[LinuxBIOS] [PATCH] Support other than 0xC000 ROMs in VM86

[Alex Beregszaszi]

Hi,

this patch changes the real_mode_switch_call_vga() function to jump to
the addr given as an argument instead the hardcoded 0xc000.

Also it changes the function to inline assembly and passes the arguments
through that.

This patch includes the changes from my previous patch "[PATCH] vgabios:
use same interface for biosemu.c and vm86.c", but should be committed
incrementally.

Btw, I think it would be a good idea to export the interrupt handlers to
a common file (biosints.c?) and let it be hooked up by both vm86 and
x86emu implementations.

[Stefan Reinauer]

* Alex Beregszaszi  [070904 18:57]:
> Hi,
>
> this patch changes the real_mode_switch_call_vga() function to jump to
> the addr given as an argument instead the hardcoded 0xc000.

This is good stuff.
> Also it changes the function to inline assembly and passes the arguments
> through that.
>
> This patch includes the changes from my previous patch "[PATCH] vgabios:
> use same interface for biosemu.c and vm86.c", but should be committed
> incrementally.

Can you please rediff against the current svn and send this patch again?
> Btw, I think it would be a good idea to export the interrupt handlers to
> a common file (biosints.c?) and let it be hooked up by both vm86 and
> x86emu implementations.

Yes, definitely!
>

Stefan

[Alex Beregszaszi]

On Wed, 2007-09-05 at 03:45 +0200, Stefan Reinauer wrote:
> > Also it changes the function to inline assembly and passes the arguments
> > through that.
> >
> > This patch includes the changes from my previous patch "[PATCH] vgabios:
> > use same interface for biosemu.c and vm86.c", but should be committed
> > incrementally.
>
> Can you please rediff against the current svn and send this patch again?

Sure, attached.

[Alex Beregszaszi]

Hi,

On Wed, 2007-09-05 at 18:16 +0200, Alex Beregszaszi wrote:
> On Wed, 2007-09-05 at 03:45 +0200, Stefan Reinauer wrote:
> > > Also it changes the function to inline assembly and passes the arguments
> > > through that.
> > >
> > > This patch includes the changes from my previous patch "[PATCH] vgabios:
> > > use same interface for biosemu.c and vm86.c", but should be committed
> > > incrementally.
> >
> > Can you please rediff against the current svn and send this patch again?
>
> Sure, attached.

I hope this is not forgotten.

[Carl-Daniel Hailfinger]

On 12.09.2007 21:37, Alex Beregszaszi wrote:
> Hi,
>
> On Wed, 2007-09-05 at 18:16 +0200, Alex Beregszaszi wrote:
> > On Wed, 2007-09-05 at 03:45 +0200, Stefan Reinauer wrote:
> > > > Also it changes the function to inline assembly and passes the arguments
> > > > through that.
> > > >
> > > > This patch includes the changes from my previous patch "[PATCH] vgabios:
> > > > use same interface for biosemu.c and vm86.c", but should be committed
> > > > incrementally.
> > >
> > > Can you please rediff against the current svn and send this patch again?
> >
> > Sure, attached.
>
> I hope this is not forgotten.

Has this been committed?

Carl-Daniel

[Alex Beregszaszi]

On Sat, 2007-09-29 at 23:53 +0200, Carl-Daniel Hailfinger wrote:
> On 12.09.2007 21:37, Alex Beregszaszi wrote:
> > Hi,
> >
> > On Wed, 2007-09-05 at 18:16 +0200, Alex Beregszaszi wrote:
> > > On Wed, 2007-09-05 at 03:45 +0200, Stefan Reinauer wrote:
> > > > > Also it changes the function to inline assembly and passes the arguments
> > > > > through that.
> > > > >
> > > > > This patch includes the changes from my previous patch "[PATCH] vgabios:
> > > > > use same interface for biosemu.c and vm86.c", but should be committed
> > > > > incrementally.
> > > >
> > > > Can you please rediff against the current svn and send this patch again?
> > >
> > > Sure, attached.
> >
> > I hope this is not forgotten.
>
> Has this been committed?

Not yet, and now I have more patches regarding the BIOS emulation code.
Just my last week have been extremely busy and had no time to be active
on Linuxbios.

