[Patch net-next v2 0/5] fou: some fixes and updates

[Patch net-next v2 0/5] fou: some fixes and updates

[Cong Wang]

Patch 1~3 fix some minor bugs in net/ipv4/fou.c, the only
thing I am not sure is if it's too late to change the
byte order of FOU_ATTR_PORT, if so we have to fix iproute2
instead of kernel.

Patch 4~5 add some new features to make it complete.

v2: make fou->port be16 too

Cong Wang (5):
  fou: avoid calling udp_del_offload() twice
  fou: exit early when parsing config fails
  fou: always use be16 for port
  fou: add network namespace support
  fou: implement FOU_CMD_GET

 include/uapi/linux/fou.h |   1 +
 net/ipv4/fou.c           | 230 +++++++++++++++++++++++++++++++++++++----------
 2 files changed, 186 insertions(+), 45 deletions(-)

-- 
1.8.3.1

[Patch net-next v2 1/5] fou: avoid calling udp_del_offload() twice

[Cong Wang]

This fixes the following harmless warning:

./ip/ip fou del port 7777
[  122.907516] udp_del_offload: didn't find offload for port 7777

Cc: Tom Herbert <tom@herbertland.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
---
 net/ipv4/fou.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/net/ipv4/fou.c b/net/ipv4/fou.c
index ff069f6..c8db627 100644
--- a/net/ipv4/fou.c
+++ b/net/ipv4/fou.c
@@ -410,7 +410,8 @@ static void fou_release(struct fou *fou)
 	struct socket *sock = fou->sock;
 	struct sock *sk = sock->sk;
 
-	udp_del_offload(&fou->udp_offloads);
+	if (sk->sk_family == AF_INET)
+		udp_del_offload(&fou->udp_offloads);
 
 	list_del(&fou->list);
 
@@ -528,7 +529,6 @@ static int fou_destroy(struct net *net, struct fou_cfg *cfg)
 	spin_lock(&fou_lock);
 	list_for_each_entry(fou, &fou_list, list) {
 		if (fou->port == port) {
-			udp_del_offload(&fou->udp_offloads);
 			fou_release(fou);
 			err = 0;
 			break;
-- 
1.8.3.1

[Patch net-next v2 2/5] fou: exit early when parsing config fails

[Cong Wang]

Not a big deal, just for corretness.

Cc: Tom Herbert <tom@herbertland.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
---
 net/ipv4/fou.c | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/net/ipv4/fou.c b/net/ipv4/fou.c
index c8db627..ad0ee82 100644
--- a/net/ipv4/fou.c
+++ b/net/ipv4/fou.c
@@ -605,8 +605,11 @@ static int fou_nl_cmd_add_port(struct sk_buff *skb, struct genl_info *info)
 static int fou_nl_cmd_rm_port(struct sk_buff *skb, struct genl_info *info)
 {
 	struct fou_cfg cfg;
+	int err;
 
-	parse_nl_config(info, &cfg);
+	err = parse_nl_config(info, &cfg);
+	if (err)
+		return err;
 
 	return fou_destroy(&init_net, &cfg);
 }
-- 
1.8.3.1

[Patch net-next v2 3/5] fou: always use be16 for port

[Cong Wang]

udp_config.local_udp_port is be16. And iproute2 passes
network order for FOU_ATTR_PORT.

This doesn't fix any bug, just for consistency.

Cc: Tom Herbert <tom@herbertland.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
---
 net/ipv4/fou.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/net/ipv4/fou.c b/net/ipv4/fou.c
index ad0ee82..d61f6f9 100644
--- a/net/ipv4/fou.c
+++ b/net/ipv4/fou.c
@@ -23,7 +23,7 @@ struct fou {
 	struct socket *sock;
 	u8 protocol;
 	u8 flags;
-	u16 port;
+	__be16 port;
 	struct udp_offload udp_offloads;
 	struct list_head list;
 };
@@ -523,7 +523,7 @@ static int fou_create(struct net *net, struct fou_cfg *cfg,
 static int fou_destroy(struct net *net, struct fou_cfg *cfg)
 {
 	struct fou *fou;
-	u16 port = cfg->udp_config.local_udp_port;
+	__be16 port = cfg->udp_config.local_udp_port;
 	int err = -EINVAL;
 
 	spin_lock(&fou_lock);
@@ -573,7 +573,7 @@ static int parse_nl_config(struct genl_info *info,
 	}
 
 	if (info->attrs[FOU_ATTR_PORT]) {
-		u16 port = nla_get_u16(info->attrs[FOU_ATTR_PORT]);
+		__be16 port = nla_get_be16(info->attrs[FOU_ATTR_PORT]);
 
 		cfg->udp_config.local_udp_port = port;
 	}
-- 
1.8.3.1

[Patch net-next v2 4/5] fou: add network namespace support

[Cong Wang]

Also convert the spinlock to a mutex.

