Skip to contentSkip to conversation

2.6.2{7, 8} with compat-wireless-2008-12-3{0, 1) drops link every 2-3 minutes

ath9k-devel Joseph Cheek 3 participants 12 messages Jan 1, 2009 – Jan 15, 2009 802.11 (including cfg80211/NL80211)

lore
Share & tools
lore
#0 lore raw

Hi all,

Happy New Year! It's about 2:30 AM here and I'm up debugging my shiny new DWA-552 <AR5416+AR2133> XtremeN Desktop Adapter. Stock 2.6.28 gave me the dreaded "No ProbeResp from current AP" in syslog with a disconnect every few minutes, even after updating |IEEE80211_MONITORING_INTERVAL to (100 * HZ). It's highly annoying. I thought compat-wireless might help. It seems to do a little bit better but that's subjective. I still get a disconnect every few minutes. iwconfig and my DIR-655 show plenty of signal strength but it doesn't respond to pings and my ssh sessions all get stuck in limbo.

What can I do to debug this? This is an old Athlon XP 2000+ box running Gentoo (if it helps). I've seen a few other probs on this list recently but they don't seem similar to mine enough to warrant tacking this msg on to an existing thread.

Thanks!

Joseph Cheek
joseph at cheek.com

PS: stock 2.6.27 seemed to disconnect less often but still did. It also seemed to have memory leaks - I couldn't restart wlan0 more than a few times before I get OOM errors.
|

#1 lore raw
Quoted 7 lines · Joseph Cheek

Hi all,

Happy New Year! It's about 2:30 AM here and I'm up debugging my shiny new DWA-552 <AR5416+AR2133> XtremeN Desktop Adapter. Stock 2.6.28 gave me the dreaded "No ProbeResp from current AP" in syslog with a disconnect every few minutes, even after updating |IEEE80211_MONITORING_INTERVAL to (100 * HZ). It's highly annoying.

This should be due to the fact that the RX filter disabled beaconing from the current AP and mac80211 expects beacons from the AP. This was fixed in wireless-testing but I do not see this pushed out for 2.6.28.

Greg, this was fixed in Linus' tree via ffb826767bffda61426d964a8fc24a216a14b0bd, merged by John on Fri, 21 Nov 2008, the patch is titled:
"ath9k: enable RXing of beacons on STA/IBSS". Just in case I've provided the hunk below. Without this patch users will face this disconnect issue on noisy environments where the beacons don't reach mac80211 within a certain time frame dicated by mac80211.

Quoted 1 line

I thought compat-wireless might help.

It should resolve the "No ProbeResp from current AP" in" issue at least.

Quoted 9 lines

It seems to do a little bit better
but that's subjective. I still get a disconnect every few minutes. iwconfig and my DIR-655 show plenty of signal strength but it doesn't respond to pings and my ssh sessions all get stuck in limbo.

What can I do to debug this? This is an old Athlon XP 2000+ box running Gentoo (if it helps). I've seen a few other probs on this list recently but they don't seem similar to mine enough to warrant tacking this msg on to an existing thread.

That may be due to a separate issue, but to review that we need more information, such as the type of AP you are connecting to, encryption options used, mode of operation (11n, 11b or 11a), your type of environment (noisy or a quiet home). To review this issue you can remove stable at kernel.org from the To or Cc list.

Luis

net/wireless/ath9k/recv.c renamed from net/wireless/ath9k/recv.c.orig +2 -3
@@ -627,9 +627,8 @@
 		rfilt &= ~ATH9K_RX_FILTER_UCAST;
 	}
 
-	if (((sc->sc_ah->ah_opmode == ATH9K_M_STA) &&
-	     (sc->rx_filter & FIF_BCN_PRBRESP_PROMISC)) ||
-	    (sc->sc_ah->ah_opmode == ATH9K_M_IBSS))
+	if (sc->sc_ah->ah_opmode == ATH9K_M_STA ||
+	    sc->sc_ah->ah_opmode == ATH9K_M_IBSS)
 		rfilt |= ATH9K_RX_FILTER_BEACON;
 
 	/* If in HOSTAP mode, want to enable reception of PSPOLL frames
#2 lore raw

[ath9k-devel] 2.6.27 with compat-wireless-2008-12-31 stops transmitting every 2-3 minutes

Thanks Luis! Comments below...

