Linux Kernel Customization on Acer Aspire One

ah, ok thanks.

I have problems with shutdown with the updated grub, so perhaps it's not a good idea to do that! (I really hate this Acer version of Linpus Lite)

UPDATE: after 2nd reboot no more problems, not sure why first reboot had an issue. Use macles tab method if in doubt.

 

Just wanted to say thanks for the walk-through. Its been ages since I re-compiled a kernel and your walk-through made it painless.

Oh and so far no issues at all during start-up nor shutdown .
Thanks again!

 

Cheers. One benefit of kernel compiling on the Aspire One is, ironically, that it is pretty quick! This is due to Acer omitting so many device drivers in the config file.

Unfortunately this also means you will have to compile kernel modules for nearly all extra hardware you attach (If you stick with Linpus)

One tip:
After the initial compile, don't do 'make clean' before subsequent compiles, then you can compile extra kernel modules in seconds, simply by running 'sudo make menuconfig' (to select the device driver' and then 'sudo make modules'. This is the main purpose of kernel customisation for me, and I'm glad others are finding it useful and painless.

 

awesome thread, helped me compile my own kernel without breaking anything for the first time

quick question: has anyone tried compiling in powertop support? its this one kernel option called CONFIG_TIMER_STATS, and i'm having a heck of a time with it. I compiled the kernel no problems, it's there in the config file, but powertop ain't seeing it. anyone else try this yet?

 

annafil, re powertop, i have compiled a kernel to run with it. the .config file i placed on
http://scarydevilmonastery.net/config.2.6.27.

 

:dances: thanks Bushmills! Will try this

 

re "possible to run latest kernel" - I use 2.6.27rc5, was compiling it about 10 times to incrementally improve its operation on the Acer Aspire One (A110). I wonder whether it would make sense to start a repository of .config files, as the linpus config wouldn't work very well first (problems like no keyboard input because compiled as module, and something fishy in the linpus init. I have left linpus behind now, but that shouldn't affect kernel configuration.

As login/getty i like qingy - no need for an X display manager while still having a graphical login prompt, which can also log into console. the X server will be started on the Virtual console where login took place (not on the conventionally chosen VC7), and therefore multiple X sessions can be started as well, one per getty (ehm, per qingy). I should maybe make a series of screenshots, instead of elaboration.

 

i made some qingy screenshots on the AA1 (the L110 version).

quingy in idle mode, with screen saver enabled.


showing a not completed root login. the previous session per user is saved, and shows when user name has been typed in. root would log into a text shell on the same virtual console.


changing user name to last less-privileged used changes to his last session, gnome


the "session" section is actually a pull down, key or mouse


both text console or X session open in the console where login took place. Means, we can have an X session on VC1, text sessions on VC2 VC3, another X session on VC4 etc.

My config is now such that booting into runlevel 3 brings up qingy as only getty on VC1...4
where against runlevel 2 starts ngetty (one single process for all logins) on VC1...4, and an x display manager on VC7, as fall back.

These screen shots were taken with a digicam, therefore the perspective distortions and mediocre quality.

 

sideways, re "do report if you test such a kernel" (2.6.27..):

Working:
WLAN (ath5k module)
webcam
eth0 (rt8169)
sound (snd_hda_intel model=acer), comes out of suspend ok
suspend (using s2ram and pm-utils)
WLAN kill switch
touchpad, usb, backlight dimming per Fn-left/right keys, mixer per Fn-up/down keys (and mixer programs),
Fn-F7 (touchpad on/off), Fn-F6 (backlight off)

Not working:
WLAN LED
backlight dimming per program
suspend through acpi (closing lid) - probably a matter of config, as I can suspend per command just fine, and closing the lid turns off backlight, and locks screen, as asked by software - may also simply be a matter of requiring sufficient privileges to suspend.
SD card reader if card hasn't been inserted during boot (still running old BIOS)

Not tested:
hibernate (no restore partition large enough)
microphone

 

I don't know if that would help for the WLAN LED. This comes from a Slackware guide:

http://www.thev.net/cgi-bin/awki.cgi/_A ... 2#anchor_8

See the Wireless and LED section.

 

Thanks, these seem to need the madwifi drivers, I read somewhere. With vanilla kernel sources, and ath5k, this only results in
error: "dev.wifi0.ledpin" is an unknown key

 

I'd love to see a collection of configs to compare the differences, performance, and what all works.

I found these threads too late. I've been compiling the latest 2.6.27 snapshots and have my config pruned down quite a bit. No need for external wifi drivers anymore...

My great discovery this week was fixing my poor disk performance of sub 2MB/sec from my old kernel driver config up to:

Timing buffered disk reads: 184 MB in 3.02 seconds = 60.93 MB/sec

This is an outstanding ultraportable!

 

Care to share your config with us??

 

I will try to compile 2.6.27-rc7 this night. Will remove unnecessary module but add bluetooth stuff, zaurus module and cdc_acm (to access my phone shared file and modem), usb infrared (old stuff but I need it), maybe will add bootsplash and aufs patch if possible. I hope this kernel will fix SD/MMC hotplug.

 

My current most recent config can be found under
http://scarydevilmonastery.net/aspire/config.sep27.
Compared to the previous configuration, it adds ppp filtering, and io accounting - the latter is needed for iotop.
Used to compile vanilla kernel 2.6.27rc5 from kernel.org.

/dev/sda: (8 GB internal SSD)
Timing cached reads: 1226 MB in 2.00 seconds = 612.83 MB/sec
Timing buffered disk reads: 116 MB in 3.05 seconds = 38.06 MB/sec

 

Still no luck with SD/MMC hotplug, linux-2.6.27-rc7. I have plugged it on boot. But now I have almost compatible AA1 kernel. Bootsplash works, still not understand why the expansion storage now detected by Thunar after many compilation with SD and aufs support before, coretemp, bluetooth, usbnet works! But SSD still act like UDMA33, ~30MB/s reported by hdparm. Almost same as the stock kernel.

 

Just found an interesting info:

Code:

$ dmesg
...
ata2.00: configured for UDMA/66
ata2.00: limited to UDMA/33 due to 40-wire cable
ata2.00: configured for UDMA/33
...

Now we need a kernel patch to enable UDMA/66!

 

I'm still pruning the config file. Comments welcome, since I did this from scratch:

http://rs6.risingnet.net/~dattaway/config.txt

 

Added Aspire One to ata_piix.c short cable list to activate UDMA/66 support but this patch not help to speed up SSD data access Is it related to slow speed Intel SSD?

Will try dattaway config this night and report back the result. Thanks dattaway.

 

Maybe useful for HDD modded AA1. Change drivers/ata/ata_piix.c, code sniffed:

Code:

...
    { 0x2653, 0x1043, 0x82D8}, /*ICH6 on Asus Eee 701 */
    { 0x27c4, 0x1025, 0x015b }, /*ICH7 on Acer Aspire One */
...