s3_late_bios

s3_late_bios

[Emmanuel Thomé]

[-- Attachment #1: Type: text/plain, Size: 266 bytes --]


Hi all,

Sorry I've been off for a week. Here is the version of the s3_late_bios
patch with the segment info coded in C.

If somebody comes up with an attempt at resuming the video from the early
real mode phase by calling the pci bios, I'm all for testing it.

E.

[-- Attachment #2: s3_late_bios.patch.gz --]
[-- Type: application/x-gzip, Size: 3659 bytes --]

[-- Attachment #3: s3_late_bios_radeon.patch.gz --]
[-- Type: application/x-gzip, Size: 342 bytes --]

[-- Attachment #4: s3_late_bios-x86_64-draft.patch.gz --]
[-- Type: application/x-gzip, Size: 987 bytes --]

Re: s3_late_bios

[Pavel Machek]

Hi!

> Sorry I've been off for a week. Here is the version of the s3_late_bios
> patch with the segment info coded in C.
> 
> If somebody comes up with an attempt at resuming the video from the early
> real mode phase by calling the pci bios, I'm all for testing it.

It looks good to me...
								Pavel

> --- linux.orig/Documentation/power/video.txt	2004-08-14 23:10:42.000000000 +0200
> +++ linux/Documentation/power/video.txt	2004-11-01 14:29:51.000000000 +0100
> @@ -26,12 +26,17 @@
>    point, but it happens to work on some machines. Use
>    acpi_sleep=s3_bios (Athlon64 desktop system)
>  
> +* systems where an excursion into real mode _after_ the pci subsystem
> +  wakes up is sufficient. It has better chances of working than s3_bios.
> +  Use acpi_sleep=s3_late_bios (Dell D600, Fujitsu P2120 ; both run radeon
> +  boards, but not all radeons seem to like this).
> +
>  * radeon systems, where X can soft-boot your video card. You'll need
>    patched X, and plain text console (no vesafb or radeonfb), see
>    http://www.doesi.gmxhome.de/linux/tm800s3/s3.html. (Acer TM 800)
>  
>  Now, if you pass acpi_sleep=something, and it does not work with your
> -bios, you'll get hard crash during resume. Be carefull.
> +bios, you'll get hard crash during resume. Be careful.
>  
>  You may have system where none of above works. At that point you
>  either invent another ugly hack that works, or write proper driver for
> --- linux.orig/arch/i386/kernel/acpi/sleep.c	2004-10-31 10:55:40.000000000 +0100
> +++ linux/arch/i386/kernel/acpi/sleep.c	2004-11-06 23:57:44.000000000 +0100
> @@ -7,6 +7,8 @@
>  
>  #include <linux/acpi.h>
>  #include <linux/bootmem.h>
> +#include <asm/processor.h>
> +#include <asm/desc.h>
>  #include <asm/smp.h>
>  
>  
> @@ -63,6 +65,86 @@
>  	zap_low_mappings();
>  }
>  
> +/*
> + * acpi_vgapost - trip into real mode in order to post the vga card.
