Merge pull request #1764 from vwadekar/tf2.0-tegra-downstream-rebase-1.7.19

Tf2.0 tegra downstream rebase 1.7.19

plat/nvidia/tegra/include/drivers/smmu.h

@@ -586,12 +586,12 @@
 /*******************************************************************************
  * SMMU Global Aux. Control Register
  ******************************************************************************/
-#define SMMU_CBn_ACTLR_CPRE_BIT			(1U << 1)
+#define SMMU_CBn_ACTLR_CPRE_BIT			(1ULL << 1U)
 
 /*******************************************************************************
  * SMMU configuration constants
  ******************************************************************************/
-#define ID1_PAGESIZE				(1U << 31)
+#define ID1_PAGESIZE				(1U << 31U)
 #define ID1_NUMPAGENDXB_SHIFT			28U
 #define ID1_NUMPAGENDXB_MASK			7U
 #define ID1_NUMS2CB_SHIFT			16U
@@ -705,5 +705,6 @@ typedef struct smmu_regs {
 void tegra_smmu_init(void);
 void tegra_smmu_save_context(uint64_t smmu_ctx_addr);
 smmu_regs_t *plat_get_smmu_ctx(void);
+uint32_t plat_get_num_smmu_devices(void);
 
 #endif /* SMMU_H */

plat/nvidia/tegra/include/lib/profiler.h

+/*
+ * Copyright (c) 2017, ARM Limited and Contributors. All rights reserved.
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ */
+
+#ifndef __PROFILER_H__
+#define __PROFILER_H__
+
+/*******************************************************************************
+ * Number of bytes of memory used by the profiler on Tegra
+ ******************************************************************************/
+#define PROFILER_SIZE_BYTES	U(0x1000)
+
+void boot_profiler_init(uint64_t shmem_base, uint32_t tmr_base);
+void boot_profiler_add_record(const char *str);
+void boot_profiler_deinit(void);
+
+#endif /* __PROFILER_H__ */

plat/nvidia/tegra/include/t132/tegra_def.h

@@ -83,6 +83,9 @@
  ******************************************************************************/
 #define TEGRA_MC_BASE			U(0x70019000)
 
+/* Memory Controller Interrupt Status */
+#define MC_INTSTATUS			0x00U
+
 /* TZDRAM carveout configuration registers */
 #define MC_SECURITY_CFG0_0		U(0x70)
 #define MC_SECURITY_CFG1_0		U(0x74)

plat/nvidia/tegra/include/t186/tegra_def.h

@@ -112,8 +112,13 @@
 #define TSA_CONFIG_STATIC0_CSW_XUSB_HOSTW		U(0x15018)
 #define  TSA_CONFIG_STATIC0_CSW_XUSB_HOSTW_RESET	U(0x1100)
 
-#define TSA_CONFIG_CSW_MEMTYPE_OVERRIDE_MASK		(U(0x3) << 11)
-#define TSA_CONFIG_CSW_MEMTYPE_OVERRIDE_PASTHRU		(U(0) << 11)
+#define TSA_CONFIG_CSW_MEMTYPE_OVERRIDE_MASK		(ULL(0x3) << 11)
+#define TSA_CONFIG_CSW_MEMTYPE_OVERRIDE_PASTHRU		(ULL(0) << 11)
+
+/*******************************************************************************
+ * Tegra General Purpose Centralised DMA constants
+ ******************************************************************************/
+#define TEGRA_GPCDMA_BASE		ULL(0x2610000)
 
 /*******************************************************************************
  * Tegra Memory Controller constants
@@ -124,7 +129,7 @@
 /* General Security Carveout register macros */
 #define MC_GSC_CONFIG_REGS_SIZE		U(0x40)
 #define MC_GSC_LOCK_CFG_SETTINGS_BIT	(U(1) << 1)
-#define MC_GSC_ENABLE_TZ_LOCK_BIT	(U(1) << 0)
+#define MC_GSC_ENABLE_TZ_LOCK_BIT	(ULL(1) << 0)
 #define MC_GSC_SIZE_RANGE_4KB_SHIFT	U(27)
 #define MC_GSC_BASE_LO_SHIFT		U(12)
 #define MC_GSC_BASE_LO_MASK		U(0xFFFFF)
@@ -136,6 +141,10 @@
 #define MC_SECURITY_CFG1_0		U(0x74)
 #define MC_SECURITY_CFG3_0		U(0x9BC)
 
+#define MC_SECURITY_BOM_MASK		(U(0xFFF) << 20)
+#define MC_SECURITY_SIZE_MB_MASK	(U(0x1FFF) << 0)
+#define MC_SECURITY_BOM_HI_MASK		(U(0x3) << 0)
+
 /* Video Memory carveout configuration registers */
 #define MC_VIDEO_PROTECT_BASE_HI	U(0x978)
 #define MC_VIDEO_PROTECT_BASE_LO	U(0x648)
@@ -198,6 +207,8 @@
 #define TEGRA_CAR_RESET_BASE		U(0x05000000)
 #define TEGRA_GPU_RESET_REG_OFFSET	U(0x30)
 #define  GPU_RESET_BIT			(U(1) << 0)
+#define TEGRA_GPCDMA_RST_SET_REG_OFFSET	U(0x6A0004)
+#define TEGRA_GPCDMA_RST_CLR_REG_OFFSET	U(0x6A0008)
 
 /*******************************************************************************
  * Tegra micro-seconds timer constants

plat/nvidia/tegra/include/t210/tegra_def.h

@@ -89,6 +89,16 @@
 #define TEGRA_RST_DEV_CLR_V		U(0x434)
 #define TEGRA_CLK_ENB_V			U(0x440)
 
+/* SE Clock Offsets */
+#define TEGRA_RST_DEVICES_V		0x358UL
+#define  SE_RESET_BIT 			(0x1UL << 31)
+#define TEGRA_RST_DEVICES_W		 0x35CUL
+#define  ENTROPY_CLK_ENB_BIT		(0x1UL << 21)
+#define TEGRA_CLK_OUT_ENB_V		0x360UL
+#define  SE_CLK_ENB_BIT			(0x1UL << 31)
+#define TEGRA_CLK_OUT_ENB_W		0x364UL
+#define  ENTROPY_RESET_BIT 		(0x1UL << 21)
+
 /*******************************************************************************
  * Tegra Flow Controller constants
  ******************************************************************************/
@@ -124,6 +134,16 @@
 #define TEGRA_UARTD_BASE		U(0x70006300)
 #define TEGRA_UARTE_BASE		U(0x70006400)
 
+/*******************************************************************************
+ * Tegra Fuse Controller related constants
+ ******************************************************************************/
+#define TEGRA_FUSE_BASE			0x7000F800UL
+#define FUSE_BOOT_SECURITY_INFO		0x268UL
+#define FUSE_ATOMIC_SAVE_CARVEOUT_EN	(0x1U << 7)
+#define FUSE_JTAG_SECUREID_VALID	(0x104UL)
+#define ECID_VALID			(0x1UL)
+
+
 /*******************************************************************************
  * Tegra Power Mgmt Controller constants
  ******************************************************************************/
@@ -143,6 +163,9 @@
  ******************************************************************************/
 #define TEGRA_MC_BASE			U(0x70019000)
 
+/* Memory Controller Interrupt Status */
+#define MC_INTSTATUS			0x00U
+
 /* TZDRAM carveout configuration registers */
 #define MC_SECURITY_CFG0_0		U(0x70)
 #define MC_SECURITY_CFG1_0		U(0x74)
@@ -153,6 +176,10 @@
 #define MC_VIDEO_PROTECT_BASE_LO	U(0x648)
 #define MC_VIDEO_PROTECT_SIZE_MB	U(0x64c)
 
+/* SMMU configuration registers*/
+#define MC_SMMU_PPCS_ASID_0		0x270U
+#define  PPCS_SMMU_ENABLE		(0x1U << 31)
+
 /*******************************************************************************
  * Tegra SE constants
  ******************************************************************************/
@@ -168,4 +195,10 @@
 #define TEGRA_TZRAM_BASE		U(0x7C010000)
 #define TEGRA_TZRAM_SIZE		U(0x10000)
 
+/*******************************************************************************
+ * Tegra TZRAM carveout constants
+ ******************************************************************************/
+#define TEGRA_TZRAM_CARVEOUT_BASE	U(0x7C04C000)
+#define TEGRA_TZRAM_CARVEOUT_SIZE	U(0x4000)
+
 #endif /* TEGRA_DEF_H */

plat/nvidia/tegra/include/tegra_platform.h

@@ -46,7 +46,6 @@ bool tegra_chipid_is_t186(void);
 bool tegra_chipid_is_t210(void);
 bool tegra_chipid_is_t210_b01(void);
 
-
 /*
  * Tegra platform identifiers
  */

plat/nvidia/tegra/include/tegra_private.h

@@ -22,6 +22,16 @@
 #define TEGRA_DRAM_BASE		ULL(0x80000000)
 #define TEGRA_DRAM_END		ULL(0x27FFFFFFF)
 
+/*******************************************************************************
+ * Implementation defined ACTLR_EL1 bit definitions
+ ******************************************************************************/
+#define ACTLR_EL1_PMSTATE_MASK		(ULL(0xF) << 0)
+
+/*******************************************************************************
+ * Implementation defined ACTLR_EL2 bit definitions
+ ******************************************************************************/
+#define ACTLR_EL2_PMSTATE_MASK		(ULL(0xF) << 0)
+
 /*******************************************************************************
  * Struct for parameters received from BL2
  ******************************************************************************/
@@ -31,9 +41,11 @@ typedef struct plat_params_from_bl2 {
 	/* TZ memory base */
 	uint64_t tzdram_base;
 	/* UART port ID */
-	int uart_id;
+	int32_t uart_id;
 	/* L2 ECC parity protection disable flag */
-	int l2_ecc_parity_prot_dis;
+	int32_t l2_ecc_parity_prot_dis;
+	/* SHMEM base address for storing the boot logs */
+	uint64_t boot_profiler_shmem_base;
 } plat_params_from_bl2_t;
 
 /*******************************************************************************
@@ -82,7 +94,30 @@ extern uint8_t tegra_fake_system_suspend;
 
 void tegra_pm_system_suspend_entry(void);
 void tegra_pm_system_suspend_exit(void);
-int tegra_system_suspended(void);
+int32_t tegra_system_suspended(void);
+int32_t tegra_soc_pwr_domain_suspend(const psci_power_state_t *target_state);
+int32_t tegra_soc_pwr_domain_on(u_register_t mpidr);
+int32_t tegra_soc_pwr_domain_off(const psci_power_state_t *target_state);
+int32_t tegra_soc_pwr_domain_on_finish(const psci_power_state_t *target_state);
+int32_t tegra_soc_pwr_domain_power_down_wfi(const psci_power_state_t *target_state);
+int32_t tegra_soc_prepare_system_reset(void);
+__dead2 void tegra_soc_prepare_system_off(void);
+plat_local_state_t tegra_soc_get_target_pwr_state(uint32_t lvl,
+					     const plat_local_state_t *states,
+					     uint32_t ncpu);
+void tegra_get_sys_suspend_power_state(psci_power_state_t *req_state);
+void tegra_cpu_standby(plat_local_state_t cpu_state);
+int32_t tegra_pwr_domain_on(u_register_t mpidr);
+void tegra_pwr_domain_off(const psci_power_state_t *target_state);
+void tegra_pwr_domain_suspend(const psci_power_state_t *target_state);
+void __dead2 tegra_pwr_domain_power_down_wfi(const psci_power_state_t *target_state);
+void tegra_pwr_domain_on_finish(const psci_power_state_t *target_state);
+void tegra_pwr_domain_suspend_finish(const psci_power_state_t *target_state);
+__dead2 void tegra_system_off(void);
+__dead2 void tegra_system_reset(void);
+int32_t tegra_validate_power_state(uint32_t power_state,
+				   psci_power_state_t *req_state);
+int32_t tegra_validate_ns_entrypoint(uintptr_t entrypoint);
 
 /* Declarations for tegraXXX_pm.c */
 int tegra_prepare_cpu_suspend(unsigned int id, unsigned int afflvl);
@@ -90,7 +125,7 @@ int tegra_prepare_cpu_on_finish(unsigned long mpidr);
 
 /* Declarations for tegra_bl31_setup.c */
 plat_params_from_bl2_t *bl31_get_plat_params(void);
-int bl31_check_ns_address(uint64_t base, uint64_t size_in_bytes);
+int32_t bl31_check_ns_address(uint64_t base, uint64_t size_in_bytes);
 void plat_early_platform_setup(void);
 
 /* Declarations for tegra_delay_timer.c */

plat/nvidia/tegra/platform.mk

@@ -12,6 +12,8 @@ $(eval $(call add_define,CRASH_REPORTING))
 
 # enable assert() for release/debug builds
 ENABLE_ASSERTIONS	:=	1
+PLAT_LOG_LEVEL_ASSERT	:=	40
+$(eval $(call add_define,PLAT_LOG_LEVEL_ASSERT))
 
 # enable dynamic memory mapping
 PLAT_XLAT_TABLES_DYNAMIC :=	1
@@ -29,6 +31,9 @@ USE_COHERENT_MEM	:=	0
 # do not enable SVE
 ENABLE_SVE_FOR_NS	:=	0
 
+# enable D-cache early during CPU warmboot
+WARMBOOT_ENABLE_DCACHE_EARLY := 1
+
 include plat/nvidia/tegra/common/tegra_common.mk
 include ${SOC_DIR}/platform_${TARGET_SOC}.mk
 

plat/nvidia/tegra/soc/t132/plat_psci_handlers.c

@@ -98,19 +98,24 @@ int tegra_soc_pwr_domain_on_finish(const psci_power_state_t *target_state)
 
 int tegra_soc_pwr_domain_off(const psci_power_state_t *target_state)
 {
+	uint64_t val;
+
 	tegra_fc_cpu_off(read_mpidr() & MPIDR_CPU_MASK);
 
 	/* Disable DCO operations */
 	denver_disable_dco();
 
 	/* Power down the CPU */
-	write_actlr_el1(DENVER_CPU_STATE_POWER_DOWN);
+	val = read_actlr_el1() & ~ACTLR_EL1_PMSTATE_MASK;
+	write_actlr_el1(val | DENVER_CPU_STATE_POWER_DOWN);
 
 	return PSCI_E_SUCCESS;
 }
 
 int tegra_soc_pwr_domain_suspend(const psci_power_state_t *target_state)
 {
+	uint64_t val;
+
 #if ENABLE_ASSERTIONS
 	int cpu = read_mpidr() & MPIDR_CPU_MASK;
 
@@ -128,7 +133,8 @@ int tegra_soc_pwr_domain_suspend(const psci_power_state_t *target_state)
 	denver_disable_dco();
 
 	/* Program the suspend state ID */
-	write_actlr_el1(target_state->pwr_domain_state[PLAT_MAX_PWR_LVL]);
+	val = read_actlr_el1() & ~ACTLR_EL1_PMSTATE_MASK;
+	write_actlr_el1(val | target_state->pwr_domain_state[PLAT_MAX_PWR_LVL]);
 
 	return PSCI_E_SUCCESS;
 }

plat/nvidia/tegra/soc/t186/drivers/mce/ari.c

@@ -151,7 +151,7 @@ int32_t ari_enter_cstate(uint32_t ari_base, uint32_t state, uint32_t wake_time)
 
