All of lore.kernel.org
 help / color / mirror / Atom feed
From: Arik Nemtsov <arik@wizery.com>
To: <linux-wireless@vger.kernel.org>
Cc: Luciano Coelho <coelho@ti.com>,
	Johannes Berg <johannes@sipsolutions.net>,
	Arik Nemtsov <arik@wizery.com>
Subject: [PATCH v2 01/10] wl12xx: fix potential race condition with TX queue watermark
Date: Mon, 31 Jan 2011 13:16:20 +0200	[thread overview]
Message-ID: <1296472589-26401-2-git-send-email-arik@wizery.com> (raw)
In-Reply-To: <1296472589-26401-1-git-send-email-arik@wizery.com>

Check the conditions for the high/low TX queue watermarks when the
spin-lock is taken. This prevents race conditions as tx_queue_count and
the flag checked are only modified when the spin-lock is taken.

The following race was in mind:
- Queues are almost full and wl1271_op_tx() will stop the queues, but it
  doesn't get the spin-lock yet.
- (on another CPU) tx_work_locked() dequeues 15 skbs from this queue and
  tx_queue_count is updated to reflect this
- wl1271_op_tx() does not check tx_queue_count after taking the
  spin-lock and incorrectly stops the queue.

Signed-off-by: Arik Nemtsov <arik@wizery.com>
---
 drivers/net/wireless/wl12xx/main.c |   24 +++++++++++-------------
 drivers/net/wireless/wl12xx/tx.c   |    2 +-
 2 files changed, 12 insertions(+), 14 deletions(-)

diff --git a/drivers/net/wireless/wl12xx/main.c b/drivers/net/wireless/wl12xx/main.c
index dfab21e..cf11bda 100644
--- a/drivers/net/wireless/wl12xx/main.c
+++ b/drivers/net/wireless/wl12xx/main.c
@@ -995,6 +995,17 @@ static int wl1271_op_tx(struct ieee80211_hw *hw, struct sk_buff *skb)
 	}
 #endif
 	wl->tx_queue_count++;
+
+	/*
+	 * The workqueue is slow to process the tx_queue and we need stop
+	 * the queue here, otherwise the queue will get too long.
+	 */
+	if (wl->tx_queue_count >= WL1271_TX_QUEUE_HIGH_WATERMARK) {
+		wl1271_debug(DEBUG_TX, "op_tx: stopping queues");
+		ieee80211_stop_queues(wl->hw);
+		__set_bit(WL1271_FLAG_TX_QUEUE_STOPPED, &wl->flags);
+	}
+
 	spin_unlock_irqrestore(&wl->wl_lock, flags);
 
 	/* queue the packet */
@@ -1009,19 +1020,6 @@ static int wl1271_op_tx(struct ieee80211_hw *hw, struct sk_buff *skb)
 	if (!test_bit(WL1271_FLAG_FW_TX_BUSY, &wl->flags))
 		ieee80211_queue_work(wl->hw, &wl->tx_work);
 
-	/*
-	 * The workqueue is slow to process the tx_queue and we need stop
-	 * the queue here, otherwise the queue will get too long.
-	 */
-	if (wl->tx_queue_count >= WL1271_TX_QUEUE_HIGH_WATERMARK) {
-		wl1271_debug(DEBUG_TX, "op_tx: stopping queues");
-
-		spin_lock_irqsave(&wl->wl_lock, flags);
-		ieee80211_stop_queues(wl->hw);
-		set_bit(WL1271_FLAG_TX_QUEUE_STOPPED, &wl->flags);
-		spin_unlock_irqrestore(&wl->wl_lock, flags);
-	}
-
 	return NETDEV_TX_OK;
 }
 
diff --git a/drivers/net/wireless/wl12xx/tx.c b/drivers/net/wireless/wl12xx/tx.c
index 3507c81..ce9a19b 100644
--- a/drivers/net/wireless/wl12xx/tx.c
+++ b/drivers/net/wireless/wl12xx/tx.c
@@ -288,7 +288,7 @@ static void handle_tx_low_watermark(struct wl1271 *wl)
 		/* firmware buffer has space, restart queues */
 		spin_lock_irqsave(&wl->wl_lock, flags);
 		ieee80211_wake_queues(wl->hw);
-		clear_bit(WL1271_FLAG_TX_QUEUE_STOPPED, &wl->flags);
+		__clear_bit(WL1271_FLAG_TX_QUEUE_STOPPED, &wl->flags);
 		spin_unlock_irqrestore(&wl->wl_lock, flags);
 	}
 }
-- 
1.7.1


  reply	other threads:[~2011-01-31 11:16 UTC|newest]

Thread overview: 16+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2011-01-31 11:16 [PATCH v2 00/10] wl12xx: AP-mode per link PSM Arik Nemtsov
2011-01-31 11:16 ` Arik Nemtsov [this message]
2011-01-31 11:16 ` [PATCH v2 02/10] wl12xx: AP-mode - fix race condition on sta connection Arik Nemtsov
2011-01-31 11:16 ` [PATCH v2 03/10] wl12xx: AP-mode - TX queue per link in AC Arik Nemtsov
2011-01-31 11:16 ` [PATCH v2 04/10] mac80211: do not calc frame duration when using HW rate-control Arik Nemtsov
2011-01-31 11:16 ` [PATCH v2 05/10] wl12xx: report invalid TX rate when returning non-TX-ed skbs Arik Nemtsov
2011-01-31 11:16 ` [PATCH v2 06/10] mac80211: add HW flag for disabling auto link-PS in AP mode Arik Nemtsov
2011-01-31 11:24   ` Johannes Berg
2011-01-31 11:49     ` Arik Nemtsov
2011-01-31 12:11       ` Johannes Berg
2011-01-31 20:06         ` Luciano Coelho
2011-01-31 20:16           ` Arik Nemtsov
2011-01-31 11:16 ` [PATCH v2 07/10] wl12xx: AP-mode - support HW based link PS monitoring Arik Nemtsov
2011-01-31 11:16 ` [PATCH v2 08/10] wl12xx: AP mode - fix bug in cleanup of wl1271_op_sta_add() Arik Nemtsov
2011-01-31 11:16 ` [PATCH v2 09/10] wl12xx: AP-mode - count free FW TX blocks per link Arik Nemtsov
2011-01-31 11:16 ` [PATCH v2 10/10] wl12xx: AP-mode - management of links in PS-mode Arik Nemtsov

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=1296472589-26401-2-git-send-email-arik@wizery.com \
    --to=arik@wizery.com \
    --cc=coelho@ti.com \
    --cc=johannes@sipsolutions.net \
    --cc=linux-wireless@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.