Skip to contentSkip to conversation

ath9k: Clear DMA addresses when buffers are freed.

PATCH New greearb at candelatech.com 4 participants 6 messages Oct 15, 2010 – Oct 29, 2010 Qualcomm Atheros ATH9K wireless driver

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ath9k: Clear DMA addresses when buffers are freed.

From: Ben Greear <[email protected]>

Previous patches cleared the pointer to the SKB, but did not clear the DMA addresses. This patch moves the clear logic into a method and makes sure that the DMA addresses are cleared as well.

This does not obviously fix any bug, but may make it harder to cause bugs in the future.

---
:100644 100644 0c917e5... 8fba13d... M drivers/net/wireless/ath/ath9k/ath9k.h
:100644 100644 97d471f... cc406f9... M drivers/net/wireless/ath/ath9k/beacon.c
:100644 100644 62294da... 1560280... M drivers/net/wireless/ath/ath9k/main.c
:100644 100644 fd778d2... 5afb46f... M drivers/net/wireless/ath/ath9k/recv.c
:100644 100644 a5e5f27... e86f59c... M drivers/net/wireless/ath/ath9k/xmit.c

 drivers/net/wireless/ath/ath9k/ath9k.h  |    2 ++
 drivers/net/wireless/ath/ath9k/beacon.c |   14 +++++---------
 drivers/net/wireless/ath/ath9k/main.c   |   11 +++++++++++
 drivers/net/wireless/ath/ath9k/recv.c   |   12 ++++--------
 drivers/net/wireless/ath/ath9k/xmit.c   |    6 ++----
 5 files changed, 24 insertions(+), 21 deletions(-)
drivers/net/wireless/ath/ath9k/ath9k.h +2 -0
index 0c917e5..8fba13d 100644
@@ -737,4 +737,6 @@ bool ath_mac80211_start_queue(struct ath_softc *sc, u16 skb_queue);
 void ath_start_rfkill_poll(struct ath_softc *sc);
 extern void ath9k_rfkill_poll_state(struct ieee80211_hw *hw);
 
+void ath_clear_dma_ptrs(struct ath_buf *bf);
+
 #endif /* ATH9K_H */
drivers/net/wireless/ath/ath9k/beacon.c +5 -9
index 97d471f..cc406f9 100644
@@ -139,7 +139,7 @@ static struct ath_buf *ath_beacon_generate(struct ieee80211_hw *hw,
 		dma_unmap_single(sc->dev, bf->bf_buf_addr,
 				 skb->len, DMA_TO_DEVICE);
 		dev_kfree_skb_any(skb);
-		bf->bf_buf_addr = 0;
+		ath_clear_dma_ptrs(bf);
 	}
 
 	/* Get a new beacon from mac80211 */
@@ -167,8 +167,7 @@ static struct ath_buf *ath_beacon_generate(struct ieee80211_hw *hw,
 					 skb->len, DMA_TO_DEVICE);
 	if (unlikely(dma_mapping_error(sc->dev, bf->bf_buf_addr))) {
 		dev_kfree_skb_any(skb);
-		bf->bf_mpdu = NULL;
-		bf->bf_buf_addr = 0;
+		ath_clear_dma_ptrs(bf);
 		ath_print(common, ATH_DBG_FATAL,
 			  "dma_mapping_error on beaconing\n");
 		return NULL;
@@ -256,8 +255,7 @@ int ath_beacon_alloc(struct ath_wiphy *aphy, struct ieee80211_vif *vif)
 		dma_unmap_single(sc->dev, bf->bf_buf_addr,
 				 skb->len, DMA_TO_DEVICE);
 		dev_kfree_skb_any(skb);
-		bf->bf_mpdu = NULL;
-		bf->bf_buf_addr = 0;
+		ath_clear_dma_ptrs(bf);
 	}
 
 	/* NB: the beacon data buffer must be 32-bit aligned. */
@@ -302,8 +300,7 @@ int ath_beacon_alloc(struct ath_wiphy *aphy, struct ieee80211_vif *vif)
 					 skb->len, DMA_TO_DEVICE);
 	if (unlikely(dma_mapping_error(sc->dev, bf->bf_buf_addr))) {
 		dev_kfree_skb_any(skb);
-		bf->bf_mpdu = NULL;
-		bf->bf_buf_addr = 0;
+		ath_clear_dma_ptrs(bf);
 		ath_print(common, ATH_DBG_FATAL,
 			  "dma_mapping_error on beacon alloc\n");
 		return -ENOMEM;
@@ -329,8 +326,7 @@ void ath_beacon_return(struct ath_softc *sc, struct ath_vif *avp)
 			dma_unmap_single(sc->dev, bf->bf_buf_addr,
 					 skb->len, DMA_TO_DEVICE);
 			dev_kfree_skb_any(skb);
-			bf->bf_mpdu = NULL;
-			bf->bf_buf_addr = 0;
+			ath_clear_dma_ptrs(bf);
 		}
 		list_add_tail(&bf->list, &sc->beacon.bbuf);
 
drivers/net/wireless/ath/ath9k/main.c +11 -0
index 62294da..1560280 100644
@@ -213,6 +213,17 @@ static void ath_update_survey_stats(struct ath_softc *sc)
 	ath_update_survey_nf(sc, pos);
 }
 
+void ath_clear_dma_ptrs(struct ath_buf *bf)
+{
+	struct ath_desc *ds = bf->bf_desc;
+	bf->bf_buf_addr = 0;
+	bf->bf_mpdu = NULL;
+	if (ds) {
+		ds->ds_data = 0;
+		ds->ds_vdata = NULL;
+	}
+}
+
 /*
  * Set/change channels.  If the channel is really being changed, it's done
  * by reseting the chip.  To accomplish this we must first cleanup any pending
drivers/net/wireless/ath/ath9k/recv.c +4 -8
index fd778d2..5afb46f 100644
@@ -269,8 +269,7 @@ static int ath_rx_edma_init(struct ath_softc *sc, int nbufs)
 		if (unlikely(dma_mapping_error(sc->dev,
 						bf->bf_buf_addr))) {
 				dev_kfree_skb_any(skb);
-				bf->bf_mpdu = NULL;
-				bf->bf_buf_addr = 0;
+				ath_clear_dma_ptrs(bf);
 				ath_print(common, ATH_DBG_FATAL,
 					"dma_mapping_error() on RX init\n");
 				error = -ENOMEM;
@@ -360,8 +359,7 @@ int ath_rx_init(struct ath_softc *sc, int nbufs)
 			if (unlikely(dma_mapping_error(sc->dev,
 							bf->bf_buf_addr))) {
 				dev_kfree_skb_any(skb);
-				bf->bf_mpdu = NULL;
-				bf->bf_buf_addr = 0;
+				ath_clear_dma_ptrs(bf);
 				ath_print(common, ATH_DBG_FATAL,
 					  "dma_mapping_error() on RX init\n");
 				error = -ENOMEM;
@@ -396,8 +394,7 @@ void ath_rx_cleanup(struct ath_softc *sc)
 						common->rx_bufsize,
 						DMA_FROM_DEVICE);
 				dev_kfree_skb(skb);
-				bf->bf_buf_addr = 0;
-				bf->bf_mpdu = NULL;
+				ath_clear_dma_ptrs(bf);
 			}
 		}
 
@@ -1741,8 +1738,7 @@ int ath_rx_tasklet(struct ath_softc *sc, int flush, bool hp)
 		if (unlikely(dma_mapping_error(sc->dev,
 			  bf->bf_buf_addr))) {
 			dev_kfree_skb_any(requeue_skb);
-			bf->bf_mpdu = NULL;
-			bf->bf_buf_addr = 0;
+			ath_clear_dma_ptrs(bf);
 			ath_print(common, ATH_DBG_FATAL,
 				  "dma_mapping_error() on RX\n");
 			ath_rx_send_to_mac80211(hw, sc, skb, rxs);
drivers/net/wireless/ath/ath9k/xmit.c +2 -4
index a5e5f27..e86f59c 100644
@@ -1644,8 +1644,7 @@ static int ath_tx_setup_buffer(struct ieee80211_hw *hw, struct ath_buf *bf,
 	bf->bf_buf_addr = dma_map_single(sc->dev, skb->data,
 					 skb->len, DMA_TO_DEVICE);
 	if (unlikely(dma_mapping_error(sc->dev, bf->bf_buf_addr))) {
-		bf->bf_mpdu = NULL;
-		bf->bf_buf_addr = 0;
+		ath_clear_dma_ptrs(bf);
 		ath_print(ath9k_hw_common(sc->sc_ah), ATH_DBG_FATAL,
 			  "dma_mapping_error() on TX\n");
 		return -ENOMEM;
@@ -1915,7 +1914,6 @@ static void ath_tx_complete_buf(struct ath_softc *sc, struct ath_buf *bf,
 	}
 
 	dma_unmap_single(sc->dev, bf->bf_buf_addr, skb->len, DMA_TO_DEVICE);
-	bf->bf_buf_addr = 0;
 
 	if (bf->bf_state.bfs_paprd) {
 		if (time_after(jiffies,
@@ -1931,7 +1929,7 @@ static void ath_tx_complete_buf(struct ath_softc *sc, struct ath_buf *bf,
 	/* At this point, skb (bf->bf_mpdu) is consumed...make sure we don't
 	 * accidentally reference it later.
 	 */
-	bf->bf_mpdu = NULL;
+	ath_clear_dma_ptrs(bf);
 
 	/*
 	 * Return the list of ath_buf of this mpdu to free queue
Signature · 1 line
1.7.2.2
author
#1 lore raw

PATCH 1/1 ath9k: Clear DMA addresses when buffers are freed.