Quoted 23 lines · Luis R. Rodriguez
Quoted 8 lines · Joseph Cheek

Hi all,

Happy New Year! It's about 2:30 AM here and I'm up debugging my shiny new DWA-552 <AR5416+AR2133> XtremeN Desktop Adapter. Stock 2.6.28 gave me the dreaded "No ProbeResp from current AP" in syslog with a disconnect every few minutes, even after updating |IEEE80211_MONITORING_INTERVAL to (100 * HZ). It's highly annoying.

This should be due to the fact that the RX filter disabled beaconing from the current AP and mac80211 expects beacons from the AP. This was fixed in wireless-testing but I do not see this pushed out for 2.6.28.

Greg, this was fixed in Linus' tree via ffb826767bffda61426d964a8fc24a216a14b0bd, merged by John on Fri, 21 Nov 2008, the patch is titled:
"ath9k: enable RXing of beacons on STA/IBSS". Just in case I've provided the hunk below. Without this patch users will face this disconnect issue on noisy environments where the beacons don't reach mac80211 within a certain time frame dicated by mac80211.

Quoted 5 lines
Quoted 2 lines

I thought compat-wireless might help.

It should resolve the "No ProbeResp from current AP" in" issue at least.

I haven't been getting that message in syslog recently (since the 2nd). I moved back to 2.6.27 with compat-wireless-12-31 as that seems the most stable with the highest speeds. I also specified the AP on the iwconfig command line, thinking that perhaps that would help.

Quoted 11 lines
Quoted 10 lines

It seems to do a little bit better
but that's subjective. I still get a disconnect every few minutes. iwconfig and my DIR-655 show plenty of signal strength but it doesn't respond to pings and my ssh sessions all get stuck in limbo.

What can I do to debug this? This is an old Athlon XP 2000+ box running Gentoo (if it helps). I've seen a few other probs on this list recently but they don't seem similar to mine enough to warrant tacking this msg on to an existing thread.

Quoted 5 lines

That may be due to a separate issue, but to review that we need more information, such as the type of AP you are connecting to, encryption options used, mode of operation (11n, 11b or 11a), your type of environment (noisy or a quiet home). To review this issue you can remove [email protected] from the To or Cc list.

based on my recent observations, I believe you are correct - this is a separate issue. I no longer get the noProbeResp error but I still get transmission problems every 2 minutes or so under heavy loads. They don't result in actual disconnects - ifconfig shows the interface up and iwconfig shows it connected to the AP. So the link is no longer being dropped. But for about 30 seconds, I can neither ping nor ssh in or out of that box. If leave the system up and don't cancel anything, the transmission resumes after the 30 seconds and then stalls again after 2 minutes or so.

My AP is a D-Link DIR-655 running 1.21 firmware in g/n mode. No encryption is enabled currently (I do have MAC filtering on, but nothing else). The particular box with the problem is downstairs and there is some interference between it and AP, although the DIR-655 generally shows 60-80% signal and up to 81Mb/s connection speed. It doesn't happen with light traffic, I have to be watching a streaming video or scp-ing a file across the network. Then it is 100% repeatable.

If I restart the network (/etc/init.d/wlan0 restart) I can get a good connection/signal with good traffic without having to wait 30 seconds - for two minutes again 8-(.

This is at home and most of my neighbors have 80211g wireless set up. I generally see at least two other APs from iwlist wlan0 scanning. I haven't pinned either the AP or the NIC to a channel (6 is normal for US but they are connecting on 3). Not sure if that makes a difference.

