4:36 a.m.
I've recently upgraded a host from Ubuntu Precise (qemu-kvm-1.0) to
Ubuntu Trusty (qemu 2.0.0). I have a Windows 2008 32-bit OS running on a
64-bit VM that runs a 16-bit line of business application. (While I
realize that "upgrade the application" is the right answer, I can't.)
The Windows VM boots, but the business application crashes on startup.
Changing the VM to be a 32-bit VM doesn't help.
After several hours of compiling and testing intermediate qemu versions,
I accidentally stumbled into the real issue. Trusty has switched from
vgabios to seabios. I have confirmed that switching the vgabios*.bin
images back to the vgabios package (rather than seabios) fixes the
16-bit application in the guest.
Per a suggestion on the Ubuntu bug I filed, I built an updated seabios
package using the source from git (specifically, revision
60e0e55f212dadd043ab9e39bee05a48013ddd8f). It has the same problem.
I then set CONFIG_DEBUG_LEVEL=8 and booted with "-chardev
stdio,id=seabios -device isa-debugcon,iobase=0x402,chardev=seabios" per:
http://www.seabios.org/pipermail/seabios/2011-May/001718.html
The debug log is attached. For more details, including a couple of
screenshots of the NTVDM crash dialog, see:
https://bugs.launchpad.net/ubuntu/+source/qemu/+bug/1404396
What's the next step in debugging this?
--
Richard
noon
On 01/09/15 04:36, Richard Laager wrote:
I've recently upgraded a host from Ubuntu Precise
(qemu-kvm-1.0) to
Ubuntu Trusty (qemu 2.0.0). I have a Windows 2008 32-bit OS running on a
64-bit VM that runs a 16-bit line of business application. (While I
realize that "upgrade the application" is the right answer, I can't.)
The Windows VM boots, but the business application crashes on startup.
Changing the VM to be a 32-bit VM doesn't help.
After several hours of compiling and testing intermediate qemu versions,
I accidentally stumbled into the real issue. Trusty has switched from
vgabios to seabios. I have confirmed that switching the vgabios*.bin
images back to the vgabios package (rather than seabios) fixes the
16-bit application in the guest.
Per a suggestion on the Ubuntu bug I filed, I built an updated seabios
package using the source from git (specifically, revision
60e0e55f212dadd043ab9e39bee05a48013ddd8f). It has the same problem.
I then set CONFIG_DEBUG_LEVEL=8 and booted with "-chardev
stdio,id=seabios -device isa-debugcon,iobase=0x402,chardev=seabios" per:
http://www.seabios.org/pipermail/seabios/2011-May/001718.html
The debug log is attached. For more details, including a couple of
screenshots of the NTVDM crash dialog, see:
https://bugs.launchpad.net/ubuntu/+source/qemu/+bug/1404396
http://en.wikipedia.org/wiki/NTVDM
Very roughly, it's windows' 16-bit emulator. It parses real mode code
and emulates the instructions. The SeaVGBABIOS binary apparently
contains at least one instruction that, albeit valid, confuses NTVDM and
causes it to crash.
This has happened several times before. Not just with NTVDM but also
x86emu -- search this list for "x86emu". x86emu is free software with a
similar role, and one version or another of the X server uses it to
"execute" 16-bit VBE code.
The original vgabios was written in assembly, which (probably) made its
maintenance hell, but it provided full control over the instructions in
the final binary (so issues like this had never been encountered or
quickly fixed). SeaVGBABIOS is (mostly) written in C and sometimes gcc
generates "sophisticated" stuff that confuse old emulators. Then usually
Kevin tracks it down and does some magic to make it go away (check out
"scripts/vgafixup.py").
What's the next step in debugging this?
The offending instruction should be found.
The NTVDM crash info is not directly useful because that pinpoints (?) a
location in the NTVDM code (for which you don't have the source). The
problematic SeaVGABIOS instruction counts as data for NTVDM.
You could try to bisect SeaVGABIOS, and/or build it with an older gcc.
(I probably made several errors in the above; corrections more than
welcome...)
Thanks
Laszlo
5:58 p.m.
On Thu, Jan 08, 2015 at 09:36:44PM -0600, Richard Laager wrote:
I've recently upgraded a host from Ubuntu Precise
(qemu-kvm-1.0) to
Ubuntu Trusty (qemu 2.0.0). I have a Windows 2008 32-bit OS running on a
64-bit VM that runs a 16-bit line of business application. (While I
realize that "upgrade the application" is the right answer, I can't.)
