[PATCH 2.6.39] ath9k: fix a regression in PS frame filter handling

[PATCH 2.6.39] ath9k: fix a regression in PS frame filter handling

[Felix Fietkau]

Only leave filtering enabled for AP or VLAN interfaces, clear the
destination mask for all other interfaces.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Reported-by: Ben Greear <greearb@candelatech.com>
---
 drivers/net/wireless/ath/ath9k/main.c |    5 +++++
 drivers/net/wireless/ath/ath9k/xmit.c |    6 +++++-
 2 files changed, 10 insertions(+), 1 deletions(-)

diff --git a/drivers/net/wireless/ath/ath9k/main.c b/drivers/net/wireless/ath/ath9k/main.c
index 3de115d..3d999b7 100644
--- a/drivers/net/wireless/ath/ath9k/main.c
+++ b/drivers/net/wireless/ath/ath9k/main.c
@@ -1778,6 +1778,11 @@ static int ath9k_sta_add(struct ieee80211_hw *hw,
 	struct ieee80211_key_conf ps_key = { };
 
 	ath_node_attach(sc, sta);
+
+	if (vif->type != NL80211_IFTYPE_AP &&
+	    vif->type != NL80211_IFTYPE_AP_VLAN)
+		return 0;
+
 	an->ps_key = ath_key_config(common, vif, sta, &ps_key);
 
 	return 0;
diff --git a/drivers/net/wireless/ath/ath9k/xmit.c b/drivers/net/wireless/ath/ath9k/xmit.c
index 744779d..ee71d38 100644
--- a/drivers/net/wireless/ath/ath9k/xmit.c
+++ b/drivers/net/wireless/ath/ath9k/xmit.c
@@ -1795,6 +1795,7 @@ static void ath_tx_start_dma(struct ath_softc *sc, struct ath_buf *bf,
 	struct sk_buff *skb = bf->bf_mpdu;
 	struct ieee80211_tx_info *tx_info = IEEE80211_SKB_CB(skb);
 	struct ieee80211_hdr *hdr = (struct ieee80211_hdr *)skb->data;
+	struct ieee80211_vif *vif = tx_info->control.vif;
 	struct list_head bf_head;
 	struct ath_atx_tid *tid = NULL;
 	u8 tidno;
@@ -1829,7 +1830,10 @@ static void ath_tx_start_dma(struct ath_softc *sc, struct ath_buf *bf,
 		if (txctl->paprd)
 			bf->bf_state.bfs_paprd_timestamp = jiffies;
 
-		if (tx_info->flags & IEEE80211_TX_CTL_CLEAR_PS_FILT)
+		if ((vif && vif->type != NL80211_IFTYPE_AP &&
+		            vif->type != NL80211_IFTYPE_AP_VLAN) ||
+		    !ieee80211_is_data(hdr->frame_control) ||
+		    (tx_info->flags & IEEE80211_TX_CTL_CLEAR_PS_FILT))
 			ath9k_hw_set_clrdmask(sc->sc_ah, bf->bf_desc, true);
 
 		ath_tx_send_normal(sc, txctl->txq, tid, &bf_head);
-- 
1.7.3.2

Re: [PATCH 2.6.39] ath9k: fix a regression in PS frame filter handling

[Ben Greear]

On 05/09/2011 01:25 PM, Felix Fietkau wrote:
> Only leave filtering enabled for AP or VLAN interfaces, clear the
> destination mask for all other interfaces.

I tried removing your previous two patches and added this one.

Unfortunately, it crashes.  I upgraded to -rc7 at the same
time, so maybe it's something else..or maybe I screwed up
the build.  But, might want to hold off on this a bit until
I can debug some more.

I assume you haven't seen any such problems?

