> I still have that INVLPG problem on my K5 system.
I am not fimilar with that problem, what is it? Should I be looking for
an error in the log files or what?
> The symptoms are gcc -pipe and communicator hanging during
> compilation/startup (gcc without pipe works!).
I'm right now using 2.0.31. I don't recall seeing anyproblem with gcc,
but I have seen the problem with netscape and communicator hanging when
starting. I had just assume the problem with netscape had something to do
with the kernel or something. The last I recall I think netscape
came up just fine on the 2.1.x and for some reason I thought it starting
having problems around 2.0.29 and later kernels, you might try earlier
kernel versions and 2.1.x.
I would switch to the 2.1.x series kerenls, but they seem to lock up and
2.1.60 gave me this scsi error, I'm using a ncr53c810 with the ncr53c8xx
driver and the kernel recorded this log message "end_scsi_request:
buffer-list destroyed" and I rebooted.
> Am I the only one who experiences such mysteriously failing
> programs on a K5, or is everyone using kernels compiled for
> 386 on the K5?
>
> When I compile the kernel so that it uses INV instead of INVLPG,
> everything works fine. So it's either a bug in how Linux uses
> INVLPG, or it is a CPU bug. Since an Intel P5 does not experience
> the problem in a similarly equipped system, I suspect the latter.
What are the advantages of each (from what should happen) and what does it
do?
> So if I'm not the only one with that problem, I think it should be
> documented at least. Unfortunately I don't have a small test case
> which could make it into a runtime check.
>
>
> So I'd like to get feedback from other K5 owners
> if the have similar problems or not, and what kernel configurations
> they use.
>
> Tom
>
+---------------------------------+
| David Fries |
| dfries@mail.win.org |
+---------------------------------+