If you say ack, let me commit this, as I missed my last opportunity to
commit :)

[Stefan Reinauer]

* Alex Beregszaszi  [071003 01:50]:
> Not yet, and now I have more patches regarding the BIOS emulation code.
> Just my last week have been extremely busy and had no time to be active
> on Linuxbios.

we're looking forward to your patches a lot.
> If you say ack, let me commit this, as I missed my last opportunity to
> commit :)

go ahead.

[Carl-Daniel Hailfinger]

Stefan? I think you wanted to Ack this as well. Alex will then commit.

Alex Beregszaszi wrote:
> If you say ack, let me commit this, as I missed my last opportunity to
> commit :)
> this patch changes the real_mode_switch_call_vga() function to jump to
> the addr given as an argument instead the hardcoded 0xc000.
>
> Also it changes the function to inline assembly and passes the arguments
> through that.

Can you commit the changes above as separate revisions? First the inline
assembly change, then the variable addr stuff. This makes the changes
easier to follow.
> Signed-off-by: Alex Beregszaszi

Acked-by: Carl-Daniel Hailfinger

Carl-Daniel
> Index: util/x86emu/vm86.c
>
> --- util/x86emu/vm86.c	(revision 494)
> +++ util/x86emu/vm86.c	(working copy)
> @@ -5,6 +5,7 @@
> Copyright (C) 2001 University of California.  LA-CC Number 01-67.
> Copyright (C) 2005 [e-mail]
> Copyright (C) 2007 coresystems GmbH
> Copyright (C) 2007 Siqon Ltd. (written by Alex Beregszaszi)
> This program is free software; you can redistribute it and/or modify
> it under the terms of the GNU General Public License as published by
>
> @@ -98,7 +99,7 @@
>  );
>
> /* The address arguments to this function are PHYSICAL ADDRESSES */ 
> -static void real_mode_switch_call_vga(unsigned long devfn)
> +static void real_mode_switch_call_vga(unsigned long devfn, unsigned long addr)
>  {
>    __asm__ __volatile__ (
>    	/* paranoia -- does ecx get saved? not sure. 
> @@ -106,16 +107,12 @@
>    	 */
>    	"	pushal			\n"
>    	/* save the stack */
> "	mov 	%esp, __stack	\n"
> "	mov 	%%esp, __stack	\n"
> "	jmp 	1f		\n"
> "__stack: .long 0		\n"
> "1:\n"
> /* get devfn into %ecx */
> "	movl    %esp, %ebp	\n"
> // FIXME: why is this 8? 
> "	movl    8(%ebp), %ecx	\n"
> /* load 'our' gdt */
> "	lgdt	%cs:__mygdtaddr	\n"
> "	lgdt	%%cs:__mygdtaddr	\n"
>
> /*  This configures CS properly for real mode. */
> "	ljmp	$0x28, $__rms_16bit\n"
>
> @@ -128,17 +125,17 @@
>    	 * configurations (limits, writability, etc.) once 
>    	 * protected mode is turned off. 
>    	 */
> "	mov	$0x30, %ax	\n"
> "	mov	%ax, %ds       	\n"
> "	mov	%ax, %es       	\n"
> "	mov	%ax, %fs       	\n"
> "	mov	%ax, %gs       	\n"
> "	mov	%ax, %ss       	\n"
> "	mov	$0x30, %%ax	\n"
> "	mov	%%ax, %%ds       	\n"
> "	mov	%%ax, %%es       	\n"
> "	mov	%%ax, %%fs       	\n"
> "	mov	%%ax, %%gs       	\n"
> "	mov	%%ax, %%ss       	\n"
>
> /* Turn off protection (bit 0 in CR0) */
> "	movl	%cr0, %eax	\n"
> "	andl	$0xFFFFFFFE, %eax \n"
> "	movl	%eax, %cr0	\n"
> "	movl	%%cr0, %%eax	\n"
