[Qemu-devel] qemu and OS/2

[Qemu-devel] qemu and OS/2

[Sami Haahtinen]

Has anyone successfully tried the combination of qemu and OS/2. I think 
OS/2 uses some less used instruction as only VirtualPC used to run OS/2. 
I don't know about the recent releases atleast the note of being able to 
run it has been removed. Bochs complains about some instruction call, 
and VMWare failed with something alike.

What i've gotten so far is a failed boot from CD-Rom (qemu starts but 
exists, apparently right after bios), floppy works a bit better (it 
hangs, but doesn't exit)

Regards, Sami

Re: [Qemu-devel] qemu and OS/2

[J. Mayer]

On Mon, 2004-03-15 at 22:11, Sami Haahtinen wrote:
> Has anyone successfully tried the combination of qemu and OS/2. I think 
> OS/2 uses some less used instruction as only VirtualPC used to run OS/2. 
> I don't know about the recent releases atleast the note of being able to 
> run it has been removed. Bochs complains about some instruction call, 
> and VMWare failed with something alike.
> 
> What i've gotten so far is a failed boot from CD-Rom (qemu starts but 
> exists, apparently right after bios), floppy works a bit better (it 
> hangs, but doesn't exit)

Hi,
I try for time to time to install OS/2 Warp 3,
but it crashes while executing code on disk1
(second boot disk).
I only try from floppies, because my CDROM isn't bootable.
I never tried it under another emulator...
What version are you trying to execute ?
If the floppy hangs, it may be due to a floppy emulation and/or DMA
problem.
So, I'd like to know more...

-- 
J. Mayer <l_indien@magic.fr>
Never organized

Re: [Qemu-devel] qemu and OS/2

[Fabrice Bellard]

Hi,

Can you try the CVS ? I made a specific VGA fix for the OS/2 install 
(you can see the OS/2 & IBM logo), but there are still problems after.

If OS/2 does not rely on segment limit checks or weird FPU exceptions, 
then there is a chance that QEMU can launch it soon.

Fabrice.

Sami Haahtinen wrote:
> Has anyone successfully tried the combination of qemu and OS/2. I think 
> OS/2 uses some less used instruction as only VirtualPC used to run OS/2. 
> I don't know about the recent releases atleast the note of being able to 
> run it has been removed. Bochs complains about some instruction call, 
> and VMWare failed with something alike.
> 
> What i've gotten so far is a failed boot from CD-Rom (qemu starts but 
> exists, apparently right after bios), floppy works a bit better (it 
> hangs, but doesn't exit)
> 
> Regards, Sami

[Qemu-devel] Re: qemu and OS/2

[Sami Haahtinen]

J. Mayer wrote:
> On Mon, 2004-03-15 at 22:11, Sami Haahtinen wrote: 
>>Has anyone successfully tried the combination of qemu and OS/2.
>
> but it crashes while executing code on disk1

I just finished building from CVS, and my execution fails in disk1 too.

> I only try from floppies, because my CDROM isn't bootable.

I believe that merlin (Warp4) should be bootable, but booting fails.

> What version are you trying to execute ?

I'm trying to run Merlin (warp 4), with the current stable qemu things 
fail before the top left OS/2 logo and with CVS build it fails in disk1.

Regards, Sami

[Qemu-devel] Re: qemu and OS/2

[Sami Haahtinen]

Fabrice Bellard wrote:
> Can you try the CVS ? I made a specific VGA fix for the OS/2 install 
> (you can see the OS/2 & IBM logo), but there are still problems after.

I built the CVS version, and the installation goes quite a bit further. 
Now things fail in disk1 (which is the second disk of the installation)

> If OS/2 does not rely on segment limit checks or weird FPU exceptions, 
> then there is a chance that QEMU can launch it soon.

At the moment, things go a lot better than with any of the other 
emulators i have tried, so things are looking pretty good.

