24 Jan
2013
24 Jan
'13
2:48 p.m.
Mats Kärrman Mats.Karrman@tritech.se wrote on 2013/01/24 14:31:02:
Joakim Tjernlund/Transmode wrote on 2013/01/24 09:21:
Looking at the watchdog impl. I see it can be normal C code. This
makes
wait_ticks unsafe (even before my patch) as wait_ticks relies on r6 and r7 (and with my patch r0 too) to be unmodified.
Yes! I can see in the assembly from my watchdog_reset()
(arch/powerpc/cpu/mpc5125/
cpu.c) that it does not touch r7 and r8 but indeed r0 -- for storing lr
in
fact so the endless loop is explained!
Many thanks! I'll prepare a patch as soon as I've got one tested.
Great, just one thing though. I think the 4 resp.8 stack offsets should be 8 resp. 12 instead.
Jocke