Re: [OpenBIOS] testing Mac OS X boot on latest trunk (revision 1085) + divide by zero patch

[Programmingkid]

> On 13.01.13 16:43, Mark Cave-Ayland wrote:
> >
>
> On 12/01/13 16:05, Amadeusz Sławiński wrote:
> >
> > >
>
> I used the latest proposed divide by zero patch
> > >
>
> So it seems like it is almost there.
> > >
>
> Seems like it still can't load kernel when run with boot cd:,\:tbxi
> > >
>
> Mac OS X 10.0
> > > http://asmblr.net/scrot/2013-01-12-155737_1366x768_scrot.png
>
> Mac OS X 10.2
> > > http://asmblr.net/scrot/2013-01-12-155642_1366x768_scrot.png
>
> Mac OS X 10.4
> > > http://asmblr.net/scrot/2013-01-12-155935_1366x768_scrot.png
>
> However when I specify partition to use it loads:
> > >
>
> Mac OS X 10.0 (boot cd:9,\:tbxi)
> > > http://asmblr.net/scrot/2013-01-12-163133_1366x768_scrot.png
>
> Mac OS X 10.2 (boot cd:9,\:tbxi)
> > > http://asmblr.net/scrot/2013-01-12-163300_1366x768_scrot.png
>
> Mac OS X 10.4 (boot cd:3,\:tbxi)
> > > http://asmblr.net/scrot/2013-01-12-163419_1366x768_scrot.png
> >
>
> This should now be fixed in SVN trunk.
> >
> >
>
> I believe I have now committed all outstanding patches based upon
> >
>
> William's main GSOC patchset, and performed a QEMU CDROM boot test
> >
>
> across all of SPARC32, SPARC64 and PPC to confirm that I haven't
> >
>
> introduced any regressions across any of my test images. Thank you
> >
>
> everyone for your help, patience and testing.
> >
> >
>
> Interestingly enough, commits 1081 and 1082 appear to have fixed the key
> >
>
> issues with OpenBIOS's PPC MMU code and as a result, the following
> >
>
> kernels will now at least boot from ISO images in QEMU:
> >
> >
>
> Darwin
> >
>
> FreeBSD
> >
>
> NetBSD
> >
>
> HelenOS
> >
>
> Fedora Linux
> >
>
> SuSE Linux
> >
> >
>
> I think the remainder of the bugs during kernel initialisation are
> >
>
> related to either missing/erroneous properties in the OpenBIOS device
> >
>
> tree, or bugs in QEMU's device emulation/Mac hardware model. This will
> >
>
> likely require some assistance from the QEMU people to fix.
>
> I can confirm the statement from Mark regarding FreeBSD. So far I can
> boot but I fail due to the fact that some interrupt-parents are not
> populated. Namely for via-cuda and for ata-1/2/3. Adding these lets the
> boot continue until I hang somewhere in a cpu_idle routine. Still
> investigating.
>
> One thing the qemu people might clarify, is -M mac99 ok to use? Or what
> would be the preferred Mac model and which cpu?
> My trials go with mac99 and -cpu G4 on the 32-bit qemu.
>
> I think once I solved the spinning in the cpu_idle routine I'm near
> completing the boot.
>
> Thank you all for this effort!
>
> Andreas

Would you have a patch for this interrupt-parent property?

If it helps, here is the what I found in an G3 iMac:

via-pmu properties

0 > .properties 
name                    via-pmu
device_type             via-pmu
compatible              pmu
reg                     00016000  00002000 
interrupts              00000019 00000001 
interrupt-parent        ff90fa00 
pmu-version             00d07e0c 
AAPL,clock-id           73706920 73703331 
AAPL,clock-data         01de2000 00000044 00008000 00000000 00000000 00000044 00008000 00000044 
                        00000010 6e756c6c 6e756c6c 012ad400

I think the mac99 model might be better at making Mac OS X boot on QEMU than the Beige G3 model. I personally would stick with the G3 processor. It has less features than the G4 so it might be easier to work with and emulate.

testing Mac OS X boot on latest trunk (revision 1085) + divide by zero patch

[Andreas Tobler]