My laptop right next to the AP has no problems staying connected and transmitting and receiving with stock 2.6.27. It's running a DWL-652 (AR5416).

btw, when i do get good transmission, I only get about 1.25MB/s at the most. I can get that from my old g card 8-(.

I can try anything you need me to with respect to ath9k drivers or kernels. does it matter if bluetooth and/or ipv6 is enabled? both are enabled on this box but not on my laptop?

thanks!

Joseph Cheek

Quoted 18 lines

Luis

net/wireless/ath9k/recv.c renamed from net/wireless/ath9k/recv.c.orig +2 -3
@@ -627,9 +627,8 @@
 		rfilt &= ~ATH9K_RX_FILTER_UCAST;
 	}
 
-	if (((sc->sc_ah->ah_opmode == ATH9K_M_STA) &&
-	     (sc->rx_filter & FIF_BCN_PRBRESP_PROMISC)) ||
-	    (sc->sc_ah->ah_opmode == ATH9K_M_IBSS))
+	if (sc->sc_ah->ah_opmode == ATH9K_M_STA ||
+	    sc->sc_ah->ah_opmode == ATH9K_M_IBSS)
 		rfilt |= ATH9K_RX_FILTER_BEACON;
 
 	/* If in HOSTAP mode, want to enable reception of PSPOLL frames
#3 lore raw

[ath9k-devel] 2.6.27 with compat-wireless-2008-12-31 stops transmitting every 2-3 minutes

Quoted 80 lines · Joseph Cheek

Thanks Luis! Comments below...

Quoted 23 lines · Luis R. Rodriguez
Quoted 8 lines · Joseph Cheek

Hi all,

Happy New Year! It's about 2:30 AM here and I'm up debugging my shiny new DWA-552 <AR5416+AR2133> XtremeN Desktop Adapter. Stock 2.6.28 gave me the dreaded "No ProbeResp from current AP" in syslog with a disconnect every few minutes, even after updating |IEEE80211_MONITORING_INTERVAL to (100 * HZ). It's highly annoying.

This should be due to the fact that the RX filter disabled beaconing from the current AP and mac80211 expects beacons from the AP. This was fixed in wireless-testing but I do not see this pushed out for 2.6.28.

Greg, this was fixed in Linus' tree via ffb826767bffda61426d964a8fc24a216a14b0bd, merged by John on Fri, 21 Nov 2008, the patch is titled:
"ath9k: enable RXing of beacons on STA/IBSS". Just in case I've provided the hunk below. Without this patch users will face this disconnect issue on noisy environments where the beacons don't reach mac80211 within a certain time frame dicated by mac80211.

Quoted 5 lines
Quoted 2 lines

I thought compat-wireless might help.

It should resolve the "No ProbeResp from current AP" in" issue at least.

I haven't been getting that message in syslog recently (since the 2nd). I moved back to 2.6.27 with compat-wireless-12-31 as that seems the most stable with the highest speeds. I also specified the AP on the iwconfig command line, thinking that perhaps that would help.

Quoted 11 lines
Quoted 10 lines

It seems to do a little bit better
but that's subjective. I still get a disconnect every few minutes. iwconfig and my DIR-655 show plenty of signal strength but it doesn't respond to pings and my ssh sessions all get stuck in limbo.

What can I do to debug this? This is an old Athlon XP 2000+ box running Gentoo (if it helps). I've seen a few other probs on this list recently but they don't seem similar to mine enough to warrant tacking this msg on to an existing thread.

Quoted 5 lines

That may be due to a separate issue, but to review that we need more information, such as the type of AP you are connecting to, encryption options used, mode of operation (11n, 11b or 11a), your type of environment (noisy or a quiet home). To review this issue you can remove stable at kernel.org from the To or Cc list.

based on my recent observations, I believe you are correct - this is a separate issue. I no longer get the noProbeResp error but I still get transmission problems every 2 minutes or so under heavy loads. They don't result in actual disconnects - ifconfig shows the interface up and iwconfig shows it connected to the AP. So the link is no longer being dropped. But for about 30 seconds, I can neither ping nor ssh in or out of that box. If leave the system up and don't cancel anything, the transmission resumes after the 30 seconds and then stalls again after 2 minutes or so.

