Merge "fix(versal): replace FPD_MAINCCI* macros" into integration
diff --git a/lib/cpus/aarch64/cortex_a510.S b/lib/cpus/aarch64/cortex_a510.S
index 886e1f3..e10ebb0 100644
--- a/lib/cpus/aarch64/cortex_a510.S
+++ b/lib/cpus/aarch64/cortex_a510.S
@@ -382,7 +382,11 @@
 	bl	check_errata_2684597
 	cbz	x0, 2f
 
-	tsb	csync
+	/*
+	 * Many assemblers do not yet understand the "tsb csync" mnemonic,
+	 * so use the equivalent hint instruction.
+	 */
+	hint	#18			/* tsb csync */
 2:
 	ret	x17
 endfunc errata_cortex_a510_2684597_wa
diff --git a/plat/mediatek/drivers/apusys/apusys.c b/plat/mediatek/drivers/apusys/apusys.c
index 1d34627..c82b3a7 100644
--- a/plat/mediatek/drivers/apusys/apusys.c
+++ b/plat/mediatek/drivers/apusys/apusys.c
@@ -11,6 +11,35 @@
 #include "apusys.h"
 #include "apusys_power.h"
 #include <lib/mtk_init/mtk_init.h>
+#include <mtk_sip_svc.h>
+
+static u_register_t apusys_kernel_handler(u_register_t x1,
+					  u_register_t x2,
+					  u_register_t x3,
+					  u_register_t x4,
+					  void *handle,
+					  struct smccc_res *smccc_ret)
+{
+	uint32_t request_ops;
+	int32_t ret = -1;
+
+	request_ops = (uint32_t)x1;
+
+	switch (request_ops) {
+	case MTK_APUSYS_KERNEL_OP_APUSYS_PWR_TOP_ON:
+		ret = apusys_kernel_apusys_pwr_top_on();
+		break;
+	case MTK_APUSYS_KERNEL_OP_APUSYS_PWR_TOP_OFF:
+		ret = apusys_kernel_apusys_pwr_top_off();
+		break;
+	default:
+		ERROR(MODULE_TAG "%s unknown request_ops = %x\n", MODULE_TAG, request_ops);
+		break;
+	}
+
+	return ret;
+}
+DECLARE_SMC_HANDLER(MTK_SIP_APUSYS_CONTROL, apusys_kernel_handler);
 
 int apusys_init(void)
 {
diff --git a/plat/mediatek/drivers/apusys/apusys.h b/plat/mediatek/drivers/apusys/apusys.h
index 5fdd2ec..1592cff 100644
--- a/plat/mediatek/drivers/apusys/apusys.h
+++ b/plat/mediatek/drivers/apusys/apusys.h
@@ -9,4 +9,10 @@
 
 #define MODULE_TAG "[APUSYS]"
 
+enum MTK_APUSYS_KERNEL_OP {
+	MTK_APUSYS_KERNEL_OP_APUSYS_PWR_TOP_ON,	/*  0 */
+	MTK_APUSYS_KERNEL_OP_APUSYS_PWR_TOP_OFF,/*  1 */
+	MTK_APUSYS_KERNEL_OP_NUM,
+};
+
 #endif
diff --git a/plat/mediatek/drivers/apusys/mt8188/apusys_power.c b/plat/mediatek/drivers/apusys/mt8188/apusys_power.c
index d7b0d24..ac62f2f 100644
--- a/plat/mediatek/drivers/apusys/mt8188/apusys_power.c
+++ b/plat/mediatek/drivers/apusys/mt8188/apusys_power.c
@@ -10,6 +10,7 @@
 #include <common/debug.h>
 #include <drivers/delay_timer.h>
 #include <lib/mmio.h>
+#include <lib/spinlock.h>
 #include <lib/utils_def.h>
 #include <lib/xlat_tables/xlat_tables_v2.h>
 
@@ -18,6 +19,9 @@
 #include "apusys_power.h"
 #include <mtk_mmap_pool.h>
 
+static spinlock_t apu_lock;
+static bool apusys_top_on;
+
 static int apu_poll(uintptr_t reg, uint32_t mask, uint32_t value, uint32_t timeout_us)
 {
 	uint32_t reg_val, count;
@@ -43,6 +47,135 @@
 	return -1;
 }
 