Cc: Tom Herbert <tom@herbertland.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
---
 net/ipv4/fou.c | 106 ++++++++++++++++++++++++++++++++++++---------------------
 1 file changed, 67 insertions(+), 39 deletions(-)

diff --git a/net/ipv4/fou.c b/net/ipv4/fou.c
index d61f6f9..c244b1a 100644
--- a/net/ipv4/fou.c
+++ b/net/ipv4/fou.c
@@ -16,9 +16,6 @@
 #include <uapi/linux/fou.h>
 #include <uapi/linux/genetlink.h>
 
-static DEFINE_SPINLOCK(fou_lock);
-static LIST_HEAD(fou_list);
-
 struct fou {
 	struct socket *sock;
 	u8 protocol;
@@ -37,6 +34,13 @@ struct fou_cfg {
 	struct udp_port_cfg udp_config;
 };
 
+static unsigned int fou_net_id;
+
+struct fou_net {
+	struct list_head fou_list;
+	struct mutex fou_lock;
+};
+
 static inline struct fou *fou_from_sock(struct sock *sk)
 {
 	return sk->sk_user_data;
@@ -387,20 +391,21 @@ static int gue_gro_complete(struct sk_buff *skb, int nhoff,
 	return err;
 }
 
-static int fou_add_to_port_list(struct fou *fou)
+static int fou_add_to_port_list(struct net *net, struct fou *fou)
 {
+	struct fou_net *fn = net_generic(net, fou_net_id);
 	struct fou *fout;
 
-	spin_lock(&fou_lock);
-	list_for_each_entry(fout, &fou_list, list) {
+	mutex_lock(&fn->fou_lock);
+	list_for_each_entry(fout, &fn->fou_list, list) {
 		if (fou->port == fout->port) {
-			spin_unlock(&fou_lock);
+			mutex_unlock(&fn->fou_lock);
 			return -EALREADY;
 		}
 	}
 
-	list_add(&fou->list, &fou_list);
-	spin_unlock(&fou_lock);
+	list_add(&fou->list, &fn->fou_list);
+	mutex_unlock(&fn->fou_lock);
 
 	return 0;
 }
@@ -412,13 +417,8 @@ static void fou_release(struct fou *fou)
 
 	if (sk->sk_family == AF_INET)
 		udp_del_offload(&fou->udp_offloads);
-
 	list_del(&fou->list);
-
-	/* Remove hooks into tunnel socket */
-	sk->sk_user_data = NULL;
-
-	sock_release(sock);
+	udp_tunnel_sock_release(sock);
 
 	kfree(fou);
 }
@@ -448,10 +448,10 @@ static int gue_encap_init(struct sock *sk, struct fou *fou, struct fou_cfg *cfg)
 static int fou_create(struct net *net, struct fou_cfg *cfg,
 		      struct socket **sockp)
 {
-	struct fou *fou = NULL;
-	int err;
 	struct socket *sock = NULL;
+	struct fou *fou = NULL;
 	struct sock *sk;
+	int err;
 
 	/* Open UDP socket */
 	err = udp_sock_create(net, &cfg->udp_config, &sock);
@@ -503,7 +503,7 @@ static int fou_create(struct net *net, struct fou_cfg *cfg,
 			goto error;
 	}
 
-	err = fou_add_to_port_list(fou);
+	err = fou_add_to_port_list(net, fou);
 	if (err)
 		goto error;
 
@@ -515,26 +515,27 @@ static int fou_create(struct net *net, struct fou_cfg *cfg,
 error:
 	kfree(fou);
 	if (sock)
-		sock_release(sock);
+		udp_tunnel_sock_release(sock);
 
 	return err;
 }
 
 static int fou_destroy(struct net *net, struct fou_cfg *cfg)
 {
-	struct fou *fou;
+	struct fou_net *fn = net_generic(net, fou_net_id);
 	__be16 port = cfg->udp_config.local_udp_port;
 	int err = -EINVAL;
+	struct fou *fou;
 
