[PATCH] wireless extensions: play with netns

[PATCH] wireless extensions: play with netns

[Johannes Berg]

This makes wireless extensions netns aware.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
---
Is this ok, or is this racy? I guess what I'm asking is -- will
for_each_net() stop iterating over a netns that is going away before the
pernet exit op is called? If yes, this should be fine.

 include/net/net_namespace.h |    3 +++
 net/wireless/wext.c         |   30 ++++++++++++++++++++++--------
 2 files changed, 25 insertions(+), 8 deletions(-)

--- wireless-testing.orig/include/net/net_namespace.h	2009-06-17 20:20:47.000000000 +0200
+++ wireless-testing/include/net/net_namespace.h	2009-06-17 20:20:51.000000000 +0200
@@ -78,6 +78,9 @@ struct net {
 #ifdef CONFIG_XFRM
 	struct netns_xfrm	xfrm;
 #endif
+#ifdef CONFIG_WIRELESS_EXT
+	struct sk_buff_head	wext_nlevents;
+#endif
 	struct net_generic	*gen;
 };
 
--- wireless-testing.orig/net/wireless/wext.c	2009-06-17 20:20:47.000000000 +0200
+++ wireless-testing/net/wireless/wext.c	2009-06-17 20:20:51.000000000 +0200
@@ -1273,11 +1273,25 @@ int compat_wext_handle_ioctl(struct net 
  * Jean II
  */
 
-static struct sk_buff_head wireless_nlevent_queue;
+static int __net_init wext_pernet_init(struct net *net)
+{
+	skb_queue_head_init(&net->wext_nlevents);
+	return 0;
+}
+
+static void __net_exit wext_pernet_exit(struct net *net)
+{
+	skb_queue_purge(&net->wext_nlevents);
+}
+
+static struct pernet_operations wext_pernet_ops = {
+	.init = wext_pernet_init,
+	.exit = wext_pernet_exit,
+};
 
 static int __init wireless_nlevent_init(void)
 {
-	skb_queue_head_init(&wireless_nlevent_queue);
+	return register_pernet_subsys(&wext_pernet_ops);
 	return 0;
 }
 
@@ -1286,9 +1300,12 @@ subsys_initcall(wireless_nlevent_init);
 static void wireless_nlevent_process(unsigned long data)
 {
 	struct sk_buff *skb;
+	struct net *net;
 
-	while ((skb = skb_dequeue(&wireless_nlevent_queue)))
-		rtnl_notify(skb, &init_net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
+	for_each_net(net)
+		while ((skb = skb_dequeue(&net->wext_nlevents)))
+			rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL,
+				    GFP_ATOMIC);
 }
 
 static DECLARE_TASKLET(wireless_nlevent_tasklet, wireless_nlevent_process, 0);
@@ -1341,9 +1358,6 @@ static void rtmsg_iwinfo(struct net_devi
 	struct sk_buff *skb;
 	int err;
 
-	if (!net_eq(dev_net(dev), &init_net))
-		return;
-
 	skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
 	if (!skb)
 		return;
@@ -1356,7 +1370,7 @@ static void rtmsg_iwinfo(struct net_devi
 	}
 
 	NETLINK_CB(skb).dst_group = RTNLGRP_LINK;
-	skb_queue_tail(&wireless_nlevent_queue, skb);
+	skb_queue_tail(&dev_net(dev)->wext_nlevents, skb);
 	tasklet_schedule(&wireless_nlevent_tasklet);
 }
 

[PATCH] wireless extensions: play with netns

[Johannes Berg]

This makes wireless extensions netns aware.

Signed-off-by: Johannes Berg <johannes-cdvu00un1VgdHxzADdlk8Q@public.gmane.org>
---
Is this ok, or is this racy? I guess what I'm asking is -- will
for_each_net() stop iterating over a netns that is going away before the
pernet exit op is called? If yes, this should be fine.

 include/net/net_namespace.h |    3 +++
 net/wireless/wext.c         |   30 ++++++++++++++++++++++--------
 2 files changed, 25 insertions(+), 8 deletions(-)

--- wireless-testing.orig/include/net/net_namespace.h	2009-06-17 20:20:47.000000000 +0200
+++ wireless-testing/include/net/net_namespace.h	2009-06-17 20:20:51.000000000 +0200
@@ -78,6 +78,9 @@ struct net {
 #ifdef CONFIG_XFRM
 	struct netns_xfrm	xfrm;
 #endif
+#ifdef CONFIG_WIRELESS_EXT
+	struct sk_buff_head	wext_nlevents;
+#endif
 	struct net_generic	*gen;
 };
 
--- wireless-testing.orig/net/wireless/wext.c	2009-06-17 20:20:47.000000000 +0200
+++ wireless-testing/net/wireless/wext.c	2009-06-17 20:20:51.000000000 +0200
@@ -1273,11 +1273,25 @@ int compat_wext_handle_ioctl(struct net 
  * Jean II
  */
 
-static struct sk_buff_head wireless_nlevent_queue;
+static int __net_init wext_pernet_init(struct net *net)
+{
+	skb_queue_head_init(&net->wext_nlevents);
+	return 0;
+}
+
+static void __net_exit wext_pernet_exit(struct net *net)
+{
+	skb_queue_purge(&net->wext_nlevents);
+}
+
+static struct pernet_operations wext_pernet_ops = {
+	.init = wext_pernet_init,
+	.exit = wext_pernet_exit,
+};
 
 static int __init wireless_nlevent_init(void)
 {
-	skb_queue_head_init(&wireless_nlevent_queue);
+	return register_pernet_subsys(&wext_pernet_ops);
 	return 0;
 }
 
@@ -1286,9 +1300,12 @@ subsys_initcall(wireless_nlevent_init);
 static void wireless_nlevent_process(unsigned long data)
 {
 	struct sk_buff *skb;
+	struct net *net;
 
-	while ((skb = skb_dequeue(&wireless_nlevent_queue)))
-		rtnl_notify(skb, &init_net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
+	for_each_net(net)
+		while ((skb = skb_dequeue(&net->wext_nlevents)))
+			rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL,
+				    GFP_ATOMIC);
 }
 
 static DECLARE_TASKLET(wireless_nlevent_tasklet, wireless_nlevent_process, 0);
@@ -1341,9 +1358,6 @@ static void rtmsg_iwinfo(struct net_devi
 	struct sk_buff *skb;
 	int err;
 
-	if (!net_eq(dev_net(dev), &init_net))
-		return;

Re: [PATCH] wireless extensions: play with netns

[Eric W. Biederman]

Johannes Berg <johannes@sipsolutions.net> writes:

> This makes wireless extensions netns aware.
>
> Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
> ---
> Is this ok, or is this racy? I guess what I'm asking is -- will
> for_each_net() stop iterating over a netns that is going away before the
> pernet exit op is called? If yes, this should be fine.

for_each_net requires the rtnl_lock or the net_mutex to be safe.
You aren't taking either so your code is racy.

A dying network namespace will be removed from the net_namespace_list
(aka for_each_net) before the per net exit methods are called.

Is grabbing the rtnl_lock safe in your workqueue and is that something
we want to do?

