On Tue, 13 Sep 2011, Joshua Kinard wrote:
Date: Tue, 13 Sep 2011 08:43:39 -0400
From: Joshua Kinard <firstname.lastname@example.org>
Cc: email@example.com, firstname.lastname@example.org,
Subject: Re: [PATCH] Impact video driver for SGI Indigo2
On 09/11/2011 07:14, peter fuerst wrote:
+ if (ctype && strstr(ctype, "impact=early"))
I think we can gut all the Impact early printk stuff. Per Skylark, to use
it, you have to drop a function call into arch/mips/kernel/setup.c and
It's the setup_impact_earlycons() call in ip22-setup.c above. In so far the
patch is complete, one can leave setup.c and printk.c alone.
Depending on the situation, ip22-setup.c may not be early enough. The
setup_impact_earlycons-call could be moved near to the beginning of
start_kernel() (init/main.c), it doesn't depend on non-static kernel setup.
But this had to be done by hand, it won't be in the regular source.
So, given the intended purpose of the early console - providing helpfull
information simply by a boot option to a potential bug report by an user,
who has no inclination to kernel hacking, just wants to use the kernel out
of the box - early console may or may not be usefull.
so I don't think this is an easily-used feature. In my
patchset for Gentoo's mips-sources, I actually broke both Impact and
Odyssey's early console bits out as separate drivers applied as patches if
one forces a debug mode on.
Do you intend to submit these drivers? Anyway i will separate the early
console stuff from the "basic" Impact-patch, when i resubmit the patches,
so it's optional and can be discussed independently.
Because they're not really even real
consoles....just a way to get output out of offbeat systems like Octane,
which hides all the good stuff behind the broken (horifically broken) IOC3
Unless you've come up with a better way to do early printk that can work
shortly after the CPUs are brought up?
No regular/every-day usable method, only debug-hackery:
Moving the setup_impact_earlycons-call to start_kernel() or using a specially
tailored ARCS-console, which in addition allows to capture kernel output via
serial line (and helped me debugging for long).
(BTW: in recent kernels any ARCS-console must be switched off very early,
while in e.g. 2.6.14 it could be used until Impact or IP22Zilog take over)
"The past tempts us, the present confuses us, the future frightens us. And
our lives slip away, moment by moment, lost in that vast, terrible in-between."
--Emperor Turhan, Centauri Republic