My AP is a D-Link DIR-655 running 1.21 firmware in g/n mode. No encryption is enabled currently (I do have MAC filtering on, but nothing else). The particular box with the problem is downstairs and there is some interference between it and AP, although the DIR-655 generally shows 60-80% signal and up to 81Mb/s connection speed. It doesn't happen with light traffic, I have to be watching a streaming video or scp-ing a file across the network. Then it is 100% repeatable.

If I restart the network (/etc/init.d/wlan0 restart) I can get a good connection/signal with good traffic without having to wait 30 seconds - for two minutes again 8-(.

This is at home and most of my neighbors have 80211g wireless set up. I generally see at least two other APs from iwlist wlan0 scanning. I haven't pinned either the AP or the NIC to a channel (6 is normal for US but they are connecting on 3). Not sure if that makes a difference.

Quoted 6 lines

My laptop right next to the AP has no problems staying connected and transmitting and receiving with stock 2.6.27. It's running a DWL-652 (AR5416).

btw, when i do get good transmission, I only get about 1.25MB/s at the most. I can get that from my old g card 8-(.

Thanks for the details, this seems to be related to noise and therefore could be due to some issues with how we deal with ANI. We will have to try to reproduce this ourselves now and try to fix it.

Quoted 3 lines

I can try anything you need me to with respect to ath9k drivers or kernels. does it matter if bluetooth and/or ipv6 is enabled? both are enabled on this box but not on my laptop?

Anything communication on the same spectrum affects your 802.11 connection, so using the microwave oven, using bluetooth, or a 2.4 GHz home phone should affet your connection. Not sure the freq used by Xbox / PS3 controllers... Bluetooth particularly should affect your 802.11 connection and we only recently added bluetooth coexistance to help with this but this is only available for more modern cards so your card does not have this feature.

If you are using bluetooth then it should affect your connection if using the 2.4 GHz spectrum on your wirless AP.

Luis

#4 lore raw

[ath9k-devel] 2.6.27 with compat-wireless-2008-12-31 stops transmitting every 2-3 minutes

Is this page up to date? I found the latest 802.11n NIC that I could find for my machine...

http://wireless.kernel.org/en/users/Drivers/ath9k#ProductsintheretailmarketcontainingAtheros11nsolutions

I bought my card after specifically consulting this page.

thanks!

Joseph Cheek

Quoted 14 lines · Luis R. Rodriguez

Anything communication on the same spectrum affects your 802.11 connection,
so using the microwave oven, using bluetooth, or a 2.4 GHz home phone should affet your connection. Not sure the freq used by Xbox / PS3 controllers... Bluetooth particularly should affect your 802.11 connection and we only recently added bluetooth coexistance to help with this but this is only available for more modern cards so your card does not have this feature.

If you are using bluetooth then it should affect your connection if using the 2.4 GHz spectrum on your wirless AP.

Luis

#6 lore raw

[ath9k-devel] 2.6.27 with compat-wireless-2008-12-31 stops transmitting every 2-3 minutes

Quoted 9 lines · Luis Rodriguez
Quoted 6 lines · Joseph Cheek

Is this page up to date? I found the latest 802.11n NIC that I could find for my machine...

http://wireless.kernel.org/en/users/Drivers/ath9k#ProductsintheretailmarketcontainingAtheros11nsolutions

I bought my card after specifically consulting this page.

Yes.

Just FYI the Bluetooth coexistance stuff seems to be a feature of at least AR9280 revision 10 and later so if you have an older card than you don't have BT coexistance, at least from what I can tell.