Eric


>  include/net/net_namespace.h |    3 +++
>  net/wireless/wext.c         |   30 ++++++++++++++++++++++--------
>  2 files changed, 25 insertions(+), 8 deletions(-)
>
> --- wireless-testing.orig/include/net/net_namespace.h	2009-06-17 20:20:47.000000000 +0200
> +++ wireless-testing/include/net/net_namespace.h	2009-06-17 20:20:51.000000000 +0200
> @@ -78,6 +78,9 @@ struct net {
>  #ifdef CONFIG_XFRM
>  	struct netns_xfrm	xfrm;
>  #endif
> +#ifdef CONFIG_WIRELESS_EXT
> +	struct sk_buff_head	wext_nlevents;
> +#endif
>  	struct net_generic	*gen;
>  };
>  
> --- wireless-testing.orig/net/wireless/wext.c	2009-06-17 20:20:47.000000000 +0200
> +++ wireless-testing/net/wireless/wext.c	2009-06-17 20:20:51.000000000 +0200
> @@ -1273,11 +1273,25 @@ int compat_wext_handle_ioctl(struct net 
>   * Jean II
>   */
>  
> -static struct sk_buff_head wireless_nlevent_queue;
> +static int __net_init wext_pernet_init(struct net *net)
> +{
> +	skb_queue_head_init(&net->wext_nlevents);
> +	return 0;
> +}
> +
> +static void __net_exit wext_pernet_exit(struct net *net)
> +{
> +	skb_queue_purge(&net->wext_nlevents);
> +}
> +
> +static struct pernet_operations wext_pernet_ops = {
> +	.init = wext_pernet_init,
> +	.exit = wext_pernet_exit,
> +};
>  
>  static int __init wireless_nlevent_init(void)
>  {
> -	skb_queue_head_init(&wireless_nlevent_queue);
> +	return register_pernet_subsys(&wext_pernet_ops);
>  	return 0;
>  }
>  
> @@ -1286,9 +1300,12 @@ subsys_initcall(wireless_nlevent_init);
>  static void wireless_nlevent_process(unsigned long data)
>  {
>  	struct sk_buff *skb;
> +	struct net *net;
>  
> -	while ((skb = skb_dequeue(&wireless_nlevent_queue)))
> -		rtnl_notify(skb, &init_net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
> +	for_each_net(net)
> +		while ((skb = skb_dequeue(&net->wext_nlevents)))
> +			rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL,
> +				    GFP_ATOMIC);
>  }
>  
>  static DECLARE_TASKLET(wireless_nlevent_tasklet, wireless_nlevent_process, 0);
> @@ -1341,9 +1358,6 @@ static void rtmsg_iwinfo(struct net_devi
>  	struct sk_buff *skb;
>  	int err;
>  
> -	if (!net_eq(dev_net(dev), &init_net))
> -		return;
> -
>  	skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
>  	if (!skb)
>  		return;
> @@ -1356,7 +1370,7 @@ static void rtmsg_iwinfo(struct net_devi
>  	}
>  
>  	NETLINK_CB(skb).dst_group = RTNLGRP_LINK;
> -	skb_queue_tail(&wireless_nlevent_queue, skb);
> +	skb_queue_tail(&dev_net(dev)->wext_nlevents, skb);
>  	tasklet_schedule(&wireless_nlevent_tasklet);
>  }
>  
>
>
> --
> To unsubscribe from this list: send the line "unsubscribe netdev" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] wireless extensions: play with netns

[Eric W. Biederman]

Johannes Berg <johannes-cdvu00un1VgdHxzADdlk8Q@public.gmane.org> writes:

> This makes wireless extensions netns aware.
>
> Signed-off-by: Johannes Berg <johannes-cdvu00un1VgdHxzADdlk8Q@public.gmane.org>
> ---
> Is this ok, or is this racy? I guess what I'm asking is -- will
> for_each_net() stop iterating over a netns that is going away before the
> pernet exit op is called? If yes, this should be fine.

for_each_net requires the rtnl_lock or the net_mutex to be safe.
You aren't taking either so your code is racy.

A dying network namespace will be removed from the net_namespace_list
(aka for_each_net) before the per net exit methods are called.

Is grabbing the rtnl_lock safe in your workqueue and is that something
we want to do?

Eric


>  include/net/net_namespace.h |    3 +++
>  net/wireless/wext.c         |   30 ++++++++++++++++++++++--------
>  2 files changed, 25 insertions(+), 8 deletions(-)
>
> --- wireless-testing.orig/include/net/net_namespace.h	2009-06-17 20:20:47.000000000 +0200
> +++ wireless-testing/include/net/net_namespace.h	2009-06-17 20:20:51.000000000 +0200
> @@ -78,6 +78,9 @@ struct net {
>  #ifdef CONFIG_XFRM
>  	struct netns_xfrm	xfrm;
>  #endif
> +#ifdef CONFIG_WIRELESS_EXT
> +	struct sk_buff_head	wext_nlevents;
> +#endif
>  	struct net_generic	*gen;
>  };
>  
> --- wireless-testing.orig/net/wireless/wext.c	2009-06-17 20:20:47.000000000 +0200
> +++ wireless-testing/net/wireless/wext.c	2009-06-17 20:20:51.000000000 +0200
> @@ -1273,11 +1273,25 @@ int compat_wext_handle_ioctl(struct net 
>   * Jean II
>   */
>  
> -static struct sk_buff_head wireless_nlevent_queue;
> +static int __net_init wext_pernet_init(struct net *net)
> +{
> +	skb_queue_head_init(&net->wext_nlevents);
> +	return 0;
> +}
> +
> +static void __net_exit wext_pernet_exit(struct net *net)
> +{
> +	skb_queue_purge(&net->wext_nlevents);
> +}
> +
> +static struct pernet_operations wext_pernet_ops = {
> +	.init = wext_pernet_init,
> +	.exit = wext_pernet_exit,
> +};
>  
>  static int __init wireless_nlevent_init(void)
>  {
> -	skb_queue_head_init(&wireless_nlevent_queue);
> +	return register_pernet_subsys(&wext_pernet_ops);
>  	return 0;
>  }
>  
> @@ -1286,9 +1300,12 @@ subsys_initcall(wireless_nlevent_init);
>  static void wireless_nlevent_process(unsigned long data)
>  {
>  	struct sk_buff *skb;
> +	struct net *net;
>  
> -	while ((skb = skb_dequeue(&wireless_nlevent_queue)))
> -		rtnl_notify(skb, &init_net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
> +	for_each_net(net)
> +		while ((skb = skb_dequeue(&net->wext_nlevents)))
> +			rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL,
> +				    GFP_ATOMIC);
>  }
>  
>  static DECLARE_TASKLET(wireless_nlevent_tasklet, wireless_nlevent_process, 0);
> @@ -1341,9 +1358,6 @@ static void rtmsg_iwinfo(struct net_devi
>  	struct sk_buff *skb;
>  	int err;
>  
> -	if (!net_eq(dev_net(dev), &init_net))
> -		return;
> -
>  	skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
>  	if (!skb)
>  		return;
> @@ -1356,7 +1370,7 @@ static void rtmsg_iwinfo(struct net_devi
>  	}
>  
>  	NETLINK_CB(skb).dst_group = RTNLGRP_LINK;
> -	skb_queue_tail(&wireless_nlevent_queue, skb);
> +	skb_queue_tail(&dev_net(dev)->wext_nlevents, skb);
>  	tasklet_schedule(&wireless_nlevent_tasklet);
>  }
>  
>
>
> --
> To unsubscribe from this list: send the line "unsubscribe netdev" in
> the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] wireless extensions: play with netns

[Johannes Berg]

[-- Attachment #1: Type: text/plain, Size: 948 bytes --]

On Wed, 2009-06-17 at 13:46 -0700, Eric W. Biederman wrote:
> Johannes Berg <johannes@sipsolutions.net> writes:
> 
> > This makes wireless extensions netns aware.
> >
> > Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
> > ---
> > Is this ok, or is this racy? I guess what I'm asking is -- will
> > for_each_net() stop iterating over a netns that is going away before the
> > pernet exit op is called? If yes, this should be fine.
> 
> for_each_net requires the rtnl_lock or the net_mutex to be safe.
> You aren't taking either so your code is racy.

Ah ok!

> A dying network namespace will be removed from the net_namespace_list
> (aka for_each_net) before the per net exit methods are called.
> 
> Is grabbing the rtnl_lock safe in your workqueue and is that something
> we want to do?