+static void apu_xpu2apusys_d4_slv_en(enum APU_D4_SLV_CTRL en)
+{
+	switch (en) {
+	case D4_SLV_OFF:
+		mmio_setbits_32(BCRM_FMEM_PDN_BASE + INFRA_FMEM_BUS_u_SI21_CTRL_0,
+				INFRA_FMEM_BUS_u_SI21_CTRL_EN);
+		mmio_setbits_32(BCRM_FMEM_PDN_BASE + INFRA_FMEM_BUS_u_SI22_CTRL_0,
+				INFRA_FMEM_BUS_u_SI22_CTRL_EN);
+		mmio_setbits_32(BCRM_FMEM_PDN_BASE + INFRA_FMEM_BUS_u_SI11_CTRL_0,
+				INFRA_FMEM_BUS_u_SI11_CTRL_EN);
+		mmio_setbits_32(BCRM_FMEM_PDN_BASE + INFRA_FMEM_M6M7_BUS_u_SI24_CTRL_0,
+				INFRA_FMEM_M6M7_BUS_u_SI24_CTRL_EN);
+		break;
+	case D4_SLV_ON:
+		mmio_clrbits_32(BCRM_FMEM_PDN_BASE + INFRA_FMEM_BUS_u_SI21_CTRL_0,
+				INFRA_FMEM_BUS_u_SI21_CTRL_EN);
+		mmio_clrbits_32(BCRM_FMEM_PDN_BASE + INFRA_FMEM_BUS_u_SI22_CTRL_0,
+				INFRA_FMEM_BUS_u_SI22_CTRL_EN);
+		mmio_clrbits_32(BCRM_FMEM_PDN_BASE + INFRA_FMEM_BUS_u_SI11_CTRL_0,
+				INFRA_FMEM_BUS_u_SI11_CTRL_EN);
+		mmio_clrbits_32(BCRM_FMEM_PDN_BASE + INFRA_FMEM_M6M7_BUS_u_SI24_CTRL_0,
+				INFRA_FMEM_M6M7_BUS_u_SI24_CTRL_EN);
+		break;
+	default:
+		ERROR(MODULE_TAG "%s invalid op: %d\n", __func__, en);
+		break;
+	}
+}
+
+static void apu_pwr_flow_remote_sync(uint32_t cfg)
+{
+	mmio_write_32(APU_MBOX0_BASE + PWR_FLOW_SYNC_REG, (cfg & 0x1));
+}
+
+int apusys_kernel_apusys_pwr_top_on(void)
+{
+	int ret;
+
+	spin_lock(&apu_lock);
+
+	if (apusys_top_on == true) {
+		INFO(MODULE_TAG "%s: APUSYS already powered on!\n", __func__);
+		spin_unlock(&apu_lock);
+		return 0;
+	}
+
+	apu_pwr_flow_remote_sync(1);
+
+	mmio_setbits_32(APU_RPC_BASE + APU_RPC_TOP_SEL_1, AFC_ENA);
+
+	mmio_write_32(APU_RPC_BASE + APU_RPC_TOP_CON, REG_WAKEUP_SET);
+
+	ret = apu_poll(APU_RPC_BASE + APU_RPC_INTF_PWR_RDY,
+		       PWR_RDY, PWR_RDY, APU_TOP_ON_POLLING_TIMEOUT_US);
+	if (ret != 0) {
+		ERROR(MODULE_TAG "%s polling RPC RDY timeout, ret %d\n", __func__, ret);
+		spin_unlock(&apu_lock);
+		return ret;
+	}
+
+	ret = apu_poll(APU_RPC_BASE + APU_RPC_STATUS,
+		       RPC_STATUS_RDY, RPC_STATUS_RDY, APU_TOP_ON_POLLING_TIMEOUT_US);
+	if (ret != 0) {
+		ERROR(MODULE_TAG "%s polling ARE FSM timeout, ret %d\n", __func__, ret);
+		spin_unlock(&apu_lock);
+		return ret;
+	}
+
+	mmio_write_32(APU_VCORE_BASE + APUSYS_VCORE_CG_CLR, CG_CLR);
+	mmio_write_32(APU_RCX_BASE + APU_RCX_CG_CLR, CG_CLR);
+
+	apu_xpu2apusys_d4_slv_en(D4_SLV_OFF);
+
+	apusys_top_on = true;
+
+	spin_unlock(&apu_lock);
+	return ret;
+}
+
+static void apu_sleep_rpc_rcx(void)
+{
+	mmio_write_32(APU_RPC_BASE + APU_RPC_TOP_CON, REG_WAKEUP_CLR);
+	udelay(10);
+
+	mmio_setbits_32(APU_RPC_BASE + APU_RPC_TOP_SEL, (RPC_CTRL | RSV10));
+	udelay(10);
+
+	mmio_setbits_32(APU_RPC_BASE + APU_RPC_TOP_CON, CLR_IRQ);
+	udelay(10);
+
+	mmio_setbits_32(APU_RPC_BASE + APU_RPC_TOP_CON, SLEEP_REQ);
+	udelay(100);
+}
+
+int apusys_kernel_apusys_pwr_top_off(void)
+{
+	int ret;
+
+	spin_lock(&apu_lock);
+
+	if (apusys_top_on == false) {
+		INFO(MODULE_TAG "%s: APUSYS already powered off!\n", __func__);
+		spin_unlock(&apu_lock);
+		return 0;
+	}
+
+	apu_xpu2apusys_d4_slv_en(D4_SLV_ON);
+
+	if (mmio_read_32(APU_MBOX0_BASE + PWR_FLOW_SYNC_REG) == 0) {
+		apu_pwr_flow_remote_sync(1);
+	} else {
+		apu_sleep_rpc_rcx();
+	}
+
+	ret = apu_poll(APU_RPC_BASE + APU_RPC_INTF_PWR_RDY,
+		       PWR_RDY, PWR_OFF, APU_TOP_OFF_POLLING_TIMEOUT_US);
+	if (ret != 0) {
+		ERROR(MODULE_TAG "%s timeout to wait RPC sleep (val:%d), ret %d\n",
+		      __func__, APU_TOP_OFF_POLLING_TIMEOUT_US, ret);
+		spin_unlock(&apu_lock);
+		return ret;
+	}
+
+	apusys_top_on = false;
+
+	spin_unlock(&apu_lock);
+	return ret;
+}
+
 static void get_pll_pcw(const uint32_t clk_rate, uint32_t *r1, uint32_t *r2)
 {
 	unsigned int fvco = clk_rate;
diff --git a/plat/mediatek/drivers/apusys/mt8188/apusys_power.h b/plat/mediatek/drivers/apusys/mt8188/apusys_power.h
index 1f68bd2..b4968d6 100644
--- a/plat/mediatek/drivers/apusys/mt8188/apusys_power.h
+++ b/plat/mediatek/drivers/apusys/mt8188/apusys_power.h
@@ -24,6 +24,11 @@
 	APU_ARE_NUM,
 };
 