On 01.02.13 21:38, Programmingkid wrote:
> > On 13.01.13 16:43, Mark Cave-Ayland wrote:
> > >
> >
> > On 12/01/13 16:05, Amadeusz Sławiński wrote:
> > >
> > > >
> >
> > I used the latest proposed divide by zero patch
> > > >
> >
> > So it seems like it is almost there.
> > > >
> >
> > Seems like it still can't load kernel when run with boot cd:,\:tbxi
> > > >
> >
> > Mac OS X 10.0
> > > > http://asmblr.net/scrot/2013-01-12-155737_1366x768_scrot.png
> >
> > Mac OS X 10.2
> > > > http://asmblr.net/scrot/2013-01-12-155642_1366x768_scrot.png
> >
> > Mac OS X 10.4
> > > > http://asmblr.net/scrot/2013-01-12-155935_1366x768_scrot.png
> >
> > However when I specify partition to use it loads:
> > > >
> >
> > Mac OS X 10.0 (boot cd:9,\:tbxi)
> > > > http://asmblr.net/scrot/2013-01-12-163133_1366x768_scrot.png
> >
> > Mac OS X 10.2 (boot cd:9,\:tbxi)
> > > > http://asmblr.net/scrot/2013-01-12-163300_1366x768_scrot.png
> >
> > Mac OS X 10.4 (boot cd:3,\:tbxi)
> > > > http://asmblr.net/scrot/2013-01-12-163419_1366x768_scrot.png
> > >
> >
> > This should now be fixed in SVN trunk.
> > >
> > >
> >
> > I believe I have now committed all outstanding patches based upon
> > >
> >
> > William's main GSOC patchset, and performed a QEMU CDROM boot test
> > >
> >
> > across all of SPARC32, SPARC64 and PPC to confirm that I haven't
> > >
> >
> > introduced any regressions across any of my test images. Thank you
> > >
> >
> > everyone for your help, patience and testing.
> > >
> > >
> >
> > Interestingly enough, commits 1081 and 1082 appear to have fixed the key
> > >
> >
> > issues with OpenBIOS's PPC MMU code and as a result, the following
> > >
> >
> > kernels will now at least boot from ISO images in QEMU:
> > >
> > >
> >
> > Darwin
> > >
> >
> > FreeBSD
> > >
> >
> > NetBSD
> > >
> >
> > HelenOS
> > >
> >
> > Fedora Linux
> > >
> >
> > SuSE Linux
> > >
> > >
> >
> > I think the remainder of the bugs during kernel initialisation are
> > >
> >
> > related to either missing/erroneous properties in the OpenBIOS device
> > >
> >
> > tree, or bugs in QEMU's device emulation/Mac hardware model. This will
> > >
> >
> > likely require some assistance from the QEMU people to fix.
> >
> > I can confirm the statement from Mark regarding FreeBSD. So far I can
> > boot but I fail due to the fact that some interrupt-parents are not
> > populated. Namely for via-cuda and for ata-1/2/3. Adding these lets the
> > boot continue until I hang somewhere in a cpu_idle routine. Still
> > investigating.
> >
> > One thing the qemu people might clarify, is -M mac99 ok to use? Or what
> > would be the preferred Mac model and which cpu?
> > My trials go with mac99 and -cpu G4 on the 32-bit qemu.
> >
> > I think once I solved the spinning in the cpu_idle routine I'm near
> > completing the boot.
> >
> > Thank you all for this effort!
> >
> > Andreas
>
> Would you have a patch for this interrupt-parent property?

Well, attached the hacky part I have. Hacky especially because of this
'7' vs. '4'. It is not really clear what I did. At least I could boot
until I get into an endless loop with cpu_spin... with FreeBSD ppc.