> "	andl	$0xFFFFFFFE, %%eax \n"
> "	movl	%%eax, %%cr0	\n"
>
> /* Now really going into real mode */
> "	ljmp	$0,  $__rms_real\n"
>
> @@ -148,33 +145,49 @@
>    	 * That way we can easily share it between real and
>    	 * protected, since the 16-bit ESP at segment 0 will 
>    	 * work for any case. */
> "	mov	$0x0, %ax	\n"
> "	mov	%ax, %ss	\n"
> "	movl	$0x1000, %eax	\n"
> "	movl	%eax, %esp	\n"
> "	mov	$0x0, %%ax	\n"
> "	mov	%%ax, %%ss	\n"
> "	movl	$0x1000, %%eax	\n"
> "	movl	%%eax, %%esp	\n"
>
> /* Load our 16 it idt */
> "	xor	%ax, %ax	\n"
> "	mov	%ax, %ds	\n"
> "	xor	%%ax, %%ax	\n"
> "	mov	%%ax, %%ds	\n"
>
> "	lidt	__myidt		\n"
>
> /* Dump zeros in the other segment registers */
> "	mov	%ax, %es       	\n"
> "	mov	%ax, %fs       	\n"
> "	mov	%ax, %gs       	\n"
> "	mov	$0x40, %ax	\n"
> "	mov	%ax, %ds	\n"
> "	mov	%cx, %ax	\n"
> "	mov	%%ax, %%es       	\n"
> "	mov	%%ax, %%fs       	\n"
> "	mov	%%ax, %%gs       	\n"
> "	mov	$0x40, %%ax	\n"
> "	mov	%%ax, %%ds	\n"
> "	mov	%%cx, %%ax	\n"
>
> /* run VGA BIOS at 0xc000:0003 */
> "	lcall	$0xc000, $0x0003\n"
>
> +/*		"	lcall	$0xc000, $0x0003   \n"*/
> /* calculate addr */
> "       mov     %%edx, %%ebx      \n"
> "       add     $3, %%ebx        \n"
> "       and     $0xffff, %%ebx   \n"
> /* calculate segment */
> "       and     $0xf0000, %%edx  \n"
> "       shr     $4, %%edx        \n"
> "       pushw   %%dx             \n"
> "       pushw   %%bx             \n"
> "       mov     %%esp, %%ebp      \n"
> /* idea from bochs bios' rom_scan */
> /* call far ss:[bp+0] */
> "       .byte 0xff, 0x5e, 0x00   \n"
> "       add     $4, %%esp         \n"
> /* If we got here, just about done.
> Need to get back to protected mode
> */
> "	movl	%cr0, %eax	\n"
> "	orl	$0x0000001, %eax\n" /* PE = 1 */
> "	movl	%eax, %cr0	\n"
> "	movl	%%cr0, %%eax	\n"
> "	orl	$0x0000001, %%eax\n" /* PE = 1 */
> "	movl	%%eax, %%cr0	\n"
>
> /* Now that we are in protected mode
> jump to a 32 bit code segment.
>
> @@ -182,21 +195,23 @@
>    	"	data32	ljmp	$0x10, $vgarestart\n"
>    	"vgarestart:\n"
>    	"	.code32\n"
> "	movw	$0x18, %ax     	\n"
> "	mov	%ax, %ds       	\n"
> "	mov	%ax, %es	\n"
> "	mov	%ax, %fs	\n"
> "	mov	%ax, %gs	\n"
> "	mov	%ax, %ss	\n"
> "	movw	$0x18, %%ax     	\n"
> "	mov	%%ax, %%ds       	\n"
> "	mov	%%ax, %%es	\n"
> "	mov	%%ax, %%fs	\n"
> "	mov	%%ax, %%gs	\n"
> "	mov	%%ax, %%ss	\n"
>
> /* restore proper gdt and idt */
> "	lgdt	%cs:gdtarg	\n"
> "	lgdt	%%cs:gdtarg	\n"
>
> "	lidt	idtarg		\n"
>
> ".globl vga_exit		\n"
> "vga_exit:			\n"
> "	mov	__stack, %esp	\n"
> "	mov	__stack, %%esp	\n"
> "	popal			\n"
> :
> : "c" (devfn), "d" (addr)
> );
>
> }
>
> @@ -315,8 +330,8 @@
>    for (i = 0x400; i < 0x500; i++) {
>    	*(unsigned char *) i = 0;
>    }
> real_mode_switch_call_vga((dev->bus->secondary << 8) | dev->path.u.pci.devfn);
> real_mode_switch_call_vga((dev->bus->secondary << 8) | dev->path.u.pci.devfn, addr);
>
> }

[Stefan Reinauer]

* Carl-Daniel Hailfinger  [071005 18:31]:
> Stefan? I think you wanted to Ack this as well. Alex will then commit.