> + *
> + * slot is typically pdev->bus->number << 8 | pdev->devfn
> + * where pdev is a struct pci_dev *
> + */
> +extern struct Xgt_desc_struct go_real_gdtr;
> +extern struct Xgt_desc_struct go_real_idtr;
> +extern u16 go_real_jmp_seg;
> +
> +extern void do_vgapost_lowlevel (void);
> +
> +int acpi_vgapost (unsigned long slot)
> +{
> +	unsigned long flags;
> +	unsigned long saved_video_flags;
> +	u32 low_mapping;
> +	struct user_desc info;
> +	struct desc_struct gdt[3];
> +
> +	if ((acpi_video_flags & ACPI_SLEEP_S3_LATE_MASK) == 0)
> +		return -1;
> +	if (num_online_cpus() != 1)	/* do you dare ? */
> +		return -EPERM;
> +
> +	/* Load the required info for the jump to real mode. The code in
> +	 * here could maybe go to acpi_reserve_bootmem instead, so that
> +	 * it's done once and for all.
> +	 */
> +
> +	low_mapping = virt_to_phys((void*)acpi_wakeup_address);
> +		
> +	go_real_jmp_seg = low_mapping >> 4;
> +
> +	/* ldt and gdt descriptors are the same, hence the ldt macros.
> +	 * however, we want the privilege flag be set to 0 (reboot is
> +	 * waiting at the corner otherwise). */
> +	memset(gdt,0,sizeof(gdt));
> +
> +	memset(&info,0,sizeof(info));
> +	info.limit = 0xffff;
> +	info.base_addr = low_mapping;
> +	info.seg_32bit = 0;
> +	
> +	info.contents = MODIFY_LDT_CONTENTS_CODE;
> +	gdt[1].a = LDT_entry_a(&info);
> +	gdt[1].b = LDT_entry_b(&info) & ~(0x6000);	/* clear dpl */
> +
> +	info.contents = MODIFY_LDT_CONTENTS_DATA;
> +	gdt[2].a = LDT_entry_a(&info);
> +	gdt[2].b = LDT_entry_b(&info) & ~(0x6000);	/* clear dpl */
> +
> +	go_real_gdtr.size = sizeof(gdt)-1;
> +	go_real_gdtr.address = (u32) &gdt;
> +	go_real_idtr.size = (1<<10)-1;
> +	go_real_idtr.address = 0;
> +
> +	/* Copy the wakeup code again (this time with the _late_ flags).  */
> +
> +	saved_video_flags = acpi_video_flags;
> +	acpi_video_flags = (slot & 0xffff) << 16 |
> +		ACPI_SLEEP_S3_LATE_BITS(acpi_video_flags);
> +	
> +	acpi_save_state_mem();
> +
> +	/* Tunnel thru real mode, call the wakeup code, and return. We're
> +	 * typically called here from pci_default_resume, hence w/ irqs
> +	 * on. */
> +	local_irq_save(flags);
> +	do_vgapost_lowlevel();
> +	local_irq_restore(flags);
> +
> +	/* Restore mapping etc */
> +	acpi_restore_state_mem();
> +
> +	/* restore normal acpi_video_flags for next resume */
> +	acpi_video_flags = saved_video_flags;
> +	return 0;
> +}
> +
>  /**
>   * acpi_reserve_bootmem - do _very_ early ACPI initialisation
>   *
> @@ -87,9 +169,17 @@
>  {
>  	while ((str != NULL) && (*str != '\0')) {
>  		if (strncmp(str, "s3_bios", 7) == 0)
> -			acpi_video_flags = 1;
