Fri, May 27, 2016 at 04:25:29AM CEST, zyjzyj2...@gmail.com wrote:
>Hi, Jiri
>
>I delved into the source code of team. And I found that the function
>team_add_slave actually works in the protection of rtnl lock. So in my
>humble opinion, it is not necessary to use the mutex lock to protect
>team_port_add again.
>
>And I made tests in ubuntu 16.04 server, without the mutex lock, the
>team_add_slave can also work well.
>The steps:
>
>1.
>nmcli con add type team con-name team0 ifname team0 config
>'{"runner":{"name":"activebackup"}}'
>2.
>nmcli con add type team-slave con-name team0-port1 ifname eno16777736
>master team0
>nmcli con add type team-slave con-name team0-port2 ifname eno33554960
>master team0
>
>The above test can work well without the mutex lock. So I made the
>following patch, please comment on this patch.Well, team lock could be replaced by rtnl and removed. Removing just this tiny usage solves nothing. I decided to use team-specific lock to do not overload rtnl at the beginning. > >Thanks a lot. >Zhu Yanjun > >From: Zhu Yanjun<zyjzyj2...@gmail.com> > >The function team_add_slave works in the context of the rtnl lock. >It is not necessary to use the mutex lock since the rtnl lock is >enough. > >Signed-off-by: Zhu Yanjun <zyjzyj2...@gmail.com> >--- > drivers/net/team/team.c | 2 -- > 1 file changed, 2 deletions(-) > >diff --git a/drivers/net/team/team.c b/drivers/net/team/team.c >index a0f64cb..ad84069 100644 >--- a/drivers/net/team/team.c >+++ b/drivers/net/team/team.c >@@ -1936,9 +1936,7 @@ static int team_add_slave(struct net_device *dev, >struct net_device *port_dev) > struct team *team = netdev_priv(dev); > int err; > >- mutex_lock(&team->lock); > err = team_port_add(team, port_dev); >- mutex_unlock(&team->lock); > return err; > } > >-- >1.9.1