Thanks,
Ben


Oops: 0000 [#1] SMP
last sysfs file: /sys/devices/virtual/net/rddVR10-p/address
Modules linked in: veth 8021q garp stp llc macvlan pktgen fuse coretemp hwmon nfs lockd fscache auth_rpcgss nfs_acl sunrpc ipv6 uinput arc4 ath9k mac]

Pid: 34, comm: kworker/u:1 Tainted: G        W   2.6.39-rc7-wl+ #68 To Be Filled By O.E.M. To Be Filled By O.E.M./To be filled by O.E.M.
EIP: 0060:[<f8d620ff>] EFLAGS: 00010206 CPU: 1
EIP is at ath_tx_start+0x640/0x6ab [ath9k]
EAX: 00000000 EBX: 00000000 ECX: 0000004d EDX: f4e59de8
ESI: f3cd8000 EDI: f269e640 EBP: f4e59dd0 ESP: f4e59d88
  DS: 007b ES: 007b FS: 00d8 GS: 0000 SS: 0068
Process kworker/u:1 (pid: 34, ti=f4e58000 task=f4f30a50 task.ti=f4e58000)
Stack:
  f46cd640 00000000 f46ce0a8 f46ce0a8 f1023c18 f46cd640 f3c34000 0000004d
  f46cd640 f4e59de8 f1023c00 00000046 00000000 f3cd8000 f3cd8000 f46cd640
  f269e640 f4e59dfc f4e59e08 f8d5bbee f46cc480 f1023c00 f3c34000 f3c3401c
Call Trace:
  [<f8d5bbee>] ath9k_tx+0x154/0x186 [ath9k]
  [<f8bf091c>] __ieee80211_tx+0x149/0x190 [mac80211]
  [<f8bf185e>] ieee80211_tx+0x7b/0x93 [mac80211]
  [<f8bf1a2d>] ieee80211_xmit+0x1b7/0x1c4 [mac80211]
  [<c04311cf>] ? get_parent_ip+0xb/0x31
  [<f8bf1a83>] ieee80211_tx_skb+0x49/0x50 [mac80211]
  [<f8bf3f6d>] ieee80211_send_probe_req+0x29/0x2e [mac80211]
  [<f8bde320>] ieee80211_scan_work+0x32d/0x44c [mac80211]
  [<c044cff9>] ? process_one_work+0x181/0x322
  [<c044d06d>] process_one_work+0x1f5/0x322
  [<f8bddff3>] ? ieee80211_request_scan+0x41/0x41 [mac80211]
  [<c044ed1c>] worker_thread+0xfc/0x203
  [<c042d894>] ? complete+0x39/0x43
  [<c044ec20>] ? manage_workers+0x169/0x169
  [<c0451c89>] kthread+0x67/0x6c
  [<c0451c22>] ? __init_kthread_worker+0x47/0x47
  [<c07f9e3a>] kernel_thread_helper+0x6/0x10
Code: 8b 56 14 8b 80 f4 07 00 00 89 45 b8 e8 a0 70 d8 ff 8b 55 dc 80 7a 10 00 74 08 a1 40 9a 9e c0 89 46 28 83 7d d4 00 74 0f 8b 4d d4 <8b> 01 83 f8
EIP: [<f8d620ff>] ath_tx_start+0x640/0x6ab [ath9k] SS:ESP 0068:f4e59d88
CR2: 000000000000004d