-	spin_lock(&fou_lock);
-	list_for_each_entry(fou, &fou_list, list) {
+	mutex_lock(&fn->fou_lock);
+	list_for_each_entry(fou, &fn->fou_list, list) {
 		if (fou->port == port) {
 			fou_release(fou);
 			err = 0;
 			break;
 		}
 	}
-	spin_unlock(&fou_lock);
+	mutex_unlock(&fn->fou_lock);
 
 	return err;
 }
@@ -592,6 +593,7 @@ static int parse_nl_config(struct genl_info *info,
 
 static int fou_nl_cmd_add_port(struct sk_buff *skb, struct genl_info *info)
 {
+	struct net *net = genl_info_net(info);
 	struct fou_cfg cfg;
 	int err;
 
@@ -599,11 +601,12 @@ static int fou_nl_cmd_add_port(struct sk_buff *skb, struct genl_info *info)
 	if (err)
 		return err;
 
-	return fou_create(&init_net, &cfg, NULL);
+	return fou_create(net, &cfg, NULL);
 }
 
 static int fou_nl_cmd_rm_port(struct sk_buff *skb, struct genl_info *info)
 {
+	struct net *net = genl_info_net(info);
 	struct fou_cfg cfg;
 	int err;
 
@@ -611,7 +614,7 @@ static int fou_nl_cmd_rm_port(struct sk_buff *skb, struct genl_info *info)
 	if (err)
 		return err;
 
-	return fou_destroy(&init_net, &cfg);
+	return fou_destroy(net, &cfg);
 }
 
 static const struct genl_ops fou_nl_ops[] = {
@@ -823,38 +826,63 @@ static void ip_tunnel_encap_del_fou_ops(void)
 
 #endif
 
+static __net_init int fou_init_net(struct net *net)
+{
+	struct fou_net *fn = net_generic(net, fou_net_id);
+
+	INIT_LIST_HEAD(&fn->fou_list);
+	mutex_init(&fn->fou_lock);
+	return 0;
+}
+
+static __net_exit void fou_exit_net(struct net *net)
+{
+	struct fou_net *fn = net_generic(net, fou_net_id);
+	struct fou *fou, *next;
+
+	/* Close all the FOU sockets */
+	mutex_lock(&fn->fou_lock);
+	list_for_each_entry_safe(fou, next, &fn->fou_list, list)
+		fou_release(fou);
+	mutex_unlock(&fn->fou_lock);
+}
+
+static struct pernet_operations fou_net_ops = {
+	.init = fou_init_net,
+	.exit = fou_exit_net,
+	.id   = &fou_net_id,
+	.size = sizeof(struct fou_net),
+};
+
 static int __init fou_init(void)
 {
 	int ret;
 
+	ret = register_pernet_device(&fou_net_ops);
+	if (ret)
+		goto exit;
+
 	ret = genl_register_family_with_ops(&fou_nl_family,
 					    fou_nl_ops);
-
 	if (ret < 0)
-		goto exit;
+		goto unregister;
 
 	ret = ip_tunnel_encap_add_fou_ops();
-	if (ret < 0)
-		genl_unregister_family(&fou_nl_family);
+	if (ret == 0)
+		return 0;
 
+	genl_unregister_family(&fou_nl_family);
+unregister:
+	unregister_pernet_device(&fou_net_ops);
 exit:
 	return ret;
 }
 
 static void __exit fou_fini(void)
 {
-	struct fou *fou, *next;
-
 	ip_tunnel_encap_del_fou_ops();
-
 	genl_unregister_family(&fou_nl_family);
-
-	/* Close all the FOU sockets */
-
-	spin_lock(&fou_lock);
-	list_for_each_entry_safe(fou, next, &fou_list, list)
-		fou_release(fou);
-	spin_unlock(&fou_lock);
+	unregister_pernet_device(&fou_net_ops);
 }
 
 module_init(fou_init);
-- 
1.8.3.1

[Patch net-next v2 5/5] fou: implement FOU_CMD_GET

[Cong Wang]

Cc: Tom Herbert <tom@herbertland.com>
Signed-off-by: Cong Wang <xiyou.wangcong@gmail.com>
---
 include/uapi/linux/fou.h |   1 +
 net/ipv4/fou.c           | 109 +++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 110 insertions(+)

diff --git a/include/uapi/linux/fou.h b/include/uapi/linux/fou.h
index c303588..d2947c5 100644
--- a/include/uapi/linux/fou.h
+++ b/include/uapi/linux/fou.h
@@ -25,6 +25,7 @@ enum {
 	FOU_CMD_UNSPEC,
 	FOU_CMD_ADD,
 	FOU_CMD_DEL,
+	FOU_CMD_GET,
 
