[PATCH]

[PATCH]

[Johannes Berg]

Subject: wext: remove seq_start/stop sparse annotations

Even though they are true, they cause sparse to complain
because it doesn't see the __acquires(dev_base_lock) on
dev_seq_start() because it is only added to the function
in net/core/dev.c, not the header file. To keep track of
the nesting correctly we should probably annotate those
functions publically, but for now let's just remove the
annotation I added to wext.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
---
I guess I should have actually run sparse... This could be part of the
wext locking fix "wext: fix get_wireless_stats locking" instead of being
standalone, if you're so inclined.

 net/wireless/wext.c |    2 --
 1 file changed, 2 deletions(-)

--- wireless-testing.orig/net/wireless/wext.c	2009-05-12 08:12:23.000000000 +0200
+++ wireless-testing/net/wireless/wext.c	2009-05-12 08:13:16.000000000 +0200
@@ -650,14 +650,12 @@ static int wireless_seq_show(struct seq_
 }
 
 static void *wireless_dev_seq_start(struct seq_file *seq, loff_t *pos)
-	__acquires(dev_base_lock)
 {
 	rtnl_lock();
 	return dev_seq_start(seq, pos);
 }
 
 static void wireless_dev_seq_stop(struct seq_file *seq, void *v)
-	__releases(dev_base_lock)
 {
 	dev_seq_stop(seq, v);
 	rtnl_unlock();

[PATCH v2] wext: remove seq_start/stop sparse annotations

[Johannes Berg]

Even though they are true, they cause sparse to complain
because it doesn't see the __acquires(dev_base_lock) on
dev_seq_start() because it is only added to the function
in net/core/dev.c, not the header file. To keep track of
the nesting correctly we should probably annotate those
functions publically, but for now let's just remove the
annotation I added to wext.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
---
Umm, now with subject.

 net/wireless/wext.c |    2 --
 1 file changed, 2 deletions(-)

--- wireless-testing.orig/net/wireless/wext.c	2009-05-12 08:12:23.000000000 +0200
+++ wireless-testing/net/wireless/wext.c	2009-05-12 08:13:16.000000000 +0200
@@ -650,14 +650,12 @@ static int wireless_seq_show(struct seq_
 }
 
 static void *wireless_dev_seq_start(struct seq_file *seq, loff_t *pos)
-	__acquires(dev_base_lock)
 {
 	rtnl_lock();
 	return dev_seq_start(seq, pos);
 }
 
 static void wireless_dev_seq_stop(struct seq_file *seq, void *v)
-	__releases(dev_base_lock)
 {
 	dev_seq_stop(seq, v);
 	rtnl_unlock();

[PATCH]

[Johannes Berg]

In commit 31ccc476b77234f6afb3 (mac80211: QoS related cleanups) I
accidentally changed a variable from int to u16 causing a warning
that a comparison for < 0 was always false. John thought this was
a missing deletion of code and removed the warning by deleting the
never executed branch of code in commit 13e5f0888caddf7a020dcd918
(wireless: fix warning introduced by "mac80211: QoS related cleanups")
but the problem really was my mistake of using a u16 variable for
the queue variable when that variable can also contain an error
code. This patch restores the original code and variable type.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
---
 net/mac80211/wme.c |   26 +++++++++++++++-----------
 1 file changed, 15 insertions(+), 11 deletions(-)

--- everything.orig/net/mac80211/wme.c	2008-05-03 00:29:07.000000000 +0200
+++ everything/net/mac80211/wme.c	2008-05-03 00:32:01.000000000 +0200
@@ -155,8 +155,7 @@ static int wme_qdiscop_enqueue(struct sk
 	unsigned short fc = le16_to_cpu(hdr->frame_control);
 	struct Qdisc *qdisc;
 	struct sta_info *sta;
-	int err;
-	u16 queue;
+	int err, queue;
 	u8 tid;
 
 	if (pkt_data->flags & IEEE80211_TXPD_REQUEUE) {
@@ -216,15 +215,20 @@ static int wme_qdiscop_enqueue(struct sk
 		rcu_read_unlock();
 	}
 
-	tid = skb->priority & QOS_CONTROL_TAG1D_MASK;
-	pkt_data->queue = (unsigned int) queue;
-	qdisc = q->queues[queue];
-	err = qdisc->enqueue(skb, qdisc);
-	if (err == NET_XMIT_SUCCESS) {
-		qd->q.qlen++;
-		qd->bstats.bytes += skb->len;
-		qd->bstats.packets++;
-		return NET_XMIT_SUCCESS;
+	if (unlikely(queue < 0)) {
+			kfree_skb(skb);
+			err = NET_XMIT_DROP;
+	} else {
+		tid = skb->priority & QOS_CONTROL_TAG1D_MASK;
+		pkt_data->queue = (unsigned int) queue;
+		qdisc = q->queues[queue];
+		err = qdisc->enqueue(skb, qdisc);
+		if (err == NET_XMIT_SUCCESS) {
+			qd->q.qlen++;
+			qd->bstats.bytes += skb->len;
+			qd->bstats.packets++;
+			return NET_XMIT_SUCCESS;
+		}
 	}
 	qd->qstats.drops++;
 	return err;

[PATCH]

[Johannes Berg]

The ieee80211_ioctl_giwrate() ioctl handler doesn't rcu_read_lock()
its access to the sta table, fix it.

Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
---
 net/mac80211/ieee80211_ioctl.c |   18 ++++++++++++------
 1 file changed, 12 insertions(+), 6 deletions(-)

--- everything.orig/net/mac80211/ieee80211_ioctl.c	2008-04-04 17:48:11.000000000 +0200
+++ everything/net/mac80211/ieee80211_ioctl.c	2008-04-04 17:48:56.000000000 +0200
@@ -586,19 +586,25 @@ static int ieee80211_ioctl_giwrate(struc
 
 	sdata = IEEE80211_DEV_TO_SUB_IF(dev);
 
-	if (sdata->vif.type == IEEE80211_IF_TYPE_STA)
-		sta = sta_info_get(local, sdata->u.sta.bssid);
-	else
+	if (sdata->vif.type != IEEE80211_IF_TYPE_STA)
 		return -EOPNOTSUPP;
-	if (!sta)
-		return -ENODEV;
 
 	sband = local->hw.wiphy->bands[local->hw.conf.channel->band];
 
-	if (sta->txrate_idx < sband->n_bitrates)
+	rcu_read_lock();
+
+	sta = sta_info_get(local, sdata->u.sta.bssid);
+
+	if (sta && sta->txrate_idx < sband->n_bitrates)
 		rate->value = sband->bitrates[sta->txrate_idx].bitrate;
 	else
 		rate->value = 0;
+
+	rcu_read_unlock();
+
+	if (!sta)
+		return -ENODEV;
+
 	rate->value *= 100000;
 
 	return 0;