The Windows VM boots, but the business application crashes on startup.
Changing the VM to be a 32-bit VM doesn't help.
After several hours of compiling and testing intermediate qemu versions,
I accidentally stumbled into the real issue. Trusty has switched from
vgabios to seabios. I have confirmed that switching the vgabios*.bin
images back to the vgabios package (rather than seabios) fixes the
16-bit application in the guest.
Hi Richard,
Thanks for the detailed report, and I'm sorry that you are having
problems.
Per a suggestion on the Ubuntu bug I filed, I built an
updated seabios
package using the source from git (specifically, revision
60e0e55f212dadd043ab9e39bee05a48013ddd8f). It has the same problem.
I then set CONFIG_DEBUG_LEVEL=8 and booted with "-chardev
stdio,id=seabios -device isa-debugcon,iobase=0x402,chardev=seabios" per:
http://www.seabios.org/pipermail/seabios/2011-May/001718.html
The debug log is attached. For more details, including a couple of
screenshots of the NTVDM crash dialog, see:
https://bugs.launchpad.net/ubuntu/+source/qemu/+bug/1404396
Does the crash occur around the time one of the debug messages is
produced, or is the crash seemingly uncorrelated? I don't see
anything suspicious in the log.
What's the next step in debugging this?
Getting a test case that we could run to reproduce on our side would
really help. Do other 16bit programs in your environment also crash?
Does the app or some part of its 16bit code run for some time before
the crash, or does the crash occur immiediately (ie, is it NTVDM
crashing or is something in the app causing NTVDM to crash)?
I'm a bit surprised that NTVDM would be directly accessing the video
bios, or would allow the 16bit programs it runs to direclty access the
bios. Can you provide some more info on the app itself - is it a dos
program, a windows 3.0 program, does it run full screen or in a
window? If it runs full screen, is it possible to run it in a window,
and does that still crash?
Ultimately what we want to try and do is find what part of seavgabios
is at issue.
There are a couple of things you could try to see if it makes any
difference - entirely disable debugging in seavgabios
(CONFIG_DEBUG_LEVEL=0) in the unlikley case that the debug port writes
themselves are causing confusion, and try disabling
CONFIG_VGA_ALLOCATE_EXTRA_STACK.
Also, can you check if the emulated cirrus vga card has the same issue
(qemu command line of "-vga cirrus" and make sure there is no "-vga
std" - not sure how one does that from libvirt).
-Kevin
9:53 a.m.
[I've reordered the quoted text.]
On Fri, 2015-01-09 at 11:58 -0500, Kevin O'Connor wrote:
Does the crash occur around the time one of the debug
messages is
produced, or is the crash seemingly uncorrelated? I don't see
anything suspicious in the log.
Sorry, I forgot to note this in my email. All of the messages are from
bootup. There are no logs from when I start the 16-bit application.
Do other 16bit programs in your environment also
crash?
This suite of programs are the only 16-bit applications we run.
Does the app or some part of its 16bit code run for
some time before
the crash, or does the crash occur immiediately (ie, is it NTVDM
crashing or is something in the app causing NTVDM to crash)?
It crashes immediately on startup.
I'm a bit surprised that NTVDM would be directly
accessing the video
bios, or would allow the 16bit programs it runs to direclty access the
bios. Can you provide some more info on the app itself - is it a dos
program, a windows 3.0 program, does it run full screen or in a
window? If it runs full screen, is it possible to run it in a window,
and does that still crash?
I'm fairly confident they're Windows programs, as opposed to DOS
programs. They are not fullscreen, they run in a window.
Getting a test case that we could run to reproduce on
our side would
really help.
Paolo Bonzini suggested I try the 16-bit SkiFree. It works on vgabios.
It crashes on seabios, immediately on start.
To be specific, 16-bit Skifree crashes with Ubuntu-packaged
seabios1.7.4-4 on 32-bit Windows Server 2008 running on Ubuntu Trusty.
It also crashes with a build from seabios git I did just now.
Also, can you check if the emulated cirrus vga card
has the same issue
(qemu command line of "-vga cirrus" and make sure there is no "-vga
std" - not sure how one does that from libvirt).
Skifree still crashes with seabios git and cirrus. I verified there is
no -vga std in the command line in the cirrus test. The only "vga" on
the command line is: -device cirrus-vga,id=video0,bus=pci.0,addr=0x2
There are a couple of things you could try to see if
it makes any
difference - entirely disable debugging in seavgabios
(CONFIG_DEBUG_LEVEL=0) in the unlikley case that the debug port writes
themselves are causing confusion, and try disabling
CONFIG_VGA_ALLOCATE_EXTRA_STACK.
