VOGONS


First post, by superfury

User metadata
Rank l33t++
Rank
l33t++

Can the I/O permission bitmap be disabled for operating systems not supporting it?

I know that the CPU can disable it by making it map out-of-range (base TSS offset out of range) of the TSS segment limits, as well as with mapping the base offset register out-of-range (by lowering the limit field to be below it).

But imagine a 80386+ operating system or program using the upper data as some scratch space (which would be bad in this situation, if the bitmap still is applied anyways).

Is there any way to turn it off in hardware somehow, to support emulation of older (say 386/486) processors?

Author of the UniPCemu emulator.
UniPCemu Git repository
UniPCemu for Android, Windows, PSP, Vita and Switch on itch.io

Reply 1 of 3, by crazyc

User metadata
Rank Member
Rank
Member

The IOPB was introduced with the 386. What older processors would be supported by disabling it?

Reply 2 of 3, by superfury

User metadata
Rank l33t++
Rank
l33t++

Ah. You're right about that. And the related bitmap below it can be disabled by clearing CR4's bit on Pentium-compatible processors (CR4.VME).
So it's actually compatible with the 80386 and below (16-bit (286) TSS have neither, so they assume it's bits are set and interrupts trapping I think (like the field is outside descriptor limits)?).

Author of the UniPCemu emulator.
UniPCemu Git repository
UniPCemu for Android, Windows, PSP, Vita and Switch on itch.io

Reply 3 of 3, by superfury

User metadata
Rank l33t++
Rank
l33t++
crazyc wrote on 2024-04-12, 23:42:

The IOPB was introduced with the 386. What older processors would be supported by disabling it?

Oh, my bad. You're right about that. I am wondering about the Interrupt Redirection bitmap in this case.

What happens if IOPB's value minus 32 reaches the IOPB offsets(it's 2 bytes after all) or below? Is that used as-is(reading IOPB or below fields as the IR bitmap) or is the end of the IOPB truncating the limit of the IR bitmap to make the IOPB start after the IOPB always (and past IOPB's value treated as ones)?
This case would happen when the IOPB is at 0x87 and below.

From what I can see in Bochs's v86_redirect_interrupt function (https://sourceforge.net/p/bochs/code/HEAD/tre … s/cpu/vm8086.cc line 203) it doesn't seem to perform any checks, other than a simple limit check?
So you can move the Interrupt Redirection table into the TSS register address space if you want (mapping the registers into it to perform weird hacks)?

Author of the UniPCemu emulator.
UniPCemu Git repository
UniPCemu for Android, Windows, PSP, Vita and Switch on itch.io