Massive browser slowdown with kernel 2.6.32 and RaLink RT61 [HACKED]

skottish · 2009-12-06 04:33:49

**********************************************************
While not a proper solution, there are two things that I found that can be done. The second one depends on the speed of your access point. As root:

iwconfig <your_network_card> power off
iwconfig <your_network_card> rate 54M fixed

***>>> kernel 2.6.32.3 has power saving disabled by default for rt2x00 based cards. <<<***

**********************************************************

I updated to the 2.6.32 kernel today and now I'm seeing a massive browser slowdown with the RaLink RT2561/RT61 wireless NIC. Pages are taking a long time to load and sometimes privoxy comes back with a connection can't be established message. There were no changes in my configuration files, no messages in the logs, dmesg is fine, searching turned up nothing, etc. It seems that there's some sort of DNS issue when inside of browsers (both gecko and webkit). pacman works normally and I can ping the Internet fine with no connectivity issues. This is only happening on the machine with this NIC. The Atheros in the other machine works normally. Any ideas?

deconstrained · 2009-12-06 04:48:39

I've experienced the same thing before in earlier kernel versions with that same piece of hardware, but I remember building/installing legacy drivers from source and getting better performance, albeit with occasional (but seldom) connection drops;
http://rt2x00.serialmonkey.com/wiki/index.php/Main_Page
If you've already tried this then I don't know what to say.

skottish · 2009-12-06 05:21:39

Thanks for the link. I'll check it out see how it works.

For a thread update, I can't get out with subversion or git, so I'm pretty crippled at the moment.

--EDIT--

It turns out that there is a bug report filed at kernel.org, but in the filers case there's error messages involved where I have none:

http://bugzilla.kernel.org/show_bug.cgi?id=14731

Bummer, this is my first hardware issue in a very long time.

deconstrained · 2009-12-06 09:07:23

I don't think you need subversion or git to build the serialmonkey driver; subversion/git are kinds of version control software that only developers need - so if you wanted to modify the driver and found you could actually improve it, joining their github would make it easier for you to contribute.

As for the source itself, the "web interface" link at the bottom of the download page will take you to the kernel.org git page, where clicking one of the "snapshot" links will automatically generate a tarball of it for you to download (but you have to be patient; it takes a little while).

Good luck!

Last edited by deconstrained (2009-12-06 09:08:00)

skottish · 2009-12-06 18:58:09

Thanks a lot for all of your help deconstrained. It not only didn't work, there was a while there were I thought that my system may be trashed. But, since I have solid backups, I just watched; A good excuse to reinstall wouldn't be so bad.

I couldn't find anything about this bug yesterday before I posted this thread, but today I found the (temporary) solution. The problem is in the power saving functionality of the card. To get the card to resolve DNS properly, power saving needs to be disabled:

#iwconfig <wireless_interface> power off

Fortunately the card in question is in my workstation, so I can live with full power until the kernel devs get this sorted out.

Resource:

http://patchwork.kernel.org/patch/65154/

C!HO · 2009-12-07 11:47:16

Thank you, I've had the same problem with 2.6.32 and rt61!

Miitch · 2009-12-07 16:52:44

Thank you very much! solved my problem with an intel wifi 5100 pro

EDIT:
I was wrong I still have te same problem

Last edited by Miitch (2009-12-07 20:21:26)

skottish · 2009-12-08 02:21:42

I'm glad that it's helping at least some people. For those that tried it, you've probably noticed by now that your card is completely powered down upon next reboot. Oddly enough, issuing the same command as above will get it to work again.

Cyrusm · 2009-12-10 04:11:53

HAHAHA!!!!! skottish is an hero. I never I've been troubleshooting my wireless for 2 days before I saw this post

peets · 2009-12-16 05:00:18

That seems to resolve my issues as well. Thanks!

descendent87 · 2009-12-16 15:01:58

Thanks I couldn't even connect with the new kernel, the power off trick solved it for me

jocheem67 · 2009-12-17 16:09:02

Same issue here and same solution. O had really fu^&*$ my install a cuple of days ago and had to revert to 2.6.31 where things work well...

So thanks a lot!

Edit: made a tiny script ( wlan0.sh ) to issue the power command as normal user. Gave permissions through the sudoers-file.
( As I'm pretty new to scripting this is something noteworthy:) )

Last edited by jocheem67 (2009-12-17 16:53:29)

skottish · 2009-12-22 22:26:16

I'm bumping this to get a little more readership.

The 2.6.32 kernel is inching towards core and there's no good solution to this problem. Since I discovered the command above, I'm still only getting about 70% stability compared to what I use to. If anyone is following the links above, or any hit on the subject (of which there are many now), you'll see that no progress has been made publicly. This is a major regression that is known to the kernel devs including the people at serialmonkey.