Turning off CONFIG_VGA_ALLOCATE_EXTRA_STACK makes Skifree work on cirrus
and vga.
I don't know how this all works, but as a guess, I further tried with
"vga" (not cirrus) using VGA_EXTRA_STACK_SIZE=256 and
VGA_EXTRA_STACK_SIZE=16. The crash recurs with both.
--
Richard
4:32 p.m.
On Mon, Jan 12, 2015 at 02:53:58AM -0600, Richard Laager wrote:
On Fri, 2015-01-09 at 11:58 -0500, Kevin O'Connor
wrote:
Interesting. Does your original business app then also work with
CONFIG_VGA_ALLOCATE_EXTRA_STACK disabled?
-Kevin
There are a couple of things you could try to see
if it makes any
difference - entirely disable debugging in seavgabios
(CONFIG_DEBUG_LEVEL=0) in the unlikley case that the debug port writes
themselves are causing confusion, and try disabling
CONFIG_VGA_ALLOCATE_EXTRA_STACK.
Turning off CONFIG_VGA_ALLOCATE_EXTRA_STACK makes Skifree work on cirrus
and vga.
4:59 p.m.
On Mon, 2015-01-12 at 10:32 -0500, Kevin O'Connor wrote:
Interesting. Does your original business app then
also work with
CONFIG_VGA_ALLOCATE_EXTRA_STACK disabled?
Yes. I couldn't be sure if I tested before, since I didn't write it
down. But I just tested now.
Obviously this extra stack feature exists for a reason. What are the
implications of turning it off?
--
Richard
5:09 p.m.
On 12/01/2015 16:59, Richard Laager wrote:
On Mon, 2015-01-12 at 10:32 -0500, Kevin O'Connor
wrote:
commit 4a8b58cb6cccc8f6431167dfdd36f3e39601ff79
Author: Kevin O'Connor <kevin(a)koconnor.net>
Date: Sat Nov 30 19:16:15 2013 -0500
vgabios: Support allocating an extra stack for vgabios calls and default on.
Add code to allocate an extra stack for the main vgabios int 0x10
entry point. The allocation is done via the PMM spec and uses a PCI
v3 permanent low memory region request. This request will work with
SeaBIOS - it is unknown how many other main BIOS implementations
support this PMM call.
The extra stack is useful for old DOS programs that call the VGABIOS
and expect it to work with very small amounts of stack space.
Signed-off-by: Kevin O'Connor <kevin(a)koconnor.net>
Paolo
Interesting. Does your original business app
then also work with
CONFIG_VGA_ALLOCATE_EXTRA_STACK disabled?
Yes. I couldn't be sure if I tested before, since I didn't write it
down. But I just tested now.
Obviously this extra stack feature exists for a reason. What are the
implications of turning it off?
5:13 p.m.
On Mon, Jan 12, 2015 at 09:59:49AM -0600, Richard Laager wrote:
On Mon, 2015-01-12 at 10:32 -0500, Kevin O'Connor
wrote:
There is a high-level description of the feature at:
http://seabios.org/Execution_and_code_flow#Extra_16bit_stack
Basically, SeaVGABIOS is capable of allocating space for an internal
stack at startup and then switching to that stack on each vgabios call
so that it can work with some really old DOS-era programs that used
ridiculously small stacks. If things work for you with the feature
off, then it's not a problem to leave it off.
I'm trying to reproduce the fault with skifree locally. If I can't,
it would be helpful if you can run some additional tests and grab some
additional logs. I'll let you know.
-Kevin
Interesting. Does your original business app
then also work with
CONFIG_VGA_ALLOCATE_EXTRA_STACK disabled?
Yes. I couldn't be sure if I tested before, since I didn't write it
down. But I just tested now.
Obviously this extra stack feature exists for a reason. What are the
implications of turning it off?
7:19 p.m.
On Mon, Jan 12, 2015 at 02:53:58AM -0600, Richard Laager wrote:
Turning off CONFIG_VGA_ALLOCATE_EXTRA_STACK makes
Skifree work on cirrus
and vga.
I was able to reproduce this locally with 16bit skifree on Windows
Vista. (Interestingly, the problem doesn't occur on winxp.)