Ok, thanks. Well, it's a tasklet currently, but can easily be converted
to a work struct and then take the rtnl.

johannes

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 801 bytes --]

[PATCH v2] wireless extensions: make netns aware

[Johannes Berg]

This makes wireless extensions netns aware. The
tasklet sending the events is converted to a work
struct so that we can rtnl_lock() in it.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
---
 include/net/net_namespace.h |    3 ++
 net/wireless/wext.c         |   61 ++++++++++++++++++++------------------------
 2 files changed, 32 insertions(+), 32 deletions(-)

--- wireless-testing.orig/include/net/net_namespace.h	2009-06-17 20:20:47.000000000 +0200
+++ wireless-testing/include/net/net_namespace.h	2009-06-17 23:24:25.000000000 +0200
@@ -78,6 +78,9 @@ struct net {
 #ifdef CONFIG_XFRM
 	struct netns_xfrm	xfrm;
 #endif
+#ifdef CONFIG_WIRELESS_EXT
+	struct sk_buff_head	wext_nlevents;
+#endif
 	struct net_generic	*gen;
 };
 
--- wireless-testing.orig/net/wireless/wext.c	2009-06-17 20:20:47.000000000 +0200
+++ wireless-testing/net/wireless/wext.c	2009-06-17 23:29:00.000000000 +0200
@@ -1250,48 +1250,48 @@ int compat_wext_handle_ioctl(struct net 
 }
 #endif
 
-/************************* EVENT PROCESSING *************************/
-/*
- * Process events generated by the wireless layer or the driver.
- * Most often, the event will be propagated through rtnetlink
- */
+static int __net_init wext_pernet_init(struct net *net)
+{
+	skb_queue_head_init(&net->wext_nlevents);
+	return 0;
+}
 
-/* ---------------------------------------------------------------- */
-/*
- * Locking...
- * ----------
- *
- * Thanks to Herbert Xu <herbert@gondor.apana.org.au> for fixing
- * the locking issue in here and implementing this code !
- *
- * The issue : wireless_send_event() is often called in interrupt context,
- * while the Netlink layer can never be called in interrupt context.
- * The fully formed RtNetlink events are queued, and then a tasklet is run
- * to feed those to Netlink.
- * The skb_queue is interrupt safe, and its lock is not held while calling
- * Netlink, so there is no possibility of dealock.
- * Jean II
- */
+static void __net_exit wext_pernet_exit(struct net *net)
+{
+	skb_queue_purge(&net->wext_nlevents);
+}
 
-static struct sk_buff_head wireless_nlevent_queue;
+static struct pernet_operations wext_pernet_ops = {
+	.init = wext_pernet_init,
+	.exit = wext_pernet_exit,
+};
 
 static int __init wireless_nlevent_init(void)
 {
-	skb_queue_head_init(&wireless_nlevent_queue);
+	return register_pernet_subsys(&wext_pernet_ops);
 	return 0;
 }
 
 subsys_initcall(wireless_nlevent_init);
 
-static void wireless_nlevent_process(unsigned long data)
+/* Process events generated by the wireless layer or the driver. */
+static void wireless_nlevent_process(struct work_struct *work)
 {
 	struct sk_buff *skb;
+	struct net *net;
 
-	while ((skb = skb_dequeue(&wireless_nlevent_queue)))
-		rtnl_notify(skb, &init_net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
+	rtnl_lock();
+
+	for_each_net(net) {
+		while ((skb = skb_dequeue(&net->wext_nlevents)))
+			rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL,
+				    GFP_KERNEL);
+	}
+
+	rtnl_unlock();
 }
 
-static DECLARE_TASKLET(wireless_nlevent_tasklet, wireless_nlevent_process, 0);
+static DECLARE_WORK(wireless_nlevent_work, wireless_nlevent_process);
 
 /* ---------------------------------------------------------------- */
 /*
@@ -1341,9 +1341,6 @@ static void rtmsg_iwinfo(struct net_devi
 	struct sk_buff *skb;
 	int err;
 
-	if (!net_eq(dev_net(dev), &init_net))
-		return;
-
 	skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
 	if (!skb)
 		return;
@@ -1356,8 +1353,8 @@ static void rtmsg_iwinfo(struct net_devi
 	}
 
 	NETLINK_CB(skb).dst_group = RTNLGRP_LINK;
-	skb_queue_tail(&wireless_nlevent_queue, skb);
-	tasklet_schedule(&wireless_nlevent_tasklet);
+	skb_queue_tail(&dev_net(dev)->wext_nlevents, skb);
+	schedule_work(&wireless_nlevent_work);
 }
 
 /* ---------------------------------------------------------------- */

[PATCH v2] wireless extensions: make netns aware

[Johannes Berg]

This makes wireless extensions netns aware. The
tasklet sending the events is converted to a work
struct so that we can rtnl_lock() in it.

Signed-off-by: Johannes Berg <johannes-cdvu00un1VgdHxzADdlk8Q@public.gmane.org>
---
 include/net/net_namespace.h |    3 ++
 net/wireless/wext.c         |   61 ++++++++++++++++++++------------------------
 2 files changed, 32 insertions(+), 32 deletions(-)

--- wireless-testing.orig/include/net/net_namespace.h	2009-06-17 20:20:47.000000000 +0200
+++ wireless-testing/include/net/net_namespace.h	2009-06-17 23:24:25.000000000 +0200
@@ -78,6 +78,9 @@ struct net {
 #ifdef CONFIG_XFRM
 	struct netns_xfrm	xfrm;
 #endif
+#ifdef CONFIG_WIRELESS_EXT
+	struct sk_buff_head	wext_nlevents;
+#endif
 	struct net_generic	*gen;
 };
 
--- wireless-testing.orig/net/wireless/wext.c	2009-06-17 20:20:47.000000000 +0200
+++ wireless-testing/net/wireless/wext.c	2009-06-17 23:29:00.000000000 +0200
@@ -1250,48 +1250,48 @@ int compat_wext_handle_ioctl(struct net 
 }
 #endif
 
-/************************* EVENT PROCESSING *************************/
-/*
- * Process events generated by the wireless layer or the driver.
- * Most often, the event will be propagated through rtnetlink
- */
+static int __net_init wext_pernet_init(struct net *net)
+{
+	skb_queue_head_init(&net->wext_nlevents);
+	return 0;
+}
 
-/* ---------------------------------------------------------------- */
-/*
- * Locking...
- * ----------
- *
- * Thanks to Herbert Xu <herbert-lOAM2aK0SrRLBo1qDEOMRrpzq4S04n8Q@public.gmane.org> for fixing
- * the locking issue in here and implementing this code !
- *
- * The issue : wireless_send_event() is often called in interrupt context,
- * while the Netlink layer can never be called in interrupt context.
- * The fully formed RtNetlink events are queued, and then a tasklet is run
- * to feed those to Netlink.
- * The skb_queue is interrupt safe, and its lock is not held while calling
- * Netlink, so there is no possibility of dealock.
- * Jean II
- */
+static void __net_exit wext_pernet_exit(struct net *net)
+{
+	skb_queue_purge(&net->wext_nlevents);
+}
 
-static struct sk_buff_head wireless_nlevent_queue;
+static struct pernet_operations wext_pernet_ops = {
+	.init = wext_pernet_init,
+	.exit = wext_pernet_exit,
+};
 
 static int __init wireless_nlevent_init(void)
 {
-	skb_queue_head_init(&wireless_nlevent_queue);
+	return register_pernet_subsys(&wext_pernet_ops);
 	return 0;
 }
 
 subsys_initcall(wireless_nlevent_init);
 
-static void wireless_nlevent_process(unsigned long data)
+/* Process events generated by the wireless layer or the driver. */
+static void wireless_nlevent_process(struct work_struct *work)
 {
 	struct sk_buff *skb;
+	struct net *net;
 
-	while ((skb = skb_dequeue(&wireless_nlevent_queue)))
-		rtnl_notify(skb, &init_net, 0, RTNLGRP_LINK, NULL, GFP_ATOMIC);
+	rtnl_lock();
+
+	for_each_net(net) {
+		while ((skb = skb_dequeue(&net->wext_nlevents)))
+			rtnl_notify(skb, net, 0, RTNLGRP_LINK, NULL,
+				    GFP_KERNEL);
+	}
+
+	rtnl_unlock();
 }
 
-static DECLARE_TASKLET(wireless_nlevent_tasklet, wireless_nlevent_process, 0);
+static DECLARE_WORK(wireless_nlevent_work, wireless_nlevent_process);
 
 /* ---------------------------------------------------------------- */
 /*
@@ -1341,9 +1341,6 @@ static void rtmsg_iwinfo(struct net_devi
 	struct sk_buff *skb;
 	int err;
 
-	if (!net_eq(dev_net(dev), &init_net))
-		return;

Re: [PATCH] wireless extensions: play with netns

[Johannes Berg]

[-- Attachment #1: Type: text/plain, Size: 689 bytes --]

On Wed, 2009-06-17 at 13:46 -0700, Eric W. Biederman wrote:
> Johannes Berg <johannes@sipsolutions.net> writes:
> 
> > This makes wireless extensions netns aware.
> >
> > Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
> > ---
> > Is this ok, or is this racy? I guess what I'm asking is -- will
> > for_each_net() stop iterating over a netns that is going away before the
> > pernet exit op is called? If yes, this should be fine.
> 
> for_each_net requires the rtnl_lock or the net_mutex to be safe.
> You aren't taking either so your code is racy.

So it looks like I can also use rcu_read_lock(), but there's no
for_each_net_rcu(), should there be?

johannes

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 801 bytes --]

Re: [PATCH] wireless extensions: play with netns

[Johannes Berg]

[-- Attachment #1: Type: text/plain, Size: 735 bytes --]

On Wed, 2009-06-17 at 13:46 -0700, Eric W. Biederman wrote:
> Johannes Berg <johannes-cdvu00un1VgdHxzADdlk8Q@public.gmane.org> writes:
> 
> > This makes wireless extensions netns aware.
> >
> > Signed-off-by: Johannes Berg <johannes-cdvu00un1VgdHxzADdlk8Q@public.gmane.org>
> > ---
> > Is this ok, or is this racy? I guess what I'm asking is -- will
> > for_each_net() stop iterating over a netns that is going away before the
> > pernet exit op is called? If yes, this should be fine.
> 
> for_each_net requires the rtnl_lock or the net_mutex to be safe.
> You aren't taking either so your code is racy.

So it looks like I can also use rcu_read_lock(), but there's no
for_each_net_rcu(), should there be?

johannes

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 801 bytes --]

Re: [PATCH] wireless extensions: play with netns

[Eric W. Biederman]

Johannes Berg <johannes@sipsolutions.net> writes:

> On Wed, 2009-06-17 at 13:46 -0700, Eric W. Biederman wrote:
>> Johannes Berg <johannes@sipsolutions.net> writes:
>> 
>> > This makes wireless extensions netns aware.
>> >
>> > Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
>> > ---
>> > Is this ok, or is this racy? I guess what I'm asking is -- will
>> > for_each_net() stop iterating over a netns that is going away before the
>> > pernet exit op is called? If yes, this should be fine.
>> 
>> for_each_net requires the rtnl_lock or the net_mutex to be safe.
>> You aren't taking either so your code is racy.
>
> So it looks like I can also use rcu_read_lock(), but there's no
> for_each_net_rcu(), should there be?

I'm not using rcu safe list manipulation.  What makes it look like
rcu_read_lock() is safe?

Eric

Re: [PATCH] wireless extensions: play with netns

[Eric W. Biederman]

Johannes Berg <johannes-cdvu00un1VgdHxzADdlk8Q@public.gmane.org> writes:

> On Wed, 2009-06-17 at 13:46 -0700, Eric W. Biederman wrote:
>> Johannes Berg <johannes-cdvu00un1VgdHxzADdlk8Q@public.gmane.org> writes:
>> 
>> > This makes wireless extensions netns aware.
>> >
>> > Signed-off-by: Johannes Berg <johannes-cdvu00un1VgdHxzADdlk8Q@public.gmane.org>
>> > ---
>> > Is this ok, or is this racy? I guess what I'm asking is -- will
>> > for_each_net() stop iterating over a netns that is going away before the
>> > pernet exit op is called? If yes, this should be fine.
>> 
>> for_each_net requires the rtnl_lock or the net_mutex to be safe.
>> You aren't taking either so your code is racy.
>
> So it looks like I can also use rcu_read_lock(), but there's no
> for_each_net_rcu(), should there be?

I'm not using rcu safe list manipulation.  What makes it look like
rcu_read_lock() is safe?

Eric
--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] wireless extensions: play with netns

[Johannes Berg]

On Wed, 2009-06-17 at 16:24 -0700, Eric W. Biederman wrote:

> > So it looks like I can also use rcu_read_lock(), but there's no
> > for_each_net_rcu(), should there be?
> 
> I'm not using rcu safe list manipulation.  What makes it look like
> rcu_read_lock() is safe?

Indeed. I was looking at rcu_barrier() only. How about the patch below?

johannes

Subject: net: make namespace iteration possible under RCU

We already call rcu_barrier(), so all we need to take
care of is using proper RCU list add/del primitives.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
---
 include/net/net_namespace.h |    3 +++
 net/core/net_namespace.c    |    7 ++++---
 2 files changed, 7 insertions(+), 3 deletions(-)

--- wireless-testing.orig/include/net/net_namespace.h	2009-06-18 01:36:26.000000000 +0200
+++ wireless-testing/include/net/net_namespace.h	2009-06-18 01:39:35.000000000 +0200
@@ -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
--- wireless-testing.orig/net/core/net_namespace.c	2009-06-18 01:36:39.000000000 +0200
+++ wireless-testing/net/core/net_namespace.c	2009-06-18 01:38:36.000000000 +0200
@@ -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>
 
@@ -134,7 +135,7 @@ struct net *copy_net_ns(unsigned long fl
 	err = setup_net(new_net);
 	if (!err) {
 		rtnl_lock();
-		list_add_tail(&new_net->list, &net_namespace_list);
+		list_add_tail_rcu(&new_net->list, &net_namespace_list);
 		rtnl_unlock();
 	}
 	mutex_unlock(&net_mutex);
@@ -163,7 +164,7 @@ 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();
 
 	/* Run all of the network namespace exit methods */
@@ -227,7 +228,7 @@ static int __init net_ns_init(void)
 	err = setup_net(&init_net);
 
 	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);

Re: [PATCH] wireless extensions: play with netns

[Johannes Berg]

On Wed, 2009-06-17 at 16:24 -0700, Eric W. Biederman wrote:

> > So it looks like I can also use rcu_read_lock(), but there's no
> > for_each_net_rcu(), should there be?
> 
> I'm not using rcu safe list manipulation.  What makes it look like
> rcu_read_lock() is safe?

Indeed. I was looking at rcu_barrier() only. How about the patch below?

johannes

Subject: net: make namespace iteration possible under RCU

We already call rcu_barrier(), so all we need to take
care of is using proper RCU list add/del primitives.

Signed-off-by: Johannes Berg <johannes-cdvu00un1VgdHxzADdlk8Q@public.gmane.org>
---
 include/net/net_namespace.h |    3 +++
 net/core/net_namespace.c    |    7 ++++---
 2 files changed, 7 insertions(+), 3 deletions(-)

--- wireless-testing.orig/include/net/net_namespace.h	2009-06-18 01:36:26.000000000 +0200
+++ wireless-testing/include/net/net_namespace.h	2009-06-18 01:39:35.000000000 +0200
@@ -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
--- wireless-testing.orig/net/core/net_namespace.c	2009-06-18 01:36:39.000000000 +0200
+++ wireless-testing/net/core/net_namespace.c	2009-06-18 01:38:36.000000000 +0200
@@ -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>
 
@@ -134,7 +135,7 @@ struct net *copy_net_ns(unsigned long fl
 	err = setup_net(new_net);
 	if (!err) {
 		rtnl_lock();
-		list_add_tail(&new_net->list, &net_namespace_list);
+		list_add_tail_rcu(&new_net->list, &net_namespace_list);
 		rtnl_unlock();
 	}
 	mutex_unlock(&net_mutex);
@@ -163,7 +164,7 @@ 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();
 
 	/* Run all of the network namespace exit methods */
@@ -227,7 +228,7 @@ static int __init net_ns_init(void)
 	err = setup_net(&init_net);
 
 	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);


--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] wireless extensions: play with netns

[Johannes Berg]

On Thu, 2009-06-18 at 01:41 +0200, Johannes Berg wrote:
> On Wed, 2009-06-17 at 16:24 -0700, Eric W. Biederman wrote:
> 
> > > So it looks like I can also use rcu_read_lock(), but there's no
> > > for_each_net_rcu(), should there be?
> > 
> > I'm not using rcu safe list manipulation.  What makes it look like
> > rcu_read_lock() is safe?
> 
> Indeed. I was looking at rcu_barrier() only. How about the patch below?

