Re: [PATCH v9 2/4] ath9k: always call ath_reset from workqueue context

[Felix Fietkau <nbd@openwrt.org>]

On 2011-08-30 6:54 PM, Rajkumar Manoharan wrote:
> On Tue, Aug 30, 2011 at 05:52:02PM +0200, Felix Fietkau wrote:
>>  On 2011-08-30 4:20 PM, Rajkumar Manoharan wrote:
>>  >On Tue, Aug 30, 2011 at 10:30:45AM +0200, Felix Fietkau wrote:
>>  >>  This makes it much easier to add further rework to avoid race conditions
>>  >>  between reset and other work items.
>>  >>  Move other functions to make ath_reset static.
>>  >>
>>  >>  Signed-off-by: Felix Fietkau<nbd@openwrt.org>
>>  >>  ---
>>  >>  diff --git a/drivers/net/wireless/ath/ath9k/xmit.c b/drivers/net/wireless/ath/ath9k/xmit.c
>>  >>  index 49b93c2..2b2a975 100644
>>  >>  --- a/drivers/net/wireless/ath/ath9k/xmit.c
>>  >>  +++ b/drivers/net/wireless/ath/ath9k/xmit.c
>>  >>  @@ -581,8 +581,10 @@ static void ath_tx_complete_aggr(struct ath_softc *sc, struct ath_txq *txq,
>>  >>
>>  >>   	rcu_read_unlock();
>>  >>
>>  >>  -	if (needreset)
>>  >>  -		ath_reset(sc, false);
>>  >>  +	if (needreset) {
>>  >>  +		sc->sc_flags |= SC_OP_HW_RESET;
>>  >>  +		ieee80211_queue_work(sc->hw,&sc->hw_reset_work);
>>  >>  +	}
>>  >>   }
>>  >>
>>  >Still ath_txq_schedule can be scheduled. Return a error code from
>>  >ath_tx_complete_aggr to abort tx process intead of SC_OP_HW_RESET.
>>  >That looks much more generic and simple.
>>  I added this to make sure that there won't be any subsequent
>>  attempts at processing the tx queue until the reset (e.g. if there
>>  was a tx status IRQ in the mean time).
>>
> If so, it would be safer not to re-enable interrupts at the end of ath9k_tasklet
> if SC_OP_HW_RESET is set.
I think tx is the primary issue here, so I only added the check to the 
tx processing functions.

- Felix