Posted: Mon Jan 11, 2016 18:23 Post subject: [DEVEL] Possible help w/5G drop-outs TL-WDR3x00, 4x00 models
This entry has been edited to remove observations and conclusions that were not repeatable. See my latest posts in this thread below.
========
TP-Link TL-WDR3500, 3600, 3700(?), 4300 issues ->
I'm going to try to post updates here occasionally with info of value from the thread below.
Major problem currently noted for these models: 5GHz band will drop out. Some users with this problem have seen it happen after approx. one hour of uptime or 5G connection time.
Prior/last reported good build is betas/2014/12-22-2014-r25697/tplink_tl-wdr3600v1 as found in reports on other threads for these models. (FWIW, I have seen minor 5G problems in the field with r25697 as well, but sporadic, far less frequent and not really troublesome. This was also true for stock TP-Link firmware.)
I'm one of a number of people with problems running TP-Link TL-WDR3500, 3600, 3700, 4300 and similar model routers. When using dd-wrt the 5G band will disappear. My model is the TL-WDR3600 v1.5.
This appears to happen under load. It may be wireless connection load or data transmission. I set one box up at home to try to reproduce the problem, identical model/version and firmware. It doesn't fail at home, only in the field. I've been using the last reported "good" build, betas_2014_12-22-2014-r25697_tplink_tl-wdr3600v1, but even that one has this problem. (From what I've read drop-outs are possibly worse on more recent builds.)
=> It's important to note that I saw occasional 5G dropouts when running stock TP-Link firmware which was upgraded to the latest version.
I've set up remote syslogging and captured entries like these:
Jan 11 13:57:13 kernel: [32219.430000] ath: phy0: Failed to stop TX DMA, queues=0x002!
Jan 11 17:15:03 kernel: [44089.540000] ath: phy0: Failed to stop TX DMA, queues=0x102!
I'm not familiar enough with dd-wrt's inner workings to know if these are helpful. No other error messages are logged. I can't even say for certain that these messages are logged at the same time as the 5G band disappears. (FYI, I reboot this production unit every midnight.)
Are any developers willing and able to look into this? I have one spare TL-WDR3600 I can donate ASAP and would be willing to round up more.
These are fairly inexpensive units and work quite well for small/medium-sized office deployments with the sole exception of the 5G band problem. Fixing this would be a BIG win for me and quite a few other dedicated dd-wrt users.
1.) If any developers are interested, please let me know by PM or replying to this. We'll settle the details at once and I'll send the unit(s) to you.
2.) If other users don't mind, a "Me, too!" reply to this thread might help gauge interest in this fix for devs. If I'm the only poor sap having trouble it's probably not worth anyone's time.
TIA
Clemmitt
Last edited by cmsigler on Fri Jan 15, 2016 20:13; edited 2 times in total
Posted: Mon Jan 11, 2016 18:31 Post subject: Re: [DEVEL] Possible help w/5G drop-outs TL-WDR3x00, 4x00 mo
cmsigler wrote:
Good afternoon,
I'm one of a number of people with problems running TP-Link TL-WDR3500, 3600, 3700, 4300 and similar model routers. When using dd-wrt the 5G band will disappear. My model is the TL-WDR3600 v1.5.
This appears to happen under load. It may be wireless connection load or data transmission. I set one box up at home to try to reproduce the problem, identical model/version and firmware. It doesn't fail at home, only in the field. I've been using the last reported "good" build, betas_2014_12-22-2014-r25697_tplink_tl-wdr3600v1, but even that one has this problem. (From what I've read drop-outs are possibly worse on more recent builds.)
=> It's important to note that I saw occasional 5G dropouts when running stock TP-Link firmware which was upgraded to the latest version.
I've set up remote syslogging and captured entries like these:
Jan 11 13:57:13 kernel: [32219.430000] ath: phy0: Failed to stop TX DMA, queues=0x002!
Jan 11 17:15:03 kernel: [44089.540000] ath: phy0: Failed to stop TX DMA, queues=0x102!
I'm not familiar enough with dd-wrt's inner workings to know if these are helpful. No other error messages are logged. I can't even say for certain that these messages are logged at the same time as the 5G band disappears. (FYI, I reboot this production unit every midnight.)
Are any developers willing and able to look into this? I have one spare TL-WDR3600 I can donate ASAP and would be willing to round up more.
These are fairly inexpensive units and work quite well for small/medium-sized office deployments with the sole exception of the 5G band problem. Fixing this would be a BIG win for me and quite a few other dedicated dd-wrt users.
1.) If any developers are interested, please let me know by PM or replying to this. We'll settle the details at once and I'll send the unit(s) to you.
2.) If other users don't mind, a "Me, too!" reply to this thread might help gauge interest in this fix for devs. If I'm the only poor sap having trouble it's probably not worth anyone's time.
TIA
Clemmitt
P.S.: I wanted to post any bug tickets already filed but haven't received my sign-up confirmation e-mail. Will post related tickets if I find any.
Here is my "Me too" contribution. I own a 4300v1.3 and do have the same problem. The only difference is that 2014_12-22-2014-r25697 is working fine for me. As I mentioned in other post, 2.4Ghz performance is not as good as I have noticed in newer builds, but I do need 5Ghz working.
The only thing I am getting is that sometimes, maybe once or twice a day, I lose internet connection when I am on the 5Ghz network. But, I just need to connect to the 2.4Ghz network and go back to the 5Ghz network, and the internet connection is working again.
I'm "me too"-ing this as well. I'm getting rather frustrated with my parents constantly running to me when their Rokus suddenly can't get a network connection.
I find that the older builds were a lot better with respect to this problem. I believe about a year ago I had never had this issue before on DD-WRT, hopefully that helps narrow it down slightly.
I'd love to see a dev knuckle down on this issue because I've always been a big DD-WRT fan, because it has a much better UI than stock and a lot more feature support.
Posted: Wed Jan 13, 2016 18:21 Post subject: Re: [DEVEL] Possible help w/5G drop-outs
This entry has been edited to remove observations and conclusions that were not repeatable. See my latest posts in this thread below.
========
Hi,
I have syslogging from my TL-WDR3600 running r28647.
[[Unrepeatable info removed]]
The log is attached. I'll add a note to ticket 5118 about this observation.
Next I'll try the newest release, 01-13-2016-r28788.
=> NOTE that there are new US firmware versions for the TL-WDR3500, TL-WDR3600 and TL-WDR4300. Not sure what the differences are, I'll ask on the forum. Thanks
Clemmitt
Last edited by cmsigler on Fri Jan 15, 2016 20:16; edited 1 time in total
Posted: Thu Jan 14, 2016 1:25 Post subject: Re: [DEVEL] Possible help w/5G drop-outs
cmsigler wrote:
Running r28788 (01-13-2016), 5G connection drops every hour exactly.
Looking at syslog, I see this when the router reboots:
<31>Jan 13 23:06:12 process_monitor[1441]: Restarting cron (time sync change)
<31>Jan 13 23:06:12 process_monitor[1441]: We need to re-update after 3600 seconds
<30>Jan 13 23:06:12 process_monitor[1441]: set timer: 3600 seconds, callback: ntp_main()
The first entry seems innocuous. But from the last two it seems obvious to me, from a n00b perspective, given that the drop-out occurs at one hour, that this may be the root of the problems.
My signal dropped out, not exactly 3600 seconds later, but close enough:
<30>Jan 14 00:06:24 hostapd: ath1: STA d8:50:e6:12:07:82 IEEE 802.11: deauthenticated due to local deauth request
<30>Jan 14 00:06:32 hostapd: ath1: STA d8:50:e6:12:07:82 IEEE 802.11: authenticated
<30>Jan 14 00:06:32 hostapd: ath1: STA d8:50:e6:12:07:82 IEEE 802.11: associated (aid 1)
<30>Jan 14 00:06:32 hostapd: ath1: STA d8:50:e6:12:07:82 RADIUS: starting accounting session 0000000B-00000001
<30>Jan 14 00:06:32 hostapd: ath1: STA d8:50:e6:12:07:82 WPA: pairwise key handshake completed (RSN)
It also dropped and reconnected at the two hour mark:
<30>Jan 14 01:06:25 hostapd: ath1: STA d8:50:e6:12:07:82 IEEE 802.11: deauthenticated due to local deauth request
<30>Jan 14 01:06:32 hostapd: ath1: STA d8:50:e6:12:07:82 IEEE 802.11: authenticated
<30>Jan 14 01:06:32 hostapd: ath1: STA d8:50:e6:12:07:82 IEEE 802.11: associated (aid 1)
<30>Jan 14 01:06:32 hostapd: ath1: STA d8:50:e6:12:07:82 RADIUS: starting accounting session 0000000B-00000002
<30>Jan 14 01:06:32 hostapd: ath1: STA d8:50:e6:12:07:82 WPA: pairwise key handshake completed (RSN)
FYI.
Clemmitt
That's awesine Clemmitt. Let's hope that devs will take a look at your findings.
Posted: Thu Jan 14, 2016 1:59 Post subject: Re: [DEVEL] Possible help w/5G drop-outs
samgu wrote:
That's awesine Clemmitt. Let's hope that devs will take a look at your findings.