Luis

#7 lore raw

[ath9k-devel] 2.6.27 with compat-wireless-2008-12-31 stops transmitting every 2-3 mins

In reply to Luis R. Rodriguez

thanks, Luis! comments below...

Quoted 4 lines · Luis R. Rodriguez

Thanks for the details, this seems to be related to noise and therefore could be due to some issues with how we deal with ANI. We will have to try to reproduce this ourselves now and try to fix it.

great! i wish you luck! would it help it i try the same test with the problematic box right next to the AP? I could move it upstairs to the same room and retry an SCP or video serve. not sure if that would give you more useful data or not, but i can certainly move it.

if the box were quiet enough, i would keep it in the same room and just connect via ethernet 8-).

Quoted 12 lines
Quoted 4 lines

I can try anything you need me to with respect to ath9k drivers or kernels. does it matter if bluetooth and/or ipv6 is enabled? both are enabled on this box but not on my laptop?

Anything communication on the same spectrum affects your 802.11 connection, so using the microwave oven, using bluetooth, or a 2.4 GHz home phone should affet your connection. Not sure the freq used by Xbox / PS3 controllers... Bluetooth particularly should affect your 802.11 connection and we only recently added bluetooth coexistance to help with this but this is only available for more modern cards so your card does not have this feature.

Quoted 3 lines

If you are using bluetooth then it should affect your connection if using the 2.4 GHz spectrum on your wirless AP.

I didn't actually have the bluetooth in use, just the module built. so that might not be the problem. the DIR-655 will do 5MHz only if it's in n-only mode, and I have too many g devices in the house to want to replace them all 8-).

Quoted 2 lines

Luis

#8 lore raw

[ath9k-devel] 2.6.27 with compat-wireless-2008-12-31 stops transmitting every 2-3 minutes

In reply to Joseph Cheek

Hi,

looks like i misspoke. my laptop, this morning, lost its connection for about 30 seconds while running a remote X app. I just sshed to another box (running a g card with a broadcom chipset 8-), exported the display back to my laptop, and started ripperX. About 2/3 into the rip the connection stalled for about 30 seconds and then came back. perhaps i didn't see it before because i don't do any video serving from my laptop? unsure. i do watch the videos there, and i would expect the bandwidth utilization to be the same. at least it seems bandwidth related.

speaking of the broadcom card, it reassociated with the access point automatically after a reboot (of the access point). both d-link/atheros cards needed some work to get it to reassociate - the desktop needed a reboot and the laptop needed a reinsertion of the pcmcia card. possibly rmmod/modprobe ath9k would have worked to get it to reassociate, but /etc/init.d/net.wlan0 restart did not work.

thanks for your help!

Joseph Cheek

Quoted 4 lines · Joseph Cheek

My laptop right next to the AP has no problems staying connected and transmitting and receiving with stock 2.6.27. It's running a DWL-652 (AR5416).

#9 lore raw
In reply to Luis R. Rodriguez
Quoted 13 lines · Luis R. Rodriguez
Quoted 7 lines · Joseph Cheek

Hi all,

Happy New Year! It's about 2:30 AM here and I'm up debugging my shiny new DWA-552 <AR5416+AR2133> XtremeN Desktop Adapter. Stock 2.6.28 gave me the dreaded "No ProbeResp from current AP" in syslog with a disconnect every few minutes, even after updating |IEEE80211_MONITORING_INTERVAL to (100 * HZ). It's highly annoying.

This should be due to the fact that the RX filter disabled beaconing from the current AP and mac80211 expects beacons from the AP. This was fixed in wireless-testing but I do not see this pushed out for 2.6.28.

Quoted 6 lines

Greg, this was fixed in Linus' tree via ffb826767bffda61426d964a8fc24a216a14b0bd, merged by John on Fri, 21 Nov 2008, the patch is titled:
"ath9k: enable RXing of beacons on STA/IBSS". Just in case I've provided the hunk below. Without this patch users will face this disconnect issue on noisy environments where the beacons don't reach mac80211 within a certain time frame dicated by mac80211.

