ath5k: fix race condition in tx desc processing

As pointed out by Benoit Papillault, there is a potential
race condition between the host and the hardware in reading
the next link in the transmit descriptor list:

cpu0              hw
                  tx for buf completed
                  raise tx_ok interrupt
process buf
buf->ds_link = 0
                  read buf->ds_link

This change checks txdp before processing a descriptor
(if there are any subsequent descriptors) to see if
hardware moved on.  We'll then process this descriptor on
the next tasklet.

Signed-off-by: Bob Copeland <me@bobcopeland.com>
Acked-by: Bruno Randolf <br1@einfach.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

1 files changed, 11 insertions, 0 deletions

diff --git a/drivers/net/wireless/ath/ath5k/base.c b/drivers/net/wireless/ath/ath5k/base.c
index 38c41e3c7988..f7f57c1cca7d 100644
--- a/drivers/net/wireless/ath/ath5k/base.c
+++ b/drivers/net/wireless/ath/ath5k/base.c
@@ -2083,6 +2083,17 @@ ath5k_tx_processq(struct ath5k_softc *sc, struct ath5k_txq *txq)
         list_for_each_entry_safe(bf, bf0, &txq->q, list) {
                 ds = bf->desc;
 
+                /*
+                 * It's possible that the hardware can say the buffer is
+                 * completed when it hasn't yet loaded the ds_link from
+                 * host memory and moved on.  If there are more TX
+                 * descriptors in the queue, wait for TXDP to change
+                 * before processing this one.
+                 */
+                if (ath5k_hw_get_txdp(sc->ah, txq->qnum) == bf->daddr &&
+                    !list_is_last(&bf->list, &txq->q))
+                        break;
+
                 ret = sc->ah->ah_proc_tx_desc(sc->ah, ds, &ts);
                 if (unlikely(ret == -EINPROGRESS))
                         break;