With that, my genl patch can look like this:

Subject: genetlink: make netns aware

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

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
---
 fs/dlm/netlink.c               |    2 
 include/net/genetlink.h        |   32 +++++--
 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        |  184 +++++++++++++++++++++++++++++++++--------
 net/tipc/netlink.c             |    2 
 net/wireless/nl80211.c         |   14 +--
 9 files changed, 191 insertions(+), 59 deletions(-)

--- wireless-testing.orig/include/net/genetlink.h	2009-06-18 01:39:35.000000000 +0200
+++ wireless-testing/include/net/genetlink.h	2009-06-18 01:41:51.000000000 +0200
@@ -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 */
@@ -62,6 +66,7 @@ struct genl_info
 	struct genlmsghdr *	genlhdr;
 	void *			userhdr;
 	struct nlattr **	attrs;
+	struct net *		net;
 };
 
 /**
@@ -96,8 +101,6 @@ 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
@@ -168,16 +171,27 @@ static inline void genlmsg_cancel(struct
 }
 
 /**
- * genlmsg_multicast - multicast a netlink message
+ * genlmsg_multicast - multicast a netlink message to all network namespaces
+ * @skb: netlink message as socket buffer
+ * @pid: own netlink pid to avoid sending to yourself
+ * @group: multicast group id
+ * @flags: allocation flags
+ */
+int genlmsg_multicast(struct sk_buff *skb, u32 pid,
+		      unsigned int group, gfp_t flags);
+
+/**
+ * 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(struct sk_buff *skb, u32 pid,
-				    unsigned int group, gfp_t 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(genl_sock, skb, pid, group, flags);
+	return nlmsg_multicast(net->genl_sock, skb, pid, group, flags);
 }
 
 /**
@@ -185,9 +199,9 @@ static inline int genlmsg_multicast(stru
  * @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);
 }
 
 /**
@@ -197,7 +211,7 @@ 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(info->net, skb, info->snd_pid);
 }
 
 /**
--- wireless-testing.orig/net/netlink/genetlink.c	2009-06-18 01:39:35.000000000 +0200
+++ wireless-testing/net/netlink/genetlink.c	2009-06-18 01:41:51.000000000 +0200
@@ -15,11 +15,10 @@
 #include <linux/skbuff.h>
 #include <linux/mutex.h>
 #include <linux/bitmap.h>
+#include <linux/rtnetlink.h>
 #include <net/sock.h>
 #include <net/genetlink.h>
 
-struct sock *genl_sock = NULL;
-
 static DEFINE_MUTEX(genl_mutex); /* serialization of message processing */
 
 static inline void genl_lock(void)
@@ -175,10 +174,30 @@ int genl_register_mc_group(struct genl_f
 		mc_groups_longs++;
 	}
 
-	err = netlink_change_ngroups(genl_sock,
-				     mc_groups_longs * BITS_PER_LONG);
-	if (err)
-		goto out;
+	if (family->netnsok) {
+		struct net *net;
+
+		rtnl_lock();
+		for_each_net(net) {
+			err = netlink_change_ngroups(net->genl_sock,
+					mc_groups_longs * BITS_PER_LONG);
+			if (err) {
+				/*
+				 * No need to roll back can only fail
+				 * if memory allocation fails and then
+				 * the number of groups is larger for
+				 * some sockets which is ok.
+				 */
+				goto out;
+			}
+		}
+		rtnl_unlock();
+	} else {
+		err = netlink_change_ngroups(init_net.genl_sock,
+					     mc_groups_longs * BITS_PER_LONG);
+		if (err)
+			goto out;
+	}
 
 	grp->id = id;
 	set_bit(id, mc_groups);
@@ -195,8 +214,14 @@ EXPORT_SYMBOL(genl_register_mc_group);
 static void __genl_unregister_mc_group(struct genl_family *family,
 				       struct genl_multicast_group *grp)
 {
+	struct net *net;
 	BUG_ON(grp->family != family);
-	netlink_clear_multicast_users(genl_sock, grp->id);
+
+	rtnl_lock();
+	for_each_net(net)
+		netlink_clear_multicast_users(net->genl_sock, grp->id);
+	rtnl_unlock();
+
 	clear_bit(grp->id, mc_groups);
 	list_del(&grp->list);
 	genl_ctrl_event(CTRL_CMD_DELMCAST_GRP, grp);
@@ -421,6 +446,7 @@ static int genl_rcv_msg(struct sk_buff *
 {
 	struct genl_ops *ops;
 	struct genl_family *family;
+	struct net *net = sock_net(skb->sk);
 	struct genl_info info;
 	struct genlmsghdr *hdr = nlmsg_data(nlh);
 	int hdrlen, err;
@@ -429,6 +455,10 @@ static int genl_rcv_msg(struct sk_buff *
 	if (family == NULL)
 		return -ENOENT;
 
+	/* this family doesn't exist in this netns */
+	if (!family->netnsok && net != &init_net)
+		return -ENOENT;
+
 	hdrlen = GENL_HDRLEN + family->hdrsize;
 	if (nlh->nlmsg_len < nlmsg_msg_size(hdrlen))
 		return -EINVAL;
@@ -446,7 +476,7 @@ static int genl_rcv_msg(struct sk_buff *
 			return -EOPNOTSUPP;
 
 		genl_unlock();
-		err = netlink_dump_start(genl_sock, skb, nlh,
+		err = netlink_dump_start(net->genl_sock, skb, nlh,
 					 ops->dumpit, ops->done);
 		genl_lock();
 		return err;
@@ -468,6 +498,7 @@ static int genl_rcv_msg(struct sk_buff *
 	info.genlhdr = nlmsg_data(nlh);
 	info.userhdr = nlmsg_data(nlh) + GENL_HDRLEN;
 	info.attrs = family->attrbuf;
+	info.net = net;
 
 	return ops->doit(skb, &info);
 }
@@ -488,6 +519,7 @@ static struct genl_family genl_ctrl = {
 	.name = "nlctrl",
 	.version = 0x2,
 	.maxattr = CTRL_ATTR_MAX,
+	.netnsok = true,
 };
 
 static int ctrl_fill_info(struct genl_family *family, u32 pid, u32 seq,
@@ -604,6 +636,7 @@ static int ctrl_dumpfamily(struct sk_buf
 
 	int i, n = 0;
 	struct genl_family *rt;
+	struct net *net = sock_net(skb->sk);
 	int chains_to_skip = cb->args[0];
 	int fams_to_skip = cb->args[1];
 
@@ -612,6 +645,8 @@ static int ctrl_dumpfamily(struct sk_buf
 			continue;
 		n = 0;
 		list_for_each_entry(rt, genl_family_chain(i), family_list) {
+			if (!rt->netnsok && net != &init_net)
+				continue;
 			if (++n < fams_to_skip)
 				continue;
 			if (ctrl_fill_info(rt, NETLINK_CB(cb->skb).pid,
@@ -683,6 +718,7 @@ static int ctrl_getfamily(struct sk_buff
 	if (info->attrs[CTRL_ATTR_FAMILY_ID]) {
 		u16 id = nla_get_u16(info->attrs[CTRL_ATTR_FAMILY_ID]);
 		res = genl_family_find_byid(id);
+		err = -ENOENT;
 	}
 
 	if (info->attrs[CTRL_ATTR_FAMILY_NAME]) {
@@ -690,51 +726,60 @@ static int ctrl_getfamily(struct sk_buff
 
 		name = nla_data(info->attrs[CTRL_ATTR_FAMILY_NAME]);
 		res = genl_family_find_byname(name);
+		err = -ENOENT;
 	}
 
-	if (res == NULL) {
-		err = -ENOENT;
-		goto errout;
+	if (res == NULL)
+		return err;
+
+	if (!res->netnsok && info->net != &init_net) {
+		/* family doesn't exist here */
+		return -ENOENT;
 	}
 
 	msg = ctrl_build_family_msg(res, info->snd_pid, info->snd_seq,
 				    CTRL_CMD_NEWFAMILY);
-	if (IS_ERR(msg)) {
-		err = PTR_ERR(msg);
-		goto errout;
-	}
+	if (IS_ERR(msg))
+		return PTR_ERR(msg);
 
