> > > I think that problem is delays in ne2000 driver or
> > > somewhere in kernel. Those delays seems to be broken due loop
> > > optimisations in cyrix cpu. I hope that problem will be fixed
> > > soon :)
> >
> > The shouldnt be. We use two kinds of delays inb_p/outb_p generate an
> > extra I/O cycle on the ISA bus. udelay precomputes timing according to the
> > CPU. Thus for example the MMX pentium does loops far faster than the pentium
> > but the timing adjusts accordingly.
>
> But in fact something working wrong :). As I wrote exchange
> cyrix cpu on intel or isa ne2000 on amd pci has solved all my problems.
> In a few days I will be able to test pci ne2000. If it will working
> fine with cyrix, then the problem is isa delays. Are you agree? Should
> I test other isa cards like 3com?
The Cyrix CPUs have a configurable delay between I/O instructions
which defaults to 32 *bus clock* cycles for the 6x86 (I think the
5x86 is the same). This should be sufficient for normal cases but
it could be wrong if (a) the motherboard jumpers aren't set correctly
and a smart BIOS tries to adjust the delay, (b) the BIOS knows
about 6x86 I/O delays, allows you to configure it, and you get it
wrong, (c) you have applied the latest cyrix patch and enabled
the fast I/O option.
In all these case there will *only* be a problem if the device
driver does not include dummy I/O instructions or delay loops
in *all* the necessary places.
This should *only* be a problem with ISA cards. Some of the
old ISA hardware just can't cope with receiving back to back
I/O cycles (there were no such things when they were designed).
Mike
-- .----------------------------------------------------------------------. | Mike Jagdis | Internet: mailto:mike@roan.co.uk | | Roan Technology Ltd. | | | 54A Peach Street, Wokingham | Telephone: +44 118 989 0403 | | RG40 1XG, ENGLAND | Fax: +44 118 989 1195 | `----------------------------------------------------------------------'