diff options
| author | Luis R. Rodriguez <lrodriguez@atheros.com> | 2008-12-03 03:35:30 -0800 |
|---|---|---|
| committer | John W. Linville <linville@tuxdriver.com> | 2008-12-05 09:35:59 -0500 |
| commit | c112d0c5b89037dd618083b5fdf4bb36b0c51d77 (patch) | |
| tree | 195a14b4482ef3506ffa51caa4ab7f6a5b58e003 /drivers/net/wireless/ath9k | |
| parent | f8316df10c4e3bec5b4c3a5a8e026c577640c3a6 (diff) | |
ath9k: Use GFP_ATOMIC when allocating TX private area
Using GFP_KERNEL was wrong and produces a 'scheduling while atomic'
bug as we're in a tasklet. Also, check for proper return values
now, in case allocation fails and be sure to stop the TX queue
in case of memory issues but gaurantee the TX queue will
eventually be woken up.
Signed-off-by: Senthil Balasubramanian <senthilkumar@atheros.com>
Signed-off-by: Sujith <Sujith.Manoharan@atheros.com>
Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
Diffstat (limited to 'drivers/net/wireless/ath9k')
| -rw-r--r-- | drivers/net/wireless/ath9k/xmit.c | 22 |
1 files changed, 20 insertions, 2 deletions
diff --git a/drivers/net/wireless/ath9k/xmit.c b/drivers/net/wireless/ath9k/xmit.c index 17fd05e2f24..9de27c681b8 100644 --- a/drivers/net/wireless/ath9k/xmit.c +++ b/drivers/net/wireless/ath9k/xmit.c @@ -1652,7 +1652,9 @@ static int ath_tx_setup_buffer(struct ath_softc *sc, struct ath_buf *bf, int hdrlen; __le16 fc; - tx_info_priv = kzalloc(sizeof(*tx_info_priv), GFP_KERNEL); + tx_info_priv = kzalloc(sizeof(*tx_info_priv), GFP_ATOMIC); + if (unlikely(!tx_info_priv)) + return -ENOMEM; tx_info->rate_driver_data[0] = tx_info_priv; hdrlen = ieee80211_get_hdrlen_from_skb(skb); fc = hdr->frame_control; @@ -1801,10 +1803,26 @@ int ath_tx_start(struct ath_softc *sc, struct sk_buff *skb, r = ath_tx_setup_buffer(sc, bf, skb, txctl); if (unlikely(r)) { - spin_lock_bh(&sc->sc_txbuflock); + struct ath_txq *txq = txctl->txq; + DPRINTF(sc, ATH_DBG_FATAL, "TX mem alloc failure\n"); + + /* upon ath_tx_processq() this TX queue will be resumed, we + * guarantee this will happen by knowing beforehand that + * we will at least have to run TX completionon one buffer + * on the queue */ + spin_lock_bh(&txq->axq_lock); + if (ath_txq_depth(sc, txq->axq_qnum) > 1) { + ieee80211_stop_queue(sc->hw, + skb_get_queue_mapping(skb)); + txq->stopped = 1; + } + spin_unlock_bh(&txq->axq_lock); + + spin_lock_bh(&sc->sc_txbuflock); list_add_tail(&bf->list, &sc->sc_txbuf); spin_unlock_bh(&sc->sc_txbuflock); + return r; } |