I created a trace of the crash (yeah, qemu exits with segfault) and it's 
available at http://ressukka.net/~ressu/merlin.log.gz (the log is 6 megs 
compressed). And in case you (or someone else) need the installation 
floppies, disk0 and disk1_cd which should be enough to reproduce the 
issue. I can provide them for testing purposes. (i don't want to put 
them up for public)

Regards, Sami

Re: [Qemu-devel] Re: qemu and OS/2

[Jocelyn Mayer]

On Tue, 2004-03-16 at 21:03, Sami Haahtinen wrote:
> Fabrice Bellard wrote:
> > Can you try the CVS ? I made a specific VGA fix for the OS/2 install 
> > (you can see the OS/2 & IBM logo), but there are still problems after.
> 
> I built the CVS version, and the installation goes quite a bit further. 
> Now things fail in disk1 (which is the second disk of the installation)
> 
> > If OS/2 does not rely on segment limit checks or weird FPU exceptions, 
> > then there is a chance that QEMU can launch it soon.
> 
> At the moment, things go a lot better than with any of the other 
> emulators i have tried, so things are looking pretty good.
> 
> I created a trace of the crash (yeah, qemu exits with segfault) and it's 
> available at http://ressukka.net/~ressu/merlin.log.gz (the log is 6 megs 
> compressed). And in case you (or someone else) need the installation 
> floppies, disk0 and disk1_cd which should be enough to reproduce the 
> issue. I can provide them for testing purposes. (i don't want to put 
> them up for public)

It would be a good thing for me and/or Fabrice to have those floppy
images. I'd like to know if the problem can be reproduced with my
current floppy emulation, to see if it could come from there.
And Fabrice will know better how to fix the crash if it persists...
You can send them to me by mail.

I just figure out this morning that OS/2 3 seems to crash because of a
BIOS bug: there is a furtive message complaining about sthing related
to INT13...
I'll try to known more soon...

-- 
Jocelyn Mayer <l_indien@magic.fr>
Never organized

Re: [Qemu-devel] Re: qemu and OS/2

[J. Mayer]

On Tue, 2004-03-16 at 21:03, Sami Haahtinen wrote:
> Fabrice Bellard wrote:
> > Can you try the CVS ? I made a specific VGA fix for the OS/2 install 
> > (you can see the OS/2 & IBM logo), but there are still problems after.
> 
> I built the CVS version, and the installation goes quite a bit further. 
> Now things fail in disk1 (which is the second disk of the installation)
> 

Qemu crashes because of buggy sb16 emulation.
I have a version where error label in command function doesn't crash so
I can see this.
To make it act this way, you just have to replace all occurences of:
msg = (char *)-1;
by sthg like:
msg = "";

Then you may be able to go on...
I can't, cause I haven't got the CDROM and so OS/2 seem to hang (I think
the lack of CDROM may be the cause, but...)

-- 
J. Mayer <l_indien@magic.fr>
Never organized

Re: [Qemu-devel] qemu and OS/2

[Greg Alexander]

Fabrice Bellard wrote:
> Hi,
> 
> Can you try the CVS ? I made a specific VGA fix for the OS/2 install 
> (you can see the OS/2 & IBM logo), but there are still problems after.
> 
> If OS/2 does not rely on segment limit checks or weird FPU exceptions, 
> then there is a chance that QEMU can launch it soon.

If I remember correctly, at least the early versions of OS/2 DID rely on 
segment limit checks.  This allowed it to do multitaking on 286 
processors, and a lot of that mechanism was still used in the 386 code.

GREG

Re: [Qemu-devel] qemu and OS/2

[Fabrice Bellard]

Greg Alexander wrote:
> 
> 
> Fabrice Bellard wrote:
> 
>> Hi,
>>
>> Can you try the CVS ? I made a specific VGA fix for the OS/2 install 
>> (you can see the OS/2 & IBM logo), but there are still problems after.
>>
>> If OS/2 does not rely on segment limit checks or weird FPU exceptions, 
>> then there is a chance that QEMU can launch it soon.
> 
> 
> If I remember correctly, at least the early versions of OS/2 DID rely on 
> segment limit checks.  This allowed it to do multitaking on 286 
> processors, and a lot of that mechanism was still used in the 386 code.

