re: Starting NetBSD/mips64 kernel up and Userland Questions

[matthew green <mrg%eterna.com.au@localhost>]

"Naruaki.Etomi" writes:
> Hello,
> Thank you everyone.
> 
> ---------------------------------------------------------------------------------------
> Dear Thorpe,
> 
> >Our 64-bit MIPS platforms use n32 binaries because, as I recall,
> >it was determined to perform better.
> I understood.
> 
> I built mips64 userland and tried it!
> 
> [  11.4114064] WARNING: clock lost 8764 days
> [  11.4595812] WARNING: using filesystem time
> [  11.5088293] WARNING: CHECK AND RESET THE DATE!
> [  11.6822271] panic: init died (signal 0, exit 1)
> [  11.7426656] cpu0: Begin traceback...
> [  11.7876988] pid -2136376384 not found
> [  11.8316821] cpu0: End traceback...
> [  11.8725321] kernel: breakpoint trap
> Stopped in pid 1.1 (init) at    ffffffff80073b2c:       jr      ra
>                 bdslot: nop
> db> tr
> 0xffffffff80a97be0: 0+ffffffff80073b2c (0,ff00,0,ffffffff807c9380) ra 0 sz 0
> User-level: pid 1.1
> 
> Well, things are not going well.
> Do you have any advice on where to investigate?
[ ... ]
> >We can't run dynamically linked n64 binaries, I doubt anyone fixed PR
> >52857.
> 
> Thank you for your advice!

these are both the same problem.  you can reproduce it by
running any non trivial N64 dynamic binary.  it is very
likely a bug in the dynamic linker.  see
src/libexec/ld.elf_so/arch/mips for the code.  look for
the ELFSIZE == 64 bits, and any missing parts..o

HTH.  it would be great if this was fixed.

also, do you plan to commit the fix for cache_r10k.c?
i have an R10K O2 and it hangs a lot, and maybe this will
help it?


.mrg.