> Hi all,
>
> Recently I saw this behaviour on linux 2.0 running on i486.
>
> Instruction invd ( supposed to invalidate cache) can lock the m/c
> completely.
>
> Put Instruction invd in assembly file, assemble and link and
> see the consequences. I am not sure whether this is a 486 bug or
> linux bug.
>
> BTW invd is a privileged instruction on Pentium I am not sure
> about 486.
>
> -pkd
> --
>
> ----------------------------------------------------------------------
> Praveen Dwivedi pkd@sequent.com
> --
>
This IS a privileged instruction. It is supposed to generate a trap
from user-mode code. I will try this at home on a 486.
The trap that is supposed to be executed, int 13h, might point to bad code
in the kernel, causing the hang. This could be due to a bad link when
compiled. If the opcode is executed in kernel mode, it will just flush the
cache with no long-standing effects, so I think that code at the trap address
might be the problem.
Cheers,
Dick Johnson
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Richard B. Johnson
Project Engineer
Analogic Corporation
Voice : (508) 977-3000 ext. 3754
Fax : (508) 532-6097
Modem : (508) 977-6870
Ftp : ftp@boneserver.analogic.com
Email : rjohnson@analogic.com, johnson@analogic.com
Penguin : Linux version 2.1.21 on an i586 machine (66.15 BogoMips).
Warning : It's hard to remain at the trailing edge of technology.
-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-