Re: Which cpus are known to work on the mips64 branch?

[Simon Burge <simonb%NetBSD.org@localhost>]

Martin Husemann wrote:

> On Thu, Jun 24, 2010 at 04:11:13PM +1000, Simon Burge wrote:
>
> > Also, what is the instruction at 0x802a108c?  You can get that with
> > "mipse?--netbsd-objdump -d netbsd" or a cross gdb if you have one built.
> 
> Duh, had a 0 in the address too much, so failed matching it when I looked at
> it before sending the original mail:
> 
> 802a1088 <_atomic_cas_32>:
> 802a1088:       c0880000        lwc0    $8,0(a0)
> 802a108c:       15050006        bne     t0,a1,802a10a8 <_atomic_cas_32+0x20>
> 802a1090:       00c0482d        0xc0482d

From the cause reg, the BD bit was set so we want the instruction
in the branch delay slot.  That it's not decoded above means that
it's probably an instruction that isn't for the current compile
mode/ABI.  0x00c0482d decodes to a DADDU.

This turns out to be a bug I fixed on HEAD a while ago, but hasn't
yet been fixed on the netbsd-5 mips64 branch.  Look at rev 1.11 of
sys/arch/mips/mips/lock_stubs.S.

Cheers,
Simon.