Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932066AbbDUPdt (ORCPT ); Tue, 21 Apr 2015 11:33:49 -0400 Received: from youngberry.canonical.com ([91.189.89.112]:57319 "EHLO youngberry.canonical.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755921AbbDUPdf (ORCPT ); Tue, 21 Apr 2015 11:33:35 -0400 From: Luis Henriques To: linux-kernel@vger.kernel.org, stable@vger.kernel.org, kernel-team@lists.ubuntu.com Cc: "Ahmed S. Darwish" , Marc Kleine-Budde , Luis Henriques Subject: [PATCH 3.16.y-ckt 052/144] can: kvaser_usb: Fix tx queue start/stop race conditions Date: Tue, 21 Apr 2015 16:30:37 +0100 Message-Id: <1429630329-21748-53-git-send-email-luis.henriques@canonical.com> X-Mailer: git-send-email 1.9.1 In-Reply-To: <1429630329-21748-1-git-send-email-luis.henriques@canonical.com> References: <1429630329-21748-1-git-send-email-luis.henriques@canonical.com> X-Extended-Stable: 3.16 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 7600 Lines: 240 3.16.7-ckt10 -stable review patch. If anyone has any objections, please let me know. ------------------ From: "Ahmed S. Darwish" commit a9dc960c37b0d4eb192598dc4c94276270454514 upstream. A number of tx queue wake-up events went missing due to the outlined scenario below. Start state is a pool of 16 tx URBs, active tx_urbs count = 15, with the netdev tx queue open. CPU #1 [softirq] CPU #2 [softirq] start_xmit() tx_acknowledge() ................ ................ atomic_inc(&tx_urbs); if (atomic_read(&tx_urbs) >= 16) { --> atomic_dec(&tx_urbs); netif_wake_queue(); return; <-- netif_stop_queue(); } At the end, the correct state expected is a 15 tx_urbs count value with the tx queue state _open_. Due to the race, we get the same tx_urbs value but with the tx queue state _stopped_. The wake-up event is completely lost. Thus avoid hand-rolled concurrency mechanisms and use a proper lock for contexts and tx queue protection. Signed-off-by: Ahmed S. Darwish Signed-off-by: Marc Kleine-Budde [ luis: backported to 3.16: adjusted context ] Signed-off-by: Luis Henriques --- drivers/net/can/usb/kvaser_usb.c | 83 ++++++++++++++++++++++++---------------- 1 file changed, 51 insertions(+), 32 deletions(-) diff --git a/drivers/net/can/usb/kvaser_usb.c b/drivers/net/can/usb/kvaser_usb.c index 66b259bd9bce..4d47f224d705 100644 --- a/drivers/net/can/usb/kvaser_usb.c +++ b/drivers/net/can/usb/kvaser_usb.c @@ -12,6 +12,7 @@ * Copyright (C) 2012 Olivier Sobrie */ +#include #include #include #include @@ -298,10 +299,11 @@ struct kvaser_usb { struct kvaser_usb_net_priv { struct can_priv can; - atomic_t active_tx_urbs; - struct usb_anchor tx_submitted; + spinlock_t tx_contexts_lock; + int active_tx_contexts; struct kvaser_usb_tx_urb_context tx_contexts[MAX_TX_URBS]; + struct usb_anchor tx_submitted; struct completion start_comp, stop_comp; struct kvaser_usb *dev; @@ -506,6 +508,7 @@ static void kvaser_usb_tx_acknowledge(const struct kvaser_usb *dev, struct can_frame *cf; u8 channel = msg->u.tx_acknowledge.channel; u8 tid = msg->u.tx_acknowledge.tid; + unsigned long flags; if (channel >= dev->nchannels) { dev_err(dev->udev->dev.parent, @@ -545,12 +548,15 @@ static void kvaser_usb_tx_acknowledge(const struct kvaser_usb *dev, stats->tx_packets++; stats->tx_bytes += context->dlc; - can_get_echo_skb(priv->netdev, context->echo_index); - context->echo_index = MAX_TX_URBS; - atomic_dec(&priv->active_tx_urbs); + spin_lock_irqsave(&priv->tx_contexts_lock, flags); + can_get_echo_skb(priv->netdev, context->echo_index); + context->echo_index = MAX_TX_URBS; + --priv->active_tx_contexts; netif_wake_queue(priv->netdev); + + spin_unlock_irqrestore(&priv->tx_contexts_lock, flags); } static void kvaser_usb_simple_msg_callback(struct urb *urb) @@ -611,17 +617,6 @@ static int kvaser_usb_simple_msg_async(struct kvaser_usb_net_priv *priv, return 0; } -static void kvaser_usb_unlink_tx_urbs(struct kvaser_usb_net_priv *priv) -{ - int i; - - usb_kill_anchored_urbs(&priv->tx_submitted); - atomic_set(&priv->active_tx_urbs, 0); - - for (i = 0; i < MAX_TX_URBS; i++) - priv->tx_contexts[i].echo_index = MAX_TX_URBS; -} - static void kvaser_usb_rx_error(const struct kvaser_usb *dev, const struct kvaser_msg *msg) { @@ -1181,6 +1176,24 @@ error: return err; } +static void kvaser_usb_reset_tx_urb_contexts(struct kvaser_usb_net_priv *priv) +{ + int i; + + priv->active_tx_contexts = 0; + for (i = 0; i < MAX_TX_URBS; i++) + priv->tx_contexts[i].echo_index = MAX_TX_URBS; +} + +/* This method might sleep. Do not call it in the atomic context + * of URB completions. + */ +static void kvaser_usb_unlink_tx_urbs(struct kvaser_usb_net_priv *priv) +{ + usb_kill_anchored_urbs(&priv->tx_submitted); + kvaser_usb_reset_tx_urb_contexts(priv); +} + static void kvaser_usb_unlink_all_urbs(struct kvaser_usb *dev) { int i; @@ -1299,6 +1312,7 @@ static netdev_tx_t kvaser_usb_start_xmit(struct sk_buff *skb, void *buf; struct kvaser_msg *msg; int i, err, ret = NETDEV_TX_OK; + unsigned long flags; if (can_dropped_invalid_skb(netdev, skb)) return NETDEV_TX_OK; @@ -1342,12 +1356,21 @@ static netdev_tx_t kvaser_usb_start_xmit(struct sk_buff *skb, if (cf->can_id & CAN_RTR_FLAG) msg->u.tx_can.flags |= MSG_FLAG_REMOTE_FRAME; + spin_lock_irqsave(&priv->tx_contexts_lock, flags); for (i = 0; i < ARRAY_SIZE(priv->tx_contexts); i++) { if (priv->tx_contexts[i].echo_index == MAX_TX_URBS) { context = &priv->tx_contexts[i]; + + context->echo_index = i; + can_put_echo_skb(skb, netdev, context->echo_index); + ++priv->active_tx_contexts; + if (priv->active_tx_contexts >= MAX_TX_URBS) + netif_stop_queue(netdev); + break; } } + spin_unlock_irqrestore(&priv->tx_contexts_lock, flags); /* This should never happen; it implies a flow control bug */ if (!context) { @@ -1359,7 +1382,6 @@ static netdev_tx_t kvaser_usb_start_xmit(struct sk_buff *skb, } context->priv = priv; - context->echo_index = i; context->dlc = cf->can_dlc; msg->u.tx_can.tid = context->echo_index; @@ -1371,18 +1393,17 @@ static netdev_tx_t kvaser_usb_start_xmit(struct sk_buff *skb, kvaser_usb_write_bulk_callback, context); usb_anchor_urb(urb, &priv->tx_submitted); - can_put_echo_skb(skb, netdev, context->echo_index); - - atomic_inc(&priv->active_tx_urbs); - - if (atomic_read(&priv->active_tx_urbs) >= MAX_TX_URBS) - netif_stop_queue(netdev); - err = usb_submit_urb(urb, GFP_ATOMIC); if (unlikely(err)) { + spin_lock_irqsave(&priv->tx_contexts_lock, flags); + can_free_echo_skb(netdev, context->echo_index); + context->echo_index = MAX_TX_URBS; + --priv->active_tx_contexts; + netif_wake_queue(netdev); + + spin_unlock_irqrestore(&priv->tx_contexts_lock, flags); - atomic_dec(&priv->active_tx_urbs); usb_unanchor_urb(urb); stats->tx_dropped++; @@ -1509,7 +1530,7 @@ static int kvaser_usb_init_one(struct usb_interface *intf, struct kvaser_usb *dev = usb_get_intfdata(intf); struct net_device *netdev; struct kvaser_usb_net_priv *priv; - int i, err; + int err; err = kvaser_usb_send_simple_msg(dev, CMD_RESET_CHIP, channel); if (err) @@ -1523,19 +1544,17 @@ static int kvaser_usb_init_one(struct usb_interface *intf, priv = netdev_priv(netdev); + init_usb_anchor(&priv->tx_submitted); init_completion(&priv->start_comp); init_completion(&priv->stop_comp); - init_usb_anchor(&priv->tx_submitted); - atomic_set(&priv->active_tx_urbs, 0); - - for (i = 0; i < ARRAY_SIZE(priv->tx_contexts); i++) - priv->tx_contexts[i].echo_index = MAX_TX_URBS; - priv->dev = dev; priv->netdev = netdev; priv->channel = channel; + spin_lock_init(&priv->tx_contexts_lock); + kvaser_usb_reset_tx_urb_contexts(priv); + priv->can.state = CAN_STATE_STOPPED; priv->can.clock.freq = CAN_USB_CLOCK; priv->can.bittiming_const = &kvaser_usb_bittiming_const; -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/