 		/* Enter the cstate, to be woken up after wake_time (TSC ticks) */
 		ret = ari_request_wait(ari_base, ARI_EVT_MASK_STANDBYWFI_BIT,
-		TEGRA_ARI_ENTER_CSTATE, state, wake_time);
+			(uint32_t)TEGRA_ARI_ENTER_CSTATE, state, wake_time);
 	}
 
 	return ret;
@@ -191,7 +191,7 @@ int32_t ari_update_cstate_info(uint32_t ari_base, uint32_t cluster, uint32_t ccp
 	}
 
 	/* set the updated cstate info */
-	return ari_request_wait(ari_base, 0U, TEGRA_ARI_UPDATE_CSTATE_INFO,
+	return ari_request_wait(ari_base, 0U, (uint32_t)TEGRA_ARI_UPDATE_CSTATE_INFO,
 				(uint32_t)val, wake_mask);
 }
 
@@ -208,8 +208,8 @@ int32_t ari_update_crossover_time(uint32_t ari_base, uint32_t type, uint32_t tim
 		ari_clobber_response(ari_base);
 
 		/* update crossover threshold time */
-		ret = ari_request_wait(ari_base, 0U, TEGRA_ARI_UPDATE_CROSSOVER,
-			type, time);
+		ret = ari_request_wait(ari_base, 0U,
+				(uint32_t)TEGRA_ARI_UPDATE_CROSSOVER, type, time);
 	}
 
 	return ret;
@@ -227,7 +227,8 @@ uint64_t ari_read_cstate_stats(uint32_t ari_base, uint32_t state)
 		/* clean the previous response state */
 		ari_clobber_response(ari_base);
 
-		ret = ari_request_wait(ari_base, 0U, TEGRA_ARI_CSTATE_STATS, state, 0U);
+		ret = ari_request_wait(ari_base, 0U,
+				(uint32_t)TEGRA_ARI_CSTATE_STATS, state, 0U);
 		if (ret != 0) {
 			result = EINVAL;
 		} else {
@@ -243,8 +244,8 @@ int32_t ari_write_cstate_stats(uint32_t ari_base, uint32_t state, uint32_t stats
 	ari_clobber_response(ari_base);
 
 	/* write the cstate stats */
-	return ari_request_wait(ari_base, 0U, TEGRA_ARI_WRITE_CSTATE_STATS, state,
-			stats);
+	return ari_request_wait(ari_base, 0U, (uint32_t)TEGRA_ARI_WRITE_CSTATE_STATS,
+			state, stats);
 }
 
 uint64_t ari_enumeration_misc(uint32_t ari_base, uint32_t cmd, uint32_t data)
@@ -261,7 +262,7 @@ uint64_t ari_enumeration_misc(uint32_t ari_base, uint32_t cmd, uint32_t data)
 		local_data = 0U;
 	}
 
-	ret = ari_request_wait(ari_base, 0U, TEGRA_ARI_MISC, cmd, local_data);
+	ret = ari_request_wait(ari_base, 0U, (uint32_t)TEGRA_ARI_MISC, cmd, local_data);
 	if (ret != 0) {
 		resp = (uint64_t)ret;
 	} else {
@@ -281,8 +282,8 @@ int32_t ari_is_ccx_allowed(uint32_t ari_base, uint32_t state, uint32_t wake_time
 	/* clean the previous response state */
 	ari_clobber_response(ari_base);
 
-	ret = ari_request_wait(ari_base, 0U, TEGRA_ARI_IS_CCX_ALLOWED, state & 0x7U,
-			wake_time);
+	ret = ari_request_wait(ari_base, 0U, (uint32_t)TEGRA_ARI_IS_CCX_ALLOWED,
+			state & 0x7U, wake_time);
 	if (ret != 0) {
 		ERROR("%s: failed (%d)\n", __func__, ret);
 		result = 0U;
@@ -307,8 +308,8 @@ int32_t ari_is_sc7_allowed(uint32_t ari_base, uint32_t state, uint32_t wake_time
 		/* clean the previous response state */
 		ari_clobber_response(ari_base);
 
-		ret = ari_request_wait(ari_base, 0U, TEGRA_ARI_IS_SC7_ALLOWED, state,
-				wake_time);
+		ret = ari_request_wait(ari_base, 0U,
+			(uint32_t)TEGRA_ARI_IS_SC7_ALLOWED, state, wake_time);
 		if (ret != 0) {
 			ERROR("%s: failed (%d)\n", __func__, ret);
 			result = 0;
@@ -346,7 +347,8 @@ int32_t ari_online_core(uint32_t ari_base, uint32_t core)
 		} else {
 			/* clean the previous response state */
 			ari_clobber_response(ari_base);
-			ret = ari_request_wait(ari_base, 0U, TEGRA_ARI_ONLINE_CORE, core, 0U);
+			ret = ari_request_wait(ari_base, 0U,
+				(uint32_t)TEGRA_ARI_ONLINE_CORE, core, 0U);
 		}
 	}
 
@@ -374,7 +376,8 @@ int32_t ari_cc3_ctrl(uint32_t ari_base, uint32_t freq, uint32_t volt, uint8_t en
 		((volt & MCE_AUTO_CC3_VTG_MASK) << MCE_AUTO_CC3_VTG_SHIFT) |\
 		((enable != 0U) ? MCE_AUTO_CC3_ENABLE_BIT : 0U));
 
-	return ari_request_wait(ari_base, 0U, TEGRA_ARI_CC3_CTRL, val, 0U);
+	return ari_request_wait(ari_base, 0U,
+			(uint32_t)TEGRA_ARI_CC3_CTRL, val, 0U);
 }
 
 int32_t ari_reset_vector_update(uint32_t ari_base)
@@ -386,7 +389,8 @@ int32_t ari_reset_vector_update(uint32_t ari_base)
 	 * Need to program the CPU reset vector one time during cold boot
 	 * and SC7 exit
 	 */
-	(void)ari_request_wait(ari_base, 0U, TEGRA_ARI_COPY_MISCREG_AA64_RST, 0U, 0U);
+	(void)ari_request_wait(ari_base, 0U,
+			(uint32_t)TEGRA_ARI_COPY_MISCREG_AA64_RST, 0U, 0U);
 
 	return 0;
 }
@@ -396,8 +400,8 @@ int32_t ari_roc_flush_cache_trbits(uint32_t ari_base)
 	/* clean the previous response state */
 	ari_clobber_response(ari_base);
 
-	return ari_request_wait(ari_base, 0U, TEGRA_ARI_ROC_FLUSH_CACHE_TRBITS,
-			0U, 0U);
+	return ari_request_wait(ari_base, 0U,
+			(uint32_t)TEGRA_ARI_ROC_FLUSH_CACHE_TRBITS, 0U, 0U);
 }
 
 int32_t ari_roc_flush_cache(uint32_t ari_base)
@@ -405,8 +409,8 @@ int32_t ari_roc_flush_cache(uint32_t ari_base)
 	/* clean the previous response state */
 	ari_clobber_response(ari_base);
 
-	return ari_request_wait(ari_base, 0U, TEGRA_ARI_ROC_FLUSH_CACHE_ONLY,
-			0U, 0U);
+	return ari_request_wait(ari_base, 0U,
+			(uint32_t)TEGRA_ARI_ROC_FLUSH_CACHE_ONLY, 0U, 0U);
 }
 
 int32_t ari_roc_clean_cache(uint32_t ari_base)
@@ -414,8 +418,8 @@ int32_t ari_roc_clean_cache(uint32_t ari_base)
 	/* clean the previous response state */
 	ari_clobber_response(ari_base);
 
-	return ari_request_wait(ari_base, 0U, TEGRA_ARI_ROC_CLEAN_CACHE_ONLY,
-			0U, 0U);
+	return ari_request_wait(ari_base, 0U,
+			(uint32_t)TEGRA_ARI_ROC_CLEAN_CACHE_ONLY, 0U, 0U);
 }
 
 uint64_t ari_read_write_mca(uint32_t ari_base, uint64_t cmd, uint64_t *data)
@@ -432,7 +436,7 @@ uint64_t ari_read_write_mca(uint32_t ari_base, uint64_t cmd, uint64_t *data)
 	ari_write_32(ari_base, (uint32_t)cmd, ARI_RESPONSE_DATA_LO);
 	ari_write_32(ari_base, (uint32_t)(cmd >> 32U), ARI_RESPONSE_DATA_HI);
 
-	ret = ari_request_wait(ari_base, 0U, TEGRA_ARI_MCA,
+	ret = ari_request_wait(ari_base, 0U, (uint32_t)TEGRA_ARI_MCA,
 			       (uint32_t)mca_arg_data,
 			       (uint32_t)(mca_arg_data >> 32U));
 	if (ret == 0) {
@@ -473,7 +477,8 @@ int32_t ari_update_ccplex_gsc(uint32_t ari_base, uint32_t gsc_idx)
 		 * the ID, from the MC registers and update the internal GSC registers
 		 * of the CCPLEX.
 		 */
-		(void)ari_request_wait(ari_base, 0U, TEGRA_ARI_UPDATE_CCPLEX_GSC, gsc_idx, 0U);
+		(void)ari_request_wait(ari_base, 0U,
+				(uint32_t)TEGRA_ARI_UPDATE_CCPLEX_GSC, gsc_idx, 0U);
 	}
 
 	return ret;
@@ -487,7 +492,8 @@ void ari_enter_ccplex_state(uint32_t ari_base, uint32_t state_idx)
 	/*
 	 * The MCE will shutdown or restart the entire system
 	 */
-	(void)ari_request_wait(ari_base, 0U, TEGRA_ARI_MISC_CCPLEX, state_idx, 0U);
+	(void)ari_request_wait(ari_base, 0U,
+			(uint32_t)TEGRA_ARI_MISC_CCPLEX, state_idx, 0U);
 }
 
 int32_t ari_read_write_uncore_perfmon(uint32_t ari_base, uint64_t req,
@@ -514,8 +520,8 @@ int32_t ari_read_write_uncore_perfmon(uint32_t ari_base, uint64_t req,
 		val = (req_cmd == UNCORE_PERFMON_CMD_WRITE) ?
 			(uint32_t)*data : 0U;
 
-		ret = ari_request_wait(ari_base, 0U, TEGRA_ARI_PERFMON, val,
-				       (uint32_t)req);
+		ret = ari_request_wait(ari_base, 0U,
+			(uint32_t)TEGRA_ARI_PERFMON, val, (uint32_t)req);
 		if (ret != 0) {
 			result = ret;
 		} else {
@@ -552,6 +558,7 @@ void ari_misc_ccplex(uint32_t ari_base, uint32_t index, uint32_t value)
 	} else {
 		/* clean the previous response state */
 		ari_clobber_response(ari_base);
-		(void)ari_request_wait(ari_base, 0U, TEGRA_ARI_MISC_CCPLEX, index, value);
+		(void)ari_request_wait(ari_base, 0U,
+			(uint32_t)TEGRA_ARI_MISC_CCPLEX, index, value);
 	}
 }

plat/nvidia/tegra/soc/t186/drivers/mce/mce.c

@@ -170,12 +170,12 @@ int32_t mce_command_handler(uint64_t cmd, uint64_t arg0, uint64_t arg1,
 	cpu_ari_base = mce_get_curr_cpu_ari_base();
 
 	switch (cmd) {
-	case MCE_CMD_ENTER_CSTATE:
+	case (uint64_t)MCE_CMD_ENTER_CSTATE:
 		ret = ops->enter_cstate(cpu_ari_base, arg0, arg1);
 
 		break;
 
-	case MCE_CMD_UPDATE_CSTATE_INFO:
+	case (uint64_t)MCE_CMD_UPDATE_CSTATE_INFO:
 		/*
 		 * get the parameters required for the update cstate info
 		 * command
@@ -194,12 +194,12 @@ int32_t mce_command_handler(uint64_t cmd, uint64_t arg0, uint64_t arg1,
 
 		break;
 
-	case MCE_CMD_UPDATE_CROSSOVER_TIME:
+	case (uint64_t)MCE_CMD_UPDATE_CROSSOVER_TIME:
 		ret = ops->update_crossover_time(cpu_ari_base, arg0, arg1);
 
 		break;
 
-	case MCE_CMD_READ_CSTATE_STATS:
+	case (uint64_t)MCE_CMD_READ_CSTATE_STATS:
 		ret64 = ops->read_cstate_stats(cpu_ari_base, arg0);
 
 		/* update context to return cstate stats value */
@@ -208,12 +208,12 @@ int32_t mce_command_handler(uint64_t cmd, uint64_t arg0, uint64_t arg1,
 
 		break;
 
-	case MCE_CMD_WRITE_CSTATE_STATS:
+	case (uint64_t)MCE_CMD_WRITE_CSTATE_STATS:
 		ret = ops->write_cstate_stats(cpu_ari_base, arg0, arg1);
 
 		break;
 
-	case MCE_CMD_IS_CCX_ALLOWED:
+	case (uint64_t)MCE_CMD_IS_CCX_ALLOWED:
 		ret = ops->is_ccx_allowed(cpu_ari_base, arg0, arg1);
 
 		/* update context to return CCx status value */
@@ -221,7 +221,7 @@ int32_t mce_command_handler(uint64_t cmd, uint64_t arg0, uint64_t arg1,
 
 		break;
 
-	case MCE_CMD_IS_SC7_ALLOWED:
+	case (uint64_t)MCE_CMD_IS_SC7_ALLOWED:
 		ret = ops->is_sc7_allowed(cpu_ari_base, arg0, arg1);
 
 		/* update context to return SC7 status value */
@@ -230,17 +230,17 @@ int32_t mce_command_handler(uint64_t cmd, uint64_t arg0, uint64_t arg1,
 
 		break;
 
-	case MCE_CMD_ONLINE_CORE:
+	case (uint64_t)MCE_CMD_ONLINE_CORE:
 		ret = ops->online_core(cpu_ari_base, arg0);
 
 		break;
 
-	case MCE_CMD_CC3_CTRL:
+	case (uint64_t)MCE_CMD_CC3_CTRL:
 		ret = ops->cc3_ctrl(cpu_ari_base, arg0, arg1, arg2);
 
 		break;
 
-	case MCE_CMD_ECHO_DATA:
+	case (uint64_t)MCE_CMD_ECHO_DATA:
 		ret64 = ops->call_enum_misc(cpu_ari_base, TEGRA_ARI_MISC_ECHO,
 				arg0);
 
@@ -252,7 +252,7 @@ int32_t mce_command_handler(uint64_t cmd, uint64_t arg0, uint64_t arg1,
 
 		break;
 
-	case MCE_CMD_READ_VERSIONS:
+	case (uint64_t)MCE_CMD_READ_VERSIONS:
 		ret64 = ops->call_enum_misc(cpu_ari_base, TEGRA_ARI_MISC_VERSION,
 			arg0);
 
@@ -265,7 +265,7 @@ int32_t mce_command_handler(uint64_t cmd, uint64_t arg0, uint64_t arg1,
 
 		break;
 
-	case MCE_CMD_ENUM_FEATURES:
+	case (uint64_t)MCE_CMD_ENUM_FEATURES:
 		ret64 = ops->call_enum_misc(cpu_ari_base,
 				TEGRA_ARI_MISC_FEATURE_LEAF_0, arg0);
 
@@ -274,22 +274,22 @@ int32_t mce_command_handler(uint64_t cmd, uint64_t arg0, uint64_t arg1,
 
 		break;
 
