Re: LILO and kernel bootsector will not boot off an LS-120 IDE floppy.

Erik Troan (
Wed, 24 Sep 1997 16:18:34 -0400 (EDT)

As luck would have it, we just got a pair of LS-120 drives in today, hoping
they would reduce the boot time we have during testing installs (yes, I
really do that many installations).

> First let me say that I don't know what kind of animal an LS-120 is. If
> it identifies itself as an IDE disk to the system, LILO will treat it
> as such and use some probably stupid BIOS drive number. If the LS-120
> identifies itself as drive 0 ("A:") to the BIOS, something like
> disk = /dev/hdg bios = 0

This almost worked actually. I had to hack the lilo source to:

1) Treat major number 0x22 as a normal IDE device when
determing geometry.

2) The "disk=/dev/hdg bios=0" override didn't seem to
work. I ended up modifying the source to use a bios
number of 0, which worked fine.

I'd guess that the second problem was caused by some statically sized lilo
tables which can't handle parsing 8 IDE drives, but I haven't looked into
the source to see.

Once I did these things, it booted just fine. I believe I used linear, and
I know I used compact.

> If such devices are becoming popular, a long-term solution would probably
> be to detect them in the IDE drivers of the kernel and to give them
> either their own major device number or to have some ioctl that lets me
> get their real identity, so LILO can act appropriately without requiring
> special configuration.

I don't think they really call for a new major number (jaz drives don't
have one after all), but it would be nice to have better ioctl() support;
there doesn't seem to be a way of autoejecting them for example. The
ide-floppy driver in 2.0.31 seems a bit immature still. Once they are
mounted, they work fine. When they aren't mounted though, every operation
causes the kernel to reread the partition table, which takes about 20
seconds (presumably it's reinitializing the entire device).

All of this is with a FloppyMAX controller in an Intel Zappa motherboard
(well, we got that machine a while ago so I could be wrong about that) and
a Compaq LS-120 drive from CDW.

I'm actually a bit disappointed with the drive. I was hoping it's
"5x the performance of normal floppies) would help booting our normal
boot disk. It seems to boot it at almost twice the speed of a normal
drive, but that's not overly exciting. It does read information from a
LS-120 drive much quicker then that, but it takes about 20 seconds to
read the MBR of the disk and realize it should boot it. I have no
idea what it's doing in there, but the time between the beep and 'LILO'
appearing is pretty pronounced.


| "Tarball is idiomatically related to furball and hairball" -- Carl D. Roth |
| |
| Erik Troan = = |