ath5k in AP mode works!! (but sometimes with an oops, and a panic)

ath5k in AP mode works!! (but sometimes with an oops, and a panic)

[henk]

First of all great work! AP mode works almost flawlessly.

The only things I encounter:
- At hostapd shutdown the kernel panics frequently, freezing up the system.
- At startup I encounter the trace below.
- In between the WPA2 sessions seem to work flawlessly.

Hope this helps.

- henk

[   37.939157] ------------[ cut here ]------------
[   37.939171] WARNING: at /home/henk/compat-wireless-2009-05-22/drivers/net/wireless/ath/ath5k/base.c:2167 ath5k_beacon_update_timers+0x42/0xcd [ath5k]()
[   37.939185] Hardware name: A9830IMS
[   37.939191] Modules linked in: usbhid snd_hda_codec_realtek iwlagn snd_hda_intel iwlcore snd_hda_codec ath5k ath mac80211 ehci_hcd snd_hwdep uhci_hcd snd_pcm snd_timer usbcore snd soundcore cfg80211 intelfb snd_page_alloc i2c_algo_bit cfbcopyarea cfbimgblt cfbfillrect
[   37.939263] Pid: 4191, comm: hostapd Not tainted 2.6.29.4 #1
[   37.939271] Call Trace:
[   37.939293]  [<c02238e1>] warn_slowpath+0x71/0x8d
[   37.939327]  [<fe764375>] ? ath5k_hw_set_fast_div+0xcf/0xd3 [ath5k]
[   37.939359]  [<fe760448>] ? ath5k_hw_set_imr+0x142/0x17a [ath5k]
[   37.939390]  [<fe766297>] ? ath5k_hw_set_sleep_clock+0x30f/0x313 [ath5k]
[   37.939420]  [<fe7672fe>] ? ath5k_hw_reset+0x9d4/0x9de [ath5k]
[   37.939451]  [<fe760ea0>] ? ath5k_hw_reset_tx_queue+0x794/0x79e [ath5k]
[   37.939484]  [<fe769ce9>] ath5k_beacon_update_timers+0x42/0xcd [ath5k]
[   37.939520]  [<fe769e1c>] ath5k_beacon_config+0xa8/0xbe [ath5k]
[   37.939551]  [<fe76a524>] ath5k_reset+0xdd/0xeb [ath5k]
[   37.939581]  [<fe7682ee>] ath5k_chan_set+0x29/0x30 [ath5k]
[   37.939611]  [<fe76a66c>] ath5k_config+0x31/0x75 [ath5k]
[   37.939657]  [<fe733210>] ieee80211_hw_config+0x76/0x7f [mac80211]
[   37.939713]  [<fe740b2d>] ieee80211_set_channel+0x26/0x28 [mac80211]
[   37.939752]  [<f8187159>] nl80211_set_wiphy+0x1e8/0x33b [cfg80211]
[   37.939808]  [<fe745d68>] ? ieee80211_master_start_xmit+0x332/0x346 [mac80211]
[   37.939831]  [<c03c7040>] ? nla_parse+0x67/0xae
[   37.939850]  [<c03c7b2f>] genl_rcv_msg+0x151/0x15d
[   37.939865]  [<c03c79de>] ? genl_rcv_msg+0x0/0x15d
[   37.939877]  [<c03c6bb1>] netlink_rcv_skb+0x38/0x76
[   37.939889]  [<c03c7b57>] genl_rcv+0x1c/0x29
[   37.939904]  [<c03c5b66>] netlink_unicast+0x125/0x184
[   37.939919]  [<c03c6354>] netlink_sendmsg+0x21c/0x22b
[   37.939935]  [<c03a9b4c>] sock_sendmsg+0xb8/0xcf
[   37.939948]  [<c03a9cf6>] ? sock_recvmsg+0xc2/0xda
[   37.939964]  [<c023356b>] ? autoremove_wake_function+0x0/0x33
[   37.939980]  [<c02db1fd>] ? copy_from_user+0x2c/0x4f
[   37.939993]  [<c02db1fd>] ? copy_from_user+0x2c/0x4f
[   37.940050]  [<c03b0ec4>] ? verify_iovec+0x40/0x74
[   37.940070]  [<c03ab04c>] sys_sendmsg+0x175/0x1dc
[   37.940090]  [<c024c94a>] ? find_lock_page+0x13/0x47
[   37.940107]  [<c024d4be>] ? filemap_fault+0x8a/0x299
[   37.940123]  [<c024c739>] ? unlock_page+0x1b/0x1e
[   37.940140]  [<c03b360a>] ? __dev_get_by_name+0x66/0x7a
[   37.940158]  [<c03a9789>] ? sock_destroy_inode+0x10/0x12
[   37.940175]  [<c027ae81>] ? destroy_inode+0x27/0x38
[   37.940192]  [<c027bee0>] ? generic_forget_inode+0x16e/0x175
[   37.940209]  [<c0278ebc>] ? __d_free+0x3c/0x41
[   37.940224]  [<c0278ef3>] ? d_free+0x25/0x37
[   37.940242]  [<c027dd41>] ? mntput_no_expire+0x16/0xdf
[   37.940260]  [<c03ab3a6>] sys_socketcall+0x141/0x16b
[   37.940280]  [<c026ca57>] ? sys_close+0x50/0x82
[   37.940297]  [<c0202d05>] sysenter_do_call+0x12/0x25
[   37.940310] ---[ end trace 1be5347d05de050b ]---
[   38.220995] hostapd used greatest stack depth: 5740 bytes left

