Re: [PATCH v2 0/8] target/ppc: powerpc_excp improvements [74xx] (5/n)

[Mark Cave-Ayland]

On 27/01/2022 23:08, Fabiano Rosas wrote:

> Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk> writes:
>
> > On 27/01/2022 20:11, Fabiano Rosas wrote:
> >
> > > Changes from v1:
> > >
> > > - Restored the 'sc 1' support to avoid breaking the pegasos2 machine.
> > >
> > > I tested this version in the G4 with the following OSes:
> > >
> > > - Linux 5.15 (5.16 seems to be broken, even with master)
> > > - FreeBSD 13
> > > - Mac OS 9.2
> > > - Mac OS Darwin 6.0
> > > - Mac OS X 10.4
> > > - MorphOS 3.15 (-M pegasos2 and -M mac99,via=pmu)
> > >
> > > Based on legoater/ppc-7.0
> > >
> > > v1:
> > > 20220126164200.1048677-1-farosas@linux.ibm.com">https://lore.kernel.org/r/20220126164200.1048677-1-farosas@linux.ibm.com
> > >
> > > Fabiano Rosas (8):
> > >     target/ppc: Introduce powerpc_excp_74xx
> > >     target/ppc: Simplify powerpc_excp_74xx
> > >     target/ppc: 74xx: Machine Check exception cleanup
> > >     target/ppc: 74xx: External interrupt cleanup
> > >     target/ppc: 74xx: Program exception cleanup
> > >     target/ppc: 74xx: System Call exception cleanup
> > >     target/ppc: 74xx: System Reset interrupt cleanup
> > >     target/ppc: 74xx: Set SRRs directly in exception code
> > >
> > >    target/ppc/excp_helper.c | 197 +++++++++++++++++++++++++++++++++++++++
> > >    1 file changed, 197 insertions(+)
> >
> > That certainly covers a good range MacOS images so I'm happy to give:
> >
> > Acked-by: Mark Cave-Ayland <mark.cave-ayland@ilande.co.uk>
> >
> > Just out of curiosity have you tried booting MacOS 9.2 under KVM-PR? Last time I
> > tried on my G4 Mac Mini here, OS X worked fine but 9.2 failed early after 
> > jumping
> > into the 68k emulator. Unfortunately due to the split real mode (see
> > https://github.com/torvalds/linux/commit/c01e3f66cd5cdc1f727f4c7b0c10b3e3bdb91ba7
> > where MSR_IR != MSR_DR) I can't step through with QEMU's gdbstub to see what is 
> > going
> > wrong...
>
> Never crossed my mind actually. =)
>
> I don't immediately see why the step wouldn't work due to split
> mode. Let me understand that patch first. KVM-PR is full of surprises.

IIRC from a debugging point of view the problem was that the QEMU *_debug() functions 
can't translate correctly in split mode, so if gdbstub is enabled with KVM all you 
end up seeing in gdb is junk :(


ATB,

Mark.