ARM: bL_switcher: synchronize the outbound with the inbound

Let's wait for the inbound to come up and snoop some of the outbound
cache. That should be a bit more efficient than going down right away.
Monitoring the CCI event counters could be a better approach eventually.

Signed-off-by: Nicolas Pitre <nico@linaro.org>

diff --git a/arch/arm/common/bL_switcher.c b/arch/arm/common/bL_switcher.c
index e379c918f0831c70229bdafe6fd1f218886ccde6..0de28f101b7893268fb2c183d46fca81bf6fa697 100644 (file)
--- a/arch/arm/common/bL_switcher.c
+++ b/arch/arm/common/bL_switcher.c
@@ -52,9 +52,10 @@ static int read_mpidr(void)
  * bL switcher core code.
  */
 
-static void bL_do_switch(void *_unused)
+static void bL_do_switch(void *_arg)
 {
        unsigned ib_mpidr, ib_cpu, ib_cluster;
+       long volatile handshake, **handshake_ptr = _arg;
 
        pr_debug("%s\n", __func__);
 
@@ -62,6 +63,13 @@ static void bL_do_switch(void *_unused)
        ib_cpu = MPIDR_AFFINITY_LEVEL(ib_mpidr, 0);
        ib_cluster = MPIDR_AFFINITY_LEVEL(ib_mpidr, 1);
 
+       /* Advertise our handshake location */
+       if (handshake_ptr) {
+               handshake = 0;
+               *handshake_ptr = &handshake;
+       } else
+               handshake = -1;
+
        /*
         * Our state has been saved at this point.  Let's release our
         * inbound CPU.
@@ -80,6 +88,14 @@ static void bL_do_switch(void *_unused)
         * we have none.
         */
 
+       /*
+        * Let's wait until our inbound is alive.
+        */
+       while (!handshake) {
+               wfe();
+               smp_mb();
+       }
+
        /* Let's put ourself down. */
        mcpm_cpu_power_down();
 
@@ -127,6 +143,7 @@ static int bL_switch_to(unsigned int new_cluster_id)
        unsigned int ob_mpidr, ob_cpu, ob_cluster, ib_mpidr, ib_cpu, ib_cluster;
        struct tick_device *tdev;
        enum clock_event_mode tdev_mode;
+       long volatile *handshake_ptr;
        int ret;
 
        this_cpu = smp_processor_id();
@@ -200,7 +217,7 @@ static int bL_switch_to(unsigned int new_cluster_id)
        sync_cache_w(&cpu_logical_map(this_cpu));
 
        /* Let's do the actual CPU switch. */
-       ret = cpu_suspend(0, bL_switchpoint);
+       ret = cpu_suspend((unsigned long)&handshake_ptr, bL_switchpoint);
        if (ret > 0)
                panic("%s: cpu_suspend() returned %d\n", __func__, ret);
 
@@ -222,6 +239,9 @@ static int bL_switch_to(unsigned int new_cluster_id)
        local_fiq_enable();
        local_irq_enable();
 
+       *handshake_ptr = 1;
+       dsb_sev();
+
        if (ret)
                pr_err("%s exiting with error %d\n", __func__, ret);
        return ret;