Sorry for the delay... Please go ahead and commit

Acked-by: Stefan Reinauer
> Alex Beregszaszi wrote:
> > If you say ack, let me commit this, as I missed my last opportunity to
> > commit :)
> > this patch changes the real_mode_switch_call_vga() function to jump to
> > the addr given as an argument instead the hardcoded 0xc000.
> >
> > Also it changes the function to inline assembly and passes the arguments
> > through that.
>
> Can you commit the changes above as separate revisions? First the inline
> assembly change, then the variable addr stuff. This makes the changes
> easier to follow.
> > Signed-off-by: Alex Beregszaszi
>
> Acked-by: Carl-Daniel Hailfinger
>
> Carl-Daniel
> > Index: util/x86emu/vm86.c
> >
> > --- util/x86emu/vm86.c	(revision 494)
> > +++ util/x86emu/vm86.c	(working copy)
> > @@ -5,6 +5,7 @@
> > Copyright (C) 2001 University of California.  LA-CC Number 01-67.
> > Copyright (C) 2005 [e-mail]
> > Copyright (C) 2007 coresystems GmbH
> > Copyright (C) 2007 Siqon Ltd. (written by Alex Beregszaszi)
> > This program is free software; you can redistribute it and/or modify
> > it under the terms of the GNU General Public License as published by
> >
> > @@ -98,7 +99,7 @@
> >  );
> >
> > /* The address arguments to this function are PHYSICAL ADDRESSES */ 
> > -static void real_mode_switch_call_vga(unsigned long devfn)
> > +static void real_mode_switch_call_vga(unsigned long devfn, unsigned long addr)
> >  {
> >    __asm__ __volatile__ (
> >    	/* paranoia -- does ecx get saved? not sure. 
> > @@ -106,16 +107,12 @@
> >    	 */
> >    	"	pushal			\n"
> >    	/* save the stack */
> > "	mov 	%esp, __stack	\n"
> > "	mov 	%%esp, __stack	\n"
> > "	jmp 	1f		\n"
> > "__stack: .long 0		\n"
> > "1:\n"
> > /* get devfn into %ecx */
> > "	movl    %esp, %ebp	\n"
> > // FIXME: why is this 8? 
> > "	movl    8(%ebp), %ecx	\n"
> > /* load 'our' gdt */
> > "	lgdt	%cs:__mygdtaddr	\n"
> > "	lgdt	%%cs:__mygdtaddr	\n"
> >
> > /*  This configures CS properly for real mode. */
> > "	ljmp	$0x28, $__rms_16bit\n"
> >
> > @@ -128,17 +125,17 @@
> >    	 * configurations (limits, writability, etc.) once 
> >    	 * protected mode is turned off. 
> >    	 */
> > "	mov	$0x30, %ax	\n"
> > "	mov	%ax, %ds       	\n"
> > "	mov	%ax, %es       	\n"
> > "	mov	%ax, %fs       	\n"
> > "	mov	%ax, %gs       	\n"
> > "	mov	%ax, %ss       	\n"
> > "	mov	$0x30, %%ax	\n"
> > "	mov	%%ax, %%ds       	\n"
> > "	mov	%%ax, %%es       	\n"
> > "	mov	%%ax, %%fs       	\n"
> > "	mov	%%ax, %%gs       	\n"
> > "	mov	%%ax, %%ss       	\n"
> >
> > /* Turn off protection (bit 0 in CR0) */
> > "	movl	%cr0, %eax	\n"
> > "	andl	$0xFFFFFFFE, %eax \n"
> > "	movl	%eax, %cr0	\n"
> > "	movl	%%cr0, %%eax	\n"
> > "	andl	$0xFFFFFFFE, %%eax \n"
> > "	movl	%%eax, %%cr0	\n"
> >
> > /* Now really going into real mode */
> > "	ljmp	$0,  $__rms_real\n"
> >
> > @@ -148,33 +145,49 @@
> >    	 * That way we can easily share it between real and