Quoted 8 lines · greearb at candelatech.com

From: Ben Greear<[email protected]>

Previous patches cleared the pointer to the SKB, but did not clear the DMA addresses. This patch moves the clear logic into a method and makes sure that the DMA addresses are cleared as well.

This does not obviously fix any bug, but may make it harder to cause bugs in the future.

I wanted to follow up to see if there was any interest in this patch, as it received no comment and has not been accepted upstream.

Thanks,
Ben

Quoted 176 lines

---
:100644 100644 0c917e5... 8fba13d... M drivers/net/wireless/ath/ath9k/ath9k.h
:100644 100644 97d471f... cc406f9... M drivers/net/wireless/ath/ath9k/beacon.c
:100644 100644 62294da... 1560280... M drivers/net/wireless/ath/ath9k/main.c
:100644 100644 fd778d2... 5afb46f... M drivers/net/wireless/ath/ath9k/recv.c
:100644 100644 a5e5f27... e86f59c... M drivers/net/wireless/ath/ath9k/xmit.c

  drivers/net/wireless/ath/ath9k/ath9k.h  |    2 ++
  drivers/net/wireless/ath/ath9k/beacon.c |   14 +++++---------
  drivers/net/wireless/ath/ath9k/main.c   |   11 +++++++++++
  drivers/net/wireless/ath/ath9k/recv.c   |   12 ++++--------
  drivers/net/wireless/ath/ath9k/xmit.c   |    6 ++----
  5 files changed, 24 insertions(+), 21 deletions(-)
drivers/net/wireless/ath/ath9k/ath9k.h +2 -0
index 0c917e5..8fba13d 100644
@@ -737,4 +737,6 @@ bool ath_mac80211_start_queue(struct ath_softc *sc, u16 skb_queue);
  void ath_start_rfkill_poll(struct ath_softc *sc);
  extern void ath9k_rfkill_poll_state(struct ieee80211_hw *hw);
 
+void ath_clear_dma_ptrs(struct ath_buf *bf);
+
  #endif /* ATH9K_H */
drivers/net/wireless/ath/ath9k/beacon.c +5 -9
index 97d471f..cc406f9 100644
@@ -139,7 +139,7 @@ static struct ath_buf *ath_beacon_generate(struct ieee80211_hw *hw,
  		dma_unmap_single(sc->dev, bf->bf_buf_addr,
  				 skb->len, DMA_TO_DEVICE);
  		dev_kfree_skb_any(skb);
-		bf->bf_buf_addr = 0;
+		ath_clear_dma_ptrs(bf);
  	}
 
  	/* Get a new beacon from mac80211 */
@@ -167,8 +167,7 @@ static struct ath_buf *ath_beacon_generate(struct ieee80211_hw *hw,
  					 skb->len, DMA_TO_DEVICE);
  	if (unlikely(dma_mapping_error(sc->dev, bf->bf_buf_addr))) {
  		dev_kfree_skb_any(skb);
-		bf->bf_mpdu = NULL;
-		bf->bf_buf_addr = 0;
+		ath_clear_dma_ptrs(bf);
  		ath_print(common, ATH_DBG_FATAL,
  			  "dma_mapping_error on beaconing\n");
  		return NULL;
@@ -256,8 +255,7 @@ int ath_beacon_alloc(struct ath_wiphy *aphy, struct ieee80211_vif *vif)
  		dma_unmap_single(sc->dev, bf->bf_buf_addr,
  				 skb->len, DMA_TO_DEVICE);
  		dev_kfree_skb_any(skb);
-		bf->bf_mpdu = NULL;
-		bf->bf_buf_addr = 0;
+		ath_clear_dma_ptrs(bf);
  	}
 