> +			acpi_video_flags = ACPI_SLEEP_S3_BIOS;
>  		if (strncmp(str, "s3_mode", 7) == 0)
> -			acpi_video_flags |= 2;
> +			acpi_video_flags |= ACPI_SLEEP_S3_MODE;
> +		if (strncmp(str, "s3_late_bios", 12) == 0)
> +			acpi_video_flags |= ACPI_SLEEP_S3_LATE_BIOS;
> +		/* This last one is there for ``symmetry'', but frankly I
> +		 * don't think it's any useful. Well, for vesafb lovers maybe.
> +		 * Don't forget to celebrate if it works for you (???).
> +		 */
> +		if (strncmp(str, "s3_late_mode", 12) == 0)
> +			acpi_video_flags |= ACPI_SLEEP_S3_LATE_MODE;
>  		str = strchr(str, ',');
>  		if (str != NULL)
>  			str += strspn(str, ", \t");
> --- linux.orig/arch/i386/kernel/acpi/wakeup.S	2004-10-31 10:55:40.000000000 +0100
> +++ linux/arch/i386/kernel/acpi/wakeup.S	2004-11-06 23:54:30.000000000 +0100
> @@ -2,6 +2,7 @@
>  #include <linux/linkage.h>
>  #include <asm/segment.h>
>  #include <asm/page.h>
> +#include <asm/acpi.h>
>  
>  #
>  # wakeup_code runs in real mode, and at unknown address (determined at run-time).
> @@ -41,15 +42,16 @@
>  	cmpl	$0x12345678, %eax
>  	jne	bogus_real_magic
>  
> -	testl	$1, video_flags - wakeup_code
> +	testl	$ACPI_SLEEP_S3_BIOS, video_flags - wakeup_code
>  	jz	1f
> +	movw	video_flags - wakeup_code + 2, %ax
>  	lcall   $0xc000,$3
>  	movw	%cs, %ax
>  	movw	%ax, %ds					# Bios might have played with that
>  	movw	%ax, %ss
>  1:
>  
> -	testl	$2, video_flags - wakeup_code
> +	testl	$ACPI_SLEEP_S3_MODE, video_flags - wakeup_code
>  	jz	1f
>  	mov	video_mode - wakeup_code, %ax
>  	call	mode_set
> @@ -159,6 +161,27 @@
>  
>  _setbad: jmp setbad
>  
> +#
> +#	Real mode switch - verbatim from reboot.c
> +#
> +go_real:
> +	movl	%cr0, %eax
> +	andl	$0x00000011, %eax
> +	orl	$0x60000000, %eax
> +	movl	%eax, %cr0
> +	movl	%eax, %cr3
> +	movl	%cr0, %ebx
> +	andl	$0x60000000, %ebx
> +	jz	1f
> +	wbinvd
> +1:	andb	$0x10, %al
> +	movl	%eax, %cr0
> +	.byte	0xea
> +	.word	0x0000
> +.globl go_real_jmp_seg
> +go_real_jmp_seg:
> +	.word	0x0000
> +
>  	.code32
>  	ALIGN
>  
> @@ -244,6 +267,7 @@
>  ENTRY(saved_magic)	.long	0
>  ENTRY(saved_eip)	.long	0
>  
> +.text
>  save_registers:
>  	leal	4(%esp), %eax
>  	movl	%eax, saved_context_esp
> @@ -284,6 +308,42 @@
>  	call acpi_enter_sleep_state_s4bios
>  	ret
>  
> +ENTRY(do_vgapost_lowlevel)
> +	# Save state and registers
> +	call	save_processor_state
> +	call	save_registers
> +	# Reload page table with low mapping
> +	movl	$swapper_pg_dir-__PAGE_OFFSET, %eax
> +	movl	%eax, %cr3
> +	# Load IDTR and GDTR for real mode
> +	lidt	go_real_idtr
> +	lgdt	go_real_gdtr
> +	# Load DS & al ; 0x10 = third segment desc. in gdt (data segment)
