linux-wireless.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [RFC] iwlwifi: fix scan abort
@ 2010-07-28 13:45 Stanislaw Gruszka
  2010-07-28 14:34 ` Guy, Wey-Yi
  0 siblings, 1 reply; 2+ messages in thread
From: Stanislaw Gruszka @ 2010-07-28 13:45 UTC (permalink / raw)
  To: Wey-Yi Guy, Reinette Chatre, John W. Linville; +Cc: linux-wireless

We can not call cancel_delayed_work_sync(&priv->scan_check) with 
priv->mutex locked because workqueue function iwl_bg_scan_check()
take that lock internally.

We do not need to synchronize when canceling priv->scan_check work. 
We can avoid races (sending double abort command or send no
command at all) using STATUS_SCAN_ABORT bit. Moreover
current iwl_bg_scan_check() code seems to be broken, as
we should not send abort commands when currently aborting.

I did not test patch yet, just want to know if it is theoretically
correct. Except obvious circular priv->mutex locking fix, maybe it
can help with warning in ieee80211_scan_completed, which is still
reported by the users from time to time.

diff --git a/drivers/net/wireless/iwlwifi/iwl-scan.c b/drivers/net/wireless/iwlwifi/iwl-scan.c
index 2a7c399..b0c6b04 100644
--- a/drivers/net/wireless/iwlwifi/iwl-scan.c
+++ b/drivers/net/wireless/iwlwifi/iwl-scan.c
@@ -429,11 +429,10 @@ void iwl_bg_scan_check(struct work_struct *data)
 		return;
 
 	mutex_lock(&priv->mutex);
-	if (test_bit(STATUS_SCANNING, &priv->status) ||
-	    test_bit(STATUS_SCAN_ABORTING, &priv->status)) {
-		IWL_DEBUG_SCAN(priv, "Scan completion watchdog resetting "
-			"adapter (%dms)\n",
-			jiffies_to_msecs(IWL_SCAN_CHECK_WATCHDOG));
+	if (test_bit(STATUS_SCANNING, &priv->status) &&
+	    !test_bit(STATUS_SCAN_ABORTING, &priv->status)) {
+		IWL_DEBUG_SCAN(priv, "Scan completion watchdog (%dms)\n",
+			       jiffies_to_msecs(IWL_SCAN_CHECK_WATCHDOG));
 
 		if (!test_bit(STATUS_EXIT_PENDING, &priv->status))
 			iwl_send_scan_abort(priv);
@@ -498,12 +497,11 @@ void iwl_bg_abort_scan(struct work_struct *work)
 	    !test_bit(STATUS_GEO_CONFIGURED, &priv->status))
 		return;
 
-	mutex_lock(&priv->mutex);
-
-	cancel_delayed_work_sync(&priv->scan_check);
-	set_bit(STATUS_SCAN_ABORTING, &priv->status);
-	iwl_send_scan_abort(priv);
+	cancel_delayed_work(&priv->scan_check);
 
+	mutex_lock(&priv->mutex);
+	if (test_bit(STATUS_SCAN_ABORTING, &priv->status))
+		iwl_send_scan_abort(priv);
 	mutex_unlock(&priv->mutex);
 }
 EXPORT_SYMBOL(iwl_bg_abort_scan);

^ permalink raw reply related	[flat|nested] 2+ messages in thread

* Re: [RFC] iwlwifi: fix scan abort
  2010-07-28 13:45 [RFC] iwlwifi: fix scan abort Stanislaw Gruszka
@ 2010-07-28 14:34 ` Guy, Wey-Yi
  0 siblings, 0 replies; 2+ messages in thread
From: Guy, Wey-Yi @ 2010-07-28 14:34 UTC (permalink / raw)
  To: Stanislaw Gruszka; +Cc: Chatre, Reinette, John W. Linville, linux-wireless

Hi Gruszka,

On Wed, 2010-07-28 at 06:45 -0700, Stanislaw Gruszka wrote:
> We can not call cancel_delayed_work_sync(&priv->scan_check) with 
> priv->mutex locked because workqueue function iwl_bg_scan_check()
> take that lock internally.
> 
> We do not need to synchronize when canceling priv->scan_check work. 
> We can avoid races (sending double abort command or send no
> command at all) using STATUS_SCAN_ABORT bit. Moreover
> current iwl_bg_scan_check() code seems to be broken, as
> we should not send abort commands when currently aborting.
> 
> I did not test patch yet, just want to know if it is theoretically
> correct. Except obvious circular priv->mutex locking fix, maybe it
> can help with warning in ieee80211_scan_completed, which is still
> reported by the users from time to time.
> 
> diff --git a/drivers/net/wireless/iwlwifi/iwl-scan.c b/drivers/net/wireless/iwlwifi/iwl-scan.c
> index 2a7c399..b0c6b04 100644
> --- a/drivers/net/wireless/iwlwifi/iwl-scan.c
> +++ b/drivers/net/wireless/iwlwifi/iwl-scan.c
> @@ -429,11 +429,10 @@ void iwl_bg_scan_check(struct work_struct *data)
>  		return;
>  
>  	mutex_lock(&priv->mutex);
> -	if (test_bit(STATUS_SCANNING, &priv->status) ||
> -	    test_bit(STATUS_SCAN_ABORTING, &priv->status)) {
> -		IWL_DEBUG_SCAN(priv, "Scan completion watchdog resetting "
> -			"adapter (%dms)\n",
> -			jiffies_to_msecs(IWL_SCAN_CHECK_WATCHDOG));
> +	if (test_bit(STATUS_SCANNING, &priv->status) &&
> +	    !test_bit(STATUS_SCAN_ABORTING, &priv->status)) {
> +		IWL_DEBUG_SCAN(priv, "Scan completion watchdog (%dms)\n",
> +			       jiffies_to_msecs(IWL_SCAN_CHECK_WATCHDOG));

make sense here, the code is broken, we should not abort scan if already
doing it.

>  
>  		if (!test_bit(STATUS_EXIT_PENDING, &priv->status))
>  			iwl_send_scan_abort(priv);
> @@ -498,12 +497,11 @@ void iwl_bg_abort_scan(struct work_struct *work)
>  	    !test_bit(STATUS_GEO_CONFIGURED, &priv->status))
>  		return;
>  
> -	mutex_lock(&priv->mutex);
> -
> -	cancel_delayed_work_sync(&priv->scan_check);
> -	set_bit(STATUS_SCAN_ABORTING, &priv->status);
> -	iwl_send_scan_abort(priv);
> +	cancel_delayed_work(&priv->scan_check);
>  
> +	mutex_lock(&priv->mutex);
> +	if (test_bit(STATUS_SCAN_ABORTING, &priv->status))
> +		iwl_send_scan_abort(priv);
>  	mutex_unlock(&priv->mutex);
>  }

Looks right to me, thanks a lot to catch this one.

Wey




^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2010-07-28 14:35 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2010-07-28 13:45 [RFC] iwlwifi: fix scan abort Stanislaw Gruszka
2010-07-28 14:34 ` Guy, Wey-Yi

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).