Cutting down bootup time (to 5 sec)

[VinzC]

[Jaglover]

http://www.coreboot.org/FAQ

Has anyone tried it? I'd be interested in opinions.
Long thread, sorry if this has been discussed already.

[haarp]

[Jaglover]

From what I read there I understand the board has not to be supported, it's enough if chips on that board are supported. I may give it a try when I find some spare time, I have an Athlon/ASUS box for experimenting.

[haarp]

It's definitely not that simple. Have a look at the Wiki, you will see that often in detail, many aspects still don't work. Also be careful when flashing Coreboot. If you use a broken build or something else happens during the flash, your board can get bricked.

[Shining Arcanine]

I just took a stop watch to my laptop and did a reboot. When I saw the monitor backlight go on before the BIOS was visible, I started the stop watch. I hit enter as soon as I saw grub and stopped the stop-watch when I saw the login screen. My stop watch read 16.95 seconds. This includes all phases of boot (i.e. BIOS, GRUB, Kernel, Userland and KDM).

[albright]

[Twenynge]

I just got around to migrating to OpenRC, Baselayout-2 and Dhcpcd-5 and my boot time got a bit better, but I feel like it still isn't fast enough considering my hardware. Will post a bootchart shortly, but in the meantime, can someone tell me what Blkid is? It seems to be taking up quite a bit of time.

Thanks in advance...

[mikegpitt]

[VinzC]

[shgadwa]

I applied the patches... it did shave a few seconds off my boot. I'm down to 27 instead of 31-33. It seems to be wicd that takes 2-3 secs and then I got sama too. And speaking of Samba... I can't even get the thing to work, yet it adds time to my boot process.

Anyhow, for these patches to cut boot down to 5 secs... I'd say thats not true. It really did not help that much. I wonder if I did anything wrong? I'm using 2.6.32-zen3.

[pianosaurus]

[tipp98]

My apologies if this has already been discussed, I started reading through the whole thread and then resolved to searching each page with /checkfs.

There are a few scripts that I want to add to a periodic shutdown script, such as disk cleanup, file syncs, and the item that lead me to this thread, disk checks. As I'm sure you have all experienced, this is a long process and can sometimes occur at inopportune times.

I am setting this up but have a few concerns...

1) If I move the script,
I understand that ext3 does not need to be run after an unclean dismount, but are there times when it must be run, such as with forcecheck, and if so, does the checkfs script still need to run to trigger the check, or will it automatically run before being mounted, fail to mount, mount dangerously, etc.?

2) If I copy the script,
If max count triggers during a shutdown, fine, but if it triggers during a reboot, not fine. I don't want the script to run during a reboot, but if it is left in the startup runlevel, it may. Is there a way to read the count so that I may trigger the check prior to reaching max count?

3) Reboot required,
What will happen if a check is performed and a reboot is required? will it automatically reboot instead of finishing the shutdown?

cheers,
Kyle

[tipp98]

haha that's funny. After spending a bunch of time figuring this out I decided to upgrade to ~ openrc and psych! there's the option to do just what I was asking about in /etc/conf.d/fsck. However, there is still the vulnerability of running a full disk check on reboot. So, my #2 question still stands.

[mikegpitt]

I was reading more about moblin today, after trying out their live image (which really does boot quick to a full X11 environment). There is an article here that discusses moblin fastboot technology:
http://moblin.ironhorseinteractive.com/2009/03/moblin-2-alpha%E2%80%99s-fast-boot-amazes/

It is pretty light on details, but one thing that jumped out at my was the fastinit script, which I've never heard of before. I did a little more searching and found this thread on the arch forums:
http://bbs.archlinux.org/viewtopic.php?id=69086

It would be interesting to see if such techniques would work on Gentoo... any voluenteers?

[Letharion]

I now have an SSD.
I've done a new amd64 install, with 2.6.33-rc7 and btrfs. I haven't taken the time install bootchart and get exact numbers, but the perceived time grub -> log in -> usable desktop is < 10 seconds.
And that's without any kind of tweaking what so ever, such as openrc.

All of these things are very interesting, and it would be awesome if things started even faster, but an ssd (to no-ones surprise I believe) seems to make the greatest impact.

Edit: Bah, my perception of time sucks It was more like 20. Poking around some with openrc, ifplugd and stuff I got to 15.

[Shining Arcanine]

I have something to add to this thread. I noticed that hal has a use flag for crypt, which is used for encrypted file systems. If you do not use an encrypted file system on your computer, you can compile it without crypt support (i.e. echo "sys-apps/hal -crypt" >> /etc/portage/package.use) to remove the need for about a dozen libraries that likely load with hal. I could not detect a significant difference with the change via a stopwatch, but the boot after the kernel handed things over to OpenRC felt faster.

[andretti]

I find customizing rc scripts can make a big difference, as one can code those scripts with prior knowledge about hardware, installed programs and their uses.

I'm also currently in the process of customizing inittab and bypass openrc altogether. My plan is to use kernel's devtmpfs for coldplugs and udev for hotplugs, and start X as soon as possible, and login via screensaver as I'm the only user... (ie. have init script lock the screen immediately after X is started.)

Initial result looks quite promising...

[Letharion]

[andretti]

[Xytovl]

This is impressive !
I see you can improve the kernel boot time, on my eeePC init starts at about 1.5s.
It is almost all we can compare as my boot is CPU boundand I don't want to lose features that I have in gnome, networkmanager and such heavy programs.

[Xytovl]

I have seen that during init, hwclock may take some time, almost half a second on my eeePC. In the kernel there is an option under rtc to set time during boot, making hwclock useless, I have added a "return 0" at the beginning of hwclock and nothing is changed.

[Letharion]

I note that there is an accepted GSoC project "Decreasing Gentoo boot time using Upstart"

http://socghop.appspot.com/gsoc/student_project/show/google/gsoc2010/gentoo/t127230759529

[Jim6]

[dmpogo]