-	err = genlmsg_reply(msg, info);
-errout:
-	return err;
+	return genlmsg_reply(msg, info);
 }
 
 static int genl_ctrl_event(int event, void *data)
 {
 	struct sk_buff *msg;
+	struct genl_family *family;
+	struct genl_multicast_group *grp;
 
-	if (genl_sock == NULL)
+	/* genl is still initialising */
+	if (!init_net.genl_sock)
 		return 0;
 
 	switch (event) {
 	case CTRL_CMD_NEWFAMILY:
 	case CTRL_CMD_DELFAMILY:
-		msg = ctrl_build_family_msg(data, 0, 0, event);
-		if (IS_ERR(msg))
-			return PTR_ERR(msg);
-
-		genlmsg_multicast(msg, 0, GENL_ID_CTRL, GFP_KERNEL);
+		family = data;
+		msg = ctrl_build_family_msg(family, 0, 0, event);
 		break;
 	case CTRL_CMD_NEWMCAST_GRP:
 	case CTRL_CMD_DELMCAST_GRP:
+		grp = data;
+		family = grp->family;
 		msg = ctrl_build_mcgrp_msg(data, 0, 0, event);
-		if (IS_ERR(msg))
-			return PTR_ERR(msg);
-
-		genlmsg_multicast(msg, 0, GENL_ID_CTRL, GFP_KERNEL);
 		break;
+	default:
+		return -EINVAL;
 	}
 
+	if (IS_ERR(msg))
+		return PTR_ERR(msg);
+
+	if (!family->netnsok)
+		genlmsg_multicast_netns(&init_net, msg, 0,
+					GENL_ID_CTRL, GFP_KERNEL);
+	else
+		genlmsg_multicast(msg, 0, GENL_ID_CTRL, GFP_KERNEL);
+
 	return 0;
 }
 
@@ -749,6 +794,33 @@ static struct genl_multicast_group notif
 	.name		= "notify",
 };
 
+static int __net_init genl_pernet_init(struct net *net)
+{
+	/* we'll bump the group number right afterwards */
+	net->genl_sock = netlink_kernel_create(net, NETLINK_GENERIC, 0,
+					       genl_rcv, &genl_mutex,
+					       THIS_MODULE);
+
+	if (!net->genl_sock && net == &init_net)
+		panic("GENL: Cannot initialize generic netlink\n");
+
+	if (!net->genl_sock)
+		return -ENOMEM;
+
+	return 0;
+}
+
+static void __net_exit genl_pernet_exit(struct net *net)
+{
+	netlink_kernel_release(net->genl_sock);
+	net->genl_sock = NULL;
+}
+
+static struct pernet_operations genl_pernet_ops = {
+	.init = genl_pernet_init,
+	.exit = genl_pernet_exit,
+};
+
 static int __init genl_init(void)
 {
 	int i, err;
@@ -766,11 +838,9 @@ static int __init genl_init(void)
 
 	netlink_set_nonroot(NETLINK_GENERIC, NL_NONROOT_RECV);
 
-	/* we'll bump the group number right afterwards */
-	genl_sock = netlink_kernel_create(&init_net, NETLINK_GENERIC, 0,
-					  genl_rcv, &genl_mutex, THIS_MODULE);
-	if (genl_sock == NULL)
-		panic("GENL: Cannot initialize generic netlink\n");
+	err = register_pernet_subsys(&genl_pernet_ops);
+	if (err)
+		goto errout_register;
 
 	err = genl_register_mc_group(&genl_ctrl, &notify_grp);
 	if (err < 0)
@@ -786,8 +856,54 @@ errout:
 
 subsys_initcall(genl_init);
 
-EXPORT_SYMBOL(genl_sock);
 EXPORT_SYMBOL(genl_register_ops);
 EXPORT_SYMBOL(genl_unregister_ops);
 EXPORT_SYMBOL(genl_register_family);
 EXPORT_SYMBOL(genl_unregister_family);
+
+static int genlmsg_mcast(struct sk_buff *skb, u32 pid, unsigned long group,
+			 gfp_t flags)
+{
+	struct sk_buff *tmp;
+	struct net *net, *prev = NULL;
+	int err;
+
+	for_each_net_rcu(net) {
+		if (prev) {
+			tmp = skb_clone(skb, flags);
+			if (!tmp) {
+				err = -ENOMEM;
+				goto error;
+			}
+			err = nlmsg_multicast(prev->genl_sock, tmp,
+					      pid, group, flags);
+			if (err)
+				goto error;
+		}
+
+		prev = net;
+	}
+
+	return nlmsg_multicast(net->genl_sock, skb, pid, group, flags);
+ error:
+	return err;
+}
+
+int genlmsg_multicast(struct sk_buff *skb, u32 pid,
+		      unsigned int group, gfp_t flags)
+{
+	int ret;
+
+	if (flags & GFP_ATOMIC) {
+		rcu_read_lock();
+		ret = genlmsg_mcast(skb, pid, group, flags);
+		rcu_read_unlock();
+	} else {
+		rtnl_lock();
+		ret = genlmsg_mcast(skb, pid, group, flags);
+		rtnl_unlock();
+	}
+
+	return ret;
+}
+EXPORT_SYMBOL(genlmsg_multicast);
--- wireless-testing.orig/fs/dlm/netlink.c	2009-06-18 01:39:35.000000000 +0200
+++ wireless-testing/fs/dlm/netlink.c	2009-06-18 01:41:51.000000000 +0200
@@ -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)
--- wireless-testing.orig/kernel/taskstats.c	2009-06-18 01:39:35.000000000 +0200
+++ wireless-testing/kernel/taskstats.c	2009-06-18 01:41:51.000000000 +0200
@@ -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;
--- wireless-testing.orig/net/irda/irnetlink.c	2009-06-18 01:39:35.000000000 +0200
+++ wireless-testing/net/irda/irnetlink.c	2009-06-18 01:41:51.000000000 +0200
@@ -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);
--- wireless-testing.orig/net/netfilter/ipvs/ip_vs_ctl.c	2009-06-18 01:39:35.000000000 +0200
+++ wireless-testing/net/netfilter/ipvs/ip_vs_ctl.c	2009-06-18 01:41:51.000000000 +0200
@@ -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:
--- wireless-testing.orig/net/tipc/netlink.c	2009-06-18 01:39:35.000000000 +0200
+++ wireless-testing/net/tipc/netlink.c	2009-06-18 01:41:51.000000000 +0200
@@ -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;
--- wireless-testing.orig/net/wireless/nl80211.c	2009-06-18 01:39:35.000000000 +0200
+++ wireless-testing/net/wireless/nl80211.c	2009-06-18 01:41:51.000000000 +0200
@@ -398,7 +398,7 @@ static int nl80211_get_wiphy(struct sk_b
 
 	cfg80211_put_dev(dev);
 
-	return genlmsg_unicast(msg, info->snd_pid);
+	return genlmsg_reply(msg, info);
 
  out_free:
 	nlmsg_free(msg);
@@ -723,7 +723,7 @@ static int nl80211_get_interface(struct 
 	dev_put(netdev);
 	cfg80211_put_dev(dev);
 
-	return genlmsg_unicast(msg, info->snd_pid);
+	return genlmsg_reply(msg, info);
 
  out_free:
 	nlmsg_free(msg);
@@ -1014,7 +1014,7 @@ 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:
@@ -1602,7 +1602,7 @@ 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:
@@ -2009,7 +2009,7 @@ 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:
@@ -2358,7 +2358,7 @@ 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:
@@ -2546,7 +2546,7 @@ 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:
--- wireless-testing.orig/include/net/net_namespace.h	2009-06-18 01:39:35.000000000 +0200
+++ wireless-testing/include/net/net_namespace.h	2009-06-18 01:41:51.000000000 +0200
@@ -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
@@ -63,6 +64,7 @@ struct net {
 	struct netns_packet	packet;
 	struct netns_unix	unx;
 	struct netns_ipv4	ipv4;
+	struct sock		*genl_sock;
 #if defined(CONFIG_IPV6) || defined(CONFIG_IPV6_MODULE)
 	struct netns_ipv6	ipv6;
 #endif

Re: [PATCH] wireless extensions: play with netns

[Eric W. Biederman]

Johannes Berg <johannes@sipsolutions.net> writes:

> On Wed, 2009-06-17 at 16:24 -0700, Eric W. Biederman wrote:
>
>> > So it looks like I can also use rcu_read_lock(), but there's no
>> > for_each_net_rcu(), should there be?
>> 
>> I'm not using rcu safe list manipulation.  What makes it look like
>> rcu_read_lock() is safe?
>
> Indeed. I was looking at rcu_barrier() only. How about the patch below?
>
> johannes
>
> Subject: net: make namespace iteration possible under RCU
>
> We already call rcu_barrier(), so all we need to take
> care of is using proper RCU list add/del primitives.

