summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorFelix Fietkau <nbd@nbd.name>2016-08-02 12:21:27 +0200
committerFelix Fietkau <nbd@nbd.name>2016-08-04 23:10:41 +0200
commit4701fd3190c8da52b5683beb8a609d0abdd5f8e4 (patch)
tree53236611b39a5586a123418b44c2d5b2455a738c
parent1b9dbb85324a9e179e2b332d70c9222657a253ac (diff)
downloadmtk-20170518-4701fd3190c8da52b5683beb8a609d0abdd5f8e4.zip
mtk-20170518-4701fd3190c8da52b5683beb8a609d0abdd5f8e4.tar.gz
mtk-20170518-4701fd3190c8da52b5683beb8a609d0abdd5f8e4.tar.bz2
ath9k: improve performance in tx status handling
Signed-off-by: Felix Fietkau <nbd@nbd.name>
-rw-r--r--package/kernel/mac80211/patches/335-ath9k-use-ieee80211_tx_status_noskb-where-possible.patch305
1 files changed, 305 insertions, 0 deletions
diff --git a/package/kernel/mac80211/patches/335-ath9k-use-ieee80211_tx_status_noskb-where-possible.patch b/package/kernel/mac80211/patches/335-ath9k-use-ieee80211_tx_status_noskb-where-possible.patch
new file mode 100644
index 0000000..dbb5b90
--- /dev/null
+++ b/package/kernel/mac80211/patches/335-ath9k-use-ieee80211_tx_status_noskb-where-possible.patch
@@ -0,0 +1,305 @@
+From: Felix Fietkau <nbd@nbd.name>
+Date: Tue, 2 Aug 2016 12:12:18 +0200
+Subject: [PATCH] ath9k: use ieee80211_tx_status_noskb where possible
+
+It removes the need for undoing the padding changes to skb->data and it
+improves performance by eliminating one tx status lookup per MPDU in the
+status path. It is also useful for preparing a follow-up fix to better
+handle powersave filtering.
+
+Signed-off-by: Felix Fietkau <nbd@nbd.name>
+---
+
+--- a/drivers/net/wireless/ath/ath9k/xmit.c
++++ b/drivers/net/wireless/ath/ath9k/xmit.c
+@@ -50,9 +50,11 @@ static u16 bits_per_symbol[][2] = {
+ static void ath_tx_send_normal(struct ath_softc *sc, struct ath_txq *txq,
+ struct ath_atx_tid *tid, struct sk_buff *skb);
+ static void ath_tx_complete(struct ath_softc *sc, struct sk_buff *skb,
+- int tx_flags, struct ath_txq *txq);
++ int tx_flags, struct ath_txq *txq,
++ struct ieee80211_sta *sta);
+ static void ath_tx_complete_buf(struct ath_softc *sc, struct ath_buf *bf,
+ struct ath_txq *txq, struct list_head *bf_q,
++ struct ieee80211_sta *sta,
+ struct ath_tx_status *ts, int txok);
+ static void ath_tx_txqaddbuf(struct ath_softc *sc, struct ath_txq *txq,
+ struct list_head *head, bool internal);
+@@ -77,6 +79,22 @@ enum {
+ /* Aggregation logic */
+ /*********************/
+
++static void ath_tx_status(struct ieee80211_hw *hw, struct sk_buff *skb)
++{
++ struct ieee80211_tx_info *info = IEEE80211_SKB_CB(skb);
++ struct ieee80211_sta *sta = info->status.status_driver_data[0];
++
++ if (info->flags & IEEE80211_TX_CTL_REQ_TX_STATUS) {
++ ieee80211_tx_status(hw, skb);
++ return;
++ }
++
++ if (sta)
++ ieee80211_tx_status_noskb(hw, sta, info);
++
++ dev_kfree_skb(skb);
++}
++
