[coreboot] Success in replicating a fast boot time to the browser on a x60

[Charles Devereaux]

Hello

Sorry for this late reply, I have been a bit busy lately.

On Thu, Oct 2, 2014 at 12:24 AM, Denis 'GNUtoo' Carikli <GNUtoo at no-log.org>
wrote:

> I did a simlar setup some time ago for a narrower use case:
> I needed extremly fast reflashing times.
> I only had:
> * A stock parabola(A GNU/Linux distro based on Arch)
> * Good kernel commandline(to reduce logging)
> * A well configured grub.
> * No password at all, even for getty.
>

This is a good idea. At the moment, for simple tests I have a busybox doing
just that. Too minimal though :-)


> You could gain a lot of time by obtimising DHCP [1].
>

Indeed. Since then, I have had the best results at the moment with dhcpcd
and the following options added to the bottom of debian dhcpcd.conf:

ipv6rs
#ipv6ra_own
#ipv4only
noipv4ll
noarp

The latest 2 options are the most helpful for speed. The rest is for my
IPv6 setup.

At the moment, only dnsmasq is a bit of a hassle (used by resolvconf to do
local caching + forward request to the DNS servers given by DHCP), but I
must say I'm still using debian native script which does
ExecStart=/etc/init.d/dnsmasq systemd-exec so there's obviously a lot of
optimization to do.

I've no idea about wpa_supplicant though.
>

In my case, using priority in the network blocks and the following options
is enough:
update_config=1
eapol_version=2
fast_reauth=1
## Hidden SSIDs
#ap_scan=1

You could also replace the sleep by something better: you could look if
> you can get the status of the network connection somehow:
>

IMHO, the best idea is to start the browser, then when the connection is
established,  issue a command to the browser to open the given page.

Actually, it could be interesting to provide a minimal "demo linux distro"
that would do just what we said, to demonstrate how fast a coreboot boot
can be. Something based on parabola or debian, using systemd, dhcpcd, and
xorg without any password (where one has to provide the right xorg.conf)

It could provide a good reference, a metric.

How many miliseconds would be saved by putting grub.cfg directly inside
> grub's coreboot.cfg?
>

I don't know. I should run tests :-) I remember reading that disk access
was faster.

[1]http://cafbit.com/entry/rapid_dhcp_or_how_do
>

There was also an interesting BSD SoC project to implement this RFC 4436 in
a free software modern dhcp client, but they found nobody this year:
http://www.openbsdfoundation.org/gsoc2014.html#rfc4436
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.coreboot.org/pipermail/coreboot/attachments/20141021/0718baa7/attachment.html>