Re: [PATCH][net-next] net: vrf: switch to synchronize_net() when releasing routes

[Ido Schimmel <idosch@nvidia.com>]

On Tue, Mar 10, 2026 at 06:53:31AM -0400, lirongqing wrote:
> From: Li RongQing <lirongqing@baidu.com>
> 
> Replace synchronize_rcu() with synchronize_net() in vrf_rtable_release()
> and vrf_rt6_release().
> 
> synchronize_net() is RTNL-aware and can use the more efficient
> synchronize_rcu_expedited() when the RTNL lock is held, avoiding
> potentially long wait times for a standard RCU grace period.
> 
> Since these release functions are called during device teardown where
> the RTNL lock is typically held, this change reduces the time spent
> in RTNL critical sections and speeds up VRF device deletion.
> 
> Signed-off-by: Li RongQing <lirongqing@baidu.com>
> ---
>  drivers/net/vrf.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/net/vrf.c b/drivers/net/vrf.c
> index 8c009bc..92a0f4b 100644
> --- a/drivers/net/vrf.c
> +++ b/drivers/net/vrf.c
> @@ -755,7 +755,7 @@ static void vrf_rt6_release(struct net_device *dev, struct net_vrf *vrf)
>  	struct dst_entry *dst;
>  
>  	RCU_INIT_POINTER(vrf->rt6, NULL);
> -	synchronize_rcu();
> +	synchronize_net();

I might be missing something, but why do we need synchronize_rcu() here?
The code below calls dst_release() which will only free the dst entry
after an RCU grace period (assuming we dropped the last reference).

Also, I think that instead of using net->loopback_dev the code can be
converted to use dst_dev_put() before dst_release().

>  
>  	/* move dev in dst's to loopback so this VRF device can be deleted
>  	 * - based on dst_ifdown
> @@ -1002,7 +1002,7 @@ static void vrf_rtable_release(struct net_device *dev, struct net_vrf *vrf)
>  	struct dst_entry *dst;
>  
>  	RCU_INIT_POINTER(vrf->rth, NULL);
> -	synchronize_rcu();
> +	synchronize_net();
>  
>  	/* move dev in dst's to loopback so this VRF device can be deleted
>  	 * - based on dst_ifdown
> -- 
> 2.9.4
>