regression with kernel 3.8

regression with kernel 3.8

[Fabio Rossi]

I'm using the latest wireless-testing.git and I have found a regression with 
kernel 3.8. I don't know if this is related to the driver I'm using (ath5k) but 
I have bisected the problem to the commit 
6c17b77b67587b9f9e3070fb89fe98cef3187131 (mac80211: Fix tx queue handling 
during scans). Basically the authentication and association infos don't appear 
anymore in the kernel logs during a connection to my access point.

Tell me what you need to further debug the issue and solve the problem.

Fabio

Re: regression with kernel 3.8

[Seth Forshee]

On Mon, Feb 25, 2013 at 07:18:16PM +0100, Fabio Rossi wrote:
> I'm using the latest wireless-testing.git and I have found a regression with 
> kernel 3.8. I don't know if this is related to the driver I'm using (ath5k) but 
> I have bisected the problem to the commit 
> 6c17b77b67587b9f9e3070fb89fe98cef3187131 (mac80211: Fix tx queue handling 
> during scans). Basically the authentication and association infos don't appear 
> anymore in the kernel logs during a connection to my access point.
> 
> Tell me what you need to further debug the issue and solve the problem.

That patch is after 3.8.

Johannes, this looks like a problem with the changes you made when
applying that patch. I really should have tested the result before now.

I think the off-channel frames are getting queued rather than passed
down to the driver, which obviously isn't what we want. I'll get this
fixed.

Seth

[PATCH] mac80211: Ensure off-channel frames don't get queued

[Seth Forshee]

Commit 6c17b77b67587b9f9e3070fb89fe98cef3187131 (mac80211: Fix tx queue
handling during scans) contains a bug that causes off-channel frames to
get queued when they should be handed down to the driver for transmit.
Prevent this from happening.

Reported-by: Fabio Rossi <rossi.f@inwind.it>
Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
---

Fabio, this patch should fix the problem. Can you verify?

Johannes, this prevents the off-channel frames from getting queued
without affecting the fast path. It does however make the indentation
awfully deep ...

 net/mac80211/tx.c |   56 +++++++++++++++++++++++++++++------------------------
 1 file changed, 31 insertions(+), 25 deletions(-)

diff --git a/net/mac80211/tx.c b/net/mac80211/tx.c
index 5b9602b..b1296e7 100644
--- a/net/mac80211/tx.c
+++ b/net/mac80211/tx.c
@@ -1231,34 +1231,40 @@ static bool ieee80211_tx_frags(struct ieee80211_local *local,
 		if (local->queue_stop_reasons[q] ||
 		    (!txpending && !skb_queue_empty(&local->pending[q]))) {
 			if (unlikely(info->flags &
-					IEEE80211_TX_INTFL_OFFCHAN_TX_OK &&
-				     local->queue_stop_reasons[q] &
-					~BIT(IEEE80211_QUEUE_STOP_REASON_OFFCHANNEL))) {
+				     IEEE80211_TX_INTFL_OFFCHAN_TX_OK)) {
+				if (local->queue_stop_reasons[q] &
+				    ~BIT(IEEE80211_QUEUE_STOP_REASON_OFFCHANNEL)) {
+					/*
+					 * Drop off-channel frames if queues
+					 * are stopped for any reason other
+					 * than off-channel operation. Never
+					 * queue them.
+					 */
+					spin_unlock_irqrestore(
+						&local->queue_stop_reason_lock,
+						flags);
+					ieee80211_purge_tx_queue(&local->hw,
+								 skbs);
+					return true;
+				}
+			} else {
+
 				/*
-				 * Drop off-channel frames if queues are stopped
-				 * for any reason other than off-channel
-				 * operation. Never queue them.
+				 * Since queue is stopped, queue up frames for
+				 * later transmission from the tx-pending
+				 * tasklet when the queue is woken again.
 				 */
-				spin_unlock_irqrestore(
-					&local->queue_stop_reason_lock, flags);
-				ieee80211_purge_tx_queue(&local->hw, skbs);
-				return true;
+				if (txpending)
+					skb_queue_splice_init(skbs,
+							      &local->pending[q]);
+				else
+					skb_queue_splice_tail_init(skbs,
+								   &local->pending[q]);
+
+				spin_unlock_irqrestore(&local->queue_stop_reason_lock,
+						       flags);
+				return false;
 			}
-
-			/*
-			 * Since queue is stopped, queue up frames for later
-			 * transmission from the tx-pending tasklet when the
-			 * queue is woken again.
-			 */
-			if (txpending)
-				skb_queue_splice_init(skbs, &local->pending[q]);
-			else
-				skb_queue_splice_tail_init(skbs,
-							   &local->pending[q]);
-
-			spin_unlock_irqrestore(&local->queue_stop_reason_lock,
-					       flags);
-			return false;
 		}
 		spin_unlock_irqrestore(&local->queue_stop_reason_lock, flags);
 
