Re: [PATCH] net: mac80211: tx.c: be sure of 'sdata->vif.type' must be NL80211_IFTYPE_AP when be in NL80211_IFTYPE_AP case

[Chen Gang <gang.chen.5i5j-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>]

On 12/01/2013 07:48 AM, Chen Gang wrote:
> On 12/01/2013 04:39 AM, Joe Perches wrote:
>> On Sat, 2013-11-30 at 21:08 +0100, Johannes Berg wrote:
>>> On Sat, 2013-11-30 at 22:02 +0800, Chen Gang wrote:
>>>>  - fall-through is obvious (although I did not notice it, originally).
>>>>
>>>>  - Check 'A' again just near by "case A" seems a little strange.
>>>>
>>>>  - Some compilers aren't quit smart enough to know 'chanctx_conf' is OK.
>>>
>>> I know. If you have any good ideas of how to make it more obvious to the
>>> compiler, I'm all ears, I just don't like any of the solutions offered
>>> so far (and you aren't the first to do so either) :-)
>>>
> 
> OK, thanks.
> 
>>> FWIW, I find the label to be odd because if you're familiar with the
>>> code then AP/AP_VLAN *should* be identical except for two special things
>>> that are now linearly & neatly handled in the code (the first being the
>>> 4-addr station, the second the chanctx assignment which always has to be
>>> done regardless of 4-addr). IMHO the == check after case should be
>>> enough to make a human reader take a closer look. I understand that you
>>> didn't and that's OK since you were just trying to squelch compile
>>> warnings, but I don't see that this one warrants much attention.
>>
>> The label/test could be moved to save a couple of lines
>> of duplicated code.
>>
> 
> Hmm... for me, it must be an implementation which can satisfy all of us.
> 
> If ieee80211_subif_start_xmit() is not performance sensitive (I guess
> so), we can use some short static functions instead of some code blocks
> within ieee80211_subif_start_xmit().
> 
>  - ieee80211_subif_start_xmit() is a long function (600+ lines).
> 

Oh, one typo issue (need use 400+ instead of 600+), it is a long
function (although not quite long).

>  - use short static function can share some code.
> 
>  - if code can be shared, the work flow can be more clearer too (don't
>    need fall-through or goto).
> 
> 
> And I guess, the implementation below can cause panic when
> "sdata->vif.type == NL80211_IFTYPE_AP_VLAN".  :-(
> 
>> Maybe:
>>
>>  net/mac80211/tx.c | 14 ++++++--------
>>  1 file changed, 6 insertions(+), 8 deletions(-)
>>
>> diff --git a/net/mac80211/tx.c b/net/mac80211/tx.c
>> index 70b5a05..b2160f4 100644
>> --- a/net/mac80211/tx.c
>> +++ b/net/mac80211/tx.c
>> @@ -1777,18 +1777,16 @@ netdev_tx_t ieee80211_subif_start_xmit(struct sk_buff *skb,
>>  		}
>>  		ap_sdata = container_of(sdata->bss, struct ieee80211_sub_if_data,
>>  					u.ap);
>> -		chanctx_conf = rcu_dereference(ap_sdata->vif.chanctx_conf);
>> -		if (!chanctx_conf)
>> -			goto fail_rcu;
>> -		band = chanctx_conf->def.chan->band;
>> -		if (sta)
>> -			break;
>>  		/* fall through */
>>  	case NL80211_IFTYPE_AP:
>> -		if (sdata->vif.type == NL80211_IFTYPE_AP)
>> -			chanctx_conf = rcu_dereference(sdata->vif.chanctx_conf);
>> +		chanctx_conf = rcu_dereference(ap_sdata->vif.chanctx_conf);
>>  		if (!chanctx_conf)
>>  			goto fail_rcu;
>> +		if (sdata->vif.type == NL80211_IFTYPE_AP_VLAN) {
>> +			band = chanctx_conf->def.chan->band;
>> +			if (sta)
>> +				break;
>> +		}
>>  		fc |= cpu_to_le16(IEEE80211_FCTL_FROMDS);
>>  		/* DA BSSID SA */
>>  		memcpy(hdr.addr1, skb->data, ETH_ALEN);
>>
>>
> 
> 


-- 
Chen Gang
--
To unsubscribe from this list: send the line "unsubscribe linux-wireless" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html