[PATCH net-next] net: rmnet: set NETIF_F_LLTX flag

[PATCH net-next] net: rmnet: set NETIF_F_LLTX flag

[Taehee Yoo]

The rmnet_vnd_setup(), which is the callback of ->ndo_start_xmit() is
allowed to call concurrently because it uses RCU protected data.
So, it doesn't need tx lock.

Signed-off-by: Taehee Yoo <ap420073@gmail.com>
---
 drivers/net/ethernet/qualcomm/rmnet/rmnet_vnd.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/net/ethernet/qualcomm/rmnet/rmnet_vnd.c b/drivers/net/ethernet/qualcomm/rmnet/rmnet_vnd.c
index d7c52e398e4a..d58b51d277f1 100644
--- a/drivers/net/ethernet/qualcomm/rmnet/rmnet_vnd.c
+++ b/drivers/net/ethernet/qualcomm/rmnet/rmnet_vnd.c
@@ -212,6 +212,8 @@ void rmnet_vnd_setup(struct net_device *rmnet_dev)
 	rmnet_dev->needs_free_netdev = true;
 	rmnet_dev->ethtool_ops = &rmnet_ethtool_ops;
 
+	rmnet_dev->features |= NETIF_F_LLTX;
+
 	/* This perm addr will be used as interface identifier by IPv6 */
 	rmnet_dev->addr_assign_type = NET_ADDR_RANDOM;
 	eth_random_addr(rmnet_dev->perm_addr);
-- 
2.17.1

Re: [PATCH net-next] net: rmnet: set NETIF_F_LLTX flag

[subashab]

On 2020-03-08 07:47, Taehee Yoo wrote:
> The rmnet_vnd_setup(), which is the callback of ->ndo_start_xmit() is
> allowed to call concurrently because it uses RCU protected data.
> So, it doesn't need tx lock.
> 
> Signed-off-by: Taehee Yoo <ap420073@gmail.com>
> ---
>  drivers/net/ethernet/qualcomm/rmnet/rmnet_vnd.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/drivers/net/ethernet/qualcomm/rmnet/rmnet_vnd.c
> b/drivers/net/ethernet/qualcomm/rmnet/rmnet_vnd.c
> index d7c52e398e4a..d58b51d277f1 100644
> --- a/drivers/net/ethernet/qualcomm/rmnet/rmnet_vnd.c
> +++ b/drivers/net/ethernet/qualcomm/rmnet/rmnet_vnd.c
> @@ -212,6 +212,8 @@ void rmnet_vnd_setup(struct net_device *rmnet_dev)
>  	rmnet_dev->needs_free_netdev = true;
>  	rmnet_dev->ethtool_ops = &rmnet_ethtool_ops;
> 
> +	rmnet_dev->features |= NETIF_F_LLTX;
> +
>  	/* This perm addr will be used as interface identifier by IPv6 */
>  	rmnet_dev->addr_assign_type = NET_ADDR_RANDOM;
>  	eth_random_addr(rmnet_dev->perm_addr);

Hi Taehee

It seems the flag is deprecated per documentation.

https://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next.git/tree/Documentation/networking/netdevices.txt#n73

Re: [PATCH net-next] net: rmnet: set NETIF_F_LLTX flag

[Eric Dumazet]

On 3/8/20 5:28 PM, subashab@codeaurora.org wrote:
> On 2020-03-08 07:47, Taehee Yoo wrote:
>> The rmnet_vnd_setup(), which is the callback of ->ndo_start_xmit() is
>> allowed to call concurrently because it uses RCU protected data.
>> So, it doesn't need tx lock.
>>
>> Signed-off-by: Taehee Yoo <ap420073@gmail.com>
>> ---
>>  drivers/net/ethernet/qualcomm/rmnet/rmnet_vnd.c | 2 ++
>>  1 file changed, 2 insertions(+)
>>
>> diff --git a/drivers/net/ethernet/qualcomm/rmnet/rmnet_vnd.c
>> b/drivers/net/ethernet/qualcomm/rmnet/rmnet_vnd.c
>> index d7c52e398e4a..d58b51d277f1 100644
>> --- a/drivers/net/ethernet/qualcomm/rmnet/rmnet_vnd.c
>> +++ b/drivers/net/ethernet/qualcomm/rmnet/rmnet_vnd.c
>> @@ -212,6 +212,8 @@ void rmnet_vnd_setup(struct net_device *rmnet_dev)
>>      rmnet_dev->needs_free_netdev = true;
>>      rmnet_dev->ethtool_ops = &rmnet_ethtool_ops;
>>
>> +    rmnet_dev->features |= NETIF_F_LLTX;
>> +
>>      /* This perm addr will be used as interface identifier by IPv6 */
>>      rmnet_dev->addr_assign_type = NET_ADDR_RANDOM;
>>      eth_random_addr(rmnet_dev->perm_addr);
> 
> Hi Taehee
> 
> It seems the flag is deprecated per documentation.
> 
> https://git.kernel.org/pub/scm/linux/kernel/git/netdev/net-next.git/tree/Documentation/networking/netdevices.txt#n73


This does not count for virtual drivers, like macvlan, bonding, tunnels...

If your ndo_start_xmit() calls dev_queue_xmit() after switching skb->dev to another device,
you do not need a lock.

Re: [PATCH net-next] net: rmnet: set NETIF_F_LLTX flag

[David Miller]

From: Taehee Yoo <ap420073@gmail.com>
Date: Sun,  8 Mar 2020 13:47:06 +0000

> The rmnet_vnd_setup(), which is the callback of ->ndo_start_xmit() is
> allowed to call concurrently because it uses RCU protected data.
> So, it doesn't need tx lock.
> 
> Signed-off-by: Taehee Yoo <ap420073@gmail.com>

Applied, thanks.