The issue doesn't appear to be with the SeaVGABIOS stack switching,
but with the fact that the SeaBIOS PMM code places the stack in the
e-segment. Turning off MALLOC_UPPERMEMORY in SeaBIOS allows
SeaVGABIOS to run even with CONFIG_VGA_ALLOCATE_EXTRA_STACK set.
My guess is that Windows is emulating the vgabios, but marking the
0xc0000-0x100000 region as read-only. Oddly, it doesn't appear as
Windows actually lets the code talk to the VGA hardware, as the debug
output (and presumably other in/out accesses) is suppressed. So, it's
unclear what Windows is attempting to do with its emulation.
Not sure what the best way forward is here. It seems this is a choice
between supporting some very old programs vs support for some other
very old programs. Paolo and Gerd, maybe you have some ideas?
I can think of a few options:
1 - do nothing - let users use a modified seabios/seavgabios or the
"lgpl vgabios" for this situation. Not great - specially considering
how difficult it would be to know if one is in this situation or not.
2 - default SeaVGABIOS to CONFIG_VGA_ALLOCATE_EXTRA_STACK off. Known
to break old programs - for example, DOS 1.0. SeaVGABIOS can use just
under 300 bytes of stack space for some calls.
3 - default SeaBIOS to MALLOC_UPPERMEMORY off. Unfortunately, this
wastes additional space below 640K and it's unclear what impact that
would have on old programs.
4 - Change SeaVGABIOS to allocate its stack in the EBDA instead of via
a PMM call. Unfortunately, I've seen at least one old DOS-era program
that ignores the EBDA allocations and writes to the end of 640K
memory. It's unclear how it would react to a SeaVBABIOS stack being
there.
5 - Like 4, but know that SeaBIOS doesn't use the bottom half of the
first 1K of EBDA and use that. Same problems as 4.
6 - Try to detect if the code is called in VM86 mode and don't use the
extra stack then - see patch below. The patch does make skifree work,
but I'm uncertain if it would catch other users (eg, kvm on some intel
chipsets?, some old dos program if dos is using emm386 mode).
-Kevin
--- a/vgasrc/vgaentry.S
+++ b/vgasrc/vgaentry.S
@@ -8,6 +8,7 @@
#include "asm-offsets.h" // BREGS_*
#include "config.h" // CONFIG_*
#include "entryfuncs.S" // ENTRY_*
+#include "x86.h" // CR0_PE
/****************************************************************
@@ -109,6 +110,13 @@ entry_10:
entry_10_extrastack:
cli
cld
+
+ push %ax // Don't use extra stack if in protected mode
+ smsww %ax
+ test $CR0_PE, %ax
+ pop %ax
+ jne entry_10
+
pushw %ds // Set %ds:%eax to space on ExtraStack
pushl %eax
movw %cs:ExtraStackSeg, %ds
7:25 p.m.
On 12/01/2015 19:19, Kevin O'Connor wrote:
I was able to reproduce this locally with 16bit
skifree on Windows
Vista. (Interestingly, the problem doesn't occur on winxp.)
I put "there is a 16-bit freely downloadable version of skifree" on my
personal list of trivia that ended up becoming useful. :D
5 - Like 4, but know that SeaBIOS doesn't use the
bottom half of the
first 1K of EBDA and use that. Same problems as 4.
The 1K EBDA has been around for 15 years now, hasn't it?
6 - Try to detect if the code is called in VM86 mode
and don't use the
extra stack then - see patch below. The patch does make skifree work,
but I'm uncertain if it would catch other users (eg, kvm on some intel
chipsets?, some old dos program if dos is using emm386 mode).
No, KVM hides the fact that you are in protected mode. EMM386 would be
affected, but then it is not impossible for old programs to require
disabling it.
All in all (5) or (6) both sound good.
Paolo
7:54 p.m.
On Mon, Jan 12, 2015 at 07:25:58PM +0100, Paolo Bonzini wrote:
I was under the vague impression that kvm uses VM86 mode to run 16bit
code on some Intel chipsets. The SMSW instruction isn't privileged so
I didn't think it could be hidden.
On 12/01/2015 19:19, Kevin O'Connor wrote:
Way longer than that. Not sure when it was introduced, but my Phoenix
"CBIOS" book from 1989 has it. So, at least 25 years.
I was able to reproduce this locally with 16bit
skifree on Windows
Vista. (Interestingly, the problem doesn't occur on winxp.)