> +	movl	$0x0010, %eax
> +	movl	%eax, %ss
> +	movl	%eax, %ds
> +	movl	%eax, %es
> +	movl	%eax, %fs
> +	movl	%eax, %gs
> +	# Load CS ; 0x08 = second segment desc. in gdt (code segment)
> +	ljmpl	$0x0008, $(go_real - wakeup_start)
> +	# returns occurs from ret_point in do_suspend_lowlevel. cpu
> +	# state and registers will be restored from there, and we'll
> +	# eventually return to acpi_vgapost where we came from.
> +	
> +       .align  8
> +       .word   0
> +.globl go_real_gdtr
> +go_real_gdtr:
> +       .word   0
> +       .long   0
> +
> +       .word   0
> +.globl go_real_idtr
> +go_real_idtr:
> +       .word   0
> +       .long   0
> +
>  ALIGN
>  # saved registers
>  saved_gdt:	.long	0,0
> --- linux.orig/drivers/pci/pci-driver.c	2004-10-26 10:19:42.000000000 +0200
> +++ linux/drivers/pci/pci-driver.c	2004-11-01 19:14:20.000000000 +0100
> @@ -336,6 +336,19 @@
>  	/* if the device was busmaster before the suspend, make it busmaster again */
>  	if (pci_dev->is_busmaster)
>  		pci_set_master(pci_dev);
> +#if	defined(CONFIG_X86) && defined(CONFIG_ACPI_SLEEP)
> +	if ((pci_dev->class & 0xffff00) ==  (PCI_CLASS_DISPLAY_VGA << 8)
> +		&& (pci_dev->dev.power_state != 4))
> +	{
> +		/* This is a no-op, and returns -1, in case s3_bios_late
> +		 * hasn't been specified on the kernel command line
> +		 */
> +		if (!acpi_vgapost(pci_dev->bus->number << 8 | pci_dev->devfn)) {
> +			printk(KERN_INFO "resumed pci graphics adapter %s\n",
> +				pci_name(pci_dev));
> +		}
> +	}
> +#endif
>  }
>  
>  static int pci_device_resume(struct device * dev)
> --- linux.orig/include/asm-i386/acpi.h	2004-10-26 10:19:50.000000000 +0200
> +++ linux/include/asm-i386/acpi.h	2004-11-01 19:31:28.000000000 +0100
> @@ -27,6 +27,7 @@
>  #define _ASM_ACPI_H
>  
>  #ifdef __KERNEL__
> +#ifndef __ASSEMBLY__
>  
>  #include <asm/system.h>		/* defines cmpxchg */
>  
> @@ -187,10 +188,26 @@
>  /* early initialization routine */
>  extern void acpi_reserve_bootmem(void);
>  
> +/* real mode excursion to post code from vga rom ; must come once pci is up. */
> +extern int acpi_vgapost(unsigned long);
> +
>  #endif /*CONFIG_ACPI_SLEEP*/
>  
>  extern u8 x86_acpiid_to_apicid[];
>  
> +#endif /*__ASSEMBLY__*/
> +
> +#ifdef	CONFIG_ACPI_SLEEP
> +
> +#define	ACPI_SLEEP_S3_BIOS	1
> +#define	ACPI_SLEEP_S3_MODE	2
> +#define	ACPI_SLEEP_S3_LATE_BIOS	4
> +#define	ACPI_SLEEP_S3_LATE_MODE	8
> +#define	ACPI_SLEEP_S3_LATE_MASK	12
> +#define ACPI_SLEEP_S3_LATE_BITS(X)	(((X)&12)>>2)
> +
> +#endif	/*CONFIG_ACPI_SLEEP*/
> +
>  #endif /*__KERNEL__*/
>  
>  #endif /*_ASM_ACPI_H*/

