On Mon, 26 Nov 2001, Adam Agnew wrote:
> 2) Somewhere in that ext2 code or in the polled ide code is a bug. This
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
This is exactly the reason that, int the long term, I want Linux as my
bootloader. Boot loaders are old technology, at least 30 years old. This
kind of problem always seems to crop up sooner or later with boot loaders.
What happened with Sun is a good example. On the original Sun there was a
simple bootloader. At some point the bootloader could understand the file
system. Then you basically had two file system implementations -- one in
bootloader, one in kernel. At some point, the craziness reached a cycle:
the bootloader was built by ripping bits of kernel source over into the
bootloader tree -- bits of NFS, bits of file system, other stuff. Ouch!
We're still paying the price for this type of thing. On some Alphas, the
BIOS can only boot from BSD partitions. You can't have other partition
types for the boot partition!
I am really impressed at all the things the various bootloaders can do,
with ext2, ide drivers, net drivers, etc., etc.
But here's a neat bootloader we have now: it boots over Myrinet. It will
soon boot over SCI. It boots from ext3, reiserfs, and xfs file systems. It
can use IP multicast to load kernels and initrds.
All this stuff works with this bootloader. No other bootloader can equal
this bootloader in capability. The neat bootloader is 'Linux'. I
understand the reasons for doing other bootloaders. But, it also is a
shame that we have to do them at all. They're always going to be way
behind what is possible with Linux as the bootloader.
> Altogether, this is ext2fs code, polled ide code, then everything that
> comes with etherboot when you compile for the sis900. It weighs in at
> around 24k this way. Not too shaby..
Good stuff. Please don't take what I am saying as a criticism in any way.
Thanks
ron