I put "there is a 16-bit freely downloadable version of skifree" on my
personal list of trivia that ended up becoming useful. :D
5 - Like 4, but know that SeaBIOS doesn't use
the bottom half of the
first 1K of EBDA and use that. Same problems as 4.
The 1K EBDA has been around for 15 years now, hasn't it? 6 - Try to
detect if the code is called in VM86 mode and don't use the
extra stack then - see patch below. The patch does make skifree work,
but I'm uncertain if it would catch other users (eg, kvm on some intel
chipsets?, some old dos program if dos is using emm386 mode).
No, KVM hides the fact that you are in protected mode. EMM386 would be
affected, but then it is not impossible for old programs to require
disabling it. All in all (5) or (6) both sound good.
Thanks,
-Kevin
8 p.m.
No, KVM hides
the fact that you are in protected mode. EMM386 would be
affected, but then it is not impossible for old programs to require
disabling it.
I was under the vague impression that kvm uses VM86 mode to run 16bit
code on some Intel chipsets. The SMSW instruction isn't privileged so
I didn't think it could be hidden.
8:36 p.m.
On Mon, Jan 12, 2015 at 02:00:24PM -0500, Paolo Bonzini wrote:
It isn't privileged indeed (nice trick in fact!), but that doesn't matter
for VT-x extensions.
Old processors let you run the processor in VMX non-root mode (i.e. as a VM)
only in protected mode, so KVM uses VM86 when the processor is in real mode
(and uses an interpreter while in big real mode or during real<->protected mode
transitions).
But all the bells and whistles of VMX still apply, including the ability to
fake the value of CR0 for both MOV and [LS]MSW instructions.
Okay, so it fakes real-mode by setting up a protected mode guest with
a fake CR0 that is running vm86, and so SMSW still returns a value
with PE off? (As opposed to a regular guest that itself launches a
VM86 instance, in which case CR0 from SMSW would have PE on.)
Good to know - thanks.
-Kevin
No, KVM hides the fact that you are in protected mode.
EMM386 would be
affected, but then it is not impossible for old programs to require
disabling it.
I was under the vague impression that kvm uses VM86 mode to run 16bit
code on some Intel chipsets. The SMSW instruction isn't privileged so
I didn't think it could be hidden.
8:40 p.m.
On 12/01/2015 20:36, Kevin O'Connor wrote:
Okay, so it fakes real-mode by setting up a protected
mode guest with
a fake CR0 that is running vm86, and so SMSW still returns a value
with PE off? (As opposed to a regular guest that itself launches a
VM86 instance, in which case CR0 from SMSW would have PE on.)
Yes.
Paolo
8:13 p.m.
Your patch implementing idea 6 fixes the problem for my business
application on a stock Ubuntu package (with extra stack enabled).
--
Richard
9:06 p.m.
On Mon, Jan 12, 2015 at 01:13:54PM -0600, Richard Laager wrote:
Your patch implementing idea 6 fixes the problem for
my business
application on a stock Ubuntu package (with extra stack enabled).
Thanks.
The key part of option 5 (as described in my previous email) looks
like the patch below. It also works with skifree on Vista for me.
-Kevin
--- a/vgasrc/vgaentry.S
+++ b/vgasrc/vgaentry.S
@@ -111,8 +111,10 @@ entry_10_extrastack:
cld
pushw %ds // Set %ds:%eax to space on ExtraStack
pushl %eax
- movw %cs:ExtraStackSeg, %ds
- movl $(CONFIG_VGA_EXTRA_STACK_SIZE-PUSHBREGS_size-16), %eax
+ movw $SEG_BDA, %ax
+ movw %ax, %ds
+ movw 0x0e, %ds
+ movl $(1024-PUSHBREGS_size-16), %eax
SAVEBREGS_POP_DSEAX // Save registers on extra stack
movl %esp, PUSHBREGS_size+8(%eax)
movw %ss, PUSHBREGS_size+12(%eax)
11:07 p.m.
On Mon, 2015-01-12 at 15:06 -0500, Kevin O'Connor wrote:
On Mon, Jan 12, 2015 at 01:13:54PM -0600, Richard
Laager wrote:
I may have screwed something up while applying, but when I tested with
git plus this patch, I didn't get any video output during booting.