Thanks , but I'm afraid my info may not help
Finally sat down to do some Googling and found the "We need to re-update after 3600 seconds" and "set timer: 3600 seconds, callback: ntp_main()" log messages are regarded as normal operation in a bunch of other threads reporting problems (but not related to 5G drops).
For me, when I was testing r25697 (12-22-2014) these two messages did NOT show up. But maybe they are unrelated to our problem after all(?).
Posted: Thu Jan 14, 2016 13:36 Post subject: Re: [DEVEL] Possible help w/5G drop-outs
This entry has been edited to remove observations and conclusions that were not repeatable. See my latest posts in this thread below.
========
Good morning, all,
Two more observations:
[[Unrepeatable info removed]]
2.) In other threads about WDR3600 (and also 3500 and 4300, I believe) the problem with the 5G band is described as a "drop-out." From what I've read in the threads the 5G band disappears and wireless clients can't see it any more. The fix appears as simple as hitting the "Apply Settings" button on the Wireless (Basic Settings) page (or rebooting, which is also a "fix").
[[Erroneous info removed]]
There may be a 5G "drop-out" problem still present in the newest builds under heavier load than my at-home testing. I've seen drop-out (5G disappears) happen occasionally in the field when running old build r25697 on WDR3600 v.1.5. I don't have any way to test this on the newest builds right now.
Again, HTH. Many thanks to devs, esp. BS(!), and to any other end-users playing along
Clemmitt
Last edited by cmsigler on Fri Jan 15, 2016 20:24; edited 1 time in total
Posted: Fri Jan 15, 2016 20:08 Post subject: Re: [DEVEL] Possible help w/5G drop-outs
Hi,
In all my testing I've tried to be thorough and rigorous, but yesterday afternoon I had that sinking feeling that I was on a wild goose chase. Turns out that feeling was right. I'll be editing out most of the observations and conclusions I posted above in this thread. I can't reproduce those results now, and things are working better, so I've got that going for me, which is nice. (The Lama's a big hitter....)
I'll bet any devs following the thread knew the syslog entries I recorded weren't an indication of a problem Live and learn and do some more testing....
1.) The "process_monitor[nnnn]: We need to re-update after 3600 seconds", "process_monitor[nnnn]: set timer: 3600 seconds, callback: ntp_main()" and "hostapd: ath1: STA d8:yy:zz:aa:bb:82 IEEE 802.11: deauthenticated due to local deauth request" and "WPA: group key handshake completed (RSN)" vs. "WPA: pairwise key handshake completed (RSN)" syslog messages are almost certainly red herrings. My clients aren't getting disconnected every time I get the "pairwise" message. That seems to have been a coincidence. Also, my tablet seems to have been a bad actor. After a cold reboot of it more of the problems I saw have gone away. My apologies for not being more careful!!!
2.) I'm running the latest firmware, r28788 (01-13-2016), on my WDR3600 v.1.5 right now and having no connectivity issues on the 5G band or 2.4G band.
3.) I have a very low load on this test unit because it's at home. I have no way to test it with heavy connection/data load ATM.
==> 4.) My sole recommendation is: If you have problems, log in via ssh (or telnet) and do 'erase nvram' immediately followed by 'reboot' as KrypteX recommends, then set the router up by hand from scratch (web interface). That _MAY_ have helped eliminate disconnects I was seeing. <==
It may be that the 5G signal drop-outs (which require clicking on Apply Settings on wireless page or a reboot to recover from) are only triggered with many connects/disconnects and/or moderate to heavy data transfer. My only unit in the field is running r25697 (12-22-2014) and I won't be upgrading until there's a better handle on this.
If anyone wants to test the WDR3500, 3600, 4300 I'd suggest trying the latest dd-wrt revision as per the recommendation of KrypteX, GreyDoc, et al. Please start a new thread if you see problems. Perhaps users of these units can catch a break then and begin to iron out any remaining issues
Sorry for excess noise, and thanks to devs especially and all others for your help!
I am not sure if this is good or bad news for me ;-) I still got my disconnections around one & two hour so your theory had it's charm and a possibly (near) solution. Reality bites...
Posted: Fri Jan 15, 2016 22:08 Post subject: Re: [DEVEL] Possible help w/5G drop-outs
One final, quick follow-up:
If you have a problem with a TP-Link WDR-model router with a client that disconnects once every hour at exactly an hour, CHECK your client for config problems, etc., or RELOAD the driver and reconnect the client or REBOOT the client.
I do still see disconnect/reconnects, but the problem is probably due to a flaky, weird client. It's possible that the router/client interaction is also causing a problem.
