Re: [PATCH v2 net-next] sockopt: Change getsockopt() of SO_BINDTODEVICE to return an interface name

[Brian Haley <brian.haley@hp.com>]

On 11/17/2012 12:06 AM, Pavel Emelyanov wrote:
>> @@ -4165,6 +4180,8 @@ static int dev_ifname(struct net *net, struct ifreq __user
>> *arg)
>>
>>  	strcpy(ifr.ifr_name, dev->name);
>>  	rcu_read_unlock();
>> +	if (read_seqretry(&devnet_rename_seq, seq))
>> +		goto retry;
> 
> I believe it makes sense to make the seqcount protection as a separate patch
> with description of what may happen.

I asked about that before and Dave said he "wanted all the races resolved".  At
best I could make this a series...

>> +retry:
>> +	seq = read_seqbegin(&devnet_rename_seq);
>> +	rcu_read_lock();
>> +	dev = dev_get_by_index_rcu(net, sk->sk_bound_dev_if);
> 
> The sk->sk_bound_dev_if might have changed to 0 while we did read_seqretry (or
> did the len check above, but the race window is smaller) and this code will
> report -ENODEV instead of zero lenght.

If there are two threads twiddling with the same socket like this the
application is broken in my mind.

-Brian