From mboxrd@z Thu Jan 1 00:00:00 1970 Return-path: Received: from mail30s.wh2.ocn.ne.jp ([125.206.180.198]:10356 "HELO mail30s.wh2.ocn.ne.jp" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with SMTP id S1754700Ab0IQCgq (ORCPT ); Thu, 16 Sep 2010 22:36:46 -0400 Received: from vs3000.wh2.ocn.ne.jp (125.206.180.163) by mail30s.wh2.ocn.ne.jp (RS ver 1.0.95vs) with SMTP id 3-0686641005 for ; Fri, 17 Sep 2010 11:36:45 +0900 (JST) Subject: [PATCH 09/11] ath5k: Keep last descriptor in queue To: linville@tuxdriver.com From: Bruno Randolf Cc: ath5k-devel@lists.ath5k.org, linux-wireless@vger.kernel.org Date: Fri, 17 Sep 2010 11:37:07 +0900 Message-ID: <20100917023707.24997.98021.stgit@tt-desk> In-Reply-To: <20100917023543.24997.48466.stgit@tt-desk> References: <20100917023543.24997.48466.stgit@tt-desk> MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" Sender: linux-wireless-owner@vger.kernel.org List-ID: If we return a TX descriptor to the pool of available descriptors, while a queues TXDP still points to it we could potentially run into all sorts of troube. It has been suggested that there is hardware which can set the descriptors done bit before it reads ds_link and moves on to the next descriptor. While the documentation says this is not true for newer chipsets (the descriptor contents are copied to some internal memory), we don't know about older hardware. To be safe, we always keep the last descriptor in the queue, and avoid dangling TXDP pointers. Unfortunately this does not fully resolve the problem - queues still get stuck! This is similar to what ath9k does. Signed-off-by: Bruno Randolf --- drivers/net/wireless/ath/ath5k/base.c | 64 +++++++++++++++++---------------- 1 files changed, 32 insertions(+), 32 deletions(-) diff --git a/drivers/net/wireless/ath/ath5k/base.c b/drivers/net/wireless/ath/ath5k/base.c index c4f0786..9b67cee 100644 --- a/drivers/net/wireless/ath/ath5k/base.c +++ b/drivers/net/wireless/ath/ath5k/base.c @@ -1590,44 +1590,44 @@ ath5k_tx_processq(struct ath5k_softc *sc, struct ath5k_txq *txq) spin_lock(&txq->lock); list_for_each_entry_safe(bf, bf0, &txq->q, list) { - ds = bf->desc; + + txq->txq_poll_mark = false; + + /* skb might already have been processed last time. */ + if (bf->skb != NULL) { + ds = bf->desc; + + ret = sc->ah->ah_proc_tx_desc(sc->ah, ds, &ts); + if (unlikely(ret == -EINPROGRESS)) + break; + else if (unlikely(ret)) { + ATH5K_ERR(sc, + "error %d while processing " + "queue %u\n", ret, txq->qnum); + break; + } + + skb = bf->skb; + bf->skb = NULL; + pci_unmap_single(sc->pdev, bf->skbaddr, skb->len, + PCI_DMA_TODEVICE); + ath5k_tx_frame_completed(sc, skb, &ts); + } /* * It's possible that the hardware can say the buffer is * completed when it hasn't yet loaded the ds_link from - * host memory and moved on. If there are more TX - * descriptors in the queue, wait for TXDP to change - * before processing this one. + * host memory and moved on. + * Always keep the last descriptor to avoid HW races... */ - if (ath5k_hw_get_txdp(sc->ah, txq->qnum) == bf->daddr && - !list_is_last(&bf->list, &txq->q)) - break; - ret = sc->ah->ah_proc_tx_desc(sc->ah, ds, &ts); - if (unlikely(ret == -EINPROGRESS)) - break; - else if (unlikely(ret)) { - ATH5K_ERR(sc, "error %d while processing queue %u\n", - ret, txq->qnum); - break; + if (ath5k_hw_get_txdp(sc->ah, txq->qnum) != bf->daddr) { + spin_lock(&sc->txbuflock); + list_move_tail(&bf->list, &sc->txbuf); + sc->txbuf_len++; + txq->txq_len--; + spin_unlock(&sc->txbuflock); } - - skb = bf->skb; - bf->skb = NULL; - pci_unmap_single(sc->pdev, bf->skbaddr, skb->len, - PCI_DMA_TODEVICE); - - ath5k_tx_frame_completed(sc, skb, &ts); - - spin_lock(&sc->txbuflock); - list_move_tail(&bf->list, &sc->txbuf); - sc->txbuf_len++; - txq->txq_len--; - spin_unlock(&sc->txbuflock); - - txq->txq_poll_mark = false; } - if (likely(list_empty(&txq->q))) - txq->link = NULL; spin_unlock(&txq->lock); if (txq->txq_len < ATH5K_TXQ_LEN_LOW) ieee80211_wake_queue(sc->hw, txq->qnum); @@ -2192,7 +2192,7 @@ ath5k_tx_complete_poll_work(struct work_struct *work) if (sc->txqs[i].setup) { txq = &sc->txqs[i]; spin_lock_bh(&txq->lock); - if (txq->txq_len > 0) { + if (txq->txq_len > 1) { if (txq->txq_poll_mark) { ATH5K_DBG(sc, ATH5K_DEBUG_XMIT, "TX queue stuck %d\n",