Ubuntu 10.04 LTS: The Lucid Lynx

There are a couple of intel ones that are supposed to work just fine.
Just google for "replacing wireless card aspire one".

 

For wireless Problems see solution provided here:
https://bugs.launchpad.net/ubuntu/+sour ... bug/568090

Change your access point settings from “WPA2+AES/TKIP” to “WPA2+AES”.

Edit: Mention solution in launchpad bug.

 

Thanks for the feedback. Unfortunately, I run WEP so it is compatible with some legacy kit I've got, so I don't think it is that. I am going to buy an Intel card.

 

eek. wtf.... double post

 

hi there peeps...

sorry I have not been back to my thread in ages but you know how it is with work and stuff....

It appears I have found a fix for the flaky wifi drivers in Ubuntu lucid lynx 10.04. All credits goes to Subban over on his blog..

http://subbass.blogspot.com/2010/05/howto-ubuntu-1004-lucid-post-install.html

in short, you need to install

Code:

 linux-backports-modules-wireless-lucid-generic

to do this you need to enable the backports repos from the package manager. then install it from the synaptic package manager. It will also install another file which is the actual driver for your current kernal version...

i did this a few hours ago now and wifi has been solid for me ever since... I even tried it out downloading a ' linux distro ' ;¬) with no stalls or drop outs at all...

way to go Subban !!!

He also has a little bit of a tweak to get the media slot working (the one on the right) properly. not fully tested this....

I hope this helps a few of you out... it helped me big time, I was just getting to my limits with the flaky wifi and was gonna move over to fedora as I hear wifi is solid...

Marty.

 

I tried the wireless backports, and performance seemed worse than out of the box. I wonder if various updates to 10.04 LTS have resolved wifi problems?

AA1 D150, OCZ Vertex SSD.
Ethernet controller: Atheros Communications Inc. AR5001 Wireless Network Adapter (rev 01)

Basically, I'm taking the "if it ain't broke, don't fix it" approach.

 

All this is odd for me. I installed Lucid around alpha2, and built madwifi-hal sometime in early March (_before_ 3/24 snapshot).
I needed to do the supend_modules and blacklist stuff, but LEDs work right, and I have not had any "drops"--occasional slow connections from indoors, at a distance such that Windows will never do anything besides connecting.
Maximum reported signal percent is lower (70-90% when ath5k was 90+), but reliability is better and it would connect at lower signal strength. Right now, connection is 14%.
Ibidem

 

Some news and possible fixes for the ath5k driver.
(because for me, even after installing backports-wireless connections still drop under high loads)

This has been discussed in the past few weeks in these two bug reports, and at least a cause for the problem seems to be well understood:
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/369005
https://bugs.launchpad.net/ubuntu/+source/linux/+bug/461419

If you do:

Code:

sudo lspci -vv

you should get a bit regarding the wireless card that reads (at least in 110, 150, ZG5 models):

Code:

LnkCtl:	ASPM L0s L1 Enabled;

I don't know what ASPM is, but it's bad, very bad, and we want it to be disabled. Jussi Kivilinna figured this out in the kernel bug report, and came up with a patch here (comment #18):
https://bugzilla.kernel.org/show_bug.cgi?id=13892

I applied the patch, but there are additional modifications that need to be done to get it to compile for the Lucid 10.04 kernel (see comment #19). ASPM did get disabled, but it was very hard to connect, so I'll forget about it for now... maybe it only works for later kernels. It also appears from these threads that maybe enabling CONFIG_PCIEASPM in the kernel might be enough to fix it, and no patch is needed.

POSSIBLE FIXES:

1. Apply the patch to some compat-wireless version, make changes, get it to compile... not good results for me.
2. Wait for the patch to be merged and fixed, and when that happens, compile compat-wireless (or install backports-wireless).
3. Recompile your stock kernel enabling the right option. I'll try this one soon.
4. Use a later kernel from the Ubuntu Kernel PPA. I think these do not have the wireless problem at all (though might have other problems... I think the enable_mtrr_cleanup fix doesn't work on those)
5. Some combination of the above.

EDIT:
You can also check the discussion on whether the patch is going to be accepted:
http://thread.gmane.org/gmane.linux.ker ... eral/51512

 

Recompiled the Lucid kernel and it worked.
All that was needed was enabling CONFIG_PCIEASPM in the kernel config.
Uploading 1Gb to an ssh server with no problem at all, while before there were several disconnects.

If you don't want to recompile, I reckon installing a later kernel from the next Ubuntu will do the trick as well (I prefer to stay within the supported kernel version, though). The output from "sudo lspci -vv" confirms that aspm is disabled. It now reads:

Code:

LnkCtl:	ASPM Disabled;

(of course, using Madwifi instead of ath5k is still probably the simplest option)

 

That sounds like it's pretty solid now. It would be nice to see a kernel package (.deb) with this, I think...
If anyone dares to do that, there are a few other changes from default I'd suggest:
1. Optimize for Atom
2. Lose some unnecessary stuff, like MCA, EISA, Dec ethernet, etc.
3. pciehp should be built as a module-- that will make configuration the standard way (after boot) feasible again.
Currently, pciehp is builtin, which is a pain (must reboot to set it up).

Anyone want to try?

 

Those changes would be pretty neat. I won't be looking into it soon, because I'm trying Fedora out at the moment.
In the meantime, here are the .debs of the recompiled kernel:
http://dl.dropbox.com/u/8737818/lin....6.32.15+drm33.5-noaspm-10.00.Custom_i386.deb
http://dl.dropbox.com/u/8737818/lin....6.32.15+drm33.5-noaspm-10.00.Custom_i386.deb

For those wanting to role your own, here are the commands I used:

Code:

sudo apt-get build-dep --no-install-recommends linux-image-$(uname -r)
sudo aptitude install fakeroot kernel-package
mkdir ~/src
cd ~/src
apt-get source linux-image-$(uname -r)
cd ~/src/linux-2.6.32
make oldconfig
gedit ~/src/linux-2.6.32/.config # to add CONFIG_PCIEASPM=y
export CONCURRENCY_LEVEL=3
make-kpkg clean
fakeroot make-kpkg --initrd --append-to-version=-noaspm kernel-image kernel-headers
sudo gdebi linux-image-2.6.32.15+drm33.5-noaspm_2.6.32.15+drm33.5-noaspm-10.00.Custom_i386.deb
sudo gdebi linux-headers-2.6.32.15+drm33.5-noaspm_2.6.32.15+drm33.5-noaspm-10.00.Custom_i386.deb
sudo update-initramfs -c -k 2.6.32.15+drm33.5-noaspm
sudo update-grub

 

Interesting--Madwifi works well even when ASPM is enabled, it doesn't have its own way to disable it.

Code:

03:00.0 Ethernet controller: Atheros Communications Inc. AR5001 Wireless Network Adapter (rev 01)

----snip-----------
		LnkCtl:	ASPM L0s L1 Enabled; RCB 128 bytes Disabled- Retrain- CommClk+
			ExtSynch- ClockPM- AutWidDis- BWInt- AutBWInt-
		LnkSta:	Speed 2.5GT/s, Width x1, TrErr- Train- SlotClk+ DLActive- BWMgmt- ABWMgmt-
	Capabilities: [90] MSI-X: Enable- Mask- TabSize=1
		Vector table: BAR=0 offset=00000000
		PBA: BAR=0 offset=00000000
	Capabilities: [100] Advanced Error Reporting <?>
	Capabilities: [140] Virtual Channel <?>
	Kernel driver in use: ath_pci
	Kernel modules: ath_pci, ath5k

 

I switched CONCURRENCY_LEVEL to 4 (an Atom w/ hyperthreading has 4 hardware threads, since x86 has 2 / "core"),
made changes suggested, set max cores to 2 (48kb smaller kernel), disabled Dell, Toshiba, Sony, Compaq, HP, & Compal stuff, and built it. There's still tons of cruft, removing which would reduce build time by at least an hour... it took four hours to build the debs I have.
2.6.32 has an interesting trick or two, which I tried before:
make localmodconfig |localyesconfig
Localmodconfig creates a fairly standard kernel, except it won't build modules that aren't loaded.
localyesconfig creates a kernel that should work without loading modules: whatever's loaded becomes builtin.

