Re: [PATCH] rt2x00mac: In error case stop netdev queue, free skb and return NETDEV_TX_OK

[Daniel Wagner <wagi@monom.org>]

Ivo van Doorn wrote:
> On Wednesday 23 July 2008, Daniel Wagner wrote:
>> It is not allowed to use NETDEV_TX_BUSY in tx path anymore.
> 
> If not, then why is mac80211 checking and handling the return value
> and is tx() still a function returning an int. mac80211 is actually requeueing
> the frame when the hardware fails to send it, so why should that be completely blocked?

Well, I might be completely wrong here. I got this idea from following mail on netdev:
http://marc.info/?l=linux-wireless&m=121025252321824&w=2

>> --- a/drivers/net/wireless/rt2x00/rt2x00mac.c
>> +++ b/drivers/net/wireless/rt2x00/rt2x00mac.c
>> @@ -46,7 +46,7 @@ static int rt2x00mac_tx_rts_cts(struct rt2x00_dev *rt2x00dev,
>>  	skb = dev_alloc_skb(size + rt2x00dev->hw->extra_tx_headroom);
>>  	if (!skb) {
>>  		WARNING(rt2x00dev, "Failed to create RTS/CTS frame.\n");
>> -		return NETDEV_TX_BUSY;
>> +		return -1;
> 
> I am kind of missing the point here, this patch seems to come down to:
> 	We can't return TX_BUSY so we return a random other value

Unfortunately yes. The last return in this function is still NETDEV_TX_OK. My idea
was to return 0 on success in error -1. Is this not the expected normal
'typeless' return behavior of a function, no? 

daniel