It is possible and easy to add full segment limit and right checks in 
QEMU with a runtime selectable option. It would impact mostly 16 bit 
code and the slowdown would be small: because QEMU is a dynamic 
translator, some tests can be done at translation time.

Unfortunately it is not yet on the top on my TODO list :-(

Fabrice.

[Qemu-devel] Re: qemu and OS/2

[Sami Haahtinen]

J. Mayer wrote:
> Qemu crashes because of buggy sb16 emulation.
> I have a version where error label in command function doesn't crash so
> I can see this.
> To make it act this way, you just have to replace all occurences of:
> msg = (char *)-1;
> by sthg like:
> msg = "";

Did this.

> Then you may be able to go on...

The installation still crashes qemu, only thing that is different now is 
that instead of segfault i now get abort.

> I can't, cause I haven't got the CDROM and so OS/2 seem to hang (I think
> the lack of CDROM may be the cause, but...)

I don't think there is anything special after the first 2 disks, after 
that the installation just copies the files. so if we can get through 
the first 2 disks i would guess that things should go well after that.
Did it ask you for the CD?

Regards, Sami

Re: [Qemu-devel] Re: qemu and OS/2

[J. Mayer]

On Thu, 2004-03-18 at 22:19, Sami Haahtinen wrote:
> J. Mayer wrote:
> > Qemu crashes because of buggy sb16 emulation.
...
> > To make it act this way, you just have to replace all occurences of:
> > msg = (char *)-1;
> > by sthg like:
> > msg = "";
> 
> Did this.
> 
> > Then you may be able to go on...
> 
> The installation still crashes qemu, only thing that is different now is 
> that instead of segfault i now get abort.

My fault:
I missed to tell you to comment the abort call in Fail macro (still in
sb16.c)

> > I can't, cause I haven't got the CDROM and so OS/2 seem to hang (I think
> > the lack of CDROM may be the cause, but...)
> 
> I don't think there is anything special after the first 2 disks, after 
> that the installation just copies the files. so if we can get through 
> the first 2 disks i would guess that things should go well after that.
> Did it ask you for the CD?

In fact, I retried and it's now asking for disk2, so I cannot go on.

Regards.

-- 
J. Mayer <l_indien@magic.fr>
Never organized

[Qemu-devel] Re: qemu and OS/2

[Sami Haahtinen]

J. Mayer wrote:
> On Thu, 2004-03-18 at 22:19, Sami Haahtinen wrote:
>>The installation still crashes qemu, only thing that is different now is 
>>that instead of segfault i now get abort.
> 
> My fault:
> I missed to tell you to comment the abort call in Fail macro (still in
> sb16.c)

Ok, did this, and now it just hangs, it doesn't ask me for the next 
image or anything, it hangs in the first graphical OS/2 Warp logo.

>>>I can't, cause I haven't got the CDROM and so OS/2 seem to hang (I think
>>>the lack of CDROM may be the cause, but...)
>>
>>I don't think there is anything special after the first 2 disks, after 
>>that the installation just copies the files. so if we can get through 
>>the first 2 disks i would guess that things should go well after that.
>>Did it ask you for the CD?
>  
> In fact, I retried and it's now asking for disk2, so I cannot go on.

Ok, if you wish to try one more disk, it's up at the same location where 
you got the first disks.

Regards, Sami

Re: [Qemu-devel] Re: qemu and OS/2

[J. Mayer]

On Fri, 2004-03-19 at 06:31, Sami Haahtinen wrote:
> J. Mayer wrote:
> > On Thu, 2004-03-18 at 22:19, Sami Haahtinen wrote:
> >>The installation still crashes qemu, only thing that is different now is 
> >>that instead of segfault i now get abort.
> > 
> > My fault:
> > I missed to tell you to comment the abort call in Fail macro (still in
> > sb16.c)
> 
> Ok, did this, and now it just hangs, it doesn't ask me for the next 
> image or anything, it hangs in the first graphical OS/2 Warp logo.
> 
Well, it seems to hang 1 over 2 tries, for me.