Ape · 2010-01-02 14:24:58

I just got the same issue with my rt61pci wlan card. I downgraded to 2.6.31, but I am seeking for proper fixes.

popaclay · 2010-01-02 14:43:36

Same problem here with my rt61pci wlan card. Having the same problem with Slack 13 compiled with 2.6.32.2. Tried a debian install (sidux) with 2.6.32.2 and it works fine.

skottish · 2010-01-02 23:04:12

popaclay wrote:

Same problem here with my rt61pci wlan card. Having the same problem with Slack 13 compiled with 2.6.32.2. Tried a debian install (sidux) with 2.6.32.2 and it works fine.

If it works fine there then they patched the kernel. Maybe the Arch devs will make a patch exception for this? I'll ask:

http://bugs.archlinux.org/task/17687

popaclay · 2010-01-03 05:10:39

Thanks! I tried the workaround in slack and it works there too!

ugaciaka · 2010-01-06 08:49:47

this workaraound work also with rt73usb.

Only problem is at started pc

wlan0: deauthenticating from 00:14:6c:a9:a1:9c by local choice (reason=3)
wlan0: direct probe to AP 00:14:6c:a9:a1:9c (try 1)
wlan0: direct probe responded
wlan0: authenticate with AP 00:14:6c:a9:a1:9c (try 1)
wlan0: authenticate with AP 00:14:6c:a9:a1:9c (try 2)
wlan0: authenticate with AP 00:14:6c:a9:a1:9c (try 3)
wlan0: authentication with AP 00:14:6c:a9:a1:9c timed out
wlan0: direct probe to AP 00:14:6c:a9:a1:9c (try 1)
wlan0: direct probe responded
wlan0: authenticate with AP 00:14:6c:a9:a1:9c (try 1)
wlan0: authenticate with AP 00:14:6c:a9:a1:9c (try 2)
wlan0: authenticated
wlan0: associate with AP 00:14:6c:a9:a1:9c (try 1)
wlan0: RX AssocResp from 00:14:6c:a9:a1:9c (capab=0x471 status=0 aid=1)
wlan0: associated
[9:49:32] ugaciaka@sky

but after never problem

Ape · 2010-01-06 09:00:28

Disabling power management as told fixed the main problem: there was really extreme issues with network before it. But still 2.6.31 works noticeably better than 2.6.32 and this tweak. It is visible especially when gaming. 2.6.31 had no lag at all, 2.6.32 was unplayable, and 2.6.32 with this tweak makes it playable, but there is still some lag.

Strider · 2010-01-06 20:49:15

To add my two cents worth of experience around this problem :
I have an rt73 usb dongle communicating with a wireless N/G router. I'm using kernel 2.6.32 but experienced problems with kernel 2.6.31 also.
My problem was
1) a slower response (ping times) then under windows
2) at random times (every hour or so) drop of the connection with my wireless router, especially after the computer had been idle for some time, not accessing the internet. A restart of the router and wlan0 interface resolves the problem temporarily.

At the time I was using wicd and switched to netcfg but the problem still remained. So I had too look further. I started to play around with some iwconfig options and found out that setting power off almost doubled my connection speed (ping times halved). But problem 2 still remained. After reading some threads at the serialmonkey site I found out that the current driver had problems with variable connection speeds. So I changed the rate to 54M fixed (for my dongle AND for my wireless router). However, the connection drop still occurs from time to time.

Currently I'm trying to bring the rate down to the point where my connection speed starts dropping. For the moment I'm on 6Mbit/s (and can't go any lower without decrease in connection speed).

I also notice that other distributions do suffer less from this problems so let's hope Arch devs patch our kernel

Last edited by Strider (2010-01-06 20:50:23)

skottish · 2010-01-07 03:11:36

Ape wrote:

Disabling power management as told fixed the main problem: there was really extreme issues with network before it. But still 2.6.31 works noticeably better than 2.6.32 and this tweak. It is visible especially when gaming. 2.6.31 had no lag at all, 2.6.32 was unplayable, and 2.6.32 with this tweak makes it playable, but there is still some lag.

Sadly, I'm still seeing what you are. The connection is certainly better, but I still get network lags pretty frequently. It's gotten to be so much of a hassle that I bought a new wireless adapter. I'm hoping that this one will work alright until I can finally get back to ethernet.

*** EDIT ***

The current kernel release has the disable power saving patch applied. Expect it in the repos very soon.

CHANGELOG wrote:

commit 3815270b3107d31c8b7bd69461f0bb3c350ee25c
Author: Gertjan van Wingerde <e-mail removed by skottish>
Date: Mon Dec 14 20:33:55 2009 +0100
rt2x00: Disable powersaving for rt61pci and rt2800pci.

commit 93b6bd26b74efe46b4579592560f9f1cb7b61994 upstream.