I tested with Ubuntu's seabios 1.7.4, and adding a modified version of
this patch did fix the problem. The changes seemed pretty obvious in
context. It seems that BREGS_size-8 was changed to PUSHBREGS_size-16
somewhere after 1.7.4. Here's the patch that I applied to 1.7.4:
Index: seabios-1.7.4/vgasrc/vgaentry.S
===================================================================
--- seabios-1.7.4.orig/vgasrc/vgaentry.S 2015-01-12 15:28:20.060000981 -0600
+++ seabios-1.7.4/vgasrc/vgaentry.S 2015-01-12 15:29:35.296142288 -0600
@@ -101,8 +101,10 @@
cld
pushw %ds // Set %ds:%eax to space on ExtraStack
pushl %eax
- movw %cs:ExtraStackSeg, %ds
- movl $(CONFIG_VGA_EXTRA_STACK_SIZE-BREGS_size-8), %eax
+ movw $SEG_BDA, %ax
+ movw %ax, %ds
+ movw 0x0e, %ds
+ movl $(1024-BREGS_size-8), %eax
popl BREGS_eax(%eax) // Backup registers
popw BREGS_ds(%eax)
movl %edi, BREGS_edi(%eax)
So it seems you have two working solutions (though I should retest with
git, if you want to go with option 5).
If/when you choose one of these and commit it, I'll update the Ubuntu
bug asking them to deploy the patch as an SRU (stable release update).
--
Richard
Your patch implementing idea 6 fixes the problem
for my business
application on a stock Ubuntu package (with extra stack enabled).
Thanks.
The key part of option 5 (as described in my previous email) looks
like the patch below. It also works with skifree on Vista for me.
-Kevin
--- a/vgasrc/vgaentry.S
+++ b/vgasrc/vgaentry.S
@@ -111,8 +111,10 @@ entry_10_extrastack:
cld
pushw %ds // Set %ds:%eax to space on ExtraStack
pushl %eax
- movw %cs:ExtraStackSeg, %ds
- movl $(CONFIG_VGA_EXTRA_STACK_SIZE-PUSHBREGS_size-16), %eax
+ movw $SEG_BDA, %ax
+ movw %ax, %ds
+ movw 0x0e, %ds
+ movl $(1024-PUSHBREGS_size-16), %eax
SAVEBREGS_POP_DSEAX // Save registers on extra stack
movl %esp, PUSHBREGS_size+8(%eax)
movw %ss, PUSHBREGS_size+12(%eax)
5:32 a.m.
On Mon, 2015-01-12 at 16:07 -0600, Richard Laager wrote:
I may have screwed something up while applying, but
when I tested with
git plus this patch, I didn't get any video output during booting.
...
So it seems you have two working solutions (though I
should retest
with git, if you want to go with option 5).
Have you decided on a preferred approach? If option 5, I'll retest your
patch with seabios git.
--
Richard
1:41 p.m.
On Fri, Jan 16, 2015 at 10:32:45PM -0600, Richard Laager wrote:
On Mon, 2015-01-12 at 16:07 -0600, Richard Laager
wrote:
I'm inclined to go with option 5. The real patch for that option,
however, is a little bigger. I should be able to send a more complete
patch in the next couple of days.
Thanks,
-Kevin
I may have screwed something up while applying,
but when I tested with
git plus this patch, I didn't get any video output during booting.
...
So it seems you have two working solutions
(though I should retest
with git, if you want to go with option 5).
Have you decided on a preferred approach? If option 5, I'll retest your
patch with seabios git. 
3:30 p.m.
Paolo Bonzini wrote:
That would be pretty slow on KVM, since video memory is MMIO.
slow reliable > fast unreliable
> I can think of a few options:
7 - use top
of video memory as stack? But worse, one could imagine that NTVDM blocks
a0000-bffff as well.
I doubt that? Then how would the VGA BIOS write its data?
//Peter
3:47 p.m.
On 13/01/2015 15:30, Peter Stuge wrote:
The question is more "why is NTVDM executing the VGA BIOS?" Remember
this is a 16-bit windowed application. It's possible that some old
Win16 app was doing INT 10h calls and so they have to support it in
NTVDM: they initialize the VGABIOS, but do not write to VRAM because...
nothing should be there, right?
Not writing the data to VRAM could very well be intentional.
Paolo
That would be
pretty slow on KVM, since video memory is MMIO.
slow reliable > fast unreliable
But worse, one could imagine that NTVDM blocks
a0000-bffff as well.
I doubt that? Then how would the VGA BIOS write its data?
1905
days inactive
1913
days old
22 comments
5 participants
participants (5)
-
Kevin O'Connor -
Laszlo Ersek -
Paolo Bonzini -
Peter Stuge -
Richard Laager