The A20 story - Was pdBIOS32 (still is pdBIOS32)
in [ASM]
|
Prev: Do you know where your stack is? Was pdBIOS32 (still is pdBIOS32)
Next: QUESTION: what does "offset" mean?
From: Mike Gonta on 8 May 2010 14:58 On May 8, 4:13 am, James Harris <james.harri...(a)googlemail.com> wrote: > IIRC port 0x92 (fast a20) is not safe to use on all machines. Hi James, Those are probably old hard to find machines. > Does the code carry out some sort of check before it tries the port? I've seen no examples of checking first and am not aware of the method. I've seen examples of checking first if the A20 is enabled, but I'm assuming it is not enabled in the boot sector. Also many of the A20 examples only write to port 92h, while port_92_A20 in rm2pm.asm first reads the port value, modifies bit 1 and then writes it back, there shouldn't be any issues setting or clearing port bits that are already set or cleared. > Port 0x60 is slower but, IIRC, it is easier to verify its existence. Even with a timeout value of 64K loops on a slow machine it's not noticeable on boot up. I think the slowness concern is that it was being used on all mode switches whereas pdBIOS32 only enables the A20 and leaves it enabled when switching to real mode as that aspect of compatibility is not required by the BIOS. In the latest release (pdBIOS32 20100508.02.img) I have re-ordered the A20 methods so that port_60_A20 is executed first, which leaves port_92_A20 as a backup for PC's with no KB controller. Mike Gonta look and see - many look but few see http://mikegonta.com/pdBIOS32
From: James Harris on 9 May 2010 08:13
On 8 May, 19:58, Mike Gonta <mikego...(a)gmail.com> wrote: > On May 8, 4:13 am, James Harris <james.harri...(a)googlemail.com> wrote: > > > IIRC port 0x92 (fast a20) is not safe to use on all machines. > > Hi James, > > Those are probably old hard to find machines. > > > Does the code carry out some sort of check before it tries the port? > > I've seen no examples of checking first and am not aware of the > method. > I've seen examples of checking first if the A20 is enabled, but I'm > assuming it is not enabled in the boot sector. Also many of the A20 > examples only write to port 92h, while port_92_A20 in rm2pm.asm > first reads the port value, modifies bit 1 and then writes it back, > there > shouldn't be any issues setting or clearing port bits that are already > set or cleared. > > > Port 0x60 is slower but, IIRC, it is easier to verify its existence. > > Even with a timeout value of 64K loops on a slow machine it's not > noticeable on boot up. I think the slowness concern is that it was > being used on all mode switches whereas pdBIOS32 only enables > the A20 and leaves it enabled when switching to real mode as > that aspect of compatibility is not required by the BIOS. > > In the latest release (pdBIOS32 20100508.02.img) I have re-ordered > the A20 methods so that port_60_A20 is executed first, which > leaves port_92_A20 as a backup for PC's with no KB controller. A20 is an old chestnut, isn't it. A while ago I collected notes on a number of the above issues. I'll look them out and if they seem worth sharing I'll post to alt.os.development. James |