Thanks, I'll queue this up for 2.6.28, but it doesn't apply to 2.6.27-stable. If you want it there, can you please provide a backport for it?

thanks,

greg k-h

#10 lore raw
Quoted 21 lines · Greg KH
Quoted 13 lines · Luis R. Rodriguez
Quoted 7 lines · Joseph Cheek

Hi all,

Happy New Year! It's about 2:30 AM here and I'm up debugging my shiny new DWA-552 <AR5416+AR2133> XtremeN Desktop Adapter. Stock 2.6.28 gave me the dreaded "No ProbeResp from current AP" in syslog with a disconnect every few minutes, even after updating |IEEE80211_MONITORING_INTERVAL to (100 * HZ). It's highly annoying.

This should be due to the fact that the RX filter disabled beaconing from the current AP and mac80211 expects beacons from the AP. This was fixed in wireless-testing but I do not see this pushed out for 2.6.28.

Quoted 6 lines

Greg, this was fixed in Linus' tree via ffb826767bffda61426d964a8fc24a216a14b0bd, merged by John on Fri, 21 Nov 2008, the patch is titled:
"ath9k: enable RXing of beacons on STA/IBSS". Just in case I've provided the hunk below. Without this patch users will face this disconnect issue on noisy environments where the beacons don't reach mac80211 within a certain time frame dicated by mac80211.

Quoted 3 lines

Thanks, I'll queue this up for 2.6.28, but it doesn't apply to 2.6.27-stable. If you want it there, can you please provide a backport for it?

Thanks Greg -- it seems this was actually present in 2.6.27 since the initial commit of ath9k. From the initial commit I see:

--- on ath_calcrxfilter()

        if (sc->sc_opmode == ATH9K_M_STA || sc->sc_opmode == ATH9K_M_IBSS ||
            sc->sc_scanning)
                rfilt |= ATH9K_RX_FILTER_BEACON;

---

Seems we're good on 27 for this and this actually turned out to be a regression then.

Luis

#11 lore raw
Quoted 29 lines · Luis R. Rodriguez
Quoted 21 lines · Greg KH
Quoted 13 lines · Luis R. Rodriguez
Quoted 7 lines · Joseph Cheek

Hi all,

Happy New Year! It's about 2:30 AM here and I'm up debugging my shiny new DWA-552 <AR5416+AR2133> XtremeN Desktop Adapter. Stock 2.6.28 gave me the dreaded "No ProbeResp from current AP" in syslog with a disconnect every few minutes, even after updating |IEEE80211_MONITORING_INTERVAL to (100 * HZ). It's highly annoying.

This should be due to the fact that the RX filter disabled beaconing from the current AP and mac80211 expects beacons from the AP. This was fixed in wireless-testing but I do not see this pushed out for 2.6.28.

Quoted 6 lines

Greg, this was fixed in Linus' tree via ffb826767bffda61426d964a8fc24a216a14b0bd, merged by John on Fri, 21 Nov 2008, the patch is titled:
"ath9k: enable RXing of beacons on STA/IBSS". Just in case I've provided the hunk below. Without this patch users will face this disconnect issue on noisy environments where the beacons don't reach mac80211 within a certain time frame dicated by mac80211.

Quoted 3 lines

Thanks, I'll queue this up for 2.6.28, but it doesn't apply to 2.6.27-stable. If you want it there, can you please provide a backport for it?

Thanks Greg -- it seems this was actually present in 2.6.27 since the initial commit of ath9k. From the initial commit I see:

Quoted 7 lines

--- on ath_calcrxfilter()

        if (sc->sc_opmode == ATH9K_M_STA || sc->sc_opmode == ATH9K_M_IBSS ||
            sc->sc_scanning)
                rfilt |= ATH9K_RX_FILTER_BEACON;

---

Seems we're good on 27 for this and this actually turned out to be a regression then.

Thanks for confirming it,

greg k-h