>
> Signed-off-by: Felix Fietkau<nbd@openwrt.org>
> Reported-by: Ben Greear<greearb@candelatech.com>
> ---
>   drivers/net/wireless/ath/ath9k/main.c |    5 +++++
>   drivers/net/wireless/ath/ath9k/xmit.c |    6 +++++-
>   2 files changed, 10 insertions(+), 1 deletions(-)
>
> diff --git a/drivers/net/wireless/ath/ath9k/main.c b/drivers/net/wireless/ath/ath9k/main.c
> index 3de115d..3d999b7 100644
> --- a/drivers/net/wireless/ath/ath9k/main.c
> +++ b/drivers/net/wireless/ath/ath9k/main.c
> @@ -1778,6 +1778,11 @@ static int ath9k_sta_add(struct ieee80211_hw *hw,
>   	struct ieee80211_key_conf ps_key = { };
>
>   	ath_node_attach(sc, sta);
> +
> +	if (vif->type != NL80211_IFTYPE_AP&&
> +	    vif->type != NL80211_IFTYPE_AP_VLAN)
> +		return 0;
> +
>   	an->ps_key = ath_key_config(common, vif, sta,&ps_key);
>
>   	return 0;
> diff --git a/drivers/net/wireless/ath/ath9k/xmit.c b/drivers/net/wireless/ath/ath9k/xmit.c
> index 744779d..ee71d38 100644
> --- a/drivers/net/wireless/ath/ath9k/xmit.c
> +++ b/drivers/net/wireless/ath/ath9k/xmit.c
> @@ -1795,6 +1795,7 @@ static void ath_tx_start_dma(struct ath_softc *sc, struct ath_buf *bf,
>   	struct sk_buff *skb = bf->bf_mpdu;
>   	struct ieee80211_tx_info *tx_info = IEEE80211_SKB_CB(skb);
>   	struct ieee80211_hdr *hdr = (struct ieee80211_hdr *)skb->data;
> +	struct ieee80211_vif *vif = tx_info->control.vif;
>   	struct list_head bf_head;
>   	struct ath_atx_tid *tid = NULL;
>   	u8 tidno;
> @@ -1829,7 +1830,10 @@ static void ath_tx_start_dma(struct ath_softc *sc, struct ath_buf *bf,
>   		if (txctl->paprd)
>   			bf->bf_state.bfs_paprd_timestamp = jiffies;
>
> -		if (tx_info->flags&  IEEE80211_TX_CTL_CLEAR_PS_FILT)
> +		if ((vif&&  vif->type != NL80211_IFTYPE_AP&&
> +		            vif->type != NL80211_IFTYPE_AP_VLAN) ||
> +		    !ieee80211_is_data(hdr->frame_control) ||
> +		    (tx_info->flags&  IEEE80211_TX_CTL_CLEAR_PS_FILT))
>   			ath9k_hw_set_clrdmask(sc->sc_ah, bf->bf_desc, true);
>
>   		ath_tx_send_normal(sc, txctl->txq, tid,&bf_head);


-- 
Ben Greear <greearb@candelatech.com>
Candela Technologies Inc  http://www.candelatech.com

Re: [PATCH 2.6.39] ath9k: fix a regression in PS frame filter handling

[John W. Linville]

This doesn't seem to apply against wireless-2.6...

On Mon, May 09, 2011 at 10:25:47PM +0200, Felix Fietkau wrote:
> Only leave filtering enabled for AP or VLAN interfaces, clear the
> destination mask for all other interfaces.
> 
> Signed-off-by: Felix Fietkau <nbd@openwrt.org>
> Reported-by: Ben Greear <greearb@candelatech.com>
> ---
>  drivers/net/wireless/ath/ath9k/main.c |    5 +++++
>  drivers/net/wireless/ath/ath9k/xmit.c |    6 +++++-
>  2 files changed, 10 insertions(+), 1 deletions(-)
> 
> diff --git a/drivers/net/wireless/ath/ath9k/main.c b/drivers/net/wireless/ath/ath9k/main.c
> index 3de115d..3d999b7 100644
> --- a/drivers/net/wireless/ath/ath9k/main.c
> +++ b/drivers/net/wireless/ath/ath9k/main.c
> @@ -1778,6 +1778,11 @@ static int ath9k_sta_add(struct ieee80211_hw *hw,
>  	struct ieee80211_key_conf ps_key = { };
>  
>  	ath_node_attach(sc, sta);
> +
> +	if (vif->type != NL80211_IFTYPE_AP &&
> +	    vif->type != NL80211_IFTYPE_AP_VLAN)
> +		return 0;
> +
>  	an->ps_key = ath_key_config(common, vif, sta, &ps_key);
>  
>  	return 0;
> diff --git a/drivers/net/wireless/ath/ath9k/xmit.c b/drivers/net/wireless/ath/ath9k/xmit.c
> index 744779d..ee71d38 100644
> --- a/drivers/net/wireless/ath/ath9k/xmit.c
> +++ b/drivers/net/wireless/ath/ath9k/xmit.c
> @@ -1795,6 +1795,7 @@ static void ath_tx_start_dma(struct ath_softc *sc, struct ath_buf *bf,
>  	struct sk_buff *skb = bf->bf_mpdu;
>  	struct ieee80211_tx_info *tx_info = IEEE80211_SKB_CB(skb);
>  	struct ieee80211_hdr *hdr = (struct ieee80211_hdr *)skb->data;
> +	struct ieee80211_vif *vif = tx_info->control.vif;
>  	struct list_head bf_head;
>  	struct ath_atx_tid *tid = NULL;
>  	u8 tidno;
> @@ -1829,7 +1830,10 @@ static void ath_tx_start_dma(struct ath_softc *sc, struct ath_buf *bf,
>  		if (txctl->paprd)
>  			bf->bf_state.bfs_paprd_timestamp = jiffies;
>  
> -		if (tx_info->flags & IEEE80211_TX_CTL_CLEAR_PS_FILT)
> +		if ((vif && vif->type != NL80211_IFTYPE_AP &&
> +		            vif->type != NL80211_IFTYPE_AP_VLAN) ||
> +		    !ieee80211_is_data(hdr->frame_control) ||
> +		    (tx_info->flags & IEEE80211_TX_CTL_CLEAR_PS_FILT))
>  			ath9k_hw_set_clrdmask(sc->sc_ah, bf->bf_desc, true);
>  
>  		ath_tx_send_normal(sc, txctl->txq, tid, &bf_head);
> -- 
> 1.7.3.2
> 
> 

