ath9k: fix a DMA related race condition on reset

When ath_drain_all_txq fails to stop DMA, it issues a hw reset. This reset
happens at a very problematic point in time, when the hardware rx path has
not been stopped yet. This could lead to memory corruption, hardware hangs
or other issues.
To fix these issues, simply remove the reset entirely and check the tx DMA
stop status to prevent problems with fast channel changes.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: John W. Linville <linville@tuxdriver.com>

3 files changed, 10 insertions, 19 deletions

diff --git a/drivers/net/wireless/ath/ath9k/ath9k.h b/drivers/net/wireless/ath/ath9k/ath9k.h
index 0d0bec3628ec..0963071e8f90 100644
--- a/drivers/net/wireless/ath/ath9k/ath9k.h
+++ b/drivers/net/wireless/ath/ath9k/ath9k.h
@@ -329,7 +329,7 @@ int ath_rx_tasklet(struct ath_softc *sc, int flush, bool hp);
 struct ath_txq *ath_txq_setup(struct ath_softc *sc, int qtype, int subtype);
 void ath_tx_cleanupq(struct ath_softc *sc, struct ath_txq *txq);
 int ath_tx_setup(struct ath_softc *sc, int haltype);
-void ath_drain_all_txq(struct ath_softc *sc, bool retry_tx);
+bool ath_drain_all_txq(struct ath_softc *sc, bool retry_tx);
 void ath_draintxq(struct ath_softc *sc,
                      struct ath_txq *txq, bool retry_tx);
 void ath_tx_node_init(struct ath_softc *sc, struct ath_node *an);
diff --git a/drivers/net/wireless/ath/ath9k/main.c b/drivers/net/wireless/ath/ath9k/main.c
index dace215b693e..928ef68ab40c 100644
--- a/drivers/net/wireless/ath/ath9k/main.c
+++ b/drivers/net/wireless/ath/ath9k/main.c
@@ -244,11 +244,12 @@ int ath_set_channel(struct ath_softc *sc, struct ieee80211_hw *hw,
          * the relevant bits of the h/w.
          */
         ath9k_hw_set_interrupts(ah, 0);
-        ath_drain_all_txq(sc, false);
+        stopped = ath_drain_all_txq(sc, false);
 
         spin_lock_bh(&sc->rx.pcu_lock);
 
-        stopped = ath_stoprecv(sc);
+        if (!ath_stoprecv(sc))
+                stopped = false;
 
         /* XXX: do not flush receive queue here. We don't want
          * to flush data frames already in queue because of
diff --git a/drivers/net/wireless/ath/ath9k/xmit.c b/drivers/net/wireless/ath/ath9k/xmit.c
index f2ade2402ce2..aff04789f794 100644
--- a/drivers/net/wireless/ath/ath9k/xmit.c
+++ b/drivers/net/wireless/ath/ath9k/xmit.c
@@ -1120,7 +1120,7 @@ void ath_draintxq(struct ath_softc *sc, struct ath_txq *txq, bool retry_tx)
         }
 }
 
-void ath_drain_all_txq(struct ath_softc *sc, bool retry_tx)
+bool ath_drain_all_txq(struct ath_softc *sc, bool retry_tx)
 {
         struct ath_hw *ah = sc->sc_ah;
         struct ath_common *common = ath9k_hw_common(sc->sc_ah);
@@ -1128,7 +1128,7 @@ void ath_drain_all_txq(struct ath_softc *sc, bool retry_tx)
         int i, npend = 0;
 
         if (sc->sc_flags & SC_OP_INVALID)
-                return;
+                return true;
 
         /* Stop beacon queue */
         ath9k_hw_stoptxdma(sc->sc_ah, sc->beacon.beaconq);
@@ -1142,25 +1142,15 @@ void ath_drain_all_txq(struct ath_softc *sc, bool retry_tx)
                 }
         }
 
-        if (npend) {
-                int r;
-
-                ath_print(common, ATH_DBG_FATAL,
-                          "Failed to stop TX DMA. Resetting hardware!\n");
-
-                spin_lock_bh(&sc->sc_resetlock);
-                r = ath9k_hw_reset(ah, sc->sc_ah->curchan, ah->caldata, false);
-                if (r)
-                        ath_print(common, ATH_DBG_FATAL,
-                                  "Unable to reset hardware; reset status %d\n",
-                                  r);
-                spin_unlock_bh(&sc->sc_resetlock);
-        }
+        if (npend)
+                ath_print(common, ATH_DBG_FATAL, "Failed to stop TX DMA!\n");
 
         for (i = 0; i < ATH9K_NUM_TX_QUEUES; i++) {
                 if (ATH_TXQ_SETUP(sc, i))
                         ath_draintxq(sc, &sc->tx.txq[i], retry_tx);
         }
+
+        return !npend;
 }
 
 void ath_tx_cleanupq(struct ath_softc *sc, struct ath_txq *txq)