des@dtg.nsc.com (Desmond Young) (03/05/90)
Hi,
just a few simple (silly) questions.
How does Minix exit protected mode ? Does it ever need to ?
To enter protected mode I believe you execute the appropriate
instruction. But, how do you exit it ?
I thought AT's sent an instruction to the keyboard to issue a
reset, with the program having already stored a flag in the
CMOS RAM. When re-booting, the program checks the RAM flag, and
if set appropriately, does not perform a full reset/re-boot.
Since this can be slow..... (reset time non-trivial), there must
be a better way. Is this a closely guarded secret at Microsoft?
The only way I can think of is to cause a multiple MMU fault.
For example, access a location that is invalid, and has an
invalid page entry in local and global descriptor tables. The
286 datasheet does not define what happens in this case. Does
the chip reset?
Does Minix ever need to get out of protected mode?
In case you are wondering why these questions. If you have an
"XT" machine with a 286 processor, you do not have a keyboard
to give you a reset, and you do not have a CMOS RAM to store a
flag. So, the questions are to determine whether you may run
Minix in protected mode on a 286-based non-AT machine.
Thanks in advance for any advice,
Des.
des@dtg.nsc.com or des@gpo.nsc.com
No opinions - no disclaimer.evans@ditsyda.oz (Bruce Evans) (03/08/90)
In article <726@blenheim.nsc.com> des@dtg.nsc.com (Desmond Young) writes: > How does Minix exit protected mode ? By resetting or turning off the processor. > Does it ever need to ? Only to reboot. >[How to reset on ATs using the keyboard controller.] > Since this can be slow..... (reset time non-trivial), there must > be a better way. Is this a closely guarded secret at Microsoft? I don't think so (on a 286). On a 386, it is easy. I used it a lot while developing the Minix protected mode kernel, running users in protected mode and the system in real mode. > The only way I can think of is to cause a multiple MMU fault. A documented way is to load the interrupt descriptor table with a bad value, the cause an interrupt. The processor "shuts down". I don't know if this means that it resets. I must have shut it down many times while debugging, and it always seemed to reset. > In case you are wondering why these questions. If you have an > "XT" machine with a 286 processor, you do not have a keyboard > to give you a reset, and you do not have a CMOS RAM to store a > flag. So, the questions are to determine whether you may run > Minix in protected mode on a 286-based non-AT machine. Do you have such a strange machine, and volunteer to make rebooting work? :-) In practice the reset button is a good way to reboot. I have had strange results rebooting an AT in real mode. My debugger catches an illegal opcode at 0xFFFF:0xFFF0. This is the usual reset vector, and examination shows the instruction there is the nornal, legal jump. Jumping to this location (which is what the kernel does if the keyboard controller doesn't reset) works fine. The debugger doesn't work in protected mode (on a 286) and the reboot looks normal. Maybe it causes a pile of exceptions and the processor shuts down? Everything works as I expect in all modes on a 386. -- Bruce Evans evans@ditsyda.oz.au