13 Aug
2004
13 Aug
'04
6:26 p.m.
On Fri, 2004-08-13 at 15:48, Stephen Hassard wrote:
Something that helps to is to never remove your known
working kernel
entry from your bootloader, and make your new kernel the secondary
manually selected kernel on your first boot. This way if your kernel is
hosed for some reason, you should always be able to recover your system
with minimal pain by rebooting into your default working kernel.
later,
Steve
Andrew Dahlin wrote:
99% of the time, if you compile a new kernel and it won't boot, it's
because you forgot to enable some driver that's required to boot your
machine. So, using the vendor's .config is a pretty good way to ensure
that if your old kernel worked, the new one will.
Actually, now there's an even better way. I think this feature is only
available on 2.6 kernels. The .config used to compile the running
kernel is available in /proc/config.gz, so you can do:
zcat /proc/config.gz > .config
and be pretty much guaranteed to get the exact same working config that
you have running.
Of course it's still a good idea to keep the old kernel accessible. But
you can reduce your failure rate to almost zero.
Lee
honestly, compiling a kernel is trivial once you
build one successfully.
the first time will take a few hours to get it right
(tops), but by the time
you're done, you know exactly what you're doing. i can configure a kernel
from scratch in about 20 minutes... and i'm not a hacker or anything... i
just know what i need. i'm not saying everyone should build their own
kernel, but if you install your distro's sources (with their .config), it's
fairly risk free. just back up your old kernel. [OT: if you run lilo, don't
forget to run lilo after copying over the new kernel. TONS of people don't
realize this their first time]
>