+enum APU_D4_SLV_CTRL {
+	D4_SLV_OFF = 0,
+	D4_SLV_ON,
+};
+
 #define APU_POLL_STEP_US			(5)
 
 #define OUT_CLK_FREQ_MIN			(1500)
@@ -40,20 +45,26 @@
 #define APU_ARE_POLLING_TIMEOUT_US		(10000)
 
 /* APU related reg */
+#define APU_VCORE_BASE				(APU_RCX_VCORE_CONFIG)
+#define APU_RCX_BASE				(APU_RCX_CONFIG)
 #define APU_RPC_BASE				(APU_RPCTOP)
 #define APU_PCU_BASE				(APU_PCUTOP)
 #define APU_ARE0_BASE				(APU_ARETOP_ARE0)
 #define APU_ARE1_BASE				(APU_ARETOP_ARE1)
 #define APU_ARE2_BASE				(APU_ARETOP_ARE2)
+#define APU_MBOX0_BASE				(APU_MBOX0)
 #define APU_AO_CTL_BASE				(APU_AO_CTRL)
 #define APU_PLL_BASE				(APU_PLL)
 #define APU_ACC_BASE				(APU_ACC)
 #define APU_ACX0_RPC_LITE_BASE			(APU_ACX0_RPC_LITE)
 
 /* RPC offset define */
+#define APU_RPC_TOP_CON				(0x0000)
 #define APU_RPC_TOP_SEL				(0x0004)
+#define APU_RPC_STATUS				(0x0014)
 #define APU_RPC_TOP_SEL_1			(0x0018)
 #define APU_RPC_HW_CON				(0x001c)
+#define APU_RPC_INTF_PWR_RDY			(0x0044)
 #define APU_RPC_SW_TYPE0			(0x0200)
 
 /* RPC control */
@@ -68,6 +79,15 @@
 #define RPC_CTRL				(0x0000009e)
 #define RPC_TOP_CTRL				(0x0800501e)
 #define RPC_TOP_CTRL1				BIT(20)
+#define AFC_ENA					BIT(16)
+#define REG_WAKEUP_SET				BIT(8)
+#define REG_WAKEUP_CLR				BIT(12)
+#define PWR_RDY					BIT(0)
+#define PWR_OFF					(0)
+#define RPC_STATUS_RDY				BIT(29)
+#define RSV10					BIT(10)
+#define CLR_IRQ					(0x6)
+#define SLEEP_REQ				BIT(0)
 
 /* PLL offset define */
 #define PLL4H_PLL1_CON1				(0x000c)
@@ -137,6 +157,12 @@
 #define ARE_ENTRY1_SRAM_L_INIT			(0x76543210)
 #define ARE_CONFG_INI				BIT(2)
 
+/* VCORE offset define */
+#define APUSYS_VCORE_CG_CLR			(0x0008)
+
+/* RCX offset define */
+#define APU_RCX_CG_CLR				(0x0008)
+
 /* SPM offset define */
 #define APUSYS_BUCK_ISOLATION			(0x03ec)
 
@@ -149,6 +175,18 @@
 /* apu_rcx_ao_ctrl control */
 #define VCORE_ARE_REQ				BIT(2)
 
+/* xpu2apusys */
+#define INFRA_FMEM_BUS_u_SI21_CTRL_0		(0x002c)
+#define INFRA_FMEM_BUS_u_SI22_CTRL_0		(0x0044)
+#define INFRA_FMEM_BUS_u_SI11_CTRL_0		(0x0048)
+#define INFRA_FMEM_M6M7_BUS_u_SI24_CTRL_0	(0x01d0)
+
+/* xpu2apusys */
+#define INFRA_FMEM_BUS_u_SI21_CTRL_EN		BIT(12)
+#define INFRA_FMEM_BUS_u_SI22_CTRL_EN		BIT(13)
+#define INFRA_FMEM_BUS_u_SI11_CTRL_EN		BIT(11)
+#define INFRA_FMEM_M6M7_BUS_u_SI24_CTRL_EN	BIT(15)
+
 /* PCU offset define */
 #define APU_PCU_CTRL_SET			(0x0000)
 #define APU_PCU_BUCK_STEP_SEL			(0x0030)
@@ -188,6 +226,13 @@
 #define APU_RPC_SW_TYPE8			(0x0220)
 #define APU_RPC_SW_TYPE9			(0x0224)
 
+/* power flow sync */
+#define PWR_FLOW_SYNC_REG			(0x0440)
+
+#define CG_CLR					(0xffffffff)
+
 int apusys_power_init(void);
+int apusys_kernel_apusys_pwr_top_on(void);
+int apusys_kernel_apusys_pwr_top_off(void);
 
 #endif /* APUSYS_POWER_H */
diff --git a/plat/mediatek/mt8188/include/platform_def.h b/plat/mediatek/mt8188/include/platform_def.h
index 576dc3d..fc9725e 100644
--- a/plat/mediatek/mt8188/include/platform_def.h
+++ b/plat/mediatek/mt8188/include/platform_def.h
@@ -28,6 +28,9 @@
  * APUSYS related constants
  ******************************************************************************/
 #define BCRM_FMEM_PDN_BASE	(IO_PHYS + 0x00276000)
+#define APU_RCX_CONFIG		(IO_PHYS + 0x09020000)
+#define APU_RCX_VCORE_CONFIG	(IO_PHYS + 0x090e0000)
+#define APU_MBOX0		(IO_PHYS + 0x090e1000)
 #define APU_RPCTOP		(IO_PHYS + 0x090f0000)
 #define APU_PCUTOP		(IO_PHYS + 0x090f1000)
 #define APU_AO_CTRL		(IO_PHYS + 0x090f2000)
diff --git a/plat/qemu/qemu/include/platform_def.h b/plat/qemu/qemu/include/platform_def.h
index 803f8e2..98b8254 100644
--- a/plat/qemu/qemu/include/platform_def.h
+++ b/plat/qemu/qemu/include/platform_def.h
@@ -23,15 +23,14 @@
 #define PLATFORM_CLUSTER0_CORE_COUNT	PLATFORM_MAX_CPUS_PER_CLUSTER
 #define PLATFORM_CLUSTER1_CORE_COUNT	U(0)
 #else
-#define PLATFORM_MAX_CPUS_PER_CLUSTER	U(4)
 /*
  * Define the number of cores per cluster used in calculating core position.
  * The cluster number is shifted by this value and added to the core ID,
  * so its value represents log2(cores/cluster).
- * Default is 2**(2) = 4 cores per cluster.
+ * Default is 2**(4) = 16 cores per cluster.
  */
-#define PLATFORM_CPU_PER_CLUSTER_SHIFT	U(2)
-
+#define PLATFORM_CPU_PER_CLUSTER_SHIFT	U(4)
+#define PLATFORM_MAX_CPUS_PER_CLUSTER	(U(1) << PLATFORM_CPU_PER_CLUSTER_SHIFT)
 #define PLATFORM_CLUSTER_COUNT		U(2)
 #define PLATFORM_CLUSTER0_CORE_COUNT	PLATFORM_MAX_CPUS_PER_CLUSTER
 #define PLATFORM_CLUSTER1_CORE_COUNT	PLATFORM_MAX_CPUS_PER_CLUSTER