[cooker-amd64] Re: Fw: RC1 and SATA Sil 3512 controller

[Duncan]

timduru posted <20040212225956.1d1d2b88.tim1@timduru.org>, excerpted
below,  on Thu, 12 Feb 2004 22:59:56 +0100:

> Has anyone been able to get it to work ? From what I've read
> the silicon driver is in the 2.6 kernel for 2.4 you need a 3rd party
> driver.
> 
> I've tried installing redhat 9, using the driver they provide on
> silicon's site and it works fine.
> Not sure about MDK and 64 bit arch though
> 
> Is it planned to include the driver for the installation process on MDK
> 9.2 for AMD ?

Last I looked at a 2.4 kernel, there was a driver there, but you are
correct, as it had marked beside it (Broken).

I'm using all 2.6 kernels now, here, self-compiled.  I started out with
the Mdk kernels as in i586 cooker (the ones that don't directly compile
for amd64), extracted the tarball sources, make menuconfig'ed my own
options (sans the supermount stuff that seems to cause so many problems,
including the blockage keeping the package from directly compiling for
amd64 as is, and with my root fs, reiserfs, built-in, so I don't have to
use an initrd/initramfs, among MANY other mods), then ran make (make dep
no longer needed with 2.6, simple "make" makes both bzimage and modules),
make modules_install, and make install.  After then correcting the
lilo.conf it set up, I ran lilo -p (since I use restricted logins).

However, I've now switched to kernels d/led and gpgv'ed directly off of
kernel.org, and have deleted all my Mdk kernels, 2.4 AND 2.6, entirely.
I'm currently running the official 2.6.3-rc2 kernel, the latest as of
yesterday, anyway.

I don't know about using that driver, as I don't have any SATA drives here
to run on, but I know it isn't listed as broken any more.  I believe last
I read, it works in SATA std mode, but don't expect working RAID, yet.

What that means re Mdk for AMD64 if you plan to run primarily on SATA
drives using that driver is this:  Procure a cheap IDE PATA drive large
enough for original installation, then compile a 2.6 kernel containing the
appropriate drivers, and switch the installation over to it.  Or.. keep /
on the PATA drive and put everything else on the SATA drive. IMO, if you
aren't able to handle that, Mdk for AMD64 isn't for you anyway, since it
still has far more bugs to deal with and far less software available for
it than for i586, so those prepared to run it should be of a certain
expertise and prepared to handle significantly more problems than your
typical i586 admin should need to worry about.

**ALSO** note that according to a recent LWN article, kernel 2.4 on AMD64
is already being depreciated, AND that devfs on AMD64 has known issues,
INCLUDING MEMORY CORRUPTION, and that Mdk relies on devfs for all their
distribs, INCLUDING the AMD64 one, further complicating things by relying
on it for supermount.

This was in a review of Mdk 9.2 for AMD64, where the author was rather
disappointed, he said, that it still used a 2.4 kernel, given the above. I
understand why it does, that it's now rather outdated, because it is based
on the now dated 9.2 i586 release and that used the 2.4 kernel, so I
consider his reaction a bit overboard.  However, IMO, if that is the case,
it is STILL inexcusable for Mdk not to have a 2.6 kernel (with devfs
disabled) available via updates, by now.  (I'm going to post something
about this to a new threat later 2nite, with appropriate links and etc.,
but haven't gotten to it just yet. Thus, please keep discussion of this
angle to that thread, if possible.)

That's yet another reason I'd not recommend Mdk for AMD64 at this point,
except for "advanced" users, willing to deal with the problems of a system
not quite up to normal reliability standards, yet.

-- 
Duncan - List replies preferred.   No HTML msgs.
"They that can give up essential liberty to obtain a little
temporary safety, deserve neither liberty nor safety." --
Benjamin Franklin