[PATCH] wwan: core: Fix missing RTM_NEWLINK event

By default there is no rtnetlink event generated when registering a
netdev with rtnl_link_ops until its rtnl_link_state is switched to
initialized (RTNL_LINK_INITIALIZED). This causes issues with user
tools like NetworkManager which relies on such event to manage links.

Fix that by setting link to initialized (via rtnl_configure_link).

Cc: stable@vger.kernel.org
Fixes: 88b710532e53 ("wwan: add interface creation support")
Signed-off-by: Loic Poulain <redacted>
---
 drivers/net/wwan/wwan_core.c | 4 ++++
 1 file changed, 4 insertions(+)

diff --git a/drivers/net/wwan/wwan_core.c b/drivers/net/wwan/wwan_core.c
index 3e16c31..409caf4 100644
--- a/drivers/net/wwan/wwan_core.c
+++ b/drivers/net/wwan/wwan_core.c

@@ -866,6 +866,10 @@ static int wwan_rtnl_newlink(struct net *src_net, struct net_device *dev,
 	else
 		ret = register_netdevice(dev);
 
+	/* Link initialized, notify new link */
+	if (!ret)
+		rtnl_configure_link(dev, NULL);
+
 out:
 	/* release the reference */
 	put_device(&wwandev->dev);

-- 
2.7.4

Hello Loic,

On Thu, Jul 22, 2021 at 6:39 PM Loic Poulain [off-list ref] wrote:

By default there is no rtnetlink event generated when registering a
netdev with rtnl_link_ops until its rtnl_link_state is switched to
initialized (RTNL_LINK_INITIALIZED). This causes issues with user
tools like NetworkManager which relies on such event to manage links.

Fix that by setting link to initialized (via rtnl_configure_link).

Shouldn't the __rtnl_newlink() function call rtnl_configure_link()
just after the newlink() callback invocation? Or I missed something?

-- 
Sergey

On Thu, 22 Jul 2021 at 18:14, Sergey Ryazanov [off-list ref] wrote:

Hello Loic,

On Thu, Jul 22, 2021 at 6:39 PM Loic Poulain [off-list ref] wrote:

quoted

By default there is no rtnetlink event generated when registering a
netdev with rtnl_link_ops until its rtnl_link_state is switched to
initialized (RTNL_LINK_INITIALIZED). This causes issues with user
tools like NetworkManager which relies on such event to manage links.

Fix that by setting link to initialized (via rtnl_configure_link).

Shouldn't the __rtnl_newlink() function call rtnl_configure_link()
just after the newlink() callback invocation? Or I missed something?

Ah right, but the first call of rtnl_configure_link() (uninitialized)
does not cause RTM_NEWLINK event (cf __dev_notify_flags). It however
seems to work for other link types (e,g, rmnet), so probably something
to clarify here.

Regards,
Loic

--
Sergey

On Thu, Jul 22, 2021 at 7:44 PM Loic Poulain [off-list ref] wrote:

On Thu, 22 Jul 2021 at 18:14, Sergey Ryazanov [off-list ref] wrote:

quoted

On Thu, Jul 22, 2021 at 6:39 PM Loic Poulain [off-list ref] wrote:

quoted

By default there is no rtnetlink event generated when registering a
netdev with rtnl_link_ops until its rtnl_link_state is switched to
initialized (RTNL_LINK_INITIALIZED). This causes issues with user
tools like NetworkManager which relies on such event to manage links.

Fix that by setting link to initialized (via rtnl_configure_link).

Shouldn't the __rtnl_newlink() function call rtnl_configure_link()
just after the newlink() callback invocation? Or I missed something?

Ah right, but the first call of rtnl_configure_link() (uninitialized)
does not cause RTM_NEWLINK event (cf __dev_notify_flags). It however
seems to work for other link types (e,g, rmnet), so probably something
to clarify here.

Just check additional netdev creation with hwsim:

# ip link add wwan0.3 parentdev wwan0 type wwan linkid 3

On the other console:

# ip -d mon
6: wwan0.3: <POINTOPOINT,NOARP> mtu 1500 qdisc noop state DOWN group default
    link/none  promiscuity 0 minmtu 68 maxmtu 65535
    wwan numtxqueues 1 numrxqueues 1 gso_max_size 65536 gso_max_segs 65535

But I saw no notification at the moment of wwan_hwsim module loading.
This happens since I missed the rtnl_configure_link() call in the
wwan_create_default_link() after the default link successful creation
:(

So we need your fix at least in the default link creation routine to
fix ca374290aaad ("wwan: core: support default netdev creation").
Something like this:

diff --git a/drivers/net/wwan/wwan_core.c b/drivers/net/wwan/wwan_core.c
index 3e16c318e705..374aa2cc884c 100644
--- a/drivers/net/wwan/wwan_core.c
+++ b/drivers/net/wwan/wwan_core.c

@@ -984,6 +984,8 @@ static void wwan_create_default_link(struct

wwan_device *wwandev,
  goto unlock;
  }

+ rtnl_configure_link(dev, NULL); /* trigger the RTM_NEWLINK event */
+
 unlock:
  rtnl_unlock();

--
Sergey

On Thu, 22 Jul 2021 at 19:42, Sergey Ryazanov [off-list ref] wrote:

On Thu, Jul 22, 2021 at 7:44 PM Loic Poulain [off-list ref] wrote:

quoted

On Thu, 22 Jul 2021 at 18:14, Sergey Ryazanov [off-list ref] wrote:

quoted

On Thu, Jul 22, 2021 at 6:39 PM Loic Poulain [off-list ref] wrote:

quoted

By default there is no rtnetlink event generated when registering a
netdev with rtnl_link_ops until its rtnl_link_state is switched to
initialized (RTNL_LINK_INITIALIZED). This causes issues with user
tools like NetworkManager which relies on such event to manage links.

Fix that by setting link to initialized (via rtnl_configure_link).

Shouldn't the __rtnl_newlink() function call rtnl_configure_link()
just after the newlink() callback invocation? Or I missed something?

Ah right, but the first call of rtnl_configure_link() (uninitialized)
does not cause RTM_NEWLINK event (cf __dev_notify_flags). It however
seems to work for other link types (e,g, rmnet), so probably something
to clarify here.

Just check additional netdev creation with hwsim:

# ip link add wwan0.3 parentdev wwan0 type wwan linkid 3

On the other console:

# ip -d mon
6: wwan0.3: <POINTOPOINT,NOARP> mtu 1500 qdisc noop state DOWN group default
    link/none  promiscuity 0 minmtu 68 maxmtu 65535
    wwan numtxqueues 1 numrxqueues 1 gso_max_size 65536 gso_max_segs 65535

But I saw no notification at the moment of wwan_hwsim module loading.
This happens since I missed the rtnl_configure_link() call in the
wwan_create_default_link() after the default link successful creation
:(

Yep just realized that!

Testing right now.

Loic