> --- linux.orig/drivers/video/aty/radeon_pm.c	2004-10-31 10:55:40.000000000 +0100
> +++ linux/drivers/video/aty/radeon_pm.c	2004-11-01 19:32:53.000000000 +0100
> @@ -909,6 +909,12 @@
>  	if (pdev->dev.power_state == 0)
>  		return 0;
>  
> +	pci_restore_state(pdev, pdev->saved_config_space);
> +
> +	/* Only effective if enabled via acpi_sleep=s3_bios_late */
> +	if (pdev->dev.power_state != 4)
> +		acpi_vgapost(pdev->bus->number << 8 | pdev->devfn);
> +
>  	acquire_console_sem();
>  
>  	/* Wakeup chip */

> --- linux.orig/arch/x86_64/kernel/acpi/sleep.c	2004-08-14 23:10:21.000000000 +0200
> +++ linux/arch/x86_64/kernel/acpi/sleep.c	2004-11-01 19:30:44.000000000 +0100
> @@ -120,9 +120,18 @@
>  {
>  	while ((str != NULL) && (*str != '\0')) {
>  		if (strncmp(str, "s3_bios", 7) == 0)
> -			acpi_video_flags = 1;
> +			acpi_video_flags = ACPI_SLEEP_S3_BIOS;
>  		if (strncmp(str, "s3_mode", 7) == 0)
> -			acpi_video_flags |= 2;
> +			acpi_video_flags |= ACPI_SLEEP_S3_MODE;
> +#if 0
> +		/* These could be as useful as for x86, but it's totally
> +		 * untested.
> +		 */
> +		if (strncmp(str, "s3_late_bios", 12) == 0)
> +			acpi_video_flags |= ACPI_SLEEP_S3_LATE_BIOS;
> +		if (strncmp(str, "s3_late_mode", 12) == 0)
> +			acpi_video_flags |= ACPI_SLEEP_S3_LATE_MODE;
> +#endif
>  		str = strchr(str, ',');
>  		if (str != NULL)
>  			str += strspn(str, ", \t");
> --- linux.orig/arch/x86_64/kernel/acpi/wakeup.S	2004-02-18 04:58:43.000000000 +0100
> +++ linux/arch/x86_64/kernel/acpi/wakeup.S	2004-11-01 19:28:47.000000000 +0100
> @@ -3,6 +3,7 @@
>  #include <asm/segment.h>
>  #include <asm/page.h>
>  #include <asm/msr.h>
> +#include <asm/acpi.h>
>  
>  # Copyright 2003 Pavel Machek <pavel-AlSwsSmVLrQ@public.gmane.org>, distribute under GPLv2
>  #
> @@ -41,7 +42,7 @@
>  	cmpl	$0x12345678, %eax
>  	jne	bogus_real_magic
>  
> -	testl	$1, video_flags - wakeup_code
> +	testl	$ACPI_SLEEP_S3_BIOS, video_flags - wakeup_code
>  	jz	1f
>  	lcall   $0xc000,$3
>  	movw	%cs, %ax
> @@ -49,7 +50,7 @@
>  	movw	%ax, %ss
>  1:
>  
> -	testl	$2, video_flags - wakeup_code
> +	testl	$ACPI_SLEEP_S3_MODE, video_flags - wakeup_code
>  	jz	1f
>  	mov	video_mode - wakeup_code, %ax
>  	call	mode_seta
> --- linux.orig/include/asm-x86_64/acpi.h	2004-10-26 10:19:51.000000000 +0200
> +++ linux/include/asm-x86_64/acpi.h	2004-11-01 19:27:54.000000000 +0100
> @@ -27,6 +27,7 @@
>  #define _ASM_ACPI_H
>  
>  #ifdef __KERNEL__
> +#ifndef __ASSEMBLY__
>  
>  #define COMPILER_DEPENDENT_INT64   long long
>  #define COMPILER_DEPENDENT_UINT64  unsigned long long
> @@ -166,6 +167,19 @@
>  
>  extern u8 x86_acpiid_to_apicid[];
>  
> +#endif /*__ASSEMBLY__*/
> +
> +#ifdef	CONFIG_ACPI_SLEEP
> +
> +#define	ACPI_SLEEP_S3_BIOS	1
> +#define	ACPI_SLEEP_S3_MODE	2
> +#define	ACPI_SLEEP_S3_LATE_BIOS	4
> +#define	ACPI_SLEEP_S3_LATE_MODE	8
> +#define	ACPI_SLEEP_S3_LATE_MASK	12
> +#define ACPI_SLEEP_S3_LATE_BITS(X)	(((X)&12)>>2)
> +
> +#endif /*__ASSEMBLY__*/
> +
>  #endif /*__KERNEL__*/
>  
>  #endif /*_ASM_ACPI_H*/