This of course gives you a network namespace that can be found by for_each_net rcu
while the per net exit functions are running.  I think that opens up to races
that I don't want to think about.

I still haven't tracked down how I am occasionally getting time wait sockets
with an invalid network namespace.

Eric

Re: [PATCH] wireless extensions: play with netns

[Eric W. Biederman]

Johannes Berg <johannes-cdvu00un1VgdHxzADdlk8Q@public.gmane.org> writes:

> On Wed, 2009-06-17 at 16:24 -0700, Eric W. Biederman wrote:
>
>> > So it looks like I can also use rcu_read_lock(), but there's no
>> > for_each_net_rcu(), should there be?
>> 
>> I'm not using rcu safe list manipulation.  What makes it look like
>> rcu_read_lock() is safe?
>
> Indeed. I was looking at rcu_barrier() only. How about the patch below?
>
> johannes
>
> Subject: net: make namespace iteration possible under RCU
>
> We already call rcu_barrier(), so all we need to take
> care of is using proper RCU list add/del primitives.

This of course gives you a network namespace that can be found by for_each_net rcu
while the per net exit functions are running.  I think that opens up to races
that I don't want to think about.

I still haven't tracked down how I am occasionally getting time wait sockets
with an invalid network namespace.

Eric
--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] wireless extensions: play with netns

[Johannes Berg]

[-- Attachment #1: Type: text/plain, Size: 758 bytes --]

On Wed, 2009-06-17 at 16:54 -0700, Eric W. Biederman wrote:

> > Subject: net: make namespace iteration possible under RCU
> >
> > We already call rcu_barrier(), so all we need to take
> > care of is using proper RCU list add/del primitives.
> 
> This of course gives you a network namespace that can be found by for_each_net rcu
> while the per net exit functions are running.  I think that opens up to races
> that I don't want to think about.

Indeed. Can we move the rcu_barrier() up? Or we could insert a
synchronize_rcu() (which is sufficient for rcu_read_lock) before the
exit functions are run?

> I still haven't tracked down how I am occasionally getting time wait sockets
> with an invalid network namespace.

Ouch.

johannes

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 801 bytes --]

Re: [PATCH] wireless extensions: play with netns

[Johannes Berg]

[-- Attachment #1: Type: text/plain, Size: 758 bytes --]

On Wed, 2009-06-17 at 16:54 -0700, Eric W. Biederman wrote:

> > Subject: net: make namespace iteration possible under RCU
> >
> > We already call rcu_barrier(), so all we need to take
> > care of is using proper RCU list add/del primitives.
> 
> This of course gives you a network namespace that can be found by for_each_net rcu
> while the per net exit functions are running.  I think that opens up to races
> that I don't want to think about.

Indeed. Can we move the rcu_barrier() up? Or we could insert a
synchronize_rcu() (which is sufficient for rcu_read_lock) before the
exit functions are run?

> I still haven't tracked down how I am occasionally getting time wait sockets
> with an invalid network namespace.

Ouch.

johannes

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 801 bytes --]

Re: [PATCH] wireless extensions: play with netns

[Eric W. Biederman]

Johannes Berg <johannes@sipsolutions.net> writes:

> On Thu, 2009-06-18 at 01:41 +0200, Johannes Berg wrote:
>> On Wed, 2009-06-17 at 16:24 -0700, Eric W. Biederman wrote:
>> 
>> > > So it looks like I can also use rcu_read_lock(), but there's no
>> > > for_each_net_rcu(), should there be?
>> > 
>> > I'm not using rcu safe list manipulation.  What makes it look like
>> > rcu_read_lock() is safe?
>> 
>> Indeed. I was looking at rcu_barrier() only. How about the patch below?
>
> With that, my genl patch can look like this:
>
> Subject: genetlink: make netns aware
>
> This makes generic netlink network namespace aware.
> No actual generic netlink families are made namespace
> aware, they need to be checked one by one and then
> set the family->netnsok member to true.

Are skb_clone and nlmsg_multicast really guaranteed not to sleep?

That seems like a lot of code and a lot of code paths.

Eric

Re: [PATCH] wireless extensions: play with netns

[Eric W. Biederman]

Johannes Berg <johannes-cdvu00un1VgdHxzADdlk8Q@public.gmane.org> writes:

> On Thu, 2009-06-18 at 01:41 +0200, Johannes Berg wrote:
>> On Wed, 2009-06-17 at 16:24 -0700, Eric W. Biederman wrote:
>> 
>> > > So it looks like I can also use rcu_read_lock(), but there's no
>> > > for_each_net_rcu(), should there be?
>> > 
>> > I'm not using rcu safe list manipulation.  What makes it look like
>> > rcu_read_lock() is safe?
>> 
>> Indeed. I was looking at rcu_barrier() only. How about the patch below?
>
> With that, my genl patch can look like this:
>
> Subject: genetlink: make netns aware
>
> This makes generic netlink network namespace aware.
> No actual generic netlink families are made namespace
> aware, they need to be checked one by one and then
> set the family->netnsok member to true.

Are skb_clone and nlmsg_multicast really guaranteed not to sleep?

That seems like a lot of code and a lot of code paths.

Eric
--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] wireless extensions: play with netns

[Eric W. Biederman]

Johannes Berg <johannes@sipsolutions.net> writes:

> On Wed, 2009-06-17 at 16:54 -0700, Eric W. Biederman wrote:
>
>> > Subject: net: make namespace iteration possible under RCU
>> >
>> > We already call rcu_barrier(), so all we need to take
>> > care of is using proper RCU list add/del primitives.
>> 
>> This of course gives you a network namespace that can be found by for_each_net rcu
>> while the per net exit functions are running.  I think that opens up to races
>> that I don't want to think about.
>
> Indeed. Can we move the rcu_barrier() up? Or we could insert a
> synchronize_rcu() (which is sufficient for rcu_read_lock) before the
> exit functions are run?

I don't think we can move the barrier.   But adding an extra synchronize_rcu
should be fine.  We are talking about the slowest of the slow paths here.
It is so slow it even gets it's own kernel thread.

Eric

Re: [PATCH] wireless extensions: play with netns

[Eric W. Biederman]

Johannes Berg <johannes-cdvu00un1VgdHxzADdlk8Q@public.gmane.org> writes:

> On Wed, 2009-06-17 at 16:54 -0700, Eric W. Biederman wrote:
>
>> > Subject: net: make namespace iteration possible under RCU
>> >
>> > We already call rcu_barrier(), so all we need to take
>> > care of is using proper RCU list add/del primitives.
>> 
>> This of course gives you a network namespace that can be found by for_each_net rcu
>> while the per net exit functions are running.  I think that opens up to races
>> that I don't want to think about.
>
> Indeed. Can we move the rcu_barrier() up? Or we could insert a
> synchronize_rcu() (which is sufficient for rcu_read_lock) before the
> exit functions are run?

I don't think we can move the barrier.   But adding an extra synchronize_rcu
should be fine.  We are talking about the slowest of the slow paths here.
It is so slow it even gets it's own kernel thread.

Eric
--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] wireless extensions: play with netns

[Johannes Berg]

[-- Attachment #1: Type: text/plain, Size: 579 bytes --]

On Wed, 2009-06-17 at 17:27 -0700, Eric W. Biederman wrote:

> > Subject: genetlink: make netns aware
> >
> > This makes generic netlink network namespace aware.
> > No actual generic netlink families are made namespace
> > aware, they need to be checked one by one and then
> > set the family->netnsok member to true.
> 
> Are skb_clone and nlmsg_multicast really guaranteed not to sleep?