Now I need to see where I can upload to.

EDIT:
Look here:http://cid-f785bf5e55218c8e.office.live.com/browse.aspx/.Public/linux/deb?uc=1&lc=1033

 

Gentlemen,

I could not boot both custom kernels, I'm guessing because I'm running ubuntu from an sd card. Although, it was strange that the grub entry did not list any initramfs associated with the kernel. Indeed, there was nothing in /boot.

I was able to solve the issue on my aspire one 110 by installing a kernel from maverick, as recommended in the comments in this bug report https://bugs.launchpad.net/ubuntu/+sour ... bug/369005

Quote:
Alright, it's working for me now, after installing a backported maverick kernel from the kernel ppa. For those of you who don't feel comfortable with recompiling the kernel this seems to be the way to go:
1) Add deb http://ppa.launchpad.net/kernel-ppa/ppa/ubuntu lucid main to your package sources
2) Install linux-lts-backport-maverick
3) Restart.

Works like a charm. Now I'm gonna remove the ppa from my sources and lock the kernel version in synaptic. Thanks everyone!

 

Just in case someone decides to try my solution out, the backported kernel actually screws the graphics performance because the mtrr cleanup thing stops working. Oh linux....

 

True... So far as I know, enable_mtrr_cleanup doesn't really do anything in the newer kernels.This seems to be the case in several different distributions...

Never bothered to check whether this is being discussed somewhere, or if anyone is working on solving it.
To me, it's quite clear that graphic performance gets worse, but I guess this really only affects AAO users.

 

It seems that to fix the mtrr thing one needs to go back to old fixes... There's this one.
(I've just tried in Fedora, but the same should work for ubuntu)

My /proc/mtrr is the same as the "bad" one in this blog post:
http://blog.jolexa.net/2009/11/22/buggy-mtrr-on-acer-aspire-one-zg5/

Then as root, I ran the script found here and placed it in rc.local:
http://www.alcancelibre.org/article.php/corrigiendo-performance-graficos-aaone
(you can download it from here: http://dl.dropbox.com/u/8737818/fixmtrr.sh)

Now my /proc/mtrr looks like the "good" one in the original post.

The crucial bit is the "write-combining" part. The base and size can be easily determined by looking at lspci -v for the video card. But if all the MTRRs are occupied, one needs to resort to this black magic of disabling them in a certain order... or whatever.

I don't actually know it this improves performance, but it's good enough for me.
Sometimes it's like Ceno said: "oh, linux...."

 

Since a lot of the posts in this thread have related to performance, I thought there might be a post here, or somewhere else on this forum about a recent Linux kernel patch that provides a noticeable boost in performance.

http://www.webupd8.org/2010/11/alternat ... patch.html

I applied the "Lennart solution" to my AA1 D150 netbook about a week ago. It's relatively easy to do, and can be backed out with little effort. In my experience, there is a significant performance improvement in scrolling, web page load speed, and video playback -- all areas where netbook performance tends to be lacking.

It's well worth the 5 minutes or so it takes to make these changes, and, as far as I can see, there are no downside risks. Also, much less work is involved that building a custom kernel. I'm sure this will eventually appear in an "official" Linux kernel, but since I am currently running 2.6.32-26 with Ubuntu 10.04 LTS, and the Phoronix article talks about it not appearing until 2.6.37 at the earliest, I didn't feel like waiting. When this feature IS supported in a new kernel, the alternative code will (probably) need to be removed.

 

AOA150 ("ZG5") here, definite improvement...I used the script (http://launchpadlibrarian.net/59511828/cgroup_patch), which takes <1 minute to do. I ran /etc/rc.local, instead of rebooting.