-- 
People were complaining that M$ turns users into beta-testers...
...jr ghea gurz vagb qrirybcref, naq gurl frrz gb yvxr vg gung jnl!


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: Re: s3_late_bios

[Nate Lawson]

Pavel Machek wrote:
>>Sorry I've been off for a week. Here is the version of the s3_late_bios
>>patch with the segment info coded in C.
>>
>>If somebody comes up with an attempt at resuming the video from the early
>>real mode phase by calling the pci bios, I'm all for testing it.
> 
> It looks good to me...

I'm not sure I understand the reason for this or how it works.  Does it 
POST VGA via the "lcall $0xc000,$3" or the "int 10h" method?  I assume 
the reason it's called by the PCI code is that you want the device to be 
set to D0 and device-specific resume methods called before POSTing it?
Why the switch to real mode to run the BIOS call?  Can't this be done 
from vm86 mode?

-- 
Nate


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: Re: s3_late_bios

[Emmanuel Thomé]

On Mon, Nov 08, 2004 at 01:53:25PM -0800, Nate Lawson wrote:
> I'm not sure I understand the reason for this or how it works.  Does it 
> POST VGA via the "lcall $0xc000,$3" or the "int 10h" method?  I assume 

lcall.

> the reason it's called by the PCI code is that you want the device to be 
> set to D0 and device-specific resume methods called before POSTing it?

acpi_vgapost is meant to hook into the device-specific resume methods.
Indeed, the pci device is set to D0 prior to calling acpi_vgapost.

Doing POST from the early resume phase seems just too early ; too many
things are unitialized at this point. Doing it later on is more
convenient.

> Why the switch to real mode to run the BIOS call?  Can't this be done 
> from vm86 mode?

Yes it can, at least I think so. You're welcome to tweak this to do it in
a vm86 mode if you like.

E.


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: Re: s3_late_bios

[Nate Lawson]

Emmanuel Thomé wrote:
> On Mon, Nov 08, 2004 at 01:53:25PM -0800, Nate Lawson wrote:
>>Why the switch to real mode to run the BIOS call?  Can't this be done 
>>from vm86 mode?
> 
> Yes it can, at least I think so. You're welcome to tweak this to do it in
> a vm86 mode if you like.

I mostly work on FreeBSD and since it has clean vm86 support routines, 
it doesn't involve any inline assembly to do the lcall there.  I assume 
Linux has something similar.  I think re-entering real mode and 
overloading the resume asm code to run just one instruction is a bit of 
a hack.

-- 
Nate


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_idU88&alloc_id\x12065&op=click

Re: Re: s3_late_bios

[Matthew Garrett]

On Tue, 2004-11-09 at 12:26 -0800, Nate Lawson wrote:

> I mostly work on FreeBSD and since it has clean vm86 support routines, 
> it doesn't involve any inline assembly to do the lcall there.  I assume 
> Linux has something similar.  I think re-entering real mode and 
> overloading the resume asm code to run just one instruction is a bit of 
> a hack.

Linux/AMD64 has no vm86 mode, as far as I know.

-- 
Matthew Garrett | mjg59-1xO5oi07KQx4cg9Nei1l7Q@public.gmane.org



-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: Re: s3_late_bios

[Emmanuel Thomé]

On Tue, Nov 09, 2004 at 12:26:14PM -0800, Nate Lawson wrote:
> I mostly work on FreeBSD and since it has clean vm86 support routines, 
> it doesn't involve any inline assembly to do the lcall there.  I assume 
> Linux has something similar.  I think re-entering real mode and 
> overloading the resume asm code to run just one instruction is a bit of 
> a hack.

I don't think linux has a vm86 api at all.

E.


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: Re: s3_late_bios

[Matthew Garrett]

On Wed, 2004-11-10 at 08:20 +0100, Emmanuel Thomé wrote:
> On Tue, Nov 09, 2004 at 12:26:14PM -0800, Nate Lawson wrote:
> > I mostly work on FreeBSD and since it has clean vm86 support routines, 
> > it doesn't involve any inline assembly to do the lcall there.  I assume 
> > Linux has something similar.  I think re-entering real mode and 
> > overloading the resume asm code to run just one instruction is a bit of 
> > a hack.
> 
> I don't think linux has a vm86 api at all.

