningful information when the message is sent to a multicast group.

#13 lore raw

PATCH v2 1/1 netns: fix NLM_F_ECHO mechanism for RTM_NEWNSID

In reply to Guillaume Nault

The flag NLM_F_ECHO aims to reply to the user the message notified to all listeners.
It was not the case with the command RTM_NEWNSID, let's fix this.

Fixes 0c7aecd4bde4 ("netns: add rtnl cmd to add and get peer netns ids")

---

v2:
fix portid and seq number of the nl msg sent by the kernel

 net/core/net_namespace.c | 17 +++++++++++------
 1 file changed, 11 insertions(+), 6 deletions(-)
net/core/net_namespace.c +11 -6
index a0e0d298c991..6d3e4821b02d 100644
@@ -245,7 +245,8 @@ static int __peernet2id(struct net *net, struct net *peer)
 	return __peernet2id_alloc(net, peer, &no);
 }
 
-static void rtnl_net_notifyid(struct net *net, int cmd, int id);
+static void rtnl_net_notifyid(struct net *net, int cmd, int id, u32 portid,
+			      struct nlmsghdr *nlh);
 /* This function returns the id of a peer netns. If no id is assigned, one will
  * be allocated and returned.
  */
@@ -268,7 +269,7 @@ int peernet2id_alloc(struct net *net, struct net *peer)
 	id = __peernet2id_alloc(net, peer, &alloc);
 	spin_unlock_bh(&net->nsid_lock);
 	if (alloc && id >= 0)
-		rtnl_net_notifyid(net, RTM_NEWNSID, id);
+		rtnl_net_notifyid(net, RTM_NEWNSID, id, 0, NULL);
 	if (alive)
 		put_net(peer);
 	return id;
@@ -532,7 +533,7 @@ static void unhash_nsid(struct net *net, struct net *last)
 			idr_remove(&tmp->netns_ids, id);
 		spin_unlock_bh(&tmp->nsid_lock);
 		if (id >= 0)
-			rtnl_net_notifyid(tmp, RTM_DELNSID, id);
+			rtnl_net_notifyid(tmp, RTM_DELNSID, id, 0, NULL);
 		if (tmp == last)
 			break;
 	}
@@ -764,7 +765,8 @@ static int rtnl_net_newid(struct sk_buff *skb, struct nlmsghdr *nlh,
 	err = alloc_netid(net, peer, nsid);
 	spin_unlock_bh(&net->nsid_lock);
 	if (err >= 0) {
-		rtnl_net_notifyid(net, RTM_NEWNSID, err);
+		rtnl_net_notifyid(net, RTM_NEWNSID, err, NETLINK_CB(skb).portid,
+				  nlh);
 		err = 0;
 	} else if (err == -ENOSPC && nsid >= 0) {
 		err = -EEXIST;
@@ -1051,9 +1053,12 @@ static int rtnl_net_dumpid(struct sk_buff *skb, struct netlink_callback *cb)
 	return err < 0 ? err : skb->len;
 }
 
-static void rtnl_net_notifyid(struct net *net, int cmd, int id)
+static void rtnl_net_notifyid(struct net *net, int cmd, int id, u32 portid,
+			      struct nlmsghdr *nlh)
 {
 	struct net_fill_args fillargs = {
+		.portid = portid,
+		.seq = nlh ? nlh->nlmsg_seq : 0,
 		.cmd = cmd,
 		.nsid = id,
 	};
@@ -1068,7 +1073,7 @@ static void rtnl_net_notifyid(struct net *net, int cmd, int id)
 	if (err < 0)
 		goto err_out;
 
-	rtnl_notify(msg, net, 0, RTNLGRP_NSID, NULL, 0);
+	rtnl_notify(msg, net, portid, RTNLGRP_NSID, nlh, 0);
 	return;
 
 err_out:
Signature · 1 line
2.23.0
#15 lore raw

Re: [PATCH net v2] netns: fix NLM_F_ECHO mechanism for RTM_NEWNSID

In reply to Nicolas Dichtel
Quoted 7 lines · Nicolas Dichtel

The flag NLM_F_ECHO aims to reply to the user the message notified to all listeners.
It was not the case with the command RTM_NEWNSID, let's fix this.

Fixes 0c7aecd4bde4 ("netns: add rtnl cmd to add and get peer netns ids")

Applied, thanks!