Yes, definitely, we call them in atomic callbacks already in various
places. There are many callers that use the existing genlmsg_multicast()
in atomic contexts.

johannes

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 801 bytes --]

Re: [PATCH] wireless extensions: play with netns

[Johannes Berg]

On Wed, 2009-06-17 at 17:30 -0700, Eric W. Biederman wrote:

> >> This of course gives you a network namespace that can be found by for_each_net rcu
> >> while the per net exit functions are running.  I think that opens up to races
> >> that I don't want to think about.
> >
> > Indeed. Can we move the rcu_barrier() up? Or we could insert a
> > synchronize_rcu() (which is sufficient for rcu_read_lock) before the
> > exit functions are run?
> 
> I don't think we can move the barrier.   But adding an extra synchronize_rcu
> should be fine.  We are talking about the slowest of the slow paths here.
> It is so slow it even gets it's own kernel thread.

Ok :) New patch below.

Yes, you're right, we can't move it, it might still be used from exit
for net_assign_generic or so.

johannes

Subject: net: make namespace iteration possible under RCU

We already call rcu_barrier(), so all we need to take
care of is using proper RCU list add/del primitives.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
---
 include/net/net_namespace.h |    3 +++
 net/core/net_namespace.c    |   10 +++++++---
 2 files changed, 10 insertions(+), 3 deletions(-)

--- wireless-testing.orig/include/net/net_namespace.h	2009-06-18 01:36:26.000000000 +0200
+++ wireless-testing/include/net/net_namespace.h	2009-06-18 02:17:14.000000000 +0200
@@ -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
--- wireless-testing.orig/net/core/net_namespace.c	2009-06-18 01:36:39.000000000 +0200
+++ wireless-testing/net/core/net_namespace.c	2009-06-18 02:03:06.000000000 +0200
@@ -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>
 
@@ -134,7 +135,7 @@ struct net *copy_net_ns(unsigned long fl
 	err = setup_net(new_net);
 	if (!err) {
 		rtnl_lock();
-		list_add_tail(&new_net->list, &net_namespace_list);
+		list_add_tail_rcu(&new_net->list, &net_namespace_list);
 		rtnl_unlock();
 	}
 	mutex_unlock(&net_mutex);
@@ -163,9 +164,12 @@ 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();
 
+	/* if somebody is rcu-iterating the list, wait */
+	synchronize_rcu();
+
 	/* Run all of the network namespace exit methods */
 	list_for_each_entry_reverse(ops, &pernet_list, list) {
 		if (ops->exit)
@@ -227,7 +231,7 @@ static int __init net_ns_init(void)
 	err = setup_net(&init_net);
 
 	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);

Re: [PATCH] wireless extensions: play with netns

[Johannes Berg]

On Wed, 2009-06-17 at 17:30 -0700, Eric W. Biederman wrote:

> >> This of course gives you a network namespace that can be found by for_each_net rcu
> >> while the per net exit functions are running.  I think that opens up to races
> >> that I don't want to think about.
> >
> > Indeed. Can we move the rcu_barrier() up? Or we could insert a
> > synchronize_rcu() (which is sufficient for rcu_read_lock) before the
> > exit functions are run?
> 
> I don't think we can move the barrier.   But adding an extra synchronize_rcu
> should be fine.  We are talking about the slowest of the slow paths here.
> It is so slow it even gets it's own kernel thread.

Ok :) New patch below.

Yes, you're right, we can't move it, it might still be used from exit
for net_assign_generic or so.

johannes

Subject: net: make namespace iteration possible under RCU

We already call rcu_barrier(), so all we need to take
care of is using proper RCU list add/del primitives.

Signed-off-by: Johannes Berg <johannes-cdvu00un1VgdHxzADdlk8Q@public.gmane.org>
---
 include/net/net_namespace.h |    3 +++
 net/core/net_namespace.c    |   10 +++++++---
 2 files changed, 10 insertions(+), 3 deletions(-)

--- wireless-testing.orig/include/net/net_namespace.h	2009-06-18 01:36:26.000000000 +0200
+++ wireless-testing/include/net/net_namespace.h	2009-06-18 02:17:14.000000000 +0200
@@ -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
--- wireless-testing.orig/net/core/net_namespace.c	2009-06-18 01:36:39.000000000 +0200
+++ wireless-testing/net/core/net_namespace.c	2009-06-18 02:03:06.000000000 +0200
@@ -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>
 
@@ -134,7 +135,7 @@ struct net *copy_net_ns(unsigned long fl
 	err = setup_net(new_net);
 	if (!err) {
 		rtnl_lock();
-		list_add_tail(&new_net->list, &net_namespace_list);
+		list_add_tail_rcu(&new_net->list, &net_namespace_list);
 		rtnl_unlock();
 	}
 	mutex_unlock(&net_mutex);
@@ -163,9 +164,12 @@ 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();
 
+	/* if somebody is rcu-iterating the list, wait */
+	synchronize_rcu();
+
 	/* Run all of the network namespace exit methods */
 	list_for_each_entry_reverse(ops, &pernet_list, list) {
 		if (ops->exit)
@@ -227,7 +231,7 @@ static int __init net_ns_init(void)
 	err = setup_net(&init_net);
 
 	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);


--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] wireless extensions: play with netns

[Eric W. Biederman]

Johannes Berg <johannes@sipsolutions.net> writes:

> On Wed, 2009-06-17 at 17:30 -0700, Eric W. Biederman wrote:
>
>> >> This of course gives you a network namespace that can be found by for_each_net rcu
>> >> while the per net exit functions are running.  I think that opens up to races
>> >> that I don't want to think about.
>> >
>> > Indeed. Can we move the rcu_barrier() up? Or we could insert a
>> > synchronize_rcu() (which is sufficient for rcu_read_lock) before the
>> > exit functions are run?
>> 
>> I don't think we can move the barrier.   But adding an extra synchronize_rcu
>> should be fine.  We are talking about the slowest of the slow paths here.
>> It is so slow it even gets it's own kernel thread.
>
> Ok :) New patch below.
>
> Yes, you're right, we can't move it, it might still be used from exit
> for net_assign_generic or so.

Except for the changelog which is wrong about not needing any rcu
primitives I can't spot any bugs.

> johannes
>
> Subject: net: make namespace iteration possible under RCU
>
> We already call rcu_barrier(), so all we need to take
> care of is using proper RCU list add/del primitives.
>
> Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
> ---
>  include/net/net_namespace.h |    3 +++
>  net/core/net_namespace.c    |   10 +++++++---
>  2 files changed, 10 insertions(+), 3 deletions(-)
>
> --- wireless-testing.orig/include/net/net_namespace.h	2009-06-18 01:36:26.000000000 +0200
> +++ wireless-testing/include/net/net_namespace.h	2009-06-18 02:17:14.000000000 +0200
> @@ -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
> --- wireless-testing.orig/net/core/net_namespace.c	2009-06-18 01:36:39.000000000 +0200
> +++ wireless-testing/net/core/net_namespace.c	2009-06-18 02:03:06.000000000 +0200
> @@ -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>
>  
> @@ -134,7 +135,7 @@ struct net *copy_net_ns(unsigned long fl
>  	err = setup_net(new_net);
>  	if (!err) {
>  		rtnl_lock();
> -		list_add_tail(&new_net->list, &net_namespace_list);
> +		list_add_tail_rcu(&new_net->list, &net_namespace_list);
>  		rtnl_unlock();
>  	}
>  	mutex_unlock(&net_mutex);
> @@ -163,9 +164,12 @@ 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();
>  
> +	/* if somebody is rcu-iterating the list, wait */
> +	synchronize_rcu();
> +
>  	/* Run all of the network namespace exit methods */
>  	list_for_each_entry_reverse(ops, &pernet_list, list) {
>  		if (ops->exit)
> @@ -227,7 +231,7 @@ static int __init net_ns_init(void)
>  	err = setup_net(&init_net);
>  
>  	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);