  	/* NB: the beacon data buffer must be 32-bit aligned. */
@@ -302,8 +300,7 @@ int ath_beacon_alloc(struct ath_wiphy *aphy, struct ieee80211_vif *vif)
  					 skb->len, DMA_TO_DEVICE);
  	if (unlikely(dma_mapping_error(sc->dev, bf->bf_buf_addr))) {
  		dev_kfree_skb_any(skb);
-		bf->bf_mpdu = NULL;
-		bf->bf_buf_addr = 0;
+		ath_clear_dma_ptrs(bf);
  		ath_print(common, ATH_DBG_FATAL,
  			  "dma_mapping_error on beacon alloc\n");
  		return -ENOMEM;
@@ -329,8 +326,7 @@ void ath_beacon_return(struct ath_softc *sc, struct ath_vif *avp)
  			dma_unmap_single(sc->dev, bf->bf_buf_addr,
  					 skb->len, DMA_TO_DEVICE);
  			dev_kfree_skb_any(skb);
-			bf->bf_mpdu = NULL;
-			bf->bf_buf_addr = 0;
+			ath_clear_dma_ptrs(bf);
  		}
  		list_add_tail(&bf->list,&sc->beacon.bbuf);
 
drivers/net/wireless/ath/ath9k/main.c +11 -0
index 62294da..1560280 100644
@@ -213,6 +213,17 @@ static void ath_update_survey_stats(struct ath_softc *sc)
  	ath_update_survey_nf(sc, pos);
  }
 
+void ath_clear_dma_ptrs(struct ath_buf *bf)
+{
+	struct ath_desc *ds = bf->bf_desc;
+	bf->bf_buf_addr = 0;
+	bf->bf_mpdu = NULL;
+	if (ds) {
+		ds->ds_data = 0;
+		ds->ds_vdata = NULL;
+	}
+}
+
  /*
   * Set/change channels.  If the channel is really being changed, it's done
   * by reseting the chip.  To accomplish this we must first cleanup any pending
drivers/net/wireless/ath/ath9k/recv.c +4 -8
index fd778d2..5afb46f 100644
@@ -269,8 +269,7 @@ static int ath_rx_edma_init(struct ath_softc *sc, int nbufs)
  		if (unlikely(dma_mapping_error(sc->dev,
  						bf->bf_buf_addr))) {
  				dev_kfree_skb_any(skb);
-				bf->bf_mpdu = NULL;
-				bf->bf_buf_addr = 0;
+				ath_clear_dma_ptrs(bf);
  				ath_print(common, ATH_DBG_FATAL,
  					"dma_mapping_error() on RX init\n");
  				error = -ENOMEM;
@@ -360,8 +359,7 @@ int ath_rx_init(struct ath_softc *sc, int nbufs)
  			if (unlikely(dma_mapping_error(sc->dev,
  							bf->bf_buf_addr))) {
  				dev_kfree_skb_any(skb);
-				bf->bf_mpdu = NULL;
-				bf->bf_buf_addr = 0;
+				ath_clear_dma_ptrs(bf);
  				ath_print(common, ATH_DBG_FATAL,
  					  "dma_mapping_error() on RX init\n");
  				error = -ENOMEM;
@@ -396,8 +394,7 @@ void ath_rx_cleanup(struct ath_softc *sc)
  						common->rx_bufsize,
  						DMA_FROM_DEVICE);
  				dev_kfree_skb(skb);
-				bf->bf_buf_addr = 0;
-				bf->bf_mpdu = NULL;
+				ath_clear_dma_ptrs(bf);
  			}
  		}
 
@@ -1741,8 +1738,7 @@ int ath_rx_tasklet(struct ath_softc *sc, int flush, bool hp)
  		if (unlikely(dma_mapping_error(sc->dev,
  			  bf->bf_buf_addr))) {
  			dev_kfree_skb_any(requeue_skb);
-			bf->bf_mpdu = NULL;
-			bf->bf_buf_addr = 0;
+			ath_clear_dma_ptrs(bf);
  			ath_print(common, ATH_DBG_FATAL,
  				  "dma_mapping_error() on RX\n");
  			ath_rx_send_to_mac80211(hw, sc, skb, rxs);
drivers/net/wireless/ath/ath9k/xmit.c +2 -4
index a5e5f27..e86f59c 100644
@@ -1644,8 +1644,7 @@ static int ath_tx_setup_buffer(struct ieee80211_hw *hw, struct ath_buf *bf,
  	bf->bf_buf_addr = dma_map_single(sc->dev, skb->data,
  					 skb->len, DMA_TO_DEVICE);
  	if (unlikely(dma_mapping_error(sc->dev, bf->bf_buf_addr))) {
-		bf->bf_mpdu = NULL;
-		bf->bf_buf_addr = 0;
+		ath_clear_dma_ptrs(bf);
  		ath_print(ath9k_hw_common(sc->sc_ah), ATH_DBG_FATAL,
  			  "dma_mapping_error() on TX\n");
  		return -ENOMEM;
@@ -1915,7 +1914,6 @@ static void ath_tx_complete_buf(struct ath_softc *sc, struct ath_buf *bf,
  	}
 