We've had many reports of rt61pci failures with powersaving enabled.
Therefore, as a stop-gap measure, disable powersaving of the rt61pci
until we have found a proper solution.
Also disable powersaving on rt2800pci as it most probably will show
the same problem.

moljac024 · 2010-01-07 10:53:47

Why not just boycott the new crappy kernels? It seems to me they butchered wireless from 2.6.31 and up

bangkok_manouel · 2010-01-07 18:09:37

skottish wrote:

Ape wrote:
Disabling power management as told fixed the main problem: there was really extreme issues with network before it. But still 2.6.31 works noticeably better than 2.6.32 and this tweak. It is visible especially when gaming. 2.6.31 had no lag at all, 2.6.32 was unplayable, and 2.6.32 with this tweak makes it playable, but there is still some lag.
Sadly, I'm still seeing what you are. The connection is certainly better, but I still get network lags pretty frequently. It's gotten to be so much of a hassle that I bought a new wireless adapter. I'm hoping that this one will work alright until I can finally get back to ethernet.
*** EDIT ***
The current kernel release has the disable power saving patch applied. Expect it in the repos very soon.
CHANGELOG wrote:
commit 3815270b3107d31c8b7bd69461f0bb3c350ee25c
Author: Gertjan van Wingerde <e-mail removed by skottish>
Date: Mon Dec 14 20:33:55 2009 +0100
rt2x00: Disable powersaving for rt61pci and rt2800pci.

commit 93b6bd26b74efe46b4579592560f9f1cb7b61994 upstream.

We've had many reports of rt61pci failures with powersaving enabled.
Therefore, as a stop-gap measure, disable powersaving of the rt61pci
until we have found a proper solution.
Also disable powersaving on rt2800pci as it most probably will show
the same problem.

just FTR, works flawlessly with 2.6.32.3

Pulson229 · 2010-03-30 07:00:18

Interesting post dude....discussion are always helpful in one way or the other. Thanks for giving out information. It's really nice and mean full.

Arch Linux

#1 2009-12-06 04:33:49

Massive browser slowdown with kernel 2.6.32 and RaLink RT61 [HACKED]

#2 2009-12-06 04:48:39

Re: Massive browser slowdown with kernel 2.6.32 and RaLink RT61 [HACKED]

#3 2009-12-06 05:21:39

Re: Massive browser slowdown with kernel 2.6.32 and RaLink RT61 [HACKED]

#4 2009-12-06 09:07:23

Re: Massive browser slowdown with kernel 2.6.32 and RaLink RT61 [HACKED]

#5 2009-12-06 18:58:09

Re: Massive browser slowdown with kernel 2.6.32 and RaLink RT61 [HACKED]

#6 2009-12-07 11:47:16

Re: Massive browser slowdown with kernel 2.6.32 and RaLink RT61 [HACKED]

#7 2009-12-07 16:52:44

Re: Massive browser slowdown with kernel 2.6.32 and RaLink RT61 [HACKED]

#8 2009-12-08 02:21:42

Re: Massive browser slowdown with kernel 2.6.32 and RaLink RT61 [HACKED]

#9 2009-12-10 04:11:53

Re: Massive browser slowdown with kernel 2.6.32 and RaLink RT61 [HACKED]

#10 2009-12-16 05:00:18

Re: Massive browser slowdown with kernel 2.6.32 and RaLink RT61 [HACKED]

#11 2009-12-16 15:01:58

Re: Massive browser slowdown with kernel 2.6.32 and RaLink RT61 [HACKED]

#12 2009-12-17 16:09:02

Re: Massive browser slowdown with kernel 2.6.32 and RaLink RT61 [HACKED]

#13 2009-12-22 22:26:16

Re: Massive browser slowdown with kernel 2.6.32 and RaLink RT61 [HACKED]

#14 2010-01-02 14:24:58

Re: Massive browser slowdown with kernel 2.6.32 and RaLink RT61 [HACKED]

#15 2010-01-02 14:43:36

Re: Massive browser slowdown with kernel 2.6.32 and RaLink RT61 [HACKED]

#16 2010-01-02 23:04:12

Re: Massive browser slowdown with kernel 2.6.32 and RaLink RT61 [HACKED]

#17 2010-01-03 05:10:39

Re: Massive browser slowdown with kernel 2.6.32 and RaLink RT61 [HACKED]

#18 2010-01-06 08:49:47

Re: Massive browser slowdown with kernel 2.6.32 and RaLink RT61 [HACKED]

#19 2010-01-06 09:00:28

Re: Massive browser slowdown with kernel 2.6.32 and RaLink RT61 [HACKED]

#20 2010-01-06 20:49:15

Re: Massive browser slowdown with kernel 2.6.32 and RaLink RT61 [HACKED]

#21 2010-01-07 03:11:36

Re: Massive browser slowdown with kernel 2.6.32 and RaLink RT61 [HACKED]

#22 2010-01-07 10:53:47

Re: Massive browser slowdown with kernel 2.6.32 and RaLink RT61 [HACKED]

#23 2010-01-07 18:09:37

Re: Massive browser slowdown with kernel 2.6.32 and RaLink RT61 [HACKED]

#24 2010-03-30 07:00:18

Re: Massive browser slowdown with kernel 2.6.32 and RaLink RT61 [HACKED]

Board footer