Re: ath5k in AP mode works!! (but sometimes with an oops, and a panic)

[Bob Copeland]

On Sun, May 24, 2009 at 6:20 PM, henk <henk@god.dyndns.org> wrote:
>
> First of all great work! AP mode works almost flawlessly.
>
> The only things I encounter:
> - At hostapd shutdown the kernel panics frequently, freezing up the system.
> - At startup I encounter the trace below.
> - In between the WPA2 sessions seem to work flawlessly.

Hi,

Apologies for the slow response, just returning from vacation today.

Any chance you could capture the panic at hostapd shutdown via netconsole?

> [   37.939171] WARNING: at /home/henk/compat-wireless-2009-05-22/drivers/net/wireless/ath/ath5k/base.c:2167 ath5k_beacon_update_timers+0x42/0xcd [ath5k]()

WARN_ON(!intval) -- which has been reported already for IBSS as well.
I should have
a patch shortly.

-- 
Bob Copeland %% www.bobcopeland.com

Re: ath5k in AP mode works!! (but sometimes with an oops, and a panic)

[henk]

On Wed, May 27, 2009 at 08:50:19AM -0400, Bob Copeland wrote:
> Any chance you could capture the panic at hostapd shutdown via netconsole?

Sure:
  http://bugzilla.kernel.org/show_bug.cgi?id=13381#c2

regards,
Henk

Re: ath5k in AP mode works!! (but sometimes with an oops, and a panic)

[henk]

I added some extra checks in ath5k/base.c
see http://bugzilla.kernel.org/show_bug.cgi?id=13381#c3

This resolves the panic for now.

- henk

Re: ath5k in AP mode works!! (but sometimes with an oops, and a panic)

[Bob Copeland]

On Wed, May 27, 2009 at 08:49:56PM +0200, henk wrote:
> 
> I added some extra checks in ath5k/base.c
> see http://bugzilla.kernel.org/show_bug.cgi?id=13381#c3
> 
> This resolves the panic for now.

Can you try this patch?  It should reduce to a warning if there is
still a bug but hopefully you won't see the warning.  

There's also a chance that it could break beaconing if you kill
hostapd and restart it a few times, but I did a little testing and
it seemed to work ok.

From: Bob Copeland <me@bobcopeland.com>
Date: Thu, 28 May 2009 19:33:50 -0400
Subject: [PATCH] ath5k: disable beacon interrupt when interface is down

When we remove the active interface, there's no need to continue
sending beacons; doing so would cause a null pointer deref in
ieee80211_beacon_get().  Disable the interrupt in remove_interface
and add a WARN_ON(!vif) in case there are other instances lurking.

Signed-off-by: Bob Copeland <me@bobcopeland.com>
---
 drivers/net/wireless/ath/ath5k/base.c |   17 ++++++++++++++++-
 1 files changed, 16 insertions(+), 1 deletions(-)

diff --git a/drivers/net/wireless/ath/ath5k/base.c b/drivers/net/wireless/ath/ath5k/base.c
index b6fb15e..f011c44 100644
--- a/drivers/net/wireless/ath/ath5k/base.c
+++ b/drivers/net/wireless/ath/ath5k/base.c
@@ -2070,6 +2070,13 @@ err_unmap:
 	return ret;
 }
 
+static void ath5k_beacon_disable(struct ath5k_softc *sc)
+{
+	sc->imask &= ~(AR5K_INT_BMISS | AR5K_INT_SWBA);
+	ath5k_hw_set_imr(sc->ah, sc->imask);
+	ath5k_hw_stop_tx_dma(sc->ah, sc->bhalq);
+}
+
 /*
  * Transmit a beacon frame at SWBA.  Dynamic updates to the
  * frame contents are done as needed and the slot time is
@@ -2758,6 +2765,7 @@ ath5k_remove_interface(struct ieee80211_hw *hw,
 
 	ath5k_hw_set_lladdr(sc->ah, mac);
 	sc->vif = NULL;
+	ath5k_beacon_disable(sc);
 end:
 	mutex_unlock(&sc->lock);
 }
@@ -3059,7 +3067,14 @@ ath5k_beacon_update(struct ieee80211_hw *hw, struct ieee80211_vif *vif)
 {
 	int ret;
 	struct ath5k_softc *sc = hw->priv;
-	struct sk_buff *skb = ieee80211_beacon_get(hw, vif);
+	struct sk_buff *skb;
+
+	if (WARN_ON(!vif)) {
+		ret = -EINVAL;
+		goto out;
+	}
+
+	skb = ieee80211_beacon_get(hw, vif);
 
 	if (!skb) {
 		ret = -ENOMEM;
-- 
1.6.0.6


-- 
Bob Copeland %% www.bobcopeland.com

Re: ath5k in AP mode works!! (but sometimes with an oops, and a panic)

[henk]

On Thu, May 28, 2009 at 07:47:02PM -0400, Bob Copeland wrote:
> On Wed, May 27, 2009 at 08:49:56PM +0200, henk wrote:
> > 
> > I added some extra checks in ath5k/base.c
> > see http://bugzilla.kernel.org/show_bug.cgi?id=13381#c3
> > 
> > This resolves the panic for now.
> 
> Can you try this patch?  It should reduce to a warning if there is
> still a bug but hopefully you won't see the warning.  
> 
Sure now compiling 

...
> +
>  /*
>   * Transmit a beacon frame at SWBA.  Dynamic updates to the
>   * frame contents are done as needed and the slot time is
> @@ -2758,6 +2765,7 @@ ath5k_remove_interface(struct ieee80211_hw *hw,
>  
>  	ath5k_hw_set_lladdr(sc->ah, mac);
>  	sc->vif = NULL;
> +	ath5k_beacon_disable(sc);

Wouldn't it be better to reverse the deinitialisation? To make sure we
dont hit an interrupt in between? Like:

	ath5k_beacon_disable(sc);
	ec->vif = NULL;

Cheers
Henk

Re: ath5k in AP mode works!! (but sometimes with an oops, and a panic)

[Bob Copeland]

On Fri, May 29, 2009 at 6:02 PM, henk <henk@god.dyndns.org> wrote:
> Wouldn't it be better to reverse the deinitialisation? To make sure we
> dont hit an interrupt in between? Like:
>
>        ath5k_beacon_disable(sc);
>        ec->vif = NULL;

Yeah, actually I gave this some thought but I can't now rationalize
the current order, so maybe I just goofed.

-- 
Bob Copeland %% www.bobcopeland.com

Re: ath5k in AP mode works!! (but sometimes with an oops, and a panic)

[henk]

On Fri, May 29, 2009 at 09:06:57PM -0400, Bob Copeland wrote:
> On Fri, May 29, 2009 at 6:02 PM, henk <henk@god.dyndns.org> wrote:
> > Wouldn't it be better to reverse the deinitialisation? To make sure we
> > dont hit an interrupt in between? Like:
> >
> > ? ? ? ?ath5k_beacon_disable(sc);
> > ? ? ? ?ec->vif = NULL;
> 
> Yeah, actually I gave this some thought but I can't now rationalize
> the current order, so maybe I just goofed.
> 
> -- 
> Bob Copeland %% www.bobcopeland.com

I have tested hostapd+ath5k (only master mode without VLANs) for a few days
now with the new patches. System seems stable no glitches found.

I have aggregated your patches against compat-wireless-2009-06-01:
	http://bugzilla.kernel.org/show_bug.cgi?id=13381#c5

I would say a new milestone in ath5k driver development has been reached 
with working AP support.

Thanks!

Henk