> >    	 * protected, since the 16-bit ESP at segment 0 will 
> >    	 * work for any case. */
> > "	mov	$0x0, %ax	\n"
> > "	mov	%ax, %ss	\n"
> > "	movl	$0x1000, %eax	\n"
> > "	movl	%eax, %esp	\n"
> > "	mov	$0x0, %%ax	\n"
> > "	mov	%%ax, %%ss	\n"
> > "	movl	$0x1000, %%eax	\n"
> > "	movl	%%eax, %%esp	\n"
> >
> > /* Load our 16 it idt */
> > "	xor	%ax, %ax	\n"
> > "	mov	%ax, %ds	\n"
> > "	xor	%%ax, %%ax	\n"
> > "	mov	%%ax, %%ds	\n"
> >
> > "	lidt	__myidt		\n"
> >
> > /* Dump zeros in the other segment registers */
> > "	mov	%ax, %es       	\n"
> > "	mov	%ax, %fs       	\n"
> > "	mov	%ax, %gs       	\n"
> > "	mov	$0x40, %ax	\n"
> > "	mov	%ax, %ds	\n"
> > "	mov	%cx, %ax	\n"
> > "	mov	%%ax, %%es       	\n"
> > "	mov	%%ax, %%fs       	\n"
> > "	mov	%%ax, %%gs       	\n"
> > "	mov	$0x40, %%ax	\n"
> > "	mov	%%ax, %%ds	\n"
> > "	mov	%%cx, %%ax	\n"
> >
> > /* run VGA BIOS at 0xc000:0003 */
> > "	lcall	$0xc000, $0x0003\n"
> >
> > +/*		"	lcall	$0xc000, $0x0003   \n"*/
> > /* calculate addr */
> > "       mov     %%edx, %%ebx      \n"
> > "       add     $3, %%ebx        \n"
> > "       and     $0xffff, %%ebx   \n"
> > /* calculate segment */
> > "       and     $0xf0000, %%edx  \n"
> > "       shr     $4, %%edx        \n"
> > "       pushw   %%dx             \n"
> > "       pushw   %%bx             \n"
> > "       mov     %%esp, %%ebp      \n"
> > /* idea from bochs bios' rom_scan */
> > /* call far ss:[bp+0] */
> > "       .byte 0xff, 0x5e, 0x00   \n"
> > "       add     $4, %%esp         \n"
> > /* If we got here, just about done.
> > Need to get back to protected mode
> > */
> > "	movl	%cr0, %eax	\n"
> > "	orl	$0x0000001, %eax\n" /* PE = 1 */
> > "	movl	%eax, %cr0	\n"
> > "	movl	%%cr0, %%eax	\n"
> > "	orl	$0x0000001, %%eax\n" /* PE = 1 */
> > "	movl	%%eax, %%cr0	\n"
> >
> > /* Now that we are in protected mode
> > jump to a 32 bit code segment.
> >
> > @@ -182,21 +195,23 @@
> >    	"	data32	ljmp	$0x10, $vgarestart\n"
> >    	"vgarestart:\n"
> >    	"	.code32\n"
> > "	movw	$0x18, %ax     	\n"
> > "	mov	%ax, %ds       	\n"
> > "	mov	%ax, %es	\n"
> > "	mov	%ax, %fs	\n"
> > "	mov	%ax, %gs	\n"
> > "	mov	%ax, %ss	\n"
> > "	movw	$0x18, %%ax     	\n"
> > "	mov	%%ax, %%ds       	\n"
> > "	mov	%%ax, %%es	\n"
> > "	mov	%%ax, %%fs	\n"
> > "	mov	%%ax, %%gs	\n"
> > "	mov	%%ax, %%ss	\n"
> >
> > /* restore proper gdt and idt */
> > "	lgdt	%cs:gdtarg	\n"
> > "	lgdt	%%cs:gdtarg	\n"
> >
> > "	lidt	idtarg		\n"
> >
> > ".globl vga_exit		\n"
> > "vga_exit:			\n"
> > "	mov	__stack, %esp	\n"
> > "	mov	__stack, %%esp	\n"
> > "	popal			\n"
> > :
> > : "c" (devfn), "d" (addr)
> > );
> >
> > }
> >
> > @@ -315,8 +330,8 @@
> >    for (i = 0x400; i < 0x500; i++) {
> >    	*(unsigned char *) i = 0;
> >    }
> > real_mode_switch_call_vga((dev->bus->secondary << 8) | dev->path.u.pci.devfn);
> > real_mode_switch_call_vga((dev->bus->secondary << 8) | dev->path.u.pci.devfn, addr);
> >
> > }