Andreas

testing Mac OS X boot on latest trunk (revision 1085) + divide by zero patch

[Programmingkid]

On Feb 3, 2013, at 12:50 PM, Andreas Tobler wrote:
> On 01.02.13 21:38, Programmingkid wrote:
> > > On 13.01.13 16:43, Mark Cave-Ayland wrote:
> > > >
> > >
> > > On 12/01/13 16:05, Amadeusz Sławiński wrote:
> > > >
> > > > >
> > >
> > > I used the latest proposed divide by zero patch
> > > > >
> > >
> > > So it seems like it is almost there.
> > > > >
> > >
> > > Seems like it still can't load kernel when run with boot cd:,\:tbxi
> > > > >
> > >
> > > Mac OS X 10.0
> > > > > http://asmblr.net/scrot/2013-01-12-155737_1366x768_scrot.png
> > >
> > > Mac OS X 10.2
> > > > > http://asmblr.net/scrot/2013-01-12-155642_1366x768_scrot.png
> > >
> > > Mac OS X 10.4
> > > > > http://asmblr.net/scrot/2013-01-12-155935_1366x768_scrot.png
> > >
> > > However when I specify partition to use it loads:
> > > > >
> > >
> > > Mac OS X 10.0 (boot cd:9,\:tbxi)
> > > > > http://asmblr.net/scrot/2013-01-12-163133_1366x768_scrot.png
> > >
> > > Mac OS X 10.2 (boot cd:9,\:tbxi)
> > > > > http://asmblr.net/scrot/2013-01-12-163300_1366x768_scrot.png
> > >
> > > Mac OS X 10.4 (boot cd:3,\:tbxi)
> > > > > http://asmblr.net/scrot/2013-01-12-163419_1366x768_scrot.png
> > > >
> > >
> > > This should now be fixed in SVN trunk.
> > > >
> > > >
> > >
> > > I believe I have now committed all outstanding patches based upon
> > > >
> > >
> > > William's main GSOC patchset, and performed a QEMU CDROM boot test
> > > >
> > >
> > > across all of SPARC32, SPARC64 and PPC to confirm that I haven't
> > > >
> > >
> > > introduced any regressions across any of my test images. Thank you
> > > >
> > >
> > > everyone for your help, patience and testing.
> > > >
> > > >
> > >
> > > Interestingly enough, commits 1081 and 1082 appear to have fixed the key
> > > >
> > >
> > > issues with OpenBIOS's PPC MMU code and as a result, the following
> > > >
> > >
> > > kernels will now at least boot from ISO images in QEMU:
> > > >
> > > >
> > >
> > > Darwin
> > > >
> > >
> > > FreeBSD
> > > >
> > >
> > > NetBSD
> > > >
> > >
> > > HelenOS
> > > >
> > >
> > > Fedora Linux
> > > >
> > >
> > > SuSE Linux
> > > >
> > > >
> > >
> > > I think the remainder of the bugs during kernel initialisation are
> > > >
> > >
> > > related to either missing/erroneous properties in the OpenBIOS device
> > > >
> > >
> > > tree, or bugs in QEMU's device emulation/Mac hardware model. This will
> > > >
> > >
> > > likely require some assistance from the QEMU people to fix.
> > >
> > > I can confirm the statement from Mark regarding FreeBSD. So far I can
> > > boot but I fail due to the fact that some interrupt-parents are not
> > > populated. Namely for via-cuda and for ata-1/2/3. Adding these lets the
> > > boot continue until I hang somewhere in a cpu_idle routine. Still
> > > investigating.
> > >
> > > One thing the qemu people might clarify, is -M mac99 ok to use? Or what
> > > would be the preferred Mac model and which cpu?
> > > My trials go with mac99 and -cpu G4 on the 32-bit qemu.
> > >
> > > I think once I solved the spinning in the cpu_idle routine I'm near
> > > completing the boot.
> > >
> > > Thank you all for this effort!
> > >
> > > Andreas
> >
> > Would you have a patch for this interrupt-parent property?
>
> Well, attached the hacky part I have. Hacky especially because of this
> '7' vs. '4'. It is not really clear what I did. At least I could boot
> until I get into an endless loop with cpu_spin... with FreeBSD ppc.
>
> Andreas
>
> <irq_parrent.diff>

I have tried out your patch with Mac OS 10.0 and Mac OS 10.2. I didn't notice any differences in behavior. The patch doesn't appear to have broken anything either. It would probably be a good idea to add the interrupt-parent property to the nodes that have it on real hardware.