There's a vm86 api on i386 - see the vm86 manpage. It seems to have been
introduced in 0.97.

-- 
Matthew Garrett | mjg59-1xO5oi07KQx4cg9Nei1l7Q@public.gmane.org



-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_idU88&alloc_id\x12065&op=click

Re: s3_late_bios

[ole.rohne-vJEk5272eHo]

>> I don't think linux has a vm86 api at all.

Matthew> There's a vm86 api on i386 - see the vm86 manpage. It seems
Matthew> to have been introduced in 0.97.

I guess that should have been: Linux doesn't have kernel-mode vm86...

Ole




-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: Re: s3_late_bios

[Nate Lawson]

Matthew Garrett wrote:
> On Wed, 2004-11-10 at 08:20 +0100, Emmanuel Thomé wrote:
> 
>>On Tue, Nov 09, 2004 at 12:26:14PM -0800, Nate Lawson wrote:
>>
>>>I mostly work on FreeBSD and since it has clean vm86 support routines, 
>>>it doesn't involve any inline assembly to do the lcall there.  I assume 
>>>Linux has something similar.  I think re-entering real mode and 
>>>overloading the resume asm code to run just one instruction is a bit of 
>>>a hack.
>>
>>I don't think linux has a vm86 api at all.
> 
> There's a vm86 api on i386 - see the vm86 manpage. It seems to have been
> introduced in 0.97.

I mean for the kernel and userland:
http://www.freebsd.org/cgi/cvsweb.cgi/src/sys/i386/i386/vm86.c?rev=1.56&content-type=text/x-cvsweb-markup
http://www.freebsd.org/cgi/cvsweb.cgi/src/sys/i386/i386/vm86bios.s?rev=1.30&content-type=text/x-cvsweb-markup
http://www.freebsd.org/cgi/cvsweb.cgi/src/sys/i386/include/vm86.h?rev=1.17&content-type=text/x-cvsweb-markup

If you import any of these, please respect the original copyright.

-- 
Nate


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_idU88&alloc_id\x12065&op=click

Re: s3_late_bios

[ole.rohne-vJEk5272eHo]

Nate> I think re-entering real mode and overloading the resume asm
Nate> code to run just one instruction is a bit of a hack.

lcall $0xc000,3 of course isn't "just one instruction", the BIOS VGA
POST has to run in real mode or some emulation thereof...

If FreeBSD has kernel vm86, we should probably look at it.

Ole




-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: s3_late_bios

[Greg]

> Sorry I've been off for a week. Here is the version of the s3_late_bios
> patch with the segment info coded in C.

This only works on my Sony VAIO Z1 if the PIC IMR is saved and
restored around the VGA POST.  For some reason, when returning from
the BIOS, bit 0 of the IMR is set.  That masks further clock
interrupts so sleeping waits hang the machine after that point (for
instance when resuming the pci devices on bus #2 like the e100). 
Reading the IMR before and after the real-mode POST shows that only
bit 0 is different.

Cheers!greg


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click

Re: s3_late_bios

[Emmanuel Thomé]

Interesting.

I'll have a look to see if the same kind of things happen for me (I still
have crashes when resuming from vesa modes).

E.

On Mon, Nov 08, 2004 at 03:09:09PM -0800, Greg wrote:
> > Sorry I've been off for a week. Here is the version of the s3_late_bios
> > patch with the segment info coded in C.
> 
> This only works on my Sony VAIO Z1 if the PIC IMR is saved and
> restored around the VGA POST.  For some reason, when returning from
> the BIOS, bit 0 of the IMR is set.  That masks further clock
> interrupts so sleeping waits hang the machine after that point (for
> instance when resuming the pci devices on bus #2 like the e100). 
> Reading the IMR before and after the real-mode POST shows that only
> bit 0 is different.
> 
> Cheers!greg


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click