Re: Can't boot new kernel

[Rin Okuyama <rokuyama.rk%gmail.com@localhost>]

On 2020/10/16 4:01, Martin Husemann wrote:
> On Thu, Oct 15, 2020 at 07:48:47PM +0200, Martin Husemann wrote:
> > I can not boot a new kernel:
> [..]
> > Dies like this:
> >
> > loading XCOFF
> >
> > tsize=f160 dsize=260 bsize=2750 entry=e00000
> > SECTIONS:
> > .text    00e00000 00e00000 0000f160 00000130
> > .data    00e10000 00e10000 00000260 0000f290
> > .bss     00e10260 00e10260 00002750 00000000
> > .gnu.att 00000000 00000000 00000014 0000f4f0
> > .ident   00000000 00000000 00000076 0000f504
> > loading .text, done..
> > loading .data, done..
> > clearing .bss, done..
> >
> > > > NetBSD/macppc OpenFirmware Boot, Revision 1.12
> > 12932396+167480 [467712+464269/
> > DEFAULT CATCH!, code=900 at   %SRR0: 00000000   %SRR1: 00083030
> >   ok
>
> Seems to be gcc 9 fallout. Rin, anything special needed? You probably tested
> GENERIC only, this kernel has options DIAGNOSTIC.

Oh, yes, I'd forgotten it. GENERIC boots, but GENERIC+DIAGNOSTIC
does not.

This seems due to kernel size growth in GCC9; GCC9 adds more unwind
tables than GCC8. You can remove the most (all?) of them by adding
-fno-asynchronous-unwind-tables and -fno-unwind-tables options.
Then, DIAGNOSTIC kernel boots again.

Alternatively, by removing unnecessary features/drivers from config
file, DIAGNOSTIC kernel also boots. Therefore, I consider this is
just a problem of kernel size.

GENERIC+DIAGNOSTIC kernel exceeds some ``soft limit''. I haven't
figured out what is this limit exactly, but it is smaller than
``hard limit'' reported in PR port-macppc/54916.

Thanks,
rin