-- 
1.7.9.5

Re: [PATCH] mac80211: Ensure off-channel frames don't get queued

[Johannes Berg]

On Mon, 2013-02-25 at 14:58 -0600, Seth Forshee wrote:
> Commit 6c17b77b67587b9f9e3070fb89fe98cef3187131 (mac80211: Fix tx queue
> handling during scans) contains a bug that causes off-channel frames to
> get queued when they should be handed down to the driver for transmit.
> Prevent this from happening.
> 
> Reported-by: Fabio Rossi <rossi.f@inwind.it>
> Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
> ---
> 
> Fabio, this patch should fix the problem. Can you verify?
> 
> Johannes, this prevents the off-channel frames from getting queued
> without affecting the fast path. It does however make the indentation
> awfully deep ...

Ugh .. I do see the bug, but the indentation is crap. I'll apply this
anyway, and then we can sort it out later.

johannes

Re: regression with kernel 3.8

[Johannes Berg]

On Mon, 2013-02-25 at 13:26 -0600, Seth Forshee wrote:
> On Mon, Feb 25, 2013 at 07:18:16PM +0100, Fabio Rossi wrote:
> > I'm using the latest wireless-testing.git and I have found a regression with 
> > kernel 3.8. I don't know if this is related to the driver I'm using (ath5k) but 
> > I have bisected the problem to the commit 
> > 6c17b77b67587b9f9e3070fb89fe98cef3187131 (mac80211: Fix tx queue handling 
> > during scans). Basically the authentication and association infos don't appear 
> > anymore in the kernel logs during a connection to my access point.
> > 
> > Tell me what you need to further debug the issue and solve the problem.
> 
> That patch is after 3.8.
> 
> Johannes, this looks like a problem with the changes you made when
> applying that patch. I really should have tested the result before now.

Hmm, crap. Sorry about that!

> I think the off-channel frames are getting queued rather than passed
> down to the driver, which obviously isn't what we want. I'll get this
> fixed.

I saw you posted it, I'll take a closer look tomorrow morning.

johannes

Re: regression with kernel 3.8

[Seth Forshee]

On Mon, Feb 25, 2013 at 11:38:38PM +0100, Johannes Berg wrote:
> On Mon, 2013-02-25 at 13:26 -0600, Seth Forshee wrote:
> > On Mon, Feb 25, 2013 at 07:18:16PM +0100, Fabio Rossi wrote:
> > > I'm using the latest wireless-testing.git and I have found a regression with 
> > > kernel 3.8. I don't know if this is related to the driver I'm using (ath5k) but 
> > > I have bisected the problem to the commit 
> > > 6c17b77b67587b9f9e3070fb89fe98cef3187131 (mac80211: Fix tx queue handling 
> > > during scans). Basically the authentication and association infos don't appear 
> > > anymore in the kernel logs during a connection to my access point.
> > > 
> > > Tell me what you need to further debug the issue and solve the problem.
> > 
> > That patch is after 3.8.
> > 
> > Johannes, this looks like a problem with the changes you made when
> > applying that patch. I really should have tested the result before now.
> 
> Hmm, crap. Sorry about that!
> 
> > I think the off-channel frames are getting queued rather than passed
> > down to the driver, which obviously isn't what we want. I'll get this
> > fixed.
> 
> I saw you posted it, I'll take a closer look tomorrow morning.

Have you had a chance to look at the patch yet? I think we ought to try
and get this fixed pretty quickly.

Seth

Re: [PATCH] mac80211: Ensure off-channel frames don't get queued

[Fabio Rossi]

On 25 Feb 2013 Seth wrote:

>Commit 6c17b77b67587b9f9e3070fb89fe98cef3187131 (mac80211: Fix tx queue
>handling during scans) contains a bug that causes off-channel frames to
>get queued when they should be handed down to the driver for transmit.
>Prevent this from happening.
>
>Reported-by: Fabio Rossi <rossi.f@inwind.it>
>Signed-off-by: Seth Forshee <seth.forshee@canonical.com>
>---
>
>Fabio, this patch should fix the problem. Can you verify?

tested, now it works again

Fabio