ARM: vexpress/dcscb: handle platform coherency exit/setup and CCI

Add the required code to properly handle race free platform coherency exit
to the DCSCB power down method.

The power_up_setup callback is used to enable the CCI interface for
the cluster being brought up.  This must be done in assembly before
the kernel environment is entered.

Thanks to Achin Gupta and Nicolas Pitre for their help and
contributions.

Signed-off-by: Dave Martin <dave.martin@linaro.org>
Signed-off-by: Nicolas Pitre <nico@linaro.org>
Reviewed-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
Acked-by: Pawel Moll <pawel.moll@arm.com>

4 files changed, 98 insertions, 20 deletions

diff --git a/arch/arm/mach-vexpress/Kconfig b/arch/arm/mach-vexpress/Kconfig
index 2f46385c2819..b8bbabec6310 100644
--- a/arch/arm/mach-vexpress/Kconfig
+++ b/arch/arm/mach-vexpress/Kconfig
@@ -60,6 +60,7 @@ config ARCH_VEXPRESS_CA9X4
 config ARCH_VEXPRESS_DCSCB
         bool "Dual Cluster System Control Block (DCSCB) support"
         depends on MCPM
+        select ARM_CCI
         help
           Support for the Dual Cluster System Configuration Block (DCSCB).
           This is needed to provide CPU and cluster power management
diff --git a/arch/arm/mach-vexpress/Makefile b/arch/arm/mach-vexpress/Makefile
index 518519f57a5e..48ba89a8149f 100644
--- a/arch/arm/mach-vexpress/Makefile
+++ b/arch/arm/mach-vexpress/Makefile
@@ -6,6 +6,6 @@ ccflags-$(CONFIG_ARCH_MULTIPLATFORM) := -I$(srctree)/$(src)/include \
 
 obj-y                                   := v2m.o
 obj-$(CONFIG_ARCH_VEXPRESS_CA9X4)       += ct-ca9x4.o
-obj-$(CONFIG_ARCH_VEXPRESS_DCSCB)       += dcscb.o
+obj-$(CONFIG_ARCH_VEXPRESS_DCSCB)       += dcscb.o      dcscb_setup.o
 obj-$(CONFIG_SMP)                       += platsmp.o
 obj-$(CONFIG_HOTPLUG_CPU)               += hotplug.o
diff --git a/arch/arm/mach-vexpress/dcscb.c b/arch/arm/mach-vexpress/dcscb.c
index 44aa7b040e82..16d57a8a9d5a 100644
--- a/arch/arm/mach-vexpress/dcscb.c
+++ b/arch/arm/mach-vexpress/dcscb.c
@@ -16,6 +16,7 @@
 #include <linux/errno.h>
 #include <linux/of_address.h>
 #include <linux/vexpress.h>
+#include <linux/arm-cci.h>
 
 #include <asm/mcpm.h>
 #include <asm/proc-fns.h>
@@ -105,7 +106,10 @@ static void dcscb_power_down(void)
         pr_debug("%s: cpu %u cluster %u\n", __func__, cpu, cluster);
         BUG_ON(cpu >= 4 || cluster >= 2);
 
+        __mcpm_cpu_going_down(cpu, cluster);
+
         arch_spin_lock(&dcscb_lock);
+        BUG_ON(__mcpm_cluster_state(cluster) != CLUSTER_UP);
         dcscb_use_count[cpu][cluster]--;
         if (dcscb_use_count[cpu][cluster] == 0) {
                 rst_hold = readl_relaxed(dcscb_base + RST_HOLD0 + cluster * 4);
@@ -125,31 +129,59 @@ static void dcscb_power_down(void)
                 skip_wfi = true;
         } else
                 BUG();
-        arch_spin_unlock(&dcscb_lock);
-
-        /*
-         * Now let's clean our L1 cache and shut ourself down.
-         * If we're the last CPU in this cluster then clean L2 too.
-         */
 
-        /*
-         * A15/A7 can hit in the cache with SCTLR.C=0, so we don't need
-         * a preliminary flush here for those CPUs.  At least, that's
-         * the theory -- without the extra flush, Linux explodes on
-         * RTSM (to be investigated)..
-         */
-        flush_cache_louis();
-        set_cr(get_cr() & ~CR_C);
+        if (last_man && __mcpm_outbound_enter_critical(cpu, cluster)) {
+                arch_spin_unlock(&dcscb_lock);
 