> >>>I can't, cause I haven't got the CDROM and so OS/2 seem to hang (I think
> >>>the lack of CDROM may be the cause, but...)
> >>
> >>I don't think there is anything special after the first 2 disks, after 
> >>that the installation just copies the files. so if we can get through 
> >>the first 2 disks i would guess that things should go well after that.
> >>Did it ask you for the CD?
> >  
> > In fact, I retried and it's now asking for disk2, so I cannot go on.
> 
> Ok, if you wish to try one more disk, it's up at the same location where 
> you got the first disks.

I did launch the second disk, but merlin crashes exactly the same way
OS/2 3 does, at the same address, with almost all registers identical.
As it dumps the segment limits, I'm afraid that Fabrice may be right: it
may rely on segment limits so we won't have OS/2 support soon....

-- 
J. Mayer <l_indien@magic.fr>
Never organized

[Qemu-devel] Re: qemu and OS/2

[Sami Haahtinen]

J. Mayer wrote:
> On Fri, 2004-03-19 at 06:31, Sami Haahtinen wrote:
>>J. Mayer wrote:
>>>In fact, I retried and it's now asking for disk2, so I cannot go on.
>>
>>Ok, if you wish to try one more disk, it's up at the same location where 
>>you got the first disks.
> 
> I did launch the second disk, but merlin crashes exactly the same way
> OS/2 3 does, at the same address, with almost all registers identical.
> As it dumps the segment limits, I'm afraid that Fabrice may be right: it
> may rely on segment limits so we won't have OS/2 support soon....

Awww, too bad. i was looking forward to using OS/2 again. i don't want 
to dedicate a machine for it.

Oh well, you can't always win. Even still, qemu is the best OS VM out 
there, as any of the other VMs couldn't run OS/2 either.

Regards, Sami

Re: [Qemu-devel] Re: qemu and OS/2

[Fabrice Bellard]

Sami Haahtinen wrote:

> Awww, too bad. i was looking forward to using OS/2 again. i don't want 
> to dedicate a machine for it.
> 
> Oh well, you can't always win. Even still, qemu is the best OS VM out 
> there, as any of the other VMs couldn't run OS/2 either.

My latest patches should improve the OS/2 Warp behavior (with the 
install disks you can have the OS/2 prompt).

Note that the CVS version of Bochs is able to boot OS/2.

The most important issue now for OS/2 and some other OSes seems to be 
the RTC emulation (QEMU uses the host timer, so it cannot have a 
resolution bigger than 10 ms for timer interrupts on 2.4 host Linux 
kernels, and some guest OSes need a timer interrupt with a period < 10 ms).

Fabrice.

[Qemu-devel] Re: qemu and OS/2

[Sami Haahtinen]

Fabrice Bellard wrote:
> My latest patches should improve the OS/2 Warp behavior (with the 
> install disks you can have the OS/2 prompt).

When did the default for static build change from no to yes?
Also, when building static, it appears that SDL in debian needs -lgpm 
too. (aalib depends on it apparently)

And back to the point. The installation does get further, but it appears 
that it infact does crash now and then. The installation works better on 
some occations. First it failed to format, and just now it failed to 
install.

On the last attempt it printed this a few times:
FLOPPY ERROR: fdctrl_transfer_handler: writting sector 0
FLOPPY ERROR: fdctrl_transfer_handler: writting sector 0
FLOPPY ERROR: fdctrl_transfer_handler: writting sector 0

Otherwise, it might be installable. I'll keep on trying..

> Note that the CVS version of Bochs is able to boot OS/2.

Nice, it's good to see OS projects take the lead from commercial ones.

> The most important issue now for OS/2 and some other OSes seems to be 
> the RTC emulation (QEMU uses the host timer, so it cannot have a 
> resolution bigger than 10 ms for timer interrupts on 2.4 host Linux 
> kernels, and some guest OSes need a timer interrupt with a period < 10 ms).

does it work any better on 2.6? I'm currently trying out qemu on 2.6, 
could that be a source for problems?

-- Sami