Hello,
I'm working on a new board port for a P2020-based board, and I'm
having problems with my second core not starting up on 2.6.34, even
though it starts up fine on 2.6.32.
In adding various debugs to mpc85xx_kick_cpu(), it looks like the
virtual address is detected as 0x7ffff280 on both, and it seems to
ioremap it to the same address. Furthermore, both of them get the
"ack" from the other CPU almost immediately (within 1ms).
Unfortunately, after that the second core fails to rendezvous
elsewhere in the SMP code and as a result I get "Processor 1 is
stuck".
Unfortunately I've had no luck looking through "git log
v2.6.32..v2.6.34 -- arch/powerpc/" by hand, and there are enough
mechanical merge conflicts with my patchset to make "bisection" very
painful (IE: git bisect, git cherry-pick, test, git reset --hard
HEAD^, repeat)
Any additional suggestions on where to look would be much appreciated.
Cheers,
Kyle Moffett
Oops, put the old linuxppc list on the CC, sorry!
On Thu, Jun 24, 2010 at 23:32, Kyle Moffett <[email protected]> wrote:
> Hello,
>
> I'm working on a new board port for a P2020-based board, and I'm
> having problems with my second core not starting up on 2.6.34, even
> though it starts up fine on 2.6.32.
>
> In adding various debugs to mpc85xx_kick_cpu(), it looks like the
> virtual address is detected as 0x7ffff280 on both, and it seems to
> ioremap it to the same address. Furthermore, both of them get the
> "ack" from the other CPU almost immediately (within 1ms).
> Unfortunately, after that the second core fails to rendezvous
> elsewhere in the SMP code and as a result I get "Processor 1 is
> stuck".
>
> Unfortunately I've had no luck looking through "git log
> v2.6.32..v2.6.34 -- arch/powerpc/" by hand, and there are enough
> mechanical merge conflicts with my patchset to make "bisection" very
> painful (IE: git bisect, git cherry-pick, test, git reset --hard
> HEAD^, repeat)
>
> Any additional suggestions on where to look would be much appreciated.
>
> Cheers,
> Kyle Moffett
>