Skip to contentSkip to conversation

ath5k: consistently use rx_bufsize for RX DMA

PATCH In review Bruno Randolf 2 participants 4 messages May 14, 2010 – May 15, 2010 Atheros ATH5K wireless driver

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ath5k: consistently use rx_bufsize for RX DMA

We should use the same buffer size we set up for DMA also in the hardware descriptor. Previously we used common->rx_bufsize for setting up the DMA mapping, but skb_tailroom(skb) for the size we tell to the hardware in the descriptor itself. The problem is that skb_tailroom(skb) can give us a larger value than the size we set up for DMA before: In my case rx_bufsize is 2528, and we allocated an skb of 2559 bytes length, including padding for cache alignment, but sbk_tailroom() was 2592. Just consistently use rx_bufsize for all RX DMA memory sizes.

Also check the return value of setup function.

---

PS: I'm not sure if this is enough to fix bug 15861, but it's definetly one possibility where the hardware could write out of bounds. But the bug report looks too similar to the ath9k one, so i suspect a common bug between ath5k and ath9k. (?)

---
 drivers/net/wireless/ath/ath5k/base.c |    7 ++++---
 1 files changed, 4 insertions(+), 3 deletions(-)
drivers/net/wireless/ath/ath5k/base.c +4 -3
index 9f42af7..5f24e4a 100644
@@ -1217,6 +1217,7 @@ ath5k_rxbuf_setup(struct ath5k_softc *sc, struct ath5k_buf *bf)
 	struct ath5k_hw *ah = sc->ah;
 	struct sk_buff *skb = bf->skb;
 	struct ath5k_desc *ds;
+	int ret;
 
 	if (!skb) {
 		skb = ath5k_rx_skb_alloc(sc, &bf->skbaddr);
@@ -1243,9 +1244,9 @@ ath5k_rxbuf_setup(struct ath5k_softc *sc, struct ath5k_buf *bf)
 	ds = bf->desc;
 	ds->ds_link = bf->daddr;	/* link to self */
 	ds->ds_data = bf->skbaddr;
-	ah->ah_setup_rx_desc(ah, ds,
-		skb_tailroom(skb),	/* buffer size */
-		0);
+	ret = ah->ah_setup_rx_desc(ah, ds, ah->common.rx_bufsize, 0);
+	if (ret)
+		return ret;
 
 	if (sc->rxlink != NULL)
 		*sc->rxlink = bf->daddr;
#1 lore raw
Quoted 9 lines · Bruno Randolf

We should use the same buffer size we set up for DMA also in the hardware descriptor. Previously we used common->rx_bufsize for setting up the DMA mapping, but skb_tailroom(skb) for the size we tell to the hardware in the descriptor itself. The problem is that skb_tailroom(skb) can give us a larger value than the size we set up for DMA before: In my case rx_bufsize is 2528, and we allocated an skb of 2559 bytes length, including padding for cache alignment, but sbk_tailroom() was 2592. Just consistently use rx_bufsize for all RX DMA memory sizes.

Quoted 3 lines

Also check the return value of setup function.

Cc stable?

Is that other bug reproducible, can the user test this to cure it?

Luis

#2 lore raw
Quoted 10 lines · On Saturday 15 May 2010 01:04:34 Luis R. Rodriguez wrote:
Quoted 9 lines · Bruno Randolf

We should use the same buffer size we set up for DMA also in the hardware descriptor. Previously we used common->rx_bufsize for setting up the DMA mapping, but skb_tailroom(skb) for the size we tell to the hardware in the descriptor itself. The problem is that skb_tailroom(skb) can give us a larger value than the size we set up for DMA before: In my case rx_bufsize is 2528, and we allocated an skb of 2559 bytes length, including padding for cache alignment, but sbk_tailroom() was 2592. Just consistently use rx_bufsize for all RX DMA memory sizes.

Quoted 5 lines
Quoted 3 lines

Also check the return value of setup function.

Cc stable?

might be useful. i just would like some review before that.

Quoted 1 line

Is that other bug reproducible, can the user test this to cure it?

not sure. seems like he can with running kismet for a few hours (i'm doing the same over the weekend). i doubt that this is "the" bug though... a) because the bug on ath5k and ath9k are so similar and b) because we see the beginning of packets. rather looks like the next pointer of the descriptor is pointing to somewhere it shouldnt, sometimes? (guessing)

bruno

Reviewed-by
#3 lore raw
Quoted 11 lines · Bruno Randolf
Quoted 10 lines · On Saturday 15 May 2010 01:04:34 Luis R. Rodriguez wrote:
Quoted 9 lines · Bruno Randolf

We should use the same buffer size we set up for DMA also in the hardware descriptor. Previously we used common->rx_bufsize for setting up the DMA mapping, but skb_tailroom(skb) for the size we tell to the hardware in the descriptor itself. The problem is that skb_tailroom(skb) can give us a larger value than the size we set up for DMA before: In my case rx_bufsize is 2528, and we allocated an skb of 2559 bytes length, including padding for cache alignment, but sbk_tailroom() was 2592. Just consistently use rx_bufsize for all RX DMA memory sizes.

Quoted 7 lines
Quoted 5 lines
Quoted 3 lines

Also check the return value of setup function.

Cc stable?

might be useful. i just would like some review before that.

FWIW

Quoted 4 lines
Quoted 1 line

Is that other bug reproducible, can the user test this to cure it?

not sure. seems like he can with running kismet for a few hours (i'm doing the same over the weekend). i doubt that this is "the" bug though...

Oh I'm not assuming it is, but it still looks like a valid stable bug fix to me.

Luis