Revert a previous commit - this is causing hardware errors.

I'm not sure why this is failing.  The holding descriptor should be being
re-read when starting DMA of the next frame.  Obviously something here
isn't totally correct.

I'll review the TX queue handling and see if I can figure out why this
is failing.  I'll then re-revert this patch out and use the holding
descriptor again.

1 files changed, 13 insertions, 14 deletions

diff --git a/sys/dev/ath/if_ath.c b/sys/dev/ath/if_ath.c
index f0b5328c9820..a71b3f813556 100644
--- a/sys/dev/ath/if_ath.c
+++ b/sys/dev/ath/if_ath.c
@@ -4000,20 +4000,19 @@ ath_tx_processq(struct ath_softc *sc, struct ath_txq *txq, int dosched)
 			break;
 		}
 		ATH_TXQ_REMOVE(txq, bf, bf_list);
-
-		/*
-		 * Always mark the last buffer in this list as busy.
-		 *
-		 * The hardware may re-read the holding descriptor
-		 * even if we hit the end of the list and try writing
-		 * a new TxDP.
-		 *
-		 * If there's no holding descriptor then this is the
-		 * last buffer in the list of buffers after a fresh
-		 * reset; it'll soon become the holding buffer.
-		 */
-		bf->bf_last->bf_flags |= ATH_BUF_BUSY;
-
+		if (txq->axq_depth > 0) {
+			/*
+			 * More frames follow.  Mark the buffer busy
+			 * so it's not re-used while the hardware may
+			 * still re-read the link field in the descriptor.
+			 *
+			 * Use the last buffer in an aggregate as that
+			 * is where the hardware may be - intermediate
+			 * descriptors won't be "busy".
+			 */
+			bf->bf_last->bf_flags |= ATH_BUF_BUSY;
+		} else
+			txq->axq_link = NULL;
 		if (bf->bf_state.bfs_aggr)
 			txq->axq_aggr_depth--;