-- 
John W. Linville		Someday the world will need a hero, and you
linville@tuxdriver.com			might be all we have.  Be ready.

Re: [PATCH 2.6.39] ath9k: fix a regression in PS frame filter handling

[Felix Fietkau]

On 2011-05-10 9:39 PM, John W. Linville wrote:
> This doesn't seem to apply against wireless-2.6...
>
> On Mon, May 09, 2011 at 10:25:47PM +0200, Felix Fietkau wrote:
>>  Only leave filtering enabled for AP or VLAN interfaces, clear the
>>  destination mask for all other interfaces.
>>
>>  Signed-off-by: Felix Fietkau<nbd@openwrt.org>
>>  Reported-by: Ben Greear<greearb@candelatech.com>

Ben: You mentioned to me that you bisected the regression to my AP 
powersave related change with 2.6.39-rc6, however this change does not 
appear to be in the 2.6.39 tree at all, so you were probably running a 
bisect on w-t, which has a lot more stuff than w-2.6.

John: Doesn't seem to be relevant for .39 after all, just queue it for 
.40 then.

- Felix

Re: [PATCH 2.6.39] ath9k: fix a regression in PS frame filter handling

[Ben Greear]

On 05/10/2011 01:00 PM, Felix Fietkau wrote:
> On 2011-05-10 9:39 PM, John W. Linville wrote:
>> This doesn't seem to apply against wireless-2.6...
>>
>> On Mon, May 09, 2011 at 10:25:47PM +0200, Felix Fietkau wrote:
>>> Only leave filtering enabled for AP or VLAN interfaces, clear the
>>> destination mask for all other interfaces.
>>>
>>> Signed-off-by: Felix Fietkau<nbd@openwrt.org>
>>> Reported-by: Ben Greear<greearb@candelatech.com>
>
> Ben: You mentioned to me that you bisected the regression to my AP
> powersave related change with 2.6.39-rc6, however this change does not
> appear to be in the 2.6.39 tree at all, so you were probably running a
> bisect on w-t, which has a lot more stuff than w-2.6.
>
> John: Doesn't seem to be relevant for .39 after all, just queue it for
> .40 then.

I did run bisect on w-t, and it seems I was confused about it being in
mainline -rcX.

Sorry for the trouble.

Makes me feel a little better though...glad it wasn't in real kernels
and still I was the only one hitting it :P

Ben

>
> - Felix


-- 
Ben Greear <greearb@candelatech.com>
Candela Technologies Inc  http://www.candelatech.com