WARNIG: This patch does't work! I'm sending the RFC mainly because I need help understanding why this is the case. I've tried reversing the loop, that is let CPU1 wait for CPU0 to write a message, and that works: the kernel boots both CPUs in Hyp mode in that case. A theroy could be that it is related to CPU0 being in secure mode and CPU1 being in non-secure mode. I have also tried using a standard DRAM address for the signal writes, but with no luck. Note, that the previous commit works without this follow-on, so you can use that for testing and fun stuff for now. Signed-off-by: Jeremy C. Andrus <jeremya@xxxxxxxxxxxxxxx> Signed-off-by: Christoffer Dall <cdall@xxxxxxxxxxxxxxx> --- board/samsung/smdk5250/smp.S | 11 +++++++++++ 1 file changed, 11 insertions(+) diff --git a/board/samsung/smdk5250/smp.S b/board/samsung/smdk5250/smp.S index 998d2ab..c17f3a2 100644 --- a/board/samsung/smdk5250/smp.S +++ b/board/samsung/smdk5250/smp.S @@ -24,6 +24,9 @@ #include <config.h> +__smp_notify: + .word 0 + __smp_pen: wfe ldr r1, [r0] @@ -43,6 +46,9 @@ enter_smp_pen: ldr r0, =CONFIG_SYSFLAGS_ADDR str r2, [r0] + mov r3, #1 + str r3, __smp_notify + /* * Make instruction copy coherent */ @@ -66,4 +72,9 @@ smp_kick_secondary: str r1, [r0] dsb +1: ldr r1, __smp_notify + cmp r1, #0 + dsb + beq 1b + mov pc, lr -- 1.7.9.5 _______________________________________________ kvmarm mailing list kvmarm@xxxxxxxxxxxxxxxxxxxxx https://lists.cs.columbia.edu/cucslists/listinfo/kvmarm