-        if (!last_man) {
-                flush_cache_louis();
-        } else {
+                /*
+                 * Flush all cache levels for this cluster.
+                 *
+                 * A15/A7 can hit in the cache with SCTLR.C=0, so we don't need
+                 * a preliminary flush here for those CPUs.  At least, that's
+                 * the theory -- without the extra flush, Linux explodes on
+                 * RTSM (to be investigated).
+                 */
+                flush_cache_all();
+                set_cr(get_cr() & ~CR_C);
                 flush_cache_all();
+
+                /*
+                 * This is a harmless no-op.  On platforms with a real
+                 * outer cache this might either be needed or not,
+                 * depending on where the outer cache sits.
+                 */
                 outer_flush_all();
+
+                /* Disable local coherency by clearing the ACTLR "SMP" bit: */
+                set_auxcr(get_auxcr() & ~(1 << 6));
+
+                /*
+                 * Disable cluster-level coherency by masking
+                 * incoming snoops and DVM messages:
+                 */
+                cci_disable_port_by_cpu(mpidr);
+
+                __mcpm_outbound_leave_critical(cluster, CLUSTER_DOWN);
+        } else {
+                arch_spin_unlock(&dcscb_lock);
+
+                /*
+                 * Flush the local CPU cache.
+                 *
+                 * A15/A7 can hit in the cache with SCTLR.C=0, so we don't need
+                 * a preliminary flush here for those CPUs.  At least, that's
+                 * the theory -- without the extra flush, Linux explodes on
+                 * RTSM (to be investigated).
+                 */
+                flush_cache_louis();
+                set_cr(get_cr() & ~CR_C);
+                flush_cache_louis();
+
+                /* Disable local coherency by clearing the ACTLR "SMP" bit: */
+                set_auxcr(get_auxcr() & ~(1 << 6));
         }
 
-        /* Disable local coherency by clearing the ACTLR "SMP" bit: */
-        set_auxcr(get_auxcr() & ~(1 << 6));
+        __mcpm_cpu_down(cpu, cluster);
 
         /* Now we are prepared for power-down, do it: */
         dsb();
@@ -177,12 +209,17 @@ static void __init dcscb_usage_count_init(void)
         dcscb_use_count[cpu][cluster] = 1;
 }
 
+extern void dcscb_power_up_setup(unsigned int affinity_level);
+
 static int __init dcscb_init(void)
 {
         struct device_node *node;
         unsigned int cfg;
         int ret;
 
+        if (!cci_probed())
+                return -ENODEV;
+
         node = of_find_compatible_node(NULL, NULL, "arm,rtsm,dcscb");
         if (!node)
                 return -ENODEV;
@@ -195,6 +232,8 @@ static int __init dcscb_init(void)
         dcscb_usage_count_init();
 
         ret = mcpm_platform_register(&dcscb_power_ops);
+        if (!ret)
+                ret = mcpm_sync_init(dcscb_power_up_setup);
         if (ret) {
                 iounmap(dcscb_base);
                 return ret;
diff --git a/arch/arm/mach-vexpress/dcscb_setup.S b/arch/arm/mach-vexpress/dcscb_setup.S
new file mode 100644
index 000000000000..4bb7fbe0f621
--- /dev/null
+++ b/arch/arm/mach-vexpress/dcscb_setup.S
@@ -0,0 +1,38 @@
+/*
+ * arch/arm/include/asm/dcscb_setup.S
+ *
+ * Created by:  Dave Martin, 2012-06-22
+ * Copyright:   (C) 2012-2013  Linaro Limited
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ */
+
+#include <linux/linkage.h>
+
+
+ENTRY(dcscb_power_up_setup)
+
+        cmp     r0, #0                  @ check affinity level
+        beq     2f
+
+/*
+ * Enable cluster-level coherency, in preparation for turning on the MMU.
+ * The ACTLR SMP bit does not need to be set here, because cpu_resume()
+ * already restores that.
+ *
+ * A15/A7 may not require explicit L2 invalidation on reset, dependent
+ * on hardware integration decisions.
+ * For now, this code assumes that L2 is either already invalidated,
+ * or invalidation is not required.
+ */
+
+        b       cci_enable_port_for_self
+
+2:      @ Implementation-specific local CPU setup operations should go here,
+        @ if any.  In this case, there is nothing to do.
+
+        bx      lr
+
+ENDPROC(dcscb_power_up_setup)