 	__FOU_CMD_MAX,
 };
diff --git a/net/ipv4/fou.c b/net/ipv4/fou.c
index c244b1a..2637102 100644
--- a/net/ipv4/fou.c
+++ b/net/ipv4/fou.c
@@ -21,6 +21,7 @@ struct fou {
 	u8 protocol;
 	u8 flags;
 	__be16 port;
+	u16 type;
 	struct udp_offload udp_offloads;
 	struct list_head list;
 };
@@ -487,6 +488,8 @@ static int fou_create(struct net *net, struct fou_cfg *cfg,
 		goto error;
 	}
 
+	fou->type = cfg->type;
+
 	udp_sk(sk)->encap_type = 1;
 	udp_encap_enable();
 
@@ -617,6 +620,106 @@ static int fou_nl_cmd_rm_port(struct sk_buff *skb, struct genl_info *info)
 	return fou_destroy(net, &cfg);
 }
 
+static int fou_fill_info(struct fou *fou, struct sk_buff *msg)
+{
+	if (nla_put_u8(msg, FOU_ATTR_AF, fou->sock->sk->sk_family) ||
+	    nla_put_be16(msg, FOU_ATTR_PORT, fou->port) ||
+	    nla_put_u8(msg, FOU_ATTR_IPPROTO, fou->protocol) ||
+	    nla_put_u8(msg, FOU_ATTR_TYPE, fou->type))
+		return -1;
+
+	if (fou->flags & FOU_F_REMCSUM_NOPARTIAL)
+		if (nla_put_flag(msg, FOU_ATTR_REMCSUM_NOPARTIAL))
+			return -1;
+	return 0;
+}
+
+static int fou_dump_info(struct fou *fou, u32 portid, u32 seq,
+			 u32 flags, struct sk_buff *skb, u8 cmd)
+{
+	void *hdr;
+
+	hdr = genlmsg_put(skb, portid, seq, &fou_nl_family, flags, cmd);
+	if (!hdr)
+		return -ENOMEM;
+
+	if (fou_fill_info(fou, skb) < 0)
+		goto nla_put_failure;
+
+	genlmsg_end(skb, hdr);
+	return 0;
+
+nla_put_failure:
+	genlmsg_cancel(skb, hdr);
+	return -EMSGSIZE;
+}
+
+static int fou_nl_cmd_get_port(struct sk_buff *skb, struct genl_info *info)
+{
+	struct net *net = genl_info_net(info);
+	struct fou_net *fn = net_generic(net, fou_net_id);
+	struct sk_buff *msg;
+	struct fou_cfg cfg;
+	struct fou *fout;
+	__be16 port;
+	int ret;
+
+	ret = parse_nl_config(info, &cfg);
+	if (ret)
+		return ret;
+	port = cfg.udp_config.local_udp_port;
+	if (port == 0)
+		return -EINVAL;
+
+	msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
+	if (!msg)
+		return -ENOMEM;
+
+	ret = -ESRCH;
+	mutex_lock(&fn->fou_lock);
+	list_for_each_entry(fout, &fn->fou_list, list) {
+		if (port == fout->port) {
+			ret = fou_dump_info(fout, info->snd_portid,
+					    info->snd_seq, 0, msg,
+					    info->genlhdr->cmd);
+			break;
+		}
+	}
+	mutex_unlock(&fn->fou_lock);
+	if (ret < 0)
+		goto out_free;
+
+	return genlmsg_reply(msg, info);
+
+out_free:
+	nlmsg_free(msg);
+	return ret;
+}
+
+static int fou_nl_dump(struct sk_buff *skb, struct netlink_callback *cb)
+{
+	struct net *net = sock_net(skb->sk);
+	struct fou_net *fn = net_generic(net, fou_net_id);
+	struct fou *fout;
+	int idx = 0, ret;
+
+	mutex_lock(&fn->fou_lock);
+	list_for_each_entry(fout, &fn->fou_list, list) {
+		if (idx++ < cb->args[0])
+			continue;
+		ret = fou_dump_info(fout, NETLINK_CB(cb->skb).portid,
+				    cb->nlh->nlmsg_seq, NLM_F_MULTI,
+				    skb, FOU_CMD_GET);
+		if (ret)
+			goto done;
+	}
+	mutex_unlock(&fn->fou_lock);
+
+done:
+	cb->args[0] = idx;
+	return skb->len;
+}
+
 static const struct genl_ops fou_nl_ops[] = {
 	{
 		.cmd = FOU_CMD_ADD,
@@ -630,6 +733,12 @@ static const struct genl_ops fou_nl_ops[] = {
 		.policy = fou_nl_policy,
 		.flags = GENL_ADMIN_PERM,
 	},
+	{
+		.cmd = FOU_CMD_GET,
+		.doit = fou_nl_cmd_get_port,
+		.dumpit = fou_nl_dump,
+		.policy = fou_nl_policy,
+	},
 };
 
 size_t fou_encap_hlen(struct ip_tunnel_encap *e)
-- 
1.8.3.1

Re: [Patch net-next v2 0/5] fou: some fixes and updates

[Tom Herbert]

