[RESEND net-next PATCH] net: rtnetlink: Enslave device before bringing it up

[RESEND net-next PATCH] net: rtnetlink: Enslave device before bringing it up

[Phil Sutter]

Unlike with bridges, one can't add an interface to a bond and set it up
at the same time:

| # ip link set dummy0 down
| # ip link set dummy0 master bond0 up
| Error: Device can not be enslaved while up.

Of all drivers with ndo_add_slave callback, bond and team decline if
IFF_UP flag is set, vrf cycles the interface (i.e., sets it down and
immediately up again) and the others just don't care.

Support the common notion of setting the interface up after enslaving it
by sorting the operations accordingly.

Signed-off-by: Phil Sutter <phil@nwl.cc>
---
Resubmitting this after review, concerns uttered in original discussion
three years ago do not apply: Any flag changes happening during
do_set_master() call are preserved, unless user space intended to change
them. I verified that a team device in promisc mode still correctly
propagates the flag to a new slave after applying this patch.
---
 net/core/rtnetlink.c | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
index ac45328607f77..2e15f744544e5 100644
--- a/net/core/rtnetlink.c
+++ b/net/core/rtnetlink.c
@@ -2776,13 +2776,6 @@ static int do_setlink(const struct sk_buff *skb,
 		call_netdevice_notifiers(NETDEV_CHANGEADDR, dev);
 	}
 
-	if (ifm->ifi_flags || ifm->ifi_change) {
-		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
-				       extack);
-		if (err < 0)
-			goto errout;
-	}
-
 	if (tb[IFLA_MASTER]) {
 		err = do_set_master(dev, nla_get_u32(tb[IFLA_MASTER]), extack);
 		if (err)
@@ -2790,6 +2783,13 @@ static int do_setlink(const struct sk_buff *skb,
 		status |= DO_SETLINK_MODIFIED;
 	}
 
+	if (ifm->ifi_flags || ifm->ifi_change) {
+		err = dev_change_flags(dev, rtnl_dev_combine_flags(dev, ifm),
+				       extack);
+		if (err < 0)
+			goto errout;
+	}
+
 	if (tb[IFLA_CARRIER]) {
 		err = dev_change_carrier(dev, nla_get_u8(tb[IFLA_CARRIER]));
 		if (err)
-- 
2.34.1

Re: [RESEND net-next PATCH] net: rtnetlink: Enslave device before bringing it up

[David Ahern]

On 9/14/22 9:06 AM, Phil Sutter wrote:
> Unlike with bridges, one can't add an interface to a bond and set it up
> at the same time:
> 
> | # ip link set dummy0 down
> | # ip link set dummy0 master bond0 up
> | Error: Device can not be enslaved while up.
> 
> Of all drivers with ndo_add_slave callback, bond and team decline if
> IFF_UP flag is set, vrf cycles the interface (i.e., sets it down and
> immediately up again) and the others just don't care.
> 
> Support the common notion of setting the interface up after enslaving it
> by sorting the operations accordingly.
> 
> Signed-off-by: Phil Sutter <phil@nwl.cc>
> ---
> Resubmitting this after review, concerns uttered in original discussion
> three years ago do not apply: Any flag changes happening during
> do_set_master() call are preserved, unless user space intended to change
> them. I verified that a team device in promisc mode still correctly
> propagates the flag to a new slave after applying this patch.
> ---
>  net/core/rtnetlink.c | 14 +++++++-------
>  1 file changed, 7 insertions(+), 7 deletions(-)
> 

Reviewed-by: David Ahern <dsahern@kernel.org>

Re: [RESEND net-next PATCH] net: rtnetlink: Enslave device before bringing it up

[patchwork-bot+netdevbpf]

Hello:

This patch was applied to netdev/net-next.git (master)
by Jakub Kicinski <kuba@kernel.org>:

On Wed, 14 Sep 2022 17:06:23 +0200 you wrote:
> Unlike with bridges, one can't add an interface to a bond and set it up
> at the same time:
> 
> | # ip link set dummy0 down
> | # ip link set dummy0 master bond0 up
> | Error: Device can not be enslaved while up.
> 
> [...]

Here is the summary with links:
  - [RESEND,net-next] net: rtnetlink: Enslave device before bringing it up
    https://git.kernel.org/netdev/net-next/c/a4abfa627c38

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html

Re: [RESEND net-next PATCH] net: rtnetlink: Enslave device before bringing it up

[Jiri Pirko]

Wed, Sep 14, 2022 at 05:06:23PM CEST, phil@nwl.cc wrote:
>Unlike with bridges, one can't add an interface to a bond and set it up
>at the same time:
>
>| # ip link set dummy0 down
>| # ip link set dummy0 master bond0 up
>| Error: Device can not be enslaved while up.
>
>Of all drivers with ndo_add_slave callback, bond and team decline if
>IFF_UP flag is set, vrf cycles the interface (i.e., sets it down and
>immediately up again) and the others just don't care.
>
>Support the common notion of setting the interface up after enslaving it
>by sorting the operations accordingly.
>
>Signed-off-by: Phil Sutter <phil@nwl.cc>

Reviewed-by: Jiri Pirko <jiri@nvidia.com>