Re: iwlagn is getting even worse with 3.3-rc1

[Emmanuel Grumbach <egrumbach@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 849 bytes --]

Hi Norbert,

>
> syslog output is at http://www.logic.at/people/preining/syslog.log
> this is after a suspend and resume.
>
> first connection after resume succeeded at around kernel time 378.859794
> then it worked for say 20sec or so, hard to say when exactely it dropped,
> but defeinitely at kernel time 451 it was finished.
>
> After that nothing went out or in again.
>
> Hope you can make something out of that.
>

Unfortunately, not really. I can see you are hitting PASSIVE_NO_RX,
meaning that the uCode complains that you are trying to Tx on a
passive channel on which it hasn't heard any beacon yet. But then we
finally get a beacon there, and open the queue again.

Anyway, can you try the attached patch ?
Note that I have no real way to test it (I have to admit I haven't
even compiled... I am not in my office).
I hope it might help.

[-- Attachment #2: tid.patch --]
[-- Type: application/octet-stream, Size: 1998 bytes --]

diff --git a/drivers/net/wireless/iwlwifi/iwl-agn-tx.c b/drivers/net/wireless/iwlwifi/iwl-agn-tx.c
index d9d758e..1b70048 100644
--- a/drivers/net/wireless/iwlwifi/iwl-agn-tx.c
+++ b/drivers/net/wireless/iwlwifi/iwl-agn-tx.c
@@ -95,6 +95,7 @@ static void iwlagn_tx_cmd_build_basic(struct iwl_priv *priv,
 			tx_flags |= TX_CMD_FLG_SEQ_CTL_MSK;
 		else
 			tx_flags &= ~TX_CMD_FLG_SEQ_CTL_MSK;
+		tx_cmd->tid_tspec = IWL_MAX_TID_COUNT;
 	}
 
 	iwlagn_tx_cmd_protection(priv, info, fc, &tx_flags);
@@ -808,6 +809,8 @@ static void iwl_rx_reply_tx_agg(struct iwl_priv *priv,
 	u32 status = le16_to_cpu(tx_resp->status.status);
 	int i;
 
+        WARN_ON(tid == IWL_MAX_TID_COUNT);
+
 	if (agg->wait_for_ba)
 		IWL_DEBUG_TX_REPLY(priv,
 			"got tx response w/o block-ack\n");
@@ -1035,10 +1038,12 @@ int iwlagn_rx_reply_tx(struct iwl_priv *priv, struct iwl_rx_mem_buffer *rxb,
 		}
 
 		__skb_queue_head_init(&skbs);
-		priv->tid_data[sta_id][tid].next_reclaimed = next_reclaimed;
 
-		IWL_DEBUG_TX_REPLY(priv, "Next reclaimed packet:%d",
-					  next_reclaimed);
+		if (tid != IWL_MAX_TID_COUNT) {
+			priv->tid_data[sta_id][tid].next_reclaimed = next_reclaimed;
+			IWL_DEBUG_TX_REPLY(priv, "Next reclaimed packet:%d",
+						  next_reclaimed);
+		}
 
 		/*we can free until ssn % q.n_bd not inclusive */
 		WARN_ON(iwl_trans_reclaim(trans(priv), sta_id, tid, txq_id,
diff --git a/drivers/net/wireless/iwlwifi/iwl-trans-pcie.c b/drivers/net/wireless/iwlwifi/iwl-trans-pcie.c
index aa87994..fa1b369 100644
--- a/drivers/net/wireless/iwlwifi/iwl-trans-pcie.c
+++ b/drivers/net/wireless/iwlwifi/iwl-trans-pcie.c
@@ -1551,7 +1551,7 @@ static int iwl_trans_pcie_reclaim(struct iwl_trans *trans, int sta_id, int tid,
 
 	txq->time_stamp = jiffies;
 
-	if (unlikely(txq_id >= IWLAGN_FIRST_AMPDU_QUEUE &&
+	if (unlikely(tid != IWL_MAX_TID_COUNT && txq_id >= IWLAGN_FIRST_AMPDU_QUEUE &&
 		     txq_id != trans_pcie->agg_txq[sta_id][tid])) {
 		/*
 		 * FIXME: this is a uCode bug which need to be addressed,