On Fri, Apr 10, 2015 at 12:00 PM, Cong Wang <xiyou.wangcong@gmail.com> wrote:
> Patch 1~3 fix some minor bugs in net/ipv4/fou.c, the only
> thing I am not sure is if it's too late to change the
> byte order of FOU_ATTR_PORT, if so we have to fix iproute2
> instead of kernel.
>
Is this a relic from previous patch set, or is there a reason you want
to make this hbo?

> Patch 4~5 add some new features to make it complete.
>
> v2: make fou->port be16 too
>
> Cong Wang (5):
>   fou: avoid calling udp_del_offload() twice
>   fou: exit early when parsing config fails
>   fou: always use be16 for port
>   fou: add network namespace support
>   fou: implement FOU_CMD_GET
>

I assume you didn't see any adverse performance impact with these patches?

Acked-by: Tom Herbert <tom@herbertland.com>

Thanks!
Tom

>  include/uapi/linux/fou.h |   1 +
>  net/ipv4/fou.c           | 230 +++++++++++++++++++++++++++++++++++++----------
>  2 files changed, 186 insertions(+), 45 deletions(-)
>
> --
> 1.8.3.1
>

Re: [Patch net-next v2 0/5] fou: some fixes and updates

[Cong Wang]

On Fri, Apr 10, 2015 at 1:38 PM, Tom Herbert <tom@herbertland.com> wrote:
> On Fri, Apr 10, 2015 at 12:00 PM, Cong Wang <xiyou.wangcong@gmail.com> wrote:
>> Patch 1~3 fix some minor bugs in net/ipv4/fou.c, the only
>> thing I am not sure is if it's too late to change the
>> byte order of FOU_ATTR_PORT, if so we have to fix iproute2
>> instead of kernel.
>>
> Is this a relic from previous patch set, or is there a reason you want
> to make this hbo?

I was referring to this change:

-               u16 port = nla_get_u16(info->attrs[FOU_ATTR_PORT]);
+               __be16 port = nla_get_be16(info->attrs[FOU_ATTR_PORT]);

I meant I am not sure if this breaks ABI here even in v2, we
just use the raw value passed from user-space, so probably
we are safe.

>
>> Patch 4~5 add some new features to make it complete.
>>
>> v2: make fou->port be16 too
>>
>> Cong Wang (5):
>>   fou: avoid calling udp_del_offload() twice
>>   fou: exit early when parsing config fails
>>   fou: always use be16 for port
>>   fou: add network namespace support
>>   fou: implement FOU_CMD_GET
>>
>
> I assume you didn't see any adverse performance impact with these patches?
>

I don't think I touch any hot path, foo_create() and foo_destroy() are
both slow path.

Thanks.

Re: [Patch net-next v2 0/5] fou: some fixes and updates

[David Miller]

From: Cong Wang <cwang@twopensource.com>
Date: Fri, 10 Apr 2015 13:50:55 -0700

> On Fri, Apr 10, 2015 at 1:38 PM, Tom Herbert <tom@herbertland.com> wrote:
>> On Fri, Apr 10, 2015 at 12:00 PM, Cong Wang <xiyou.wangcong@gmail.com> wrote:
>>> Patch 1~3 fix some minor bugs in net/ipv4/fou.c, the only
>>> thing I am not sure is if it's too late to change the
>>> byte order of FOU_ATTR_PORT, if so we have to fix iproute2
>>> instead of kernel.
>>>
>> Is this a relic from previous patch set, or is there a reason you want
>> to make this hbo?
> 
> I was referring to this change:
> 
> -               u16 port = nla_get_u16(info->attrs[FOU_ATTR_PORT]);
> +               __be16 port = nla_get_be16(info->attrs[FOU_ATTR_PORT]);
> 
> I meant I am not sure if this breaks ABI here even in v2, we
> just use the raw value passed from user-space, so probably
> we are safe.

This is fine, there is no byte swapping done by nla_get_be16() so the
bits used are exactly the same.