+ void ath_txq_lock(struct ath_softc *sc, struct ath_txq *txq)
+ __acquires(&txq->axq_lock)
+ {
+@@ -92,6 +110,7 @@ void ath_txq_unlock(struct ath_softc *sc
+ void ath_txq_unlock_complete(struct ath_softc *sc, struct ath_txq *txq)
+ __releases(&txq->axq_lock)
+ {
++ struct ieee80211_hw *hw = sc->hw;
+ struct sk_buff_head q;
+ struct sk_buff *skb;
+
+@@ -100,7 +119,7 @@ void ath_txq_unlock_complete(struct ath_
+ spin_unlock_bh(&txq->axq_lock);
+
+ while ((skb = __skb_dequeue(&q)))
+- ieee80211_tx_status(sc->hw, skb);
++ ath_tx_status(hw, skb);
+ }
+
+ static void ath_tx_queue_tid(struct ath_softc *sc, struct ath_txq *txq,
+@@ -268,7 +287,7 @@ static void ath_tx_flush_tid(struct ath_
+ }
+
+ list_add_tail(&bf->list, &bf_head);
+- ath_tx_complete_buf(sc, bf, txq, &bf_head, &ts, 0);
++ ath_tx_complete_buf(sc, bf, txq, &bf_head, NULL, &ts, 0);
+ }
+
+ if (sendbar) {
+@@ -333,12 +352,12 @@ static void ath_tid_drain(struct ath_sof
+ bf = fi->bf;
+
+ if (!bf) {
+- ath_tx_complete(sc, skb, ATH_TX_ERROR, txq);
++ ath_tx_complete(sc, skb, ATH_TX_ERROR, txq, NULL);
+ continue;
+ }
+
+ list_add_tail(&bf->list, &bf_head);
+- ath_tx_complete_buf(sc, bf, txq, &bf_head, &ts, 0);
++ ath_tx_complete_buf(sc, bf, txq, &bf_head, NULL, &ts, 0);
+ }
+ }
+
+@@ -441,12 +460,11 @@ static void ath_tx_count_frames(struct a
+
+ static void ath_tx_complete_aggr(struct ath_softc *sc, struct ath_txq *txq,
+ struct ath_buf *bf, struct list_head *bf_q,
++ struct ieee80211_sta *sta,
+ struct ath_tx_status *ts, int txok)
+ {
+ struct ath_node *an = NULL;
+ struct sk_buff *skb;
+- struct ieee80211_sta *sta;
+- struct ieee80211_hw *hw = sc->hw;
+ struct ieee80211_hdr *hdr;
+ struct ieee80211_tx_info *tx_info;
+ struct ath_atx_tid *tid = NULL;
+@@ -475,12 +493,7 @@ static void ath_tx_complete_aggr(struct
+ for (i = 0; i < ts->ts_rateindex; i++)
+ retries += rates[i].count;
+
+- rcu_read_lock();
+-
+- sta = ieee80211_find_sta_by_ifaddr(hw, hdr->addr1, hdr->addr2);
+ if (!sta) {
+- rcu_read_unlock();
+-
+ INIT_LIST_HEAD(&bf_head);
+ while (bf) {
+ bf_next = bf->bf_next;
+@@ -488,7 +501,7 @@ static void ath_tx_complete_aggr(struct
+ if (!bf->bf_state.stale || bf_next != NULL)
+ list_move_tail(&bf->list, &bf_head);
+
+- ath_tx_complete_buf(sc, bf, txq, &bf_head, ts, 0);
++ ath_tx_complete_buf(sc, bf, txq, &bf_head, NULL, ts, 0);
+
+ bf = bf_next;
+ }
+@@ -598,7 +611,7 @@ static void ath_tx_complete_aggr(struct
+ ts);
+ }
+
+- ath_tx_complete_buf(sc, bf, txq, &bf_head, ts,
++ ath_tx_complete_buf(sc, bf, txq, &bf_head, sta, ts,
+ !txfail);
+ } else {
+ if (tx_info->flags & IEEE80211_TX_STATUS_EOSP) {
+@@ -619,7 +632,8 @@ static void ath_tx_complete_aggr(struct
+ ath_tx_update_baw(sc, tid, seqno);
+
+ ath_tx_complete_buf(sc, bf, txq,
+- &bf_head, ts, 0);
++ &bf_head, NULL, ts,
++ 0);
+ bar_index = max_t(int, bar_index,
