netdev - [PATCH] r8169: fix early Tx queue wake-up.

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Message-ID: <20120130231641.GA5287@electric-eye.fr.zoreil.com>
Date:	Tue, 31 Jan 2012 00:16:41 +0100
From:	Francois Romieu <romieu@...zoreil.com>
To:	netdev@...r.kernel.org
Cc:	David Miller <davem@...emloft.net>,
	Eric Dumazet <eric.dumazet@...il.com>,
	Michał Mirosław <mirqus@...il.com>,
	Hayes Wang <hayeswang@...ltek.com>
Subject: [PATCH] r8169: fix early Tx queue wake-up.

With infinite gratitude to Eric Dumazet for allowing me to identify
the error.

Signed-off-by: Francois Romieu <romieu@...zoreil.com>
Cc: Eric Dumazet <eric.dumazet@...il.com>
Cc: Hayes Wang <hayeswang@...ltek.com>
---
 I'll address Michał's remarks and add unrelated comments after some sleep.

 drivers/net/ethernet/realtek/r8169.c |   19 +++++++++++++++++++
 1 files changed, 19 insertions(+), 0 deletions(-)

diff --git a/drivers/net/ethernet/realtek/r8169.c b/drivers/net/ethernet/realtek/r8169.c
index d039d39..f994606 100644
--- a/drivers/net/ethernet/realtek/r8169.c
+++ b/drivers/net/ethernet/realtek/r8169.c
@@ -5559,7 +5559,18 @@ static netdev_tx_t rtl8169_start_xmit(struct sk_buff *skb,
 	mmiowb();
 
 	if (TX_BUFFS_AVAIL(tp) < MAX_SKB_FRAGS) {
+		/* Avoid wrongly optimistic queue wake-up: rtl_tx thread must
+		 * not miss a ring update when it notices a stopped queue.
+		 */
+		smp_wmb();
 		netif_stop_queue(dev);
+		/* Sync with rtl_tx:
+		 * - publish queue status and cur_tx ring index (write barrier)
+		 * - refresh dirty_tx ring index (read barrier).
+		 * May the current thread have a pessimistic view of the ring
+		 * status and forget to wake up queue, a racing rtl_tx thread
+		 * can't.
+		 */
 		smp_mb();
 		if (TX_BUFFS_AVAIL(tp) >= MAX_SKB_FRAGS)
 			netif_wake_queue(dev);
@@ -5659,6 +5670,14 @@ static void rtl_tx(struct net_device *dev, struct rtl8169_private *tp)
 
 	if (tp->dirty_tx != dirty_tx) {
 		tp->dirty_tx = dirty_tx;
+		/*
+		 * Sync with rtl8169_start_xmit:
+		 * - publish dirty_tx ring index (write barrier)
+		 * - refresh cur_tx ring index and queue status (read barrier)
+		 * May the current thread miss the stopped queue condition,
+		 * a racing xmit thread can only have a right view of the
+		 * ring status.
+		 */
 		smp_mb();
 		if (netif_queue_stopped(dev) &&
 		    (TX_BUFFS_AVAIL(tp) >= MAX_SKB_FRAGS)) {
-- 
1.7.6.5

--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html