Re: [Qemu-devel] Where is vga-rom mapped in guest system memory?

[Jaeyong Yoo <jaeyong.yoo@samsung.com>]

> -----Original Message-----
> From: qemu-devel-bounces+jaeyong.yoo=samsung.com@nongnu.org [mailto:qemu-
> devel-bounces+jaeyong.yoo=samsung.com@nongnu.org] On Behalf Of Laszlo
> Ersek
> Sent: Thursday, May 15, 2014 11:05 PM
> To: Jaeyong Yoo; qemu-devel@nongnu.org
> Subject: Re: [Qemu-devel] Where is vga-rom mapped in guest system memory?
> 
> On 05/15/14 15:25, Jaeyong Yoo wrote:
> >> I rely on the qemu debug port rather than on serial:
> >>
> >>   -debugcon file:debug.log -global isa-debugcon.iobase=0x402
> >>
> >> Check the DEBUG_IO and DEBUG_LEVEL settings in the SeaBIOS config as
> well.
> >
> > Thanks Laszlo! Now I can see the log message in VirtualBox vgabios
> > with qemu debug port.
> >
> > By the way, I've got one more question, which is very strange for me.
> >
> > I tried to print a string and the values read by the string address do
> > not look correct. More specifically, I wrote the following code in
> VirtualBox vga-bios:
> >
> > char *msg = "Start Vgabios\n";
> >
> > for (i = msg; ++i; *i != 0)
> > {
> >     outb(0x402, *i);
> > }
> >
> > And, I expect to see the message "Start Vgabios" in the debug port,
> > but it doesn't work. If I object-dump the related parts in VirtualBox
> > vga-bios, I can see that the msg (msg = 0x7eea) properly contains
> > "Start Vgabios". And, if I print the value of i in vga-bios, it gives
> > 0x7eea (which is correct address). Then I expect that *i should be 'S'
> > but *I gives zero. I checked the vga-rom size field and it properly
> > covers the actual size of vgarom. Then, I think qemu properly map the
> entire vgarom in the proper, so it won't be an issue.
> >
> > Is there some issues with memory addressing that I'm missing?
> 
> You are probably in real mode. The above outb() loop probably translates
> to something like this:
> 
> 00000000  3E8A04            mov al,[ds:si]
> 00000003  BA0204            mov dx,0x402
> 00000006  EE                out dx,al
> 00000007  46                inc si
> 
> (objdump -S will allow you to see it exactly.)
> 
> My take is that ds is not set correctly. You might have to set ds manually,
> or at least decorate the definition of the string with some gcc section
> attribute so that the compiler places the string in a section that will be
> "automatically" matched by ds.
> 
> Or, you might be able to test it like this:
> - first, change the type of "msg" from "pointer to character" to "array of
> characters":
> 
>   char msg[] = "Start Vgabios\n"
> 
> - second, give it automatic storage duration, rather than static storage
> duration. In English, make it local to the function containing the loop.
> Hopefully the compiler will generate code that places the string on the
> stack then, and then your loop body will (hopefully) look something like
> 
> 00000000  368A02            mov al,[ss:bp+si]
> 00000003  BA0204            mov dx,0x402
> 00000006  EE                out dx,al
> 00000007  46                inc si
> 
> (Ie. work off your stack, where ss and bp should be "just right".)
> 
> I'm just guessing of course.

Thanks a lot Laszlo,
The problem was ds as you guessed! After I set ds the same as cs, I can
see log messages correctly! And, I can start porting VirtualBox vgabios
to work on qemu.

By the way, would there be no-issue of VBox vgabios (16-bit code) working
with seaBIOS? I'm guessing not, but I want to double check.


Thanks,
Jaeyong