On 13/02/14 11:52, Andrew Lunn wrote:
> On Tue, Feb 11, 2014 at 01:48:18PM +0100, Antonio Quartulli wrote:
>> From: Antonio Quartulli <antonio@open-mesh.com>
>>
>> The phydev member of a net_device can be used to get
>> information about an ethernet link like HALF/FULL_DUPLEX
>> and advertised bandwidth (e.g. 100/10Mbps).
>>
>> This information are then stored in the hard_iface object
>> to be used during the metric computation routine.
>>
>> Signed-off-by: Antonio Quartulli <antonio@open-mesh.com>
>> ---
>>  bat_v_elp.c      |  8 ++++++++
>>  bat_v_ogm.c      |  4 ++--
>>  hard-interface.c | 19 +++++++++++++++++++
>>  types.h          | 12 ++++++++++++
>>  4 files changed, 41 insertions(+), 2 deletions(-)
>>
>> diff --git a/bat_v_elp.c b/bat_v_elp.c
>> index 763113e..958843c 100644
>> --- a/bat_v_elp.c
>> +++ b/bat_v_elp.c
>> @@ -78,6 +78,14 @@ batadv_v_elp_get_throughput(struct batadv_elp_neigh_node *neigh)
>>  	if (throughput != 0)
>>  		return throughput;
>>  
>> +	/* In case of Ethernet interface, the throughput has already been
>> +	 * obtained from the phydev object in the net_device struct (see
>> +	 * batadv_hardif_activate_interface()). So return this value.
>> +	 */
>> +	throughput = hard_iface->bat_v.eth_throughput;
>> +	if (throughput != 0)
>> +		return throughput;
>> +
>>  	/* if this is a wireless device, then ask its throughput through
>>  	 * cfg80211 API
>>  	 */
>> diff --git a/bat_v_ogm.c b/bat_v_ogm.c
>> index fa3d1a1..060ce80 100644
>> --- a/bat_v_ogm.c
>> +++ b/bat_v_ogm.c
>> @@ -393,8 +393,8 @@ static uint32_t batadv_v_penalty(struct batadv_priv *bat_priv,
>>  	/* proportion to use the same value used in batman iv (x * 128 / 256) */
>>  	hop_penalty = hop_penalty * 100 / 255;
>>  
>> -	if (batadv_is_wifi_netdev(if_incoming->net_dev) &&
>> -	    metric > link_metric / 10)
>> +	if ((if_incoming->bat_v.flags & BATADV_FULL_DUPLEX) &&
>> +	    (metric > link_metric / 10))
>>  		return metric / 2;
>>  
>>  	return metric * (100 - hop_penalty) / 100;
>> diff --git a/hard-interface.c b/hard-interface.c
>> index 2a04130..58c8669 100644
>> --- a/hard-interface.c
>> +++ b/hard-interface.c
>> @@ -31,6 +31,8 @@
>>  
>>  #include <linux/if_arp.h>
>>  #include <linux/if_ether.h>
>> +#include <linux/netdevice.h>
>> +#include <linux/phy.h>
>>  
>>  void batadv_hardif_free_rcu(struct rcu_head *rcu)
>>  {
>> @@ -297,6 +299,7 @@ void batadv_update_min_mtu(struct net_device *soft_iface)
>>  static void
>>  batadv_hardif_activate_interface(struct batadv_hard_iface *hard_iface)
>>  {
>> +	struct net_device *dev = hard_iface->net_dev;
>>  	struct batadv_priv *bat_priv;
>>  	struct batadv_hard_iface *primary_if = NULL;
>>  
>> @@ -315,6 +318,22 @@ batadv_hardif_activate_interface(struct batadv_hard_iface *hard_iface)
>>  	if (!primary_if)
>>  		batadv_primary_if_select(bat_priv, hard_iface);
>>  
>> +	/* set the default values */
>> +	hard_iface->bat_v.flags &= ~BATADV_FULL_DUPLEX;
>> +	hard_iface->bat_v.eth_throughput = 0;
>> +	if (dev->phydev) {
>> +		if (dev->phydev->duplex == DUPLEX_FULL)
>> +			hard_iface->bat_v.flags |= BATADV_FULL_DUPLEX;
>> +
>> +		/* set the speed according to the phydev setting. Store the
>> +		 * value in Kbps (as done for the other throughput variables)
>> +		 */
>> +		if (dev->phydev->speed != SPEED_UNKNOWN) {
>> +			hard_iface->bat_v.eth_throughput = dev->phydev->speed;
>> +			hard_iface->bat_v.eth_throughput *= 10;
>> +		}
>> +	}
>> +
> 
> Hi Antonio
> 
> If i'm reading this correctly, you get the bandwidth once when the
> interface is added to bat0? 

it is invoked every time the interface is brought up. So if you
disconnect and then reconnect the cable (e.g. to connect the NIC to a
faster switch) this function is invoked again and the bandwidth is updated.

> Are we sure that auto-negotiation has
> finished? It can take a few seconds to complete after the interface is
> brought up. It would not be good to have batman use 10Mbps/Half Duplex
> on my gigabit links...

So it can take up to "few" seconds? I did not expect such a delay!

However this function if invoked when the NETDEV_UP event is issued by
the underlying system. I expect the event to be thrown when the
interface is ready to be used, not when the auto-negotiatin is still
going on. I will double check.

> 
> Is there a notification mechanism which could be used to keep these
> values up to date? Or is that in a patch i have not yet got to.

No there is no notification mechanism.

I see only two ways of changing this value at runtime:

1) NIC disconnected and re-connected to a different "thing" -> NETDEV_UP
is issued, we re-read the values

2) the user manually changes the bandwidth (e.g. using ethtool) -> not
handled right now (I would also not consider it a high priority item),
but maybe the NETDEV_CHANGE event is fired up in this case?


Cheers,

-- 
Antonio Quartulli