+ ATH_BA_INDEX(seq_first, seqno));
+ break;
+@@ -663,8 +677,6 @@ static void ath_tx_complete_aggr(struct
+ ath_txq_lock(sc, txq);
+ }
+
+- rcu_read_unlock();
+-
+ if (needreset)
+ ath9k_queue_reset(sc, RESET_TYPE_TX_ERROR);
+ }
+@@ -679,7 +691,10 @@ static void ath_tx_process_buffer(struct
+ struct ath_tx_status *ts, struct ath_buf *bf,
+ struct list_head *bf_head)
+ {
++ struct ieee80211_hw *hw = sc->hw;
+ struct ieee80211_tx_info *info;
++ struct ieee80211_sta *sta;
++ struct ieee80211_hdr *hdr;
+ bool txok, flush;
+
+ txok = !(ts->ts_status & ATH9K_TXERR_MASK);
+@@ -692,6 +707,10 @@ static void ath_tx_process_buffer(struct
+
+ ts->duration = ath9k_hw_get_duration(sc->sc_ah, bf->bf_desc,
+ ts->ts_rateindex);
++
++ hdr = (struct ieee80211_hdr *) bf->bf_mpdu->data;
++ sta = ieee80211_find_sta_by_ifaddr(hw, hdr->addr1, hdr->addr2);
++
+ if (!bf_isampdu(bf)) {
+ if (!flush) {
+ info = IEEE80211_SKB_CB(bf->bf_mpdu);
+@@ -700,9 +719,9 @@ static void ath_tx_process_buffer(struct
+ ath_tx_rc_status(sc, bf, ts, 1, txok ? 0 : 1, txok);
+ ath_dynack_sample_tx_ts(sc->sc_ah, bf->bf_mpdu, ts);
+ }
+- ath_tx_complete_buf(sc, bf, txq, bf_head, ts, txok);
++ ath_tx_complete_buf(sc, bf, txq, bf_head, sta, ts, txok);
+ } else
+- ath_tx_complete_aggr(sc, txq, bf, bf_head, ts, txok);
++ ath_tx_complete_aggr(sc, txq, bf, bf_head, sta, ts, txok);
+
+ if (!flush)
+ ath_txq_schedule(sc, txq);
+@@ -938,7 +957,7 @@ ath_tx_get_tid_subframe(struct ath_softc
+ list_add(&bf->list, &bf_head);
+ __skb_unlink(skb, *q);
+ ath_tx_update_baw(sc, tid, seqno);
+- ath_tx_complete_buf(sc, bf, txq, &bf_head, &ts, 0);
++ ath_tx_complete_buf(sc, bf, txq, &bf_head, NULL, &ts, 0);
+ continue;
+ }
+
+@@ -1847,6 +1866,7 @@ static void ath_drain_txq_list(struct at
+ */
+ void ath_draintxq(struct ath_softc *sc, struct ath_txq *txq)
+ {
++ rcu_read_lock();
+ ath_txq_lock(sc, txq);
+
+ if (sc->sc_ah->caps.hw_caps & ATH9K_HW_CAP_EDMA) {
+@@ -1865,6 +1885,7 @@ void ath_draintxq(struct ath_softc *sc,
+ ath_drain_txq_list(sc, txq, &txq->axq_q);
+
+ ath_txq_unlock_complete(sc, txq);
++ rcu_read_unlock();
+ }
+
+ bool ath_drain_all_txq(struct ath_softc *sc)
+@@ -2487,7 +2508,8 @@ void ath_tx_cabq(struct ieee80211_hw *hw
+ /*****************/
+
+ static void ath_tx_complete(struct ath_softc *sc, struct sk_buff *skb,
+- int tx_flags, struct ath_txq *txq)
++ int tx_flags, struct ath_txq *txq,
++ struct ieee80211_sta *sta)
+ {
+ struct ieee80211_tx_info *tx_info = IEEE80211_SKB_CB(skb);
+ struct ath_common *common = ath9k_hw_common(sc->sc_ah);
+@@ -2507,15 +2529,17 @@ static void ath_tx_complete(struct ath_s
+ tx_info->flags |= IEEE80211_TX_STAT_ACK;
+ }
+
+- padpos = ieee80211_hdrlen(hdr->frame_control);
+- padsize = padpos & 3;
+- if (padsize && skb->len>padpos+padsize) {
+- /*
+- * Remove MAC header padding before giving the frame back to
+- * mac80211.
+- */
+- memmove(skb->data + padsize, skb->data, padpos);
+- skb_pull(skb, padsize);
++ if (tx_info->flags & IEEE80211_TX_CTL_REQ_TX_STATUS) {
++ padpos = ieee80211_hdrlen(hdr->frame_control);
++ padsize = padpos & 3;
++ if (padsize && skb->len>padpos+padsize) {
++ /*
++ * Remove MAC header padding before giving the frame back to
++ * mac80211.
++ */
++ memmove(skb->data + padsize, skb->data, padpos);
++ skb_pull(skb, padsize);
++ }
+ }
+
+ spin_lock_irqsave(&sc->sc_pm_lock, flags);
+@@ -2530,12 +2554,14 @@ static void ath_tx_complete(struct ath_s
+ }
+ spin_unlock_irqrestore(&sc->sc_pm_lock, flags);
+
+- __skb_queue_tail(&txq->complete_q, skb);
+ ath_txq_skb_done(sc, txq, skb);
++ tx_info->status.status_driver_data[0] = sta;
++ __skb_queue_tail(&txq->complete_q, skb);
+ }
+
+ static void ath_tx_complete_buf(struct ath_softc *sc, struct ath_buf *bf,
+ struct ath_txq *txq, struct list_head *bf_q,
++ struct ieee80211_sta *sta,
+ struct ath_tx_status *ts, int txok)
+ {
+ struct sk_buff *skb = bf->bf_mpdu;
+@@ -2563,7 +2589,7 @@ static void ath_tx_complete_buf(struct a
+ complete(&sc->paprd_complete);
+ } else {
+ ath_debug_stat_tx(sc, bf, ts, txq, tx_flags);
+- ath_tx_complete(sc, skb, tx_flags, txq);
++ ath_tx_complete(sc, skb, tx_flags, txq, sta);
+ }
+ skip_tx_complete:
+ /* At this point, skb (bf->bf_mpdu) is consumed...make sure we don't
+@@ -2715,10 +2741,12 @@ void ath_tx_tasklet(struct ath_softc *sc
+ u32 qcumask = ((1 << ATH9K_NUM_TX_QUEUES) - 1) & ah->intr_txqs;
+ int i;
+
++ rcu_read_lock();
+ for (i = 0; i < ATH9K_NUM_TX_QUEUES; i++) {
+ if (ATH_TXQ_SETUP(sc, i) && (qcumask & (1 << i)))
+ ath_tx_processq(sc, &sc->tx.txq[i]);
+ }
++ rcu_read_unlock();
+ }
+
+ void ath_tx_edma_tasklet(struct ath_softc *sc)
+@@ -2732,6 +2760,7 @@ void ath_tx_edma_tasklet(struct ath_soft
+ struct list_head *fifo_list;
+ int status;
+
++ rcu_read_lock();
+ for (;;) {
+ if (test_bit(ATH_OP_HW_RESET, &common->op_flags))
+ break;
+@@ -2802,6 +2831,7 @@ void ath_tx_edma_tasklet(struct ath_soft
+ ath_tx_process_buffer(sc, txq, &ts, bf, &bf_head);
+ ath_txq_unlock_complete(sc, txq);
+ }
++ rcu_read_unlock();
+ }
+
+ /*****************/