  	dma_unmap_single(sc->dev, bf->bf_buf_addr, skb->len, DMA_TO_DEVICE);
-	bf->bf_buf_addr = 0;
 
  	if (bf->bf_state.bfs_paprd) {
  		if (time_after(jiffies,
@@ -1931,7 +1929,7 @@ static void ath_tx_complete_buf(struct ath_softc *sc, struct ath_buf *bf,
  	/* At this point, skb (bf->bf_mpdu) is consumed...make sure we don't
  	 * accidentally reference it later.
  	 */
-	bf->bf_mpdu = NULL;
+	ath_clear_dma_ptrs(bf);
 
  	/*
  	 * Return the list of ath_buf of this mpdu to free queue
Signature · 2 lines
Ben Greear <[email protected]>
Candela Technologies Inc  http://www.candelatech.com
#2 lore raw

PATCH 1/1 ath9k: Clear DMA addresses when buffers are freed.

Quoted 7 lines · Ben Greear
Quoted 6 lines · greearb at candelatech.com

From: Ben Greear<[email protected]>

Previous patches cleared the pointer to the SKB, but did not clear the DMA addresses. This patch moves the clear logic into a method and makes sure that the DMA addresses are cleared as well.

Quoted 5 lines
Quoted 2 lines

This does not obviously fix any bug, but may make it harder to cause bugs in the future.

I wanted to follow up to see if there was any interest in this patch, as it received no comment and has not been accepted upstream.

I'd rather not introduce code unless you know
for sure its doing something useful.

Luis

author
#3 lore raw

PATCH 1/1 ath9k: Clear DMA addresses when buffers are freed.

Quoted 8 lines · Luis R. Rodriguez
Quoted 7 lines · Ben Greear
Quoted 6 lines · greearb at candelatech.com

From: Ben Greear<[email protected]>

Previous patches cleared the pointer to the SKB, but did not clear the DMA addresses. This patch moves the clear logic into a method and makes sure that the DMA addresses are cleared as well.

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

This does not obviously fix any bug, but may make it harder to cause bugs in the future.

I wanted to follow up to see if there was any interest in this patch, as it received no comment and has not been accepted upstream.

I'd rather not introduce code unless you know
for sure its doing something useful.

I don't think it makes any real difference, so we can just drop it.

Thanks,
Ben

Quoted 2 lines

Luis

Signature · 2 lines
Ben Greear <[email protected]>
Candela Technologies Inc  http://www.candelatech.com
#4 lore raw

PATCH 1/1 ath9k: Clear DMA addresses when buffers are freed.

In reply to Luis R. Rodriguez
Quoted 7 lines · Luis R. Rodriguez
Quoted 6 lines · Ben Greear
Quoted 6 lines

From: Ben Greear<[email protected]>

Previous patches cleared the pointer to the SKB, but did not clear the DMA addresses. This patch moves the clear logic into a method and makes sure that the DMA addresses are cleared as well.

Quoted 5 lines
Quoted 5 lines
Quoted 2 lines

This does not obviously fix any bug, but may make it harder to cause bugs in the future.

I wanted to follow up to see if there was any interest in this patch, as it received no comment and has not been accepted upstream.

Quoted 2 lines

I'd rather not introduce code unless you know
for sure its doing something useful.

Cleaning up messy and/or redundant code is extremely useful.

//Peter

#5 lore raw

PATCH 1/1 ath9k: Clear DMA addresses when buffers are freed.

In reply to Ben Greear
Quoted 7 lines · Ben Greear
Quoted 6 lines · greearb at candelatech.com

From: Ben Greear<[email protected]>

Previous patches cleared the pointer to the SKB, but did not clear the DMA addresses. This patch moves the clear logic into a method and makes sure that the DMA addresses are cleared as well.

Quoted 8 lines
Quoted 2 lines

This does not obviously fix any bug, but may make it harder to cause bugs in the future.

I wanted to follow up to see if there was any interest in this patch, as it received no comment and has not been accepted upstream.

Thanks,
Ben

The patch seems fine to me, but it isn't a fix and so isn't going anywhere until wireless-next-2.6 opens again (i.e. after the merge window closes).

John

Signature · 2 lines
John W. Linville		Someday the world will need a hero, and you
linville at tuxdriver.com			might be all we have.  Be ready.