Re: [PATCH] net/bonding: Send more than one gratuitous ARP when slave takes over

[Jay Vosburgh <fubar@us.ibm.com>]

Waskiewicz Jr, Peter P <peter.p.waskiewicz.jr@intel.com> wrote:

>> @@ -104,6 +105,8 @@ struct bond_params bonding_defaults;
>>  
>> module_param(max_bonds, int, 0);
>> MODULE_PARM_DESC(max_bonds, "Max number of bonded devices");
>> +module_param(num_grat_arp, int, 0644);
>> +MODULE_PARM_DESC(num_grat_arp, "Number of gratuitous ARP packet to
>sned on failover");
>
>Check your spelling here.  s/sned/send/
>
>Also, num_grat_arp is declared int, where send_grat_arp is s8.  What
>happens if you overflow send_grat_arp?  Either your module parameter
>needs to change to match what's in the bonding struct, or you needs some
>bounds checking to validate your inputs.

	I'd vote for bounds checking, myself.  I think a legal range of
0 - 255 is appropriate.

>> @@ -1109,14 +1112,16 @@ void bond_change_active_slave(struct bon
>> 		if (new_active && bond->params.fail_over_mac)
>> 			memcpy(bond->dev->dev_addr,
>new_active->dev->dev_addr,
>> 				new_active->dev->addr_len);
>> +		bond->send_grat_arp = num_grat_arp;
>> 		if (bond->curr_active_slave &&
>>  			test_bit(__LINK_STATE_LINKWATCH_PENDING,
>> -
>&bond->curr_active_slave->dev->state)) {
>> +
>&bond->curr_active_slave->dev->state))
>> 			dprintk("delaying gratuitous arp on %s\n",
>> 				bond->curr_active_slave->dev->name);
>> -			bond->send_grat_arp = 1;
>> -		} else
>> +		else {
>> 			bond_send_gratuitous_arp(bond);
>> +			bond->send_grat_arp--;
>> +		}
>
>Don't modify the if () else () formatting here.  Kernel coding
>guidelines read if any part of your if/else statement requires braces,
>all sections need braces.  So it still should be, with your changes:

	Agreed.

[...]
>> @@ -2144,7 +2149,7 @@ static int __bond_mii_monitor(struct bon
>>  			dprintk("sending delayed gratuitous arp on on
>%s\n",
>>  				bond->curr_active_slave->dev->name);
>>  			bond_send_gratuitous_arp(bond);
>> -			bond->send_grat_arp = 0;
>> +			bond->send_grat_arp--;
>
>Can this ever cause send_grat_arp to become less than zero?  What will
>happen if it does drop below zero?  I think some bounds checking might
>be needed here.

	I don't believe that send_grat_arp can go negative, because this
block is entered on an "if (bond->send_grat_arp)" that's just above the
context in the patch.

	The patch as a whole also needs a sysfs entry to permit
modification and inspection of the num_grat_arp parameter, as is done
for other parameters.

	-J

---
	-Jay Vosburgh, IBM Linux Technology Center, fubar@us.ibm.com