-	case MCE_CMD_ROC_FLUSH_CACHE_TRBITS:
+	case (uint64_t)MCE_CMD_ROC_FLUSH_CACHE_TRBITS:
 		ret = ops->roc_flush_cache_trbits(cpu_ari_base);
 
 		break;
 
-	case MCE_CMD_ROC_FLUSH_CACHE:
+	case (uint64_t)MCE_CMD_ROC_FLUSH_CACHE:
 		ret = ops->roc_flush_cache(cpu_ari_base);
 
 		break;
 
-	case MCE_CMD_ROC_CLEAN_CACHE:
+	case (uint64_t)MCE_CMD_ROC_CLEAN_CACHE:
 		ret = ops->roc_clean_cache(cpu_ari_base);
 
 		break;
 
-	case MCE_CMD_ENUM_READ_MCA:
+	case (uint64_t)MCE_CMD_ENUM_READ_MCA:
 		ret64 = ops->read_write_mca(cpu_ari_base, arg0, &arg1);
 
 		/* update context to return MCA data/error */
@@ -299,7 +299,7 @@ int32_t mce_command_handler(uint64_t cmd, uint64_t arg0, uint64_t arg1,
 
 		break;
 
-	case MCE_CMD_ENUM_WRITE_MCA:
+	case (uint64_t)MCE_CMD_ENUM_WRITE_MCA:
 		ret64 = ops->read_write_mca(cpu_ari_base, arg0, &arg1);
 
 		/* update context to return MCA error */
@@ -309,7 +309,7 @@ int32_t mce_command_handler(uint64_t cmd, uint64_t arg0, uint64_t arg1,
 		break;
 
 #if ENABLE_CHIP_VERIFICATION_HARNESS
-	case MCE_CMD_ENABLE_LATIC:
+	case (uint64_t)MCE_CMD_ENABLE_LATIC:
 		/*
 		 * This call is not for production use. The constant value,
 		 * 0xFFFF0000, is specific to allowing for enabling LATIC on
@@ -327,14 +327,14 @@ int32_t mce_command_handler(uint64_t cmd, uint64_t arg0, uint64_t arg1,
 		break;
 #endif
 
-	case MCE_CMD_UNCORE_PERFMON_REQ:
+	case (uint64_t)MCE_CMD_UNCORE_PERFMON_REQ:
 		ret = ops->read_write_uncore_perfmon(cpu_ari_base, arg0, &arg1);
 
 		/* update context to return data */
 		write_ctx_reg(gp_regs, CTX_GPREG_X1, (arg1));
 		break;
 
-	case MCE_CMD_MISC_CCPLEX:
+	case (uint64_t)MCE_CMD_MISC_CCPLEX:
 		ops->misc_ccplex(cpu_ari_base, arg0, arg1);
 
 		break;

plat/nvidia/tegra/soc/t186/drivers/mce/nvg.c

@@ -14,10 +14,12 @@
 
 #include <mce_private.h>
 #include <t18x_ari.h>
+#include <tegra_private.h>
 
 int32_t nvg_enter_cstate(uint32_t ari_base, uint32_t state, uint32_t wake_time)
 {
 	int32_t ret = 0;
+	uint64_t val = 0ULL;
 
 	(void)ari_base;
 
@@ -28,10 +30,11 @@ int32_t nvg_enter_cstate(uint32_t ari_base, uint32_t state, uint32_t wake_time)
 		ret = EINVAL;
 	} else {
 		/* time (TSC ticks) until the core is expected to get a wake event */
-		nvg_set_request_data(TEGRA_NVG_CHANNEL_WAKE_TIME, wake_time);
+		nvg_set_request_data((uint64_t)TEGRA_NVG_CHANNEL_WAKE_TIME, wake_time);
 
 		/* set the core cstate */
-		write_actlr_el1(state);
+		val = read_actlr_el1() & ~ACTLR_EL1_PMSTATE_MASK;
+		write_actlr_el1(val | (uint64_t)state);
 	}
 
 	return ret;
@@ -78,7 +81,7 @@ int32_t nvg_update_cstate_info(uint32_t ari_base, uint32_t cluster, uint32_t ccp
 	val |= ((uint64_t)wake_mask << CSTATE_WAKE_MASK_SHIFT);
 
 	/* set the updated cstate info */
-	nvg_set_request_data(TEGRA_NVG_CHANNEL_CSTATE_INFO, val);
+	nvg_set_request_data((uint64_t)TEGRA_NVG_CHANNEL_CSTATE_INFO, val);
 
 	return 0;
 }
@@ -189,7 +192,7 @@ int32_t nvg_is_sc7_allowed(uint32_t ari_base, uint32_t state, uint32_t wake_time
 				((uint64_t)state & MCE_SC7_ALLOWED_MASK);
 
 		/* issue command to check if SC7 is allowed */
-		nvg_set_request_data(TEGRA_NVG_CHANNEL_IS_SC7_ALLOWED, val);
+		nvg_set_request_data((uint64_t)TEGRA_NVG_CHANNEL_IS_SC7_ALLOWED, val);
 
 		/* 1 = SC7 allowed, 0 = SC7 not allowed */
 		ret = (nvg_get_result() != 0ULL) ? 1 : 0;
@@ -219,7 +222,7 @@ int32_t nvg_online_core(uint32_t ari_base, uint32_t core)
 			ret = EINVAL;
 		} else {
 			/* get a core online */
-			nvg_set_request_data(TEGRA_NVG_CHANNEL_ONLINE_CORE,
+			nvg_set_request_data((uint64_t)TEGRA_NVG_CHANNEL_ONLINE_CORE,
 				((uint64_t)core & MCE_CORE_ID_MASK));
 		}
 	}
@@ -247,7 +250,7 @@ int32_t nvg_cc3_ctrl(uint32_t ari_base, uint32_t freq, uint32_t volt, uint8_t en
 		((volt & MCE_AUTO_CC3_VTG_MASK) << MCE_AUTO_CC3_VTG_SHIFT) |\
 		((enable != 0U) ? MCE_AUTO_CC3_ENABLE_BIT : 0U));
 
-	nvg_set_request_data(TEGRA_NVG_CHANNEL_CC3_CTRL, (uint64_t)val);
+	nvg_set_request_data((uint64_t)TEGRA_NVG_CHANNEL_CC3_CTRL, (uint64_t)val);
 
 	return 0;
 }

plat/nvidia/tegra/soc/t186/plat_memctrl.c

@@ -206,11 +206,11 @@ const static mc_txn_override_cfg_t tegra186_txn_override_cfgs[] = {
  ******************************************************************************/
 static tegra_mc_settings_t tegra186_mc_settings = {
 	.streamid_override_cfg = tegra186_streamid_override_regs,
-	.num_streamid_override_cfgs = ARRAY_SIZE(tegra186_streamid_override_regs),
+	.num_streamid_override_cfgs = (uint32_t)ARRAY_SIZE(tegra186_streamid_override_regs),
 	.streamid_security_cfg = tegra186_streamid_sec_cfgs,
-	.num_streamid_security_cfgs = ARRAY_SIZE(tegra186_streamid_sec_cfgs),
+	.num_streamid_security_cfgs = (uint32_t)ARRAY_SIZE(tegra186_streamid_sec_cfgs),
 	.txn_override_cfg = tegra186_txn_override_cfgs,
-	.num_txn_override_cfgs = ARRAY_SIZE(tegra186_txn_override_cfgs)
+	.num_txn_override_cfgs = (uint32_t)ARRAY_SIZE(tegra186_txn_override_cfgs)
 };
 
 /*******************************************************************************

plat/nvidia/tegra/soc/t186/plat_psci_handlers.c

@@ -20,6 +20,7 @@
 
 #include <mce.h>
 #include <smmu.h>
+#include <stdbool.h>
 #include <t18x_ari.h>
 #include <tegra_private.h>
 
@@ -27,12 +28,9 @@ extern void memcpy16(void *dest, const void *src, unsigned int length);
 
 extern void prepare_cpu_pwr_dwn(void);
 extern void tegra186_cpu_reset_handler(void);
-extern uint32_t __tegra186_cpu_reset_handler_end,
+extern uint64_t __tegra186_cpu_reset_handler_end,
 		__tegra186_smmu_context;
 
-/* TZDRAM offset for saving SMMU context */
-#define TEGRA186_SMMU_CTX_OFFSET	16UL
-
 /* state id mask */
 #define TEGRA186_STATE_ID_MASK		0xFU
 /* constants to get power state's wake time */
@@ -111,7 +109,7 @@ int32_t tegra_soc_pwr_domain_suspend(const psci_power_state_t *target_state)
 
 		/* Enter CPU idle/powerdown */
 		val = (stateid_afflvl0 == PSTATE_ID_CORE_IDLE) ?
-			TEGRA_ARI_CORE_C6 : TEGRA_ARI_CORE_C7;
+			(uint32_t)TEGRA_ARI_CORE_C6 : (uint32_t)TEGRA_ARI_CORE_C7;
 		(void)mce_command_handler((uint64_t)MCE_CMD_ENTER_CSTATE, (uint64_t)val,
 				tegra_percpu_data[cpu].wake_time, 0U);
 
@@ -132,12 +130,12 @@ int32_t tegra_soc_pwr_domain_suspend(const psci_power_state_t *target_state)
 		/* save SMMU context to TZDRAM */
 		smmu_ctx_base = params_from_bl2->tzdram_base +
 			((uintptr_t)&__tegra186_smmu_context -
-			 (uintptr_t)tegra186_cpu_reset_handler);
+			 (uintptr_t)&tegra186_cpu_reset_handler);
 		tegra_smmu_save_context((uintptr_t)smmu_ctx_base);
 
 		/* Prepare for system suspend */
-		cstate_info.cluster = TEGRA_ARI_CLUSTER_CC7;
-		cstate_info.system = TEGRA_ARI_SYSTEM_SC7;
+		cstate_info.cluster = (uint32_t)TEGRA_ARI_CLUSTER_CC7;
+		cstate_info.system = (uint32_t)TEGRA_ARI_SYSTEM_SC7;
 		cstate_info.system_state_force = 1;
 		cstate_info.update_wake_mask = 1;
 		mce_update_cstate_info(&cstate_info);
@@ -145,14 +143,14 @@ int32_t tegra_soc_pwr_domain_suspend(const psci_power_state_t *target_state)
 		do {
 			val = (uint32_t)mce_command_handler(
 					(uint64_t)MCE_CMD_IS_SC7_ALLOWED,
-					TEGRA_ARI_CORE_C7,
+					(uint64_t)TEGRA_ARI_CORE_C7,
 					MCE_CORE_SLEEP_TIME_INFINITE,
 					0U);
 		} while (val == 0U);
 
 		/* Instruct the MCE to enter system suspend state */
 		(void)mce_command_handler((uint64_t)MCE_CMD_ENTER_CSTATE,
-			TEGRA_ARI_CORE_C7, MCE_CORE_SLEEP_TIME_INFINITE, 0U);
+			(uint64_t)TEGRA_ARI_CORE_C7, MCE_CORE_SLEEP_TIME_INFINITE, 0U);
 	} else {
 		; /* do nothing */
 	}
@@ -161,32 +159,41 @@ int32_t tegra_soc_pwr_domain_suspend(const psci_power_state_t *target_state)
 }
 
 /*******************************************************************************
- * Platform handler to calculate the proper target power level at the
- * specified affinity level
+ * Helper function to check if this is the last ON CPU in the cluster
  ******************************************************************************/
-plat_local_state_t tegra_soc_get_target_pwr_state(uint32_t lvl,
-					     const plat_local_state_t *states,
-					     uint32_t ncpu)
+static bool tegra_last_cpu_in_cluster(const plat_local_state_t *states,
+			uint32_t ncpu)
 {
-	plat_local_state_t target = *states;
-	uint32_t pos = 0;
-	plat_local_state_t result = PSCI_LOCAL_STATE_RUN;
-	uint32_t cpu = plat_my_core_pos(), num_cpu = ncpu;
-	int32_t ret, cluster_powerdn = 1;
-	uint64_t core_pos = read_mpidr() & (uint64_t)MPIDR_CPU_MASK;
-	mce_cstate_info_t cstate_info = { 0 };
+	plat_local_state_t target;
+	bool last_on_cpu = true;
+	uint32_t num_cpus = ncpu, pos = 0;
+
+	do {
+		target = states[pos];
+		if (target != PLAT_MAX_OFF_STATE) {
+			last_on_cpu = false;
+		}
+		--num_cpus;
+		pos++;
+	} while (num_cpus != 0U);
 
-	/* get the power state at this level */
-	if (lvl == (uint32_t)MPIDR_AFFLVL1) {
-		target = states[core_pos];
-	}
-	if (lvl == (uint32_t)MPIDR_AFFLVL2) {
-		target = states[cpu];
-	}
+	return last_on_cpu;
+}
 
-	/* CPU suspend */
-	if ((lvl == (uint32_t)MPIDR_AFFLVL1) && (target == PSTATE_ID_CORE_POWERDN)) {
+/*******************************************************************************
+ * Helper function to get target power state for the cluster
+ ******************************************************************************/
+static plat_local_state_t tegra_get_afflvl1_pwr_state(const plat_local_state_t *states,
+			uint32_t ncpu)
+{
+	uint32_t core_pos = (uint32_t)read_mpidr() & (uint32_t)MPIDR_CPU_MASK;
+	uint32_t cpu = plat_my_core_pos();
+	int32_t ret;
+	plat_local_state_t target = states[core_pos];
+	mce_cstate_info_t cstate_info = { 0 };
 
+	/* CPU suspend */
+	if (target == PSTATE_ID_CORE_POWERDN) {
 		/* Program default wake mask */
 		cstate_info.wake_mask = TEGRA186_CORE_WAKE_MASK;
 		cstate_info.update_wake_mask = 1;
@@ -194,41 +201,30 @@ plat_local_state_t tegra_soc_get_target_pwr_state(uint32_t lvl,
 
 		/* Check if CCx state is allowed. */
 		ret = mce_command_handler((uint64_t)MCE_CMD_IS_CCX_ALLOWED,
-				TEGRA_ARI_CORE_C7, tegra_percpu_data[cpu].wake_time,
+				(uint64_t)TEGRA_ARI_CORE_C7,
+				tegra_percpu_data[cpu].wake_time,
 				0U);
-		if (ret != 0) {
-			result = PSTATE_ID_CORE_POWERDN;
+		if (ret == 0) {
+			target = PSCI_LOCAL_STATE_RUN;
 		}
 	}
 
 	/* CPU off */
-	if ((lvl == (uint32_t)MPIDR_AFFLVL1) && (target == PLAT_MAX_OFF_STATE)) {
-
-		/* find out the number of ON cpus in the cluster */
-		do {
-			target = states[pos];
-			if (target != PLAT_MAX_OFF_STATE) {
-				cluster_powerdn = 0;
-			}
-			--num_cpu;
-			pos++;
-		} while (num_cpu != 0U);
-
+	if (target == PLAT_MAX_OFF_STATE) {
 		/* Enable cluster powerdn from last CPU in the cluster */
-		if (cluster_powerdn != 0) {
-
+		if (tegra_last_cpu_in_cluster(states, ncpu)) {
 			/* Enable CC7 state and turn off wake mask */
-			cstate_info.cluster = TEGRA_ARI_CLUSTER_CC7;
+			cstate_info.cluster = (uint32_t)TEGRA_ARI_CLUSTER_CC7;
 			cstate_info.update_wake_mask = 1;
 			mce_update_cstate_info(&cstate_info);
 
 			/* Check if CCx state is allowed. */
 			ret = mce_command_handler((uint64_t)MCE_CMD_IS_CCX_ALLOWED,
-						  TEGRA_ARI_CORE_C7,
+						  (uint64_t)TEGRA_ARI_CORE_C7,
 						  MCE_CORE_SLEEP_TIME_INFINITE,
 						  0U);
-			if (ret != 0) {
-				result = PSTATE_ID_CORE_POWERDN;
+			if (ret == 0) {
+				target = PSCI_LOCAL_STATE_RUN;
 			}
 
 		} else {
@@ -236,17 +232,37 @@ plat_local_state_t tegra_soc_get_target_pwr_state(uint32_t lvl,
 			/* Turn off wake_mask */
 			cstate_info.update_wake_mask = 1;
 			mce_update_cstate_info(&cstate_info);
+			target = PSCI_LOCAL_STATE_RUN;
 		}
 	}
 
+	return target;
+}
+
+/*******************************************************************************
+ * Platform handler to calculate the proper target power level at the
+ * specified affinity level
+ ******************************************************************************/
+plat_local_state_t tegra_soc_get_target_pwr_state(uint32_t lvl,
+					     const plat_local_state_t *states,
+					     uint32_t ncpu)
+{
+	plat_local_state_t target = PSCI_LOCAL_STATE_RUN;
+	uint32_t cpu = plat_my_core_pos();
+
 	/* System Suspend */
-	if (((lvl == (uint32_t)MPIDR_AFFLVL2) || (lvl == (uint32_t)MPIDR_AFFLVL1)) &&
-	    (target == PSTATE_ID_SOC_POWERDN)) {
-		result = PSTATE_ID_SOC_POWERDN;
+	if ((lvl == (uint32_t)MPIDR_AFFLVL2) &&
+	    (states[cpu] == PSTATE_ID_SOC_POWERDN)) {
+		target = PSTATE_ID_SOC_POWERDN;
+	}
+
+	/* CPU off, CPU suspend */
+	if (lvl == (uint32_t)MPIDR_AFFLVL1) {
+		target = tegra_get_afflvl1_pwr_state(states, ncpu);
 	}
 
-	/* default state */
-	return result;
+	/* target cluster/system state */
+	return target;
 }
 
 int32_t tegra_soc_pwr_domain_power_down_wfi(const psci_power_state_t *target_state)
@@ -276,12 +292,12 @@ int32_t tegra_soc_pwr_domain_power_down_wfi(const psci_power_state_t *target_sta
 
 int32_t tegra_soc_pwr_domain_on(u_register_t mpidr)
 {
-	uint32_t target_cpu = mpidr & (uint64_t)MPIDR_CPU_MASK;
-	uint32_t target_cluster = (mpidr & MPIDR_CLUSTER_MASK) >>
-			(uint64_t)MPIDR_AFFINITY_BITS;
 	int32_t ret = PSCI_E_SUCCESS;
+	uint64_t target_cpu = mpidr & MPIDR_CPU_MASK;
+	uint64_t target_cluster = (mpidr & MPIDR_CLUSTER_MASK) >>
+			MPIDR_AFFINITY_BITS;
 
-	if (target_cluster > (uint64_t)MPIDR_AFFLVL1) {
+	if (target_cluster > MPIDR_AFFLVL1) {
 
 		ERROR("%s: unsupported CPU (0x%lx)\n", __func__, mpidr);
 		ret = PSCI_E_NOT_PRESENT;
@@ -304,14 +320,13 @@ int32_t tegra_soc_pwr_domain_on_finish(const psci_power_state_t *target_state)
 	uint64_t impl, val;
 	const plat_params_from_bl2_t *plat_params = bl31_get_plat_params();
 
-	impl = (read_midr() >> MIDR_IMPL_SHIFT) & (uint64_t)MIDR_IMPL_MASK;
+	impl = (read_midr() >> MIDR_IMPL_SHIFT) & MIDR_IMPL_MASK;
 
 	/*
 	 * Enable ECC and Parity Protection for Cortex-A57 CPUs (Tegra186
 	 * A02p and beyond).
 	 */
-	if ((plat_params->l2_ecc_parity_prot_dis != 1) &&
-	    (impl != (uint64_t)DENVER_IMPL)) {
+	if ((plat_params->l2_ecc_parity_prot_dis != 1) && (impl != DENVER_IMPL)) {
 
 		val = read_l2ctlr_el1();
 		val |= CORTEX_A57_L2_ECC_PARITY_PROTECTION_BIT;
@@ -327,7 +342,7 @@ int32_t tegra_soc_pwr_domain_on_finish(const psci_power_state_t *target_state)
 	 */
 	if (stateid_afflvl0 == PLAT_MAX_OFF_STATE) {
 
-		cstate_info.cluster = TEGRA_ARI_CLUSTER_CC1;
+		cstate_info.cluster = (uint32_t)TEGRA_ARI_CLUSTER_CC1;
 		cstate_info.update_wake_mask = 1;
 		mce_update_cstate_info(&cstate_info);
 	}
@@ -354,8 +369,8 @@ int32_t tegra_soc_pwr_domain_on_finish(const psci_power_state_t *target_state)
 		 * and SC7 for SC7 entry which may not be requested by
 		 * non-secure SW which controls idle states.
 		 */
-		cstate_info.cluster = TEGRA_ARI_CLUSTER_CC7;
-		cstate_info.system = TEGRA_ARI_SYSTEM_SC1;
+		cstate_info.cluster = (uint32_t)TEGRA_ARI_CLUSTER_CC7;
+		cstate_info.system = (uint32_t)TEGRA_ARI_SYSTEM_SC1;
 		cstate_info.update_wake_mask = 1;
 		mce_update_cstate_info(&cstate_info);
 	}
@@ -375,8 +390,8 @@ int32_t tegra_soc_pwr_domain_off(const psci_power_state_t *target_state)
 	}
 
 	/* Turn off CPU */
-	(void)mce_command_handler((uint64_t)MCE_CMD_ENTER_CSTATE, TEGRA_ARI_CORE_C7,
-			MCE_CORE_SLEEP_TIME_INFINITE, 0U);
+	(void)mce_command_handler((uint64_t)MCE_CMD_ENTER_CSTATE,
+			(uint64_t)TEGRA_ARI_CORE_C7, MCE_CORE_SLEEP_TIME_INFINITE, 0U);
 
 	return PSCI_E_SUCCESS;
 }
@@ -384,7 +399,7 @@ int32_t tegra_soc_pwr_domain_off(const psci_power_state_t *target_state)
 __dead2 void tegra_soc_prepare_system_off(void)
 {
 	/* power off the entire system */
-	mce_enter_ccplex_state(TEGRA_ARI_MISC_CCPLEX_SHUTDOWN_POWER_OFF);
+	mce_enter_ccplex_state((uint32_t)TEGRA_ARI_MISC_CCPLEX_SHUTDOWN_POWER_OFF);
 
 	wfi();
 
@@ -396,7 +411,7 @@ __dead2 void tegra_soc_prepare_system_off(void)
 
 int32_t tegra_soc_prepare_system_reset(void)
 {
-	mce_enter_ccplex_state(TEGRA_ARI_MISC_CCPLEX_SHUTDOWN_REBOOT);
+	mce_enter_ccplex_state((uint32_t)TEGRA_ARI_MISC_CCPLEX_SHUTDOWN_REBOOT);
 
 	return PSCI_E_SUCCESS;
 }

plat/nvidia/tegra/soc/t186/plat_secondary.c

@@ -49,11 +49,10 @@ void plat_secondary_setup(void)
 		cpu_reset_handler_base = params_from_bl2->tzdram_base;
 		memcpy16((void *)((uintptr_t)cpu_reset_handler_base),
 			 (void *)(uintptr_t)tegra186_cpu_reset_handler,
-			 (uintptr_t)&__tegra186_cpu_reset_handler_end -
-			 (uintptr_t)tegra186_cpu_reset_handler);
+			 (uintptr_t)&tegra186_cpu_reset_handler);
 
 	} else {
-		cpu_reset_handler_base = (uintptr_t)tegra_secure_entrypoint;
+		cpu_reset_handler_base = (uintptr_t)&tegra_secure_entrypoint;
 	}
 
 	addr_low = (uint32_t)cpu_reset_handler_base | CPU_RESET_MODE_AA64;

plat/nvidia/tegra/soc/t186/plat_sip_calls.c

@@ -23,7 +23,7 @@
 /*******************************************************************************
  * Offset to read the ref_clk counter value
  ******************************************************************************/
-#define REF_CLK_OFFSET		4
+#define REF_CLK_OFFSET		4ULL
 
 /*******************************************************************************
  * Tegra186 SiP SMCs
@@ -35,7 +35,7 @@
 #define TEGRA_SIP_MCE_CMD_READ_CSTATE_STATS		0xC2FFFF03
 #define TEGRA_SIP_MCE_CMD_WRITE_CSTATE_STATS		0xC2FFFF04
 #define TEGRA_SIP_MCE_CMD_IS_SC7_ALLOWED		0xC2FFFF05
-#define TEGRA_SIP_MCE_CMD_ONLINE_CORE			0xC2FFFF06
+
 #define TEGRA_SIP_MCE_CMD_CC3_CTRL			0xC2FFFF07
 #define TEGRA_SIP_MCE_CMD_ECHO_DATA			0xC2FFFF08
 #define TEGRA_SIP_MCE_CMD_READ_VERSIONS			0xC2FFFF09
@@ -52,7 +52,7 @@
 /*******************************************************************************
  * This function is responsible for handling all T186 SiP calls
  ******************************************************************************/
-int plat_sip_handler(uint32_t smc_fid,
+int32_t plat_sip_handler(uint32_t smc_fid,
 		     uint64_t x1,
 		     uint64_t x2,
 		     uint64_t x3,
@@ -61,24 +61,30 @@ int plat_sip_handler(uint32_t smc_fid,
 		     void *handle,
 		     uint64_t flags)
 {
-	int mce_ret;
-	int impl, cpu;
+	int32_t mce_ret, ret = 0;
+	uint32_t impl, cpu;
 	uint32_t base, core_clk_ctr, ref_clk_ctr;
+	uint32_t local_smc_fid = smc_fid;
+	uint64_t local_x1 = x1, local_x2 = x2, local_x3 = x3;
+
+	(void)x4;
+	(void)cookie;
+	(void)flags;
 
 	if (((smc_fid >> FUNCID_CC_SHIFT) & FUNCID_CC_MASK) == SMC_32) {
 		/* 32-bit function, clear top parameter bits */
 
-		x1 = (uint32_t)x1;
-		x2 = (uint32_t)x2;
-		x3 = (uint32_t)x3;
+		local_x1 = (uint32_t)x1;
+		local_x2 = (uint32_t)x2;
+		local_x3 = (uint32_t)x3;
 	}
 
 	/*
 	 * Convert SMC FID to SMC64, to support SMC32/SMC64 configurations
 	 */
-	smc_fid |= (SMC_64 << FUNCID_CC_SHIFT);
+	local_smc_fid |= (SMC_64 << FUNCID_CC_SHIFT);
 
-	switch (smc_fid) {
+	switch (local_smc_fid) {
 	/*
 	 * Micro Coded Engine (MCE) commands reside in the 0x82FFFF00 -
 	 * 0x82FFFFFF SiP SMC space
@@ -103,14 +109,13 @@ int plat_sip_handler(uint32_t smc_fid,
 	case TEGRA_SIP_MCE_CMD_MISC_CCPLEX:
 
 		/* clean up the high bits */
-		smc_fid &= MCE_CMD_MASK;
+		local_smc_fid &= MCE_CMD_MASK;
 
 		/* execute the command and store the result */
-		mce_ret = mce_command_handler(smc_fid, x1, x2, x3);
-		write_ctx_reg(get_gpregs_ctx(handle), CTX_GPREG_X0,
-			      (uint64_t)mce_ret);
-
-		return 0;
+		mce_ret = mce_command_handler(local_smc_fid, local_x1, local_x2, local_x3);
+		write_ctx_reg(get_gpregs_ctx(handle),
+			      CTX_GPREG_X0, (uint64_t)(mce_ret));
+		break;
 
 	/*
 	 * This function ID reads the Activity monitor's core/ref clock
@@ -125,28 +130,30 @@ int plat_sip_handler(uint32_t smc_fid,
 		impl = ((uint32_t)x2 >> MIDR_IMPL_SHIFT) & MIDR_IMPL_MASK;
 
 		/* sanity check target CPU number */
-		if (cpu > PLATFORM_MAX_CPUS_PER_CLUSTER)
-			return -EINVAL;
+		if (cpu > (uint32_t)PLATFORM_MAX_CPUS_PER_CLUSTER) {
+			ret = -EINVAL;
+		} else {
+			/* get the base address for the current CPU */
+			base = (impl == DENVER_IMPL) ? TEGRA_DENVER_ACTMON_CTR_BASE :
+				TEGRA_ARM_ACTMON_CTR_BASE;
+
+			/* read the clock counter values */
+			core_clk_ctr = mmio_read_32(base + (8ULL * cpu));
+			ref_clk_ctr = mmio_read_32(base + (8ULL * cpu) + REF_CLK_OFFSET);
+
+			/* return the counter values as two different parameters */
+			write_ctx_reg(get_gpregs_ctx(handle),
+				      CTX_GPREG_X1, (core_clk_ctr));
+			write_ctx_reg(get_gpregs_ctx(handle),
+				      CTX_GPREG_X2, (ref_clk_ctr));
+		}
 
-		/* get the base address for the current CPU */
-		base = (impl == DENVER_IMPL) ? TEGRA_DENVER_ACTMON_CTR_BASE :
-			TEGRA_ARM_ACTMON_CTR_BASE;
-
-		/* read the clock counter values */
-		core_clk_ctr = mmio_read_32(base + (8 * cpu));
-		ref_clk_ctr = mmio_read_32(base + (8 * cpu) + REF_CLK_OFFSET);
-
-		/* return the counter values as two different parameters */
-		write_ctx_reg(get_gpregs_ctx(handle), CTX_GPREG_X1,
-			      (uint64_t)core_clk_ctr);
-		write_ctx_reg(get_gpregs_ctx(handle), CTX_GPREG_X2,
-			      (uint64_t)ref_clk_ctr);
-
-		return 0;
+		break;
 
 	default:
+		ret = -ENOTSUP;
 		break;
 	}
 
-	return -ENOTSUP;
+	return ret;
 }

plat/nvidia/tegra/soc/t186/plat_smmu.c

@@ -9,6 +9,8 @@
 #include <smmu.h>
 #include <tegra_def.h>
 
+#define MAX_NUM_SMMU_DEVICES	U(1)
+
 /*******************************************************************************
  * Array to hold SMMU context for Tegra186
  ******************************************************************************/
@@ -305,7 +307,15 @@ static __attribute__((aligned(16))) smmu_regs_t tegra186_smmu_context[] = {
 smmu_regs_t *plat_get_smmu_ctx(void)
 {
 	/* index of _END_OF_TABLE_ */
-	tegra186_smmu_context[0].val = ARRAY_SIZE(tegra186_smmu_context) - 1;
+	tegra186_smmu_context[0].val = (uint32_t)(ARRAY_SIZE(tegra186_smmu_context)) - 1U;
 
 	return tegra186_smmu_context;
 }
+
+/*******************************************************************************
+ * Handler to return the support SMMU devices number
+ ******************************************************************************/
+uint32_t plat_get_num_smmu_devices(void)
+{
+	return MAX_NUM_SMMU_DEVICES;
+}

plat/nvidia/tegra/soc/t186/platform_t186.mk

@@ -20,9 +20,6 @@ $(eval $(call add_define,ENABLE_CHIP_VERIFICATION_HARNESS))
 ENABLE_SMMU_DEVICE			:= 1
 $(eval $(call add_define,ENABLE_SMMU_DEVICE))
 
-NUM_SMMU_DEVICES			:= 1
-$(eval $(call add_define,NUM_SMMU_DEVICES))
-
 RESET_TO_BL31				:= 1
 
 PROGRAMMABLE_RESET_ADDRESS		:= 1
@@ -50,6 +47,7 @@ PLAT_INCLUDES		+=	-I${SOC_DIR}/drivers/include
 
 BL31_SOURCES		+=	lib/cpus/aarch64/denver.S		\
 				lib/cpus/aarch64/cortex_a57.S		\
+				${COMMON_DIR}/drivers/gpcdma/gpcdma.c	\
 				${COMMON_DIR}/drivers/memctrl/memctrl_v2.c \
 				${COMMON_DIR}/drivers/smmu/smmu.c	\
 				${SOC_DIR}/drivers/mce/mce.c		\
@@ -64,3 +62,13 @@ BL31_SOURCES		+=	lib/cpus/aarch64/denver.S		\
 				${SOC_DIR}/plat_smmu.c			\
 				${SOC_DIR}/plat_trampoline.S
 
+# Enable workarounds for selected Cortex-A57 erratas.
+A57_DISABLE_NON_TEMPORAL_HINT	:=	1
+ERRATA_A57_806969		:=	1
+ERRATA_A57_813419		:=	1
+ERRATA_A57_813420		:=	1
+ERRATA_A57_826974		:=	1
+ERRATA_A57_826977		:=	1
+ERRATA_A57_828024		:=	1
+ERRATA_A57_829520		:=	1
+ERRATA_A57_833471		:=	1

plat/nvidia/tegra/soc/t210/drivers/se/se_private.h

@@ -16,14 +16,16 @@
  */
 
 /* Secure scratch registers */
-#define PMC_SECURE_SCRATCH4_OFFSET      	0xC0U
-#define PMC_SECURE_SCRATCH5_OFFSET      	0xC4U
-#define PMC_SECURE_SCRATCH6_OFFSET      	0x224U
-#define PMC_SECURE_SCRATCH7_OFFSET      	0x228U
-#define PMC_SECURE_SCRATCH120_OFFSET    	0xB38U
-#define PMC_SECURE_SCRATCH121_OFFSET    	0xB3CU
-#define PMC_SECURE_SCRATCH122_OFFSET    	0xB40U
-#define PMC_SECURE_SCRATCH123_OFFSET    	0xB44U
+#define PMC_SECURE_SCRATCH4_OFFSET		0xC0U
+#define PMC_SECURE_SCRATCH5_OFFSET		0xC4U
+#define PMC_SECURE_SCRATCH6_OFFSET		0x224U
+#define PMC_SECURE_SCRATCH7_OFFSET		0x228U
+#define PMC_SECURE_SCRATCH116_OFFSET		0xB28U
+#define PMC_SECURE_SCRATCH117_OFFSET		0xB2CU
+#define PMC_SECURE_SCRATCH120_OFFSET		0xB38U
+#define PMC_SECURE_SCRATCH121_OFFSET		0xB3CU
+#define PMC_SECURE_SCRATCH122_OFFSET		0xB40U
+#define PMC_SECURE_SCRATCH123_OFFSET		0xB44U
 
 /*
  * AHB arbitration memory write queue
@@ -32,6 +34,12 @@
 #define ARAHB_MST_ID_SE2_MASK			(0x1U << 13)
 #define ARAHB_MST_ID_SE_MASK			(0x1U << 14)
 
+/**
+ * SE registers
+ */
+#define TEGRA_SE_AES_KEYSLOT_COUNT		16
+#define SE_MAX_LAST_BLOCK_SIZE			0xFFFFF
+
 /* SE Status register */
 #define SE_STATUS_OFFSET			0x800U
 #define SE_STATUS_SHIFT				0
@@ -42,8 +50,24 @@
 #define SE_STATUS(x)	\
 		((x) & ((0x3U) << SE_STATUS_SHIFT))
 
+#define SE_MEM_INTERFACE_SHIFT			2
+#define SE_MEM_INTERFACE_IDLE			0
+#define SE_MEM_INTERFACE_BUSY			1
+#define SE_MEM_INTERFACE(x)	((x) << SE_STATUS_SHIFT)
+
+/* SE register definitions */
+#define SE_SECURITY_REG_OFFSET			0x0
+#define SE_SECURITY_TZ_LOCK_SOFT_SHIFT		5
+#define SE_SECURE				0x0
+#define SE_SECURITY_TZ_LOCK_SOFT(x)	((x) << SE_SECURITY_TZ_LOCK_SOFT_SHIFT)
+
+#define SE_SEC_ENG_DIS_SHIFT			1
+#define SE_DISABLE_FALSE			0
+#define SE_DISABLE_TRUE				1
+#define SE_SEC_ENG_DISABLE(x)((x) << SE_SEC_ENG_DIS_SHIFT)
+
 /* SE config register */
-#define SE_CONFIG_REG_OFFSET    		0x14U
+#define SE_CONFIG_REG_OFFSET			0x14U
 #define SE_CONFIG_ENC_ALG_SHIFT 		12
 #define SE_CONFIG_ENC_ALG_AES_ENC	\
 		((1U) << SE_CONFIG_ENC_ALG_SHIFT)
@@ -66,7 +90,7 @@
 #define SE_CONFIG_DEC_ALG(x)	\
 		((x) & ((0xFU) << SE_CONFIG_DEC_ALG_SHIFT))
 
-#define SE_CONFIG_DST_SHIFT     		2
+#define SE_CONFIG_DST_SHIFT	 		2
 #define SE_CONFIG_DST_MEMORY	\
 		((0U) << SE_CONFIG_DST_SHIFT)
 #define SE_CONFIG_DST_HASHREG	\
@@ -80,33 +104,75 @@
 #define SE_CONFIG_DST(x)	\
 		((x) & ((0x7U) << SE_CONFIG_DST_SHIFT))
 
+#define SE_CONFIG_ENC_MODE_SHIFT		24
+#define SE_CONFIG_ENC_MODE_KEY128	\
+			((0UL) << SE_CONFIG_ENC_MODE_SHIFT)
+#define SE_CONFIG_ENC_MODE_KEY192	\
+			((1UL) << SE_CONFIG_ENC_MODE_SHIFT)
+#define SE_CONFIG_ENC_MODE_KEY256	\
+			((2UL) << SE_CONFIG_ENC_MODE_SHIFT)
+#define SE_CONFIG_ENC_MODE_SHA1				\
+			((0UL) << SE_CONFIG_ENC_MODE_SHIFT)
+#define SE_CONFIG_ENC_MODE_SHA224	\
+			((4UL) << SE_CONFIG_ENC_MODE_SHIFT)
+#define SE_CONFIG_ENC_MODE_SHA256	\
+			((5UL) << SE_CONFIG_ENC_MODE_SHIFT)
+#define SE_CONFIG_ENC_MODE_SHA384	\
+			((6UL) << SE_CONFIG_ENC_MODE_SHIFT)
+#define SE_CONFIG_ENC_MODE_SHA512	\
+			((7UL) << SE_CONFIG_ENC_MODE_SHIFT)
+#define SE_CONFIG_ENC_MODE(x)\
+			((x) & ((0xFFUL) << SE_CONFIG_ENC_MODE_SHIFT))
+
+#define SE_CONFIG_DEC_MODE_SHIFT		16
+#define SE_CONFIG_DEC_MODE_KEY128	\
+			((0UL) << SE_CONFIG_DEC_MODE_SHIFT)
+#define SE_CONFIG_DEC_MODE_KEY192	\
+			((1UL) << SE_CONFIG_DEC_MODE_SHIFT)
+#define SE_CONFIG_DEC_MODE_KEY256	\
+			((2UL) << SE_CONFIG_DEC_MODE_SHIFT)
+#define SE_CONFIG_DEC_MODE_SHA1				\
+			((0UL) << SE_CONFIG_DEC_MODE_SHIFT)
+#define SE_CONFIG_DEC_MODE_SHA224	\
+			((4UL) << SE_CONFIG_DEC_MODE_SHIFT)
+#define SE_CONFIG_DEC_MODE_SHA256	\
+			((5UL) << SE_CONFIG_DEC_MODE_SHIFT)
+#define SE_CONFIG_DEC_MODE_SHA384	\
+			((6UL) << SE_CONFIG_DEC_MODE_SHIFT)
+#define SE_CONFIG_DEC_MODE_SHA512	\
+			((7UL) << SE_CONFIG_DEC_MODE_SHIFT)
+#define SE_CONFIG_DEC_MODE(x)\
+			((x) & ((0xFFUL) << SE_CONFIG_DEC_MODE_SHIFT))
+
+
 /* DRBG random number generator config */
 #define SE_RNG_CONFIG_REG_OFFSET		0x340
 
 #define DRBG_MODE_SHIFT				0
 #define DRBG_MODE_NORMAL		\
-		((0UL) << DRBG_MODE_SHIFT)
+		((0U) << DRBG_MODE_SHIFT)
 #define DRBG_MODE_FORCE_INSTANTION  \
-		((1UL) << DRBG_MODE_SHIFT)
+		((1U) << DRBG_MODE_SHIFT)
 #define DRBG_MODE_FORCE_RESEED	  \
-		((2UL) << DRBG_MODE_SHIFT)
+		((2U) << DRBG_MODE_SHIFT)
 #define SE_RNG_CONFIG_MODE(x)   \
-		((x) & ((0x3UL) << DRBG_MODE_SHIFT))
+		((x) & ((0x3U) << DRBG_MODE_SHIFT))
 
 #define DRBG_SRC_SHIFT				2
 #define DRBG_SRC_NONE	   \
-		((0UL) << DRBG_SRC_SHIFT)
+		((0U) << DRBG_SRC_SHIFT)
 #define DRBG_SRC_ENTROPY	\
-		((1UL) << DRBG_SRC_SHIFT)
+		((1U) << DRBG_SRC_SHIFT)
 #define DRBG_SRC_LFSR	   \
-		((2UL) << DRBG_SRC_SHIFT)
+		((2U) << DRBG_SRC_SHIFT)
 #define SE_RNG_SRC_CONFIG_MODE(x)   \
-		((x) & ((0x3UL) << DRBG_SRC_SHIFT))
+		((x) & ((0x3U) << DRBG_SRC_SHIFT))
 
 /* DRBG random number generator entropy config */
+
 #define SE_RNG_SRC_CONFIG_REG_OFFSET		0x344U
 
-#define DRBG_RO_ENT_SRC_SHIFT       		1
+#define DRBG_RO_ENT_SRC_SHIFT			1
 #define DRBG_RO_ENT_SRC_ENABLE	\
 		((1U) << DRBG_RO_ENT_SRC_SHIFT)
 #define DRBG_RO_ENT_SRC_DISABLE	\
@@ -114,7 +180,7 @@
 #define SE_RNG_SRC_CONFIG_RO_ENT_SRC(x)	\
 		((x) & ((0x1U) << DRBG_RO_ENT_SRC_SHIFT))
 
-#define DRBG_RO_ENT_SRC_LOCK_SHIFT  		0
+#define DRBG_RO_ENT_SRC_LOCK_SHIFT		0
 #define DRBG_RO_ENT_SRC_LOCK_ENABLE	\
 		((1U) << DRBG_RO_ENT_SRC_LOCK_SHIFT)
 #define DRBG_RO_ENT_SRC_LOCK_DISABLE	\
@@ -130,9 +196,97 @@
 #define SE_RNG_SRC_CONFIG_RO_ENT_IGNORE_MEM(x)	\
 		((x) & ((0x1U) << DRBG_RO_ENT_IGNORE_MEM_SHIFT))
 
+#define SE_RNG_RESEED_INTERVAL_REG_OFFSET	0x348
+
+/* SE CRYPTO */
+#define SE_CRYPTO_REG_OFFSET			0x304
+#define SE_CRYPTO_HASH_SHIFT			0
+#define SE_CRYPTO_HASH_DISABLE	\
+		((0U) << SE_CRYPTO_HASH_SHIFT)
+#define SE_CRYPTO_HASH_ENABLE	\
+		((1U) << SE_CRYPTO_HASH_SHIFT)
+
+#define SE_CRYPTO_XOR_POS_SHIFT			1
+#define SE_CRYPTO_XOR_BYPASS	\
+		((0U) << SE_CRYPTO_XOR_POS_SHIFT)
+#define SE_CRYPTO_XOR_TOP	\
+		((2U) << SE_CRYPTO_XOR_POS_SHIFT)
+#define SE_CRYPTO_XOR_BOTTOM	\
+		((3U) << SE_CRYPTO_XOR_POS_SHIFT)
+
+#define SE_CRYPTO_INPUT_SEL_SHIFT		3
+#define SE_CRYPTO_INPUT_AHB 	\
+		((0U) << SE_CRYPTO_INPUT_SEL_SHIFT)
+#define SE_CRYPTO_INPUT_RANDOM	\
+		((1U) << SE_CRYPTO_INPUT_SEL_SHIFT)
+#define SE_CRYPTO_INPUT_AESOUT	\
+		((2U) << SE_CRYPTO_INPUT_SEL_SHIFT)
+#define SE_CRYPTO_INPUT_LNR_CTR \
+		((3U) << SE_CRYPTO_INPUT_SEL_SHIFT)
+
+#define SE_CRYPTO_VCTRAM_SEL_SHIFT		 5
+#define SE_CRYPTO_VCTRAM_AHB	\
+		((0U) << SE_CRYPTO_VCTRAM_SEL_SHIFT)
+#define SE_CRYPTO_VCTRAM_AESOUT \
+		((2U) << SE_CRYPTO_VCTRAM_SEL_SHIFT)
+#define SE_CRYPTO_VCTRAM_PREVAHB	\
+		((3U) << SE_CRYPTO_VCTRAM_SEL_SHIFT)
+
+#define SE_CRYPTO_IV_SEL_SHIFT			 7
+#define SE_CRYPTO_IV_ORIGINAL	\
+		((0U) << SE_CRYPTO_IV_SEL_SHIFT)
+#define SE_CRYPTO_IV_UPDATED	\
+		((1U) << SE_CRYPTO_IV_SEL_SHIFT)
+
+#define SE_CRYPTO_CORE_SEL_SHIFT		8
+#define SE_CRYPTO_CORE_DECRYPT	\
+		((0U) << SE_CRYPTO_CORE_SEL_SHIFT)
+#define SE_CRYPTO_CORE_ENCRYPT	\
+		((1U) << SE_CRYPTO_CORE_SEL_SHIFT)
+
+#define SE_CRYPTO_KEY_INDEX_SHIFT		24
+#define SE_CRYPTO_KEY_INDEX(x) (x << SE_CRYPTO_KEY_INDEX_SHIFT)
+
+#define SE_CRYPTO_MEMIF_AHB	\
+		((0U) << SE_CRYPTO_MEMIF_SHIFT)
+#define SE_CRYPTO_MEMIF_MCCIF	\
+		((1U) << SE_CRYPTO_MEMIF_SHIFT)
+#define SE_CRYPTO_MEMIF_SHIFT			31
+
+/* KEY TABLE */
+#define SE_KEYTABLE_REG_OFFSET			0x31C
+
+/* KEYIV PKT - key slot */
+#define SE_KEYTABLE_SLOT_SHIFT			4
+#define SE_KEYTABLE_SLOT(x)	(x << SE_KEYTABLE_SLOT_SHIFT)
+
+/* KEYIV PKT - KEYIV select */
+#define SE_KEYIV_PKT_KEYIV_SEL_SHIFT		3
+#define SE_CRYPTO_KEYIV_KEY	\
+		((0U) << SE_KEYIV_PKT_KEYIV_SEL_SHIFT)
+#define SE_CRYPTO_KEYIV_IVS	\
+		((1U) << SE_KEYIV_PKT_KEYIV_SEL_SHIFT)
+
+/* KEYIV PKT - IV select */
+#define SE_KEYIV_PKT_IV_SEL_SHIFT		2
+#define SE_CRYPTO_KEYIV_IVS_OIV	\
+		((0U) << SE_KEYIV_PKT_IV_SEL_SHIFT)
+#define SE_CRYPTO_KEYIV_IVS_UIV \
+		((1U) << SE_KEYIV_PKT_IV_SEL_SHIFT)
+
+/* KEYIV PKT - key word */
+#define SE_KEYIV_PKT_KEY_WORD_SHIFT		0
+#define SE_KEYIV_PKT_KEY_WORD(x)	\
+		((x) << SE_KEYIV_PKT_KEY_WORD_SHIFT)
+
+/* KEYIV PKT - iv word */
+#define SE_KEYIV_PKT_IV_WORD_SHIFT		0
+#define SE_KEYIV_PKT_IV_WORD(x)		\
+		((x) << SE_KEYIV_PKT_IV_WORD_SHIFT)
+
 /* SE OPERATION */
 #define SE_OPERATION_REG_OFFSET 		0x8U
-#define SE_OPERATION_SHIFT      		0
+#define SE_OPERATION_SHIFT			0
 #define SE_OP_ABORT	\
 		((0x0U) << SE_OPERATION_SHIFT)
 #define SE_OP_START	\
@@ -146,11 +300,85 @@
 #define SE_OPERATION(x)	\
 		((x) & ((0x7U) << SE_OPERATION_SHIFT))
 
+/* SE CONTEXT */
+#define SE_CTX_SAVE_CONFIG_REG_OFFSET		0x70
+#define SE_CTX_SAVE_WORD_QUAD_SHIFT		0
+#define SE_CTX_SAVE_WORD_QUAD(x)	\
+		(x << SE_CTX_SAVE_WORD_QUAD_SHIFT)
+#define SE_CTX_SAVE_WORD_QUAD_KEYS_0_3	\
+		((0U) << SE_CTX_SAVE_WORD_QUAD_SHIFT)
+#define SE_CTX_SAVE_WORD_QUAD_KEYS_4_7	\
+		((1U) << SE_CTX_SAVE_WORD_QUAD_SHIFT)
+#define SE_CTX_SAVE_WORD_QUAD_ORIG_IV	\
+		((2U) << SE_CTX_SAVE_WORD_QUAD_SHIFT)
+#define SE_CTX_SAVE_WORD_QUAD_UPD_IV	\
+		((3U) << SE_CTX_SAVE_WORD_QUAD_SHIFT)
+
+#define SE_CTX_SAVE_KEY_INDEX_SHIFT		8
+#define SE_CTX_SAVE_KEY_INDEX(x)	(x << SE_CTX_SAVE_KEY_INDEX_SHIFT)
+
+#define SE_CTX_SAVE_STICKY_WORD_QUAD_SHIFT	24
+#define SE_CTX_SAVE_STICKY_WORD_QUAD_STICKY_0_3	\
+		((0U) << SE_CTX_SAVE_STICKY_WORD_QUAD_SHIFT)
+#define SE_CTX_SAVE_STICKY_WORD_QUAD_STICKY_4_7	\
+		((1U) << SE_CTX_SAVE_STICKY_WORD_QUAD_SHIFT)
+#define SE_CTX_SAVE_STICKY_WORD_QUAD(x)	\
+		(x << SE_CTX_SAVE_STICKY_WORD_QUAD_SHIFT)
+
+#define SE_CTX_SAVE_SRC_SHIFT			29
+#define SE_CTX_SAVE_SRC_STICKY_BITS	\
+		((0U) << SE_CTX_SAVE_SRC_SHIFT)
+#define SE_CTX_SAVE_SRC_RSA_KEYTABLE	\
+		((1U) << SE_CTX_SAVE_SRC_SHIFT)
+#define SE_CTX_SAVE_SRC_AES_KEYTABLE	\
+		((2U) << SE_CTX_SAVE_SRC_SHIFT)
+#define SE_CTX_SAVE_SRC_PKA1_STICKY_BITS	\
+		((3U) << SE_CTX_SAVE_SRC_SHIFT)
+#define SE_CTX_SAVE_SRC_MEM	\
+		((4U) << SE_CTX_SAVE_SRC_SHIFT)
+#define SE_CTX_SAVE_SRC_SRK	\
+		((6U) << SE_CTX_SAVE_SRC_SHIFT)
+#define SE_CTX_SAVE_SRC_PKA1_KEYTABLE	\
+		((7U) << SE_CTX_SAVE_SRC_SHIFT)
+
+#define SE_CTX_STICKY_WORD_QUAD_SHIFT		24
+#define SE_CTX_STICKY_WORD_QUAD_WORDS_0_3 \
+		((0U) << SE_CTX_STICKY_WORD_QUAD_SHIFT)
+#define SE_CTX_STICKY_WORD_QUAD_WORDS_4_7 \
+		((1U) << SE_CTX_STICKY_WORD_QUAD_SHIFT)
+#define SE_CTX_STICKY_WORD_QUAD(x)	 (x << SE_CTX_STICKY_WORD_QUAD_SHIFT)
+
+#define SE_CTX_SAVE_RSA_KEY_INDEX_SHIFT		16
+#define SE_CTX_SAVE_RSA_KEY_INDEX(x)				\
+			(x << SE_CTX_SAVE_RSA_KEY_INDEX_SHIFT)
+
+#define SE_CTX_RSA_WORD_QUAD_SHIFT		12
+#define SE_CTX_RSA_WORD_QUAD(x)	\
+			(x << SE_CTX_RSA_WORD_QUAD_SHIFT)
+
+#define SE_CTX_PKA1_WORD_QUAD_L_SHIFT		0
+#define SE_CTX_PKA1_WORD_QUAD_L_SIZE		\
+			((true ? 4:0) - \
+			(false ? 4:0) + 1)
+#define SE_CTX_PKA1_WORD_QUAD_L(x)\
+			(((x) << SE_CTX_PKA1_WORD_QUAD_L_SHIFT) & 0x1f)
+
+#define SE_CTX_PKA1_WORD_QUAD_H_SHIFT		12
+#define SE_CTX_PKA1_WORD_QUAD_H(x)\
+			((((x) >> SE_CTX_PKA1_WORD_QUAD_L_SIZE) & 0xf) \
+			<< SE_CTX_PKA1_WORD_QUAD_H_SHIFT)
+
+#define SE_RSA_KEY_INDEX_SLOT0_EXP		0
+#define SE_RSA_KEY_INDEX_SLOT0_MOD		1
+#define SE_RSA_KEY_INDEX_SLOT1_EXP		2
+#define SE_RSA_KEY_INDEX_SLOT1_MOD		3
+
+
 /* SE_CTX_SAVE_AUTO */
 #define SE_CTX_SAVE_AUTO_REG_OFFSET 		0x74U
 
 /* Enable */
-#define SE_CTX_SAVE_AUTO_ENABLE_SHIFT  		0
+#define SE_CTX_SAVE_AUTO_ENABLE_SHIFT		0
 #define SE_CTX_SAVE_AUTO_DIS	\
 		((0U) << SE_CTX_SAVE_AUTO_ENABLE_SHIFT)
 #define SE_CTX_SAVE_AUTO_EN	\
@@ -167,20 +395,22 @@
 #define SE_CTX_SAVE_AUTO_LOCK(x)	\
 		((x) & ((0x1U) << SE_CTX_SAVE_AUTO_LOCK_SHIFT))
 
-/* Current context save number of blocks  */
+/* Current context save number of blocks*/
 #define SE_CTX_SAVE_AUTO_CURR_CNT_SHIFT		16
 #define SE_CTX_SAVE_AUTO_CURR_CNT_MASK 		0x3FFU
 #define SE_CTX_SAVE_GET_BLK_COUNT(x)	\
 		(((x) >> SE_CTX_SAVE_AUTO_CURR_CNT_SHIFT) & \
 		SE_CTX_SAVE_AUTO_CURR_CNT_MASK)
 
-#define SE_CTX_SAVE_SIZE_BLOCKS_SE1      	133
-#define SE_CTX_SAVE_SIZE_BLOCKS_SE2     	646
+#define SE_CTX_SAVE_SIZE_BLOCKS_SE1		133
+#define SE_CTX_SAVE_SIZE_BLOCKS_SE2	 	646
 
 /* SE TZRAM OPERATION - only for SE1 */
-#define SE_TZRAM_OPERATION      		0x540U
+#define SE_TZRAM_OPERATION			0x540U
 
-#define SE_TZRAM_OP_MODE_SHIFT  		1
+#define SE_TZRAM_OP_MODE_SHIFT			1
+#define SE_TZRAM_OP_COMMAND_INIT		1
+#define SE_TZRAM_OP_COMMAND_SHIFT		0
 #define SE_TZRAM_OP_MODE_SAVE		\
 		((0U) << SE_TZRAM_OP_MODE_SHIFT)
 #define SE_TZRAM_OP_MODE_RESTORE	\
@@ -188,7 +418,7 @@
 #define SE_TZRAM_OP_MODE(x)		\
 		((x) & ((0x1U) << SE_TZRAM_OP_MODE_SHIFT))
 
-#define SE_TZRAM_OP_BUSY_SHIFT  		2
+#define SE_TZRAM_OP_BUSY_SHIFT			2
 #define SE_TZRAM_OP_BUSY_OFF	\
 		((0U) << SE_TZRAM_OP_BUSY_SHIFT)
 #define SE_TZRAM_OP_BUSY_ON	\
@@ -196,7 +426,7 @@
 #define SE_TZRAM_OP_BUSY(x)	\
 		((x) & ((0x1U) << SE_TZRAM_OP_BUSY_SHIFT))
 
-#define SE_TZRAM_OP_REQ_SHIFT  			0
+#define SE_TZRAM_OP_REQ_SHIFT			0
 #define SE_TZRAM_OP_REQ_IDLE	\
 		((0U) << SE_TZRAM_OP_REQ_SHIFT)
 #define SE_TZRAM_OP_REQ_INIT	\
@@ -206,7 +436,7 @@
 
 /* SE Interrupt */
 #define SE_INT_STATUS_REG_OFFSET		0x10U
-#define SE_INT_OP_DONE_SHIFT    		4
+#define SE_INT_OP_DONE_SHIFT			4
 #define SE_INT_OP_DONE_CLEAR	\
 		((0U) << SE_INT_OP_DONE_SHIFT)
 #define SE_INT_OP_DONE_ACTIVE	\
@@ -214,19 +444,186 @@
 #define SE_INT_OP_DONE(x)	\
 		((x) & ((0x1U) << SE_INT_OP_DONE_SHIFT))
 
+/* SE TZRAM SECURITY */
+#define SE_TZRAM_SEC_REG_OFFSET			0x4
+
+#define SE_TZRAM_SEC_SETTING_SHIFT		 0
+#define SE_TZRAM_SECURE		\
+		((0UL) << SE_TZRAM_SEC_SETTING_SHIFT)
+#define SE_TZRAM_NONSECURE	 \
+		((1UL) << SE_TZRAM_SEC_SETTING_SHIFT)
+#define SE_TZRAM_SEC_SETTING(x)		\
+		((x) & ((0x1UL) << SE_TZRAM_SEC_SETTING_SHIFT))
+
+/* PKA1 KEY SLOTS */
+#define TEGRA_SE_PKA1_KEYSLOT_COUNT		4
+
+
 /* SE error status */
 #define SE_ERR_STATUS_REG_OFFSET		0x804U
+#define SE_CRYPTO_KEYTABLE_DST_REG_OFFSET	0x330
+#define SE_CRYPTO_KEYTABLE_DST_WORD_QUAD_SHIFT	0
+#define SE_CRYPTO_KEYTABLE_DST_WORD_QUAD(x)	\
+			(x << SE_CRYPTO_KEYTABLE_DST_WORD_QUAD_SHIFT)
+
+#define SE_KEY_INDEX_SHIFT			8
+#define SE_CRYPTO_KEYTABLE_DST_KEY_INDEX(x)	(x << SE_KEY_INDEX_SHIFT)
+
 
 /* SE linked list (LL) register */
 #define SE_IN_LL_ADDR_REG_OFFSET		0x18U
-#define SE_OUT_LL_ADDR_REG_OFFSET  		0x24U
-#define SE_BLOCK_COUNT_REG_OFFSET  		0x318U
+#define SE_OUT_LL_ADDR_REG_OFFSET		0x24U
+#define SE_BLOCK_COUNT_REG_OFFSET		0x318U
 
 /* AES data sizes */
+#define TEGRA_SE_KEY_256_SIZE			32
+#define TEGRA_SE_KEY_192_SIZE			24
+#define TEGRA_SE_KEY_128_SIZE			16
 #define TEGRA_SE_AES_BLOCK_SIZE 		16
-#define TEGRA_SE_AES_MIN_KEY_SIZE  		16
-#define TEGRA_SE_AES_MAX_KEY_SIZE  		32
-#define TEGRA_SE_AES_IV_SIZE    		16
+#define TEGRA_SE_AES_MIN_KEY_SIZE		16
+#define TEGRA_SE_AES_MAX_KEY_SIZE		32
+#define TEGRA_SE_AES_IV_SIZE			16
+
+#define TEGRA_SE_RNG_IV_SIZE			16
+#define TEGRA_SE_RNG_DT_SIZE			16
+#define TEGRA_SE_RNG_KEY_SIZE			16
+#define TEGRA_SE_RNG_SEED_SIZE	(TEGRA_SE_RNG_IV_SIZE + \
+									TEGRA_SE_RNG_KEY_SIZE + \
+									TEGRA_SE_RNG_DT_SIZE)
+#define TEGRA_SE_RSA512_DIGEST_SIZE		64
+#define TEGRA_SE_RSA1024_DIGEST_SIZE		128
+#define TEGRA_SE_RSA1536_DIGEST_SIZE		192
+#define TEGRA_SE_RSA2048_DIGEST_SIZE		256
+
+#define SE_KEY_TABLE_ACCESS_REG_OFFSET		0x284
+#define SE_KEY_READ_DISABLE_SHIFT		0
+
+#define SE_CTX_BUFER_SIZE			1072
+#define SE_CTX_DRBG_BUFER_SIZE			2112
+
+/* SE blobs size in bytes */
+#define SE_CTX_SAVE_RSA_KEY_LENGTH		1024
+#define SE_CTX_SAVE_RANDOM_DATA_SIZE		16
+#define SE_CTX_SAVE_STICKY_BITS_SIZE		16
+#define SE2_CONTEXT_SAVE_PKA1_STICKY_BITS_LENGTH 16
+#define SE2_CONTEXT_SAVE_PKA1_KEYS_LENGTH	8192
+#define SE_CTX_KNOWN_PATTERN_SIZE		16
+#define SE_CTX_KNOWN_PATTERN_SIZE_WORDS		(SE_CTX_KNOWN_PATTERN_SIZE/4)
+
+/* SE RSA */
+#define TEGRA_SE_RSA_KEYSLOT_COUNT		2
+#define SE_RSA_KEY_SIZE_REG_OFFSET		0x404
+#define SE_RSA_EXP_SIZE_REG_OFFSET		0x408
+#define SE_RSA_MAX_EXP_BIT_SIZE			2048
+#define SE_RSA_MAX_EXP_SIZE32	\
+		(SE_RSA_MAX_EXP_BIT_SIZE >> 5)
+#define SE_RSA_MAX_MOD_BIT_SIZE			2048
+#define SE_RSA_MAX_MOD_SIZE32	\
+		(SE_RSA_MAX_MOD_BIT_SIZE >> 5)
+
+/* SE_RSA_KEYTABLE_ADDR */
+#define SE_RSA_KEYTABLE_ADDR			0x420
+#define RSA_KEY_PKT_WORD_ADDR_SHIFT		0
+#define RSA_KEY_PKT_EXPMOD_SEL_SHIFT	\
+		((6U) << RSA_KEY_PKT_WORD_ADDR_SHIFT)
+#define RSA_KEY_MOD	\
+		((1U) << RSA_KEY_PKT_EXPMOD_SEL_SHIFT)
+#define RSA_KEY_EXP	\
+		((0U) << RSA_KEY_PKT_EXPMOD_SEL_SHIFT)
+#define RSA_KEY_PKT_SLOT_SHIFT			7
+#define RSA_KEY_SLOT_1	\
+		((0U) << RSA_KEY_PKT_SLOT_SHIFT)
+#define RSA_KEY_SLOT_2	\
+		((1U) << RSA_KEY_PKT_SLOT_SHIFT)
+#define RSA_KEY_PKT_INPUT_MODE_SHIFT		8
+#define RSA_KEY_REG_INPUT	\
+		((0U) << RSA_KEY_PKT_INPUT_MODE_SHIFT)
+#define RSA_KEY_DMA_INPUT	\
+		((1U) << RSA_KEY_PKT_INPUT_MODE_SHIFT)
+
+/* SE_RSA_KEYTABLE_DATA */
+#define SE_RSA_KEYTABLE_DATA			0x424
+
+/* SE_RSA_CONFIG register */
+#define SE_RSA_CONFIG				0x400
+#define RSA_KEY_SLOT_SHIFT			24
+#define RSA_KEY_SLOT(x) \
+		((x) << RSA_KEY_SLOT_SHIFT)
+
+/*******************************************************************************
+ * Structure definition
+ ******************************************************************************/
+
+/* SE context blob */
+#pragma pack(push, 1)
+typedef struct tegra_aes_key_slot {
+	/* 0 - 7 AES key */
+	uint32_t key[8];
+	/* 8 - 11 Original IV */
+	uint32_t oiv[4];
+	/* 12 - 15 Updated IV */
+	uint32_t uiv[4];
+} tegra_se_aes_key_slot_t;
+#pragma pack(pop)
+
+#pragma pack(push, 1)
+typedef struct tegra_se_context {
+	/* random number */
+	unsigned char rand_data[SE_CTX_SAVE_RANDOM_DATA_SIZE];
+	/* Sticky bits */
+	unsigned char sticky_bits[SE_CTX_SAVE_STICKY_BITS_SIZE * 2];
+	/* AES key slots */
+	tegra_se_aes_key_slot_t key_slots[TEGRA_SE_AES_KEYSLOT_COUNT];
+	/* RSA key slots */
+	unsigned char rsa_keys[SE_CTX_SAVE_RSA_KEY_LENGTH];
+} tegra_se_context_t;
+#pragma pack(pop)
+
+/* PKA context blob */
+#pragma pack(push, 1)
+typedef struct tegra_pka_context {
+	unsigned char sticky_bits[SE2_CONTEXT_SAVE_PKA1_STICKY_BITS_LENGTH];
+	unsigned char pka_keys[SE2_CONTEXT_SAVE_PKA1_KEYS_LENGTH];
+} tegra_pka_context_t;
+#pragma pack(pop)
+
+/* SE context blob */
+#pragma pack(push, 1)
+typedef struct tegra_se_context_blob {
+	/* SE context */
+	tegra_se_context_t se_ctx;
+	/* Known Pattern */
+	unsigned char known_pattern[SE_CTX_KNOWN_PATTERN_SIZE];
+} tegra_se_context_blob_t;
+#pragma pack(pop)
+
+/* SE2 and PKA1 context blob */
+#pragma pack(push, 1)
+typedef struct tegra_se2_context_blob {
+	/* SE2 context */
+	tegra_se_context_t se_ctx;
+	/* PKA1 context */
+	tegra_pka_context_t pka_ctx;
+	/* Known Pattern */
+	unsigned char known_pattern[SE_CTX_KNOWN_PATTERN_SIZE];
+} tegra_se2_context_blob_t;
+#pragma pack(pop)
+
+/* SE AES key type 128bit, 192bit, 256bit */
+typedef enum {
+	SE_AES_KEY128,
+	SE_AES_KEY192,
+	SE_AES_KEY256,
+} tegra_se_aes_key_type_t;
+
+/* SE RSA key slot */
+typedef struct tegra_se_rsa_key_slot {
+	/* 0 - 63 exponent key */
+	uint32_t exponent[SE_RSA_MAX_EXP_SIZE32];
+	/* 64 - 127 modulus key */
+	uint32_t modulus[SE_RSA_MAX_MOD_SIZE32];
+} tegra_se_rsa_key_slot_t;
+
 
 /*******************************************************************************
  * Inline functions definition
@@ -242,8 +639,21 @@ static inline void tegra_se_write_32(const tegra_se_dev_t *dev, uint32_t offset,
 	mmio_write_32(dev->se_base + offset, val);
 }
 
+static inline uint32_t tegra_pka_read_32(tegra_pka_dev_t *dev, uint32_t offset)
+{
+	return mmio_read_32(dev->pka_base + offset);
+}
+
+static inline void tegra_pka_write_32(tegra_pka_dev_t *dev, uint32_t offset,
+uint32_t val)
+{
+	mmio_write_32(dev->pka_base + offset, val);
+}
+
 /*******************************************************************************
  * Prototypes
  ******************************************************************************/
+int tegra_se_start_normal_operation(const tegra_se_dev_t *, uint32_t);
+int tegra_se_start_ctx_save_operation(const tegra_se_dev_t *, uint32_t);
 
 #endif /* SE_PRIVATE_H */

plat/nvidia/tegra/soc/t210/drivers/se/security_engine.c

@@ -20,7 +20,8 @@
  * Constants and Macros
  ******************************************************************************/
 
-#define TIMEOUT_100MS	100UL	// Timeout in 100ms
+#define TIMEOUT_100MS	100U	// Timeout in 100ms
+#define RNG_AES_KEY_INDEX   1
 
 /*******************************************************************************
  * Data structure and global variables
@@ -67,6 +68,15 @@
  * #--------------------------------#
  */
 
+/* Known pattern data */
+static const uint32_t se_ctx_known_pattern_data[SE_CTX_KNOWN_PATTERN_SIZE_WORDS] = {
+	/* 128 bit AES block */
+	0x0C0D0E0F,
+	0x08090A0B,
+	0x04050607,
+	0x00010203,
+};
+
 /* SE input and output linked list buffers */
 static tegra_se_io_lst_t se1_src_ll_buf;
 static tegra_se_io_lst_t se1_dst_ll_buf;
@@ -78,7 +88,7 @@ static tegra_se_io_lst_t se2_dst_ll_buf;
 /* SE1 security engine device handle */
 static tegra_se_dev_t se_dev_1 = {
 	.se_num = 1,
-	/* setup base address for se */
+	/* Setup base address for se */
 	.se_base = TEGRA_SE1_BASE,
 	/* Setup context size in AES blocks */
 	.ctx_size_blks = SE_CTX_SAVE_SIZE_BLOCKS_SE1,
@@ -86,12 +96,14 @@ static tegra_se_dev_t se_dev_1 = {
 	.src_ll_buf = &se1_src_ll_buf,
 	/* Setup DST buffers for SE operations */
 	.dst_ll_buf = &se1_dst_ll_buf,
+	/* Setup context save destination */
+	.ctx_save_buf = (uint32_t *)(TEGRA_TZRAM_CARVEOUT_BASE),
 };
 
 /* SE2 security engine device handle */
 static tegra_se_dev_t se_dev_2 = {
 	.se_num = 2,
-	/* setup base address for se */
+	/* Setup base address for se */
 	.se_base = TEGRA_SE2_BASE,
 	/* Setup context size in AES blocks */
 	.ctx_size_blks = SE_CTX_SAVE_SIZE_BLOCKS_SE2,
@@ -99,8 +111,12 @@ static tegra_se_dev_t se_dev_2 = {
 	.src_ll_buf = &se2_src_ll_buf,
 	/* Setup DST buffers for SE operations */
 	.dst_ll_buf = &se2_dst_ll_buf,
+	/* Setup context save destination */
+	.ctx_save_buf = (uint32_t *)(TEGRA_TZRAM_CARVEOUT_BASE + 0x1000),
 };
 
+static bool ecid_valid;
+
 /*******************************************************************************
  * Functions Definition
  ******************************************************************************/
@@ -186,35 +202,15 @@ static int32_t tegra_se_operation_complete(const tegra_se_dev_t *se_dev)
 }
 
 /*
- * Verify the SE context save auto has been enabled.
- * SE_CTX_SAVE_AUTO.ENABLE == ENABLE
- * If the SE context save auto is not enabled, then set
- * the context save auto enable and lock the setting.
- * If the SE context save auto is not enabled and the
- * enable setting is locked, then return an error.
+ * Returns true if the SE engine is configured to perform SE context save in
+ * hardware.
  */
-static inline int32_t tegra_se_ctx_save_auto_enable(const tegra_se_dev_t *se_dev)
+static inline bool tegra_se_atomic_save_enabled(const tegra_se_dev_t *se_dev)
 {
 	uint32_t val;
-	int32_t ret = 0;
 
 	val = tegra_se_read_32(se_dev, SE_CTX_SAVE_AUTO_REG_OFFSET);
-	if (SE_CTX_SAVE_AUTO_ENABLE(val) == SE_CTX_SAVE_AUTO_DIS) {
-		if (SE_CTX_SAVE_AUTO_LOCK(val) == SE_CTX_SAVE_AUTO_LOCK_EN) {
-			ERROR("%s: ERR: Cannot enable atomic. Write locked!\n",
-					__func__);
-			ret = -EACCES;
-		}
-
-		/* Program SE_CTX_SAVE_AUTO */
-		if (ret == 0) {
-			tegra_se_write_32(se_dev, SE_CTX_SAVE_AUTO_REG_OFFSET,
-					SE_CTX_SAVE_AUTO_LOCK_EN |
-					SE_CTX_SAVE_AUTO_EN);
-		}
-	}
-
-	return ret;
+	return (SE_CTX_SAVE_AUTO_ENABLE(val) == SE_CTX_SAVE_AUTO_EN);
 }
 
 /*
@@ -259,14 +255,6 @@ static int32_t tegra_se_context_save_atomic(const tegra_se_dev_t *se_dev)
 	/* Check that previous operation is finalized */
 	ret = tegra_se_operation_prepare(se_dev);
 
-	/* Ensure HW atomic context save has been enabled
-	 * This should have been done at boot time.
-	 * SE_CTX_SAVE_AUTO.ENABLE == ENABLE
-	 */
-	if (ret == 0) {
-		ret = tegra_se_ctx_save_auto_enable(se_dev);
-	}
-
 	/* Read the context save progress counter: block_count
 	 * Ensure no previous context save has been triggered
 	 * SE_CTX_SAVE_AUTO.CURR_CNT == 0
@@ -325,7 +313,8 @@ static int32_t tegra_se_context_save_atomic(const tegra_se_dev_t *se_dev)
  * Security engine primitive operations, including normal operation
  * and the context save operation.
  */
-static int tegra_se_perform_operation(const tegra_se_dev_t *se_dev, uint32_t nbytes)
+static int tegra_se_perform_operation(const tegra_se_dev_t *se_dev, uint32_t nbytes,
+					bool context_save)
 {
 	uint32_t nblocks = nbytes / TEGRA_SE_AES_BLOCK_SIZE;
 	int ret = 0;
@@ -351,7 +340,10 @@ static int tegra_se_perform_operation(const tegra_se_dev_t *se_dev, uint32_t nby
 	tegra_se_make_data_coherent(se_dev);
 
 	/* Start hardware operation */
-	tegra_se_write_32(se_dev, SE_OPERATION_REG_OFFSET, SE_OP_START);
+	if (context_save)
+		tegra_se_write_32(se_dev, SE_OPERATION_REG_OFFSET, SE_OP_CTX_SAVE);
+	else
+		tegra_se_write_32(se_dev, SE_OPERATION_REG_OFFSET, SE_OP_START);
 
 	/* Wait for operation to finish */
 	ret = tegra_se_operation_complete(se_dev);
@@ -360,6 +352,22 @@ op_error:
 	return ret;
 }
 
+/*
+ * Normal security engine operations other than the context save
+ */
+int tegra_se_start_normal_operation(const tegra_se_dev_t *se_dev, uint32_t nbytes)
+{
+	return tegra_se_perform_operation(se_dev, nbytes, false);
+}
+
+/*
+ * Security engine context save operation
+ */
+int tegra_se_start_ctx_save_operation(const tegra_se_dev_t *se_dev, uint32_t nbytes)
+{
+	return tegra_se_perform_operation(se_dev, nbytes, true);
+}
+
 /*
  * Security Engine sequence to generat SRK
  * SE and SE2 will generate different SRK by different
@@ -381,7 +389,10 @@ static int tegra_se_generate_srk(const tegra_se_dev_t *se_dev)
 	se_dev->dst_ll_buf->last_buff_num = 0;
 
 	/* Configure random number generator */
-	val = (DRBG_MODE_FORCE_RESEED | DRBG_SRC_ENTROPY);
+	if (ecid_valid)
+		val = (DRBG_MODE_FORCE_INSTANTION | DRBG_SRC_ENTROPY);
+	else
+		val = (DRBG_MODE_FORCE_RESEED | DRBG_SRC_ENTROPY);
 	tegra_se_write_32(se_dev, SE_RNG_CONFIG_REG_OFFSET, val);
 
 	/* Configure output destination = SRK */
@@ -391,25 +402,562 @@ static int tegra_se_generate_srk(const tegra_se_dev_t *se_dev)
 	tegra_se_write_32(se_dev, SE_CONFIG_REG_OFFSET, val);
 
 	/* Perform hardware operation */
-	ret = tegra_se_perform_operation(se_dev, 0);
+	ret = tegra_se_start_normal_operation(se_dev, 0);
+
+	return ret;
+}
+
+/*
+ * Generate plain text random data to some memory location using
+ * SE/SE2's SP800-90 random number generator. The random data size
+ * must be some multiple of the AES block size (16 bytes).
+ */
+static int tegra_se_lp_generate_random_data(tegra_se_dev_t *se_dev)
+{
+	int ret = 0;
+	uint32_t val;
+
+	/* Set some arbitrary memory location to store the random data */
+	se_dev->dst_ll_buf->last_buff_num = 0;
+	if (!se_dev->ctx_save_buf) {
+		ERROR("%s: ERR: context save buffer NULL pointer!\n", __func__);
+		return PSCI_E_NOT_PRESENT;
+	}
+	se_dev->dst_ll_buf->buffer[0].addr = ((uint64_t)(&(((tegra_se_context_t *)
+					se_dev->ctx_save_buf)->rand_data)));
+	se_dev->dst_ll_buf->buffer[0].data_len = SE_CTX_SAVE_RANDOM_DATA_SIZE;
+
+
+	/* Confgure the following hardware register settings:
+	 * SE_CONFIG.DEC_ALG = NOP
+	 * SE_CONFIG.ENC_ALG = RNG
+	 * SE_CONFIG.ENC_MODE = KEY192
+	 * SE_CONFIG.DST = MEMORY
+	 */
+	val = (SE_CONFIG_ENC_ALG_RNG |
+		SE_CONFIG_DEC_ALG_NOP |
+		SE_CONFIG_ENC_MODE_KEY192 |
+		SE_CONFIG_DST_MEMORY);
+	tegra_se_write_32(se_dev, SE_CONFIG_REG_OFFSET, val);
+
+	/* Program the RNG options in SE_CRYPTO_CONFIG as follows:
+	 * XOR_POS = BYPASS
+	 * INPUT_SEL = RANDOM (Entropy or LFSR)
+	 * HASH_ENB = DISABLE
+	 */
+	val = (SE_CRYPTO_INPUT_RANDOM |
+		SE_CRYPTO_XOR_BYPASS |
+		SE_CRYPTO_CORE_ENCRYPT |
+		SE_CRYPTO_HASH_DISABLE |
+		SE_CRYPTO_KEY_INDEX(RNG_AES_KEY_INDEX) |
+		SE_CRYPTO_IV_ORIGINAL);
+	tegra_se_write_32(se_dev, SE_CRYPTO_REG_OFFSET, val);
+
+	/* Configure RNG */
+	if (ecid_valid)
+		val = (DRBG_MODE_FORCE_INSTANTION | DRBG_SRC_LFSR);
+	else
+		val = (DRBG_MODE_FORCE_RESEED | DRBG_SRC_LFSR);
+	tegra_se_write_32(se_dev, SE_RNG_CONFIG_REG_OFFSET, val);
+
+	/* SE normal operation */
+	ret = tegra_se_start_normal_operation(se_dev, SE_CTX_SAVE_RANDOM_DATA_SIZE);
+
+	return ret;
+}
+
+/*
+ * Encrypt memory blocks with SRK as part of the security engine context.
+ * The data blocks include: random data and the known pattern data, where
+ * the random data is the first block and known pattern is the last block.
+ */
+static int tegra_se_lp_data_context_save(tegra_se_dev_t *se_dev,
+		uint64_t src_addr, uint64_t dst_addr, uint32_t data_size)
+{
+	int ret = 0;
+
+	se_dev->src_ll_buf->last_buff_num = 0;
+	se_dev->dst_ll_buf->last_buff_num = 0;
+	se_dev->src_ll_buf->buffer[0].addr = src_addr;
+	se_dev->src_ll_buf->buffer[0].data_len = data_size;
+	se_dev->dst_ll_buf->buffer[0].addr = dst_addr;
+	se_dev->dst_ll_buf->buffer[0].data_len = data_size;
+
+	/* By setting the context source from memory and calling the context save
+	 * operation, the SE encrypts the memory data with SRK.
+	 */
+	tegra_se_write_32(se_dev, SE_CTX_SAVE_CONFIG_REG_OFFSET, SE_CTX_SAVE_SRC_MEM);
+
+	ret = tegra_se_start_ctx_save_operation(se_dev, data_size);
+
+	return ret;
+}
+
+/*
+ * Context save the key table access control sticky bits and
+ * security status of each key-slot. The encrypted sticky-bits are
+ * 32 bytes (2 AES blocks) and formatted as the following structure:
+ * {	bit in registers			bit in context save
+ *	SECURITY_0[4]				158
+ *	SE_RSA_KEYTABLE_ACCE4SS_1[2:0]		157:155
+ *	SE_RSA_KEYTABLE_ACCE4SS_0[2:0]		154:152
+ *	SE_RSA_SECURITY_PERKEY_0[1:0]		151:150
+ *	SE_CRYPTO_KEYTABLE_ACCESS_15[7:0]	149:142
+ *	...,
+ *	SE_CRYPTO_KEYTABLE_ACCESS_0[7:0]	29:22
+ *	SE_CRYPTO_SECURITY_PERKEY_0[15:0]	21:6
+ *	SE_TZRAM_SECURITY_0[1:0]		5:4
+ *	SE_SECURITY_0[16]			3:3
+ *	SE_SECURITY_0[2:0] }			2:0
+ */
+static int tegra_se_lp_sticky_bits_context_save(tegra_se_dev_t *se_dev)
+{
+	int ret = PSCI_E_INTERN_FAIL;
+	uint32_t val = 0;
+
+	se_dev->dst_ll_buf->last_buff_num = 0;
+	if (!se_dev->ctx_save_buf) {
+		ERROR("%s: ERR: context save buffer NULL pointer!\n", __func__);
+		return PSCI_E_NOT_PRESENT;
+	}
+	se_dev->dst_ll_buf->buffer[0].addr = ((uint64_t)(&(((tegra_se_context_t *)
+						se_dev->ctx_save_buf)->sticky_bits)));
+	se_dev->dst_ll_buf->buffer[0].data_len = SE_CTX_SAVE_STICKY_BITS_SIZE;
+
+	/*
+	 * The 1st AES block save the sticky-bits context 1 - 16 bytes (0 - 3 words).
+	 * The 2nd AES block save the sticky-bits context 17 - 32 bytes (4 - 7 words).
+	 */
+	for (int i = 0; i < 2; i++) {
+		val = SE_CTX_SAVE_SRC_STICKY_BITS |
+			SE_CTX_SAVE_STICKY_WORD_QUAD(i);
+		tegra_se_write_32(se_dev, SE_CTX_SAVE_CONFIG_REG_OFFSET, val);
+
+		/* SE context save operation */
+		ret = tegra_se_start_ctx_save_operation(se_dev,
+				SE_CTX_SAVE_STICKY_BITS_SIZE);
+		if (ret)
+			break;
+		se_dev->dst_ll_buf->buffer[0].addr += SE_CTX_SAVE_STICKY_BITS_SIZE;
+	}
+
+	return ret;
+}
+
+static int tegra_se_aeskeytable_context_save(tegra_se_dev_t *se_dev)
+{
+	uint32_t val = 0;
+	int ret = 0;
+
+	se_dev->dst_ll_buf->last_buff_num = 0;
+	if (!se_dev->ctx_save_buf) {
+		ERROR("%s: ERR: context save buffer NULL pointer!\n", __func__);
+		ret = -EINVAL;
+		goto aes_keytable_save_err;
+	}
+
+	/* AES key context save */
+	for (int slot = 0; slot < TEGRA_SE_AES_KEYSLOT_COUNT; slot++) {
+		se_dev->dst_ll_buf->buffer[0].addr = ((uint64_t)(&(
+						((tegra_se_context_t *)se_dev->
+						 ctx_save_buf)->key_slots[slot].key)));
+		se_dev->dst_ll_buf->buffer[0].data_len = TEGRA_SE_KEY_128_SIZE;
+		for (int i = 0; i < 2; i++) {
+			val = SE_CTX_SAVE_SRC_AES_KEYTABLE |
+				SE_CTX_SAVE_KEY_INDEX(slot) |
+				SE_CTX_SAVE_WORD_QUAD(i);
+			tegra_se_write_32(se_dev, SE_CTX_SAVE_CONFIG_REG_OFFSET, val);
+
+			/* SE context save operation */
+			ret = tegra_se_start_ctx_save_operation(se_dev,
+					TEGRA_SE_KEY_128_SIZE);
+			if (ret) {
+				ERROR("%s: ERR: AES key CTX_SAVE OP failed, "
+						"slot=%d, word_quad=%d.\n",
+						__func__, slot, i);
+				goto aes_keytable_save_err;
+			}
+			se_dev->dst_ll_buf->buffer[0].addr += TEGRA_SE_KEY_128_SIZE;
+		}
+
+		/* OIV context save */
+		se_dev->dst_ll_buf->last_buff_num = 0;
+		se_dev->dst_ll_buf->buffer[0].addr = ((uint64_t)(&(
+						((tegra_se_context_t *)se_dev->
+						 ctx_save_buf)->key_slots[slot].oiv)));
+		se_dev->dst_ll_buf->buffer[0].data_len = TEGRA_SE_AES_IV_SIZE;
+
+		val = SE_CTX_SAVE_SRC_AES_KEYTABLE |
+			SE_CTX_SAVE_KEY_INDEX(slot) |
+			SE_CTX_SAVE_WORD_QUAD_ORIG_IV;
+		tegra_se_write_32(se_dev, SE_CTX_SAVE_CONFIG_REG_OFFSET, val);
+
+		/* SE context save operation */
+		ret = tegra_se_start_ctx_save_operation(se_dev, TEGRA_SE_AES_IV_SIZE);
+		if (ret) {
+			ERROR("%s: ERR: OIV CTX_SAVE OP failed, slot=%d.\n",
+					__func__, slot);
+			goto aes_keytable_save_err;
+		}
+
+		/* UIV context save */
+		se_dev->dst_ll_buf->last_buff_num = 0;
+		se_dev->dst_ll_buf->buffer[0].addr = ((uint64_t)(&(
+						((tegra_se_context_t *)se_dev->
+						 ctx_save_buf)->key_slots[slot].uiv)));
+		se_dev->dst_ll_buf->buffer[0].data_len = TEGRA_SE_AES_IV_SIZE;
+
+		val = SE_CTX_SAVE_SRC_AES_KEYTABLE |
+			SE_CTX_SAVE_KEY_INDEX(slot) |
+			SE_CTX_SAVE_WORD_QUAD_UPD_IV;
+		tegra_se_write_32(se_dev, SE_CTX_SAVE_CONFIG_REG_OFFSET, val);
+
+		/* SE context save operation */
+		ret = tegra_se_start_ctx_save_operation(se_dev, TEGRA_SE_AES_IV_SIZE);
+		if (ret) {
+			ERROR("%s: ERR: UIV CTX_SAVE OP failed, slot=%d\n",
+					__func__, slot);
+			goto aes_keytable_save_err;
+		}
+	}
+
+aes_keytable_save_err:
+	return ret;
+}
+
+static int tegra_se_lp_rsakeytable_context_save(tegra_se_dev_t *se_dev)
+{
+	uint32_t val = 0;
+	int ret = 0;
+	/* First the modulus and then the exponent must be
+	 * encrypted and saved. This is repeated for SLOT 0
+	 * and SLOT 1. Hence the order:
+	 * SLOT 0 exponent : RSA_KEY_INDEX : 0
+	 * SLOT 0 modulus : RSA_KEY_INDEX : 1
+	 * SLOT 1 exponent : RSA_KEY_INDEX : 2
+	 * SLOT 1 modulus : RSA_KEY_INDEX : 3
+	 */
+	const unsigned int key_index_mod[TEGRA_SE_RSA_KEYSLOT_COUNT][2] = {
+		/* RSA key slot 0 */
+		{SE_RSA_KEY_INDEX_SLOT0_EXP, SE_RSA_KEY_INDEX_SLOT0_MOD},
+		/* RSA key slot 1 */
+		{SE_RSA_KEY_INDEX_SLOT1_EXP, SE_RSA_KEY_INDEX_SLOT1_MOD},
+	};
+
+	se_dev->dst_ll_buf->last_buff_num = 0;
+	se_dev->dst_ll_buf->buffer[0].addr = ((uint64_t)(&(
+					((tegra_se_context_t *)se_dev->
+					 ctx_save_buf)->rsa_keys)));
+	se_dev->dst_ll_buf->buffer[0].data_len = TEGRA_SE_KEY_128_SIZE;
+
+	for (int slot = 0; slot < TEGRA_SE_RSA_KEYSLOT_COUNT; slot++) {
+		/* loop for modulus and exponent */
+		for (int index = 0; index < 2; index++) {
+			for (int word_quad = 0; word_quad < 16; word_quad++) {
+				val = SE_CTX_SAVE_SRC_RSA_KEYTABLE |
+					SE_CTX_SAVE_RSA_KEY_INDEX(
+						key_index_mod[slot][index]) |
+					SE_CTX_RSA_WORD_QUAD(word_quad);
+				tegra_se_write_32(se_dev,
+					SE_CTX_SAVE_CONFIG_REG_OFFSET, val);
+
+				/* SE context save operation */
+				ret = tegra_se_start_ctx_save_operation(se_dev,
+						TEGRA_SE_KEY_128_SIZE);
+				if (ret) {
+					ERROR("%s: ERR: slot=%d.\n",
+						__func__, slot);
+					goto rsa_keytable_save_err;
+				}
+
+				/* Update the pointer to the next word quad */
+				se_dev->dst_ll_buf->buffer[0].addr +=
+					TEGRA_SE_KEY_128_SIZE;
+			}
+		}
+	}
+
+rsa_keytable_save_err:
+	return ret;
+}
+
+static int tegra_se_pkakeytable_sticky_bits_save(tegra_se_dev_t *se_dev)
+{
+	int ret = 0;
 
+	se_dev->dst_ll_buf->last_buff_num = 0;
+	se_dev->dst_ll_buf->buffer[0].addr = ((uint64_t)(&(
+					((tegra_se2_context_blob_t *)se_dev->
+					 ctx_save_buf)->pka_ctx.sticky_bits)));
+	se_dev->dst_ll_buf->buffer[0].data_len = TEGRA_SE_AES_BLOCK_SIZE;
+
+	/* PKA1 sticky bits are 1 AES block (16 bytes) */
+	tegra_se_write_32(se_dev, SE_CTX_SAVE_CONFIG_REG_OFFSET,
+			SE_CTX_SAVE_SRC_PKA1_STICKY_BITS |
+			SE_CTX_STICKY_WORD_QUAD_WORDS_0_3);
+
+	/* SE context save operation */
+	ret = tegra_se_start_ctx_save_operation(se_dev, 0);
+	if (ret) {
+		ERROR("%s: ERR: PKA1 sticky bits CTX_SAVE OP failed\n",
+				__func__);
+		goto pka_sticky_bits_save_err;
+	}
+
+pka_sticky_bits_save_err:
 	return ret;
 }
 
+static int tegra_se_pkakeytable_context_save(tegra_se_dev_t *se_dev)
+{
+	uint32_t val = 0;
+	int ret = 0;
+
+	se_dev->dst_ll_buf->last_buff_num = 0;
+	se_dev->dst_ll_buf->buffer[0].addr = ((uint64_t)(&(
+					((tegra_se2_context_blob_t *)se_dev->
+					 ctx_save_buf)->pka_ctx.pka_keys)));
+	se_dev->dst_ll_buf->buffer[0].data_len = TEGRA_SE_KEY_128_SIZE;
+
+	/* for each slot, save word quad 0-127 */
+	for (int slot = 0; slot < TEGRA_SE_PKA1_KEYSLOT_COUNT; slot++) {
+		for (int word_quad = 0; word_quad < 512/4; word_quad++) {
+			val = SE_CTX_SAVE_SRC_PKA1_KEYTABLE |
+				SE_CTX_PKA1_WORD_QUAD_L((slot * 128) +
+						word_quad) |
+				SE_CTX_PKA1_WORD_QUAD_H((slot * 128) +