Re: The lilo problem
On 2000-05-22 at 21:39 -0500, Jeff Licquia wrote:
> On Sat, May 20, 2000 at 10:38:37AM +1000, Hamish Moffatt wrote:
> > On Fri, May 19, 2000 at 05:10:55PM -0400, Mike Bilow wrote:
> > > I personally cannot think of any logical reason why it is desirable to
> > > have a zImage instead of a bzImage kernel, unless the machine is broken.
> >
> > And broken machines are a good enough reason, IMHO. My Toshiba laptop
> > requires zImage kernels, and so did my previous noname clone desktop.
>
> Still? My Dell laptop requires zImage kernels from the 2.0.x series,
> but works fine with bzImage kernels from 2.2.x.
This was really a bug in the Linux memory allocator, not the Dell.
> It was my understanding that the problem was caused by a bug in
> switching to protected mode in these machines' BIOSes, and that a
> workaround exists in the startup code for 2.2.x.
Very few BIOSes that I know about actually have broken Int 15h handlers.
My understanding was that a lot of this fell out of the licensing terms
which Award gave to manufacturers: at one time, Award could not force
manufacturers to use the latest version of their BIOS, but actually
provided financial disincentives to do so by charging higher fees for the
newer versions. As a result, many less ethical manufacturers were
shipping BIOS v3.03 which had known bugs, while Award had released through
v3.12 -- and those manufacturers had a contractural right to do so. This
was stupid and shortsighted, as Award soon realized.
AMI also had a run of bad BIOS releases in the 1988-1989 era. These
played havoc with OS/2 2.0, which was released in 1992. AMI even had
problems with keyboard controller chips and required the "F" version as a
minimum for OS/2 compatibility; the "8" version was still widely used when
OS/2 was released. By the time Windows NT was released a year later with
exactly the same problem -- there is considerable code overlap between
OS/2 and Windows NT -- AMI had already cleaned up their problems.
To the best of my knowledge, the only i386 machines which have bugs making
them unable to load a bzImage kernel dating from after the release of OS/2
2.0 in 1992 are laptops, and then only a very few such as the Toshiba
Tecra. This was because, at its peak, OS/2 had market penetration
exceeding that of MacOS on a worldwide basis, and in some places (Germany)
almost double that. This forced the Int 15h BIOS fixes.
-- Mike
Reply to: