On Tue, 26 Jan 1999, Linus Torvalds wrote:
> On Tue, 26 Jan 1999, Gerard Roudier wrote:
> >
> > I suggest to get rid of the bogus readX/writeX stuff in drivers that
> > know what they are doing, until 2.3, as follows:
>
> What?
>
> The "bogus" stuff is correct. That's how readX/writeX were _designed_ to
It is correct if we limit the physical memory to 2 GB minus something on
32 bits systems, and by the way it does a useless masking for i386 arch
on remapped address.
So, it will be correct for me forever since I will never use that amount
of memory on 32 bits Intel, but the useless masking as gone away from my
systems since months.
> work. The design was the correct one: make it easy to port old drivers to
> new architectures - that was what the thing was done for in the first
> place.
I was aware of the care of having old drivers still working and I agree
that I should not have written this stuff was 'bogus'.
Btw, I have posted about this problem weeks (perhaps months) ago but
nobody replied to my post.
> It's the _new_ restrictions that people want to have that are bogus, and
> they are stuff that will wait for 2.3.x. Changing code for 2.2.x is not an
> option.
I was not asking for this stuff to be changed in 2.2, but just suggesting
some simple and quick changes to help users go beyond the 2 GB physical
limit on 32 bit Intel if they really need or just want.
After all, everybody can implement his own patches for his personal use.
Regards,
G�rard.
-
Linux SMP list: FIRST see FAQ at http://www.irisa.fr/prive/mentre/smp-faq/
To Unsubscribe: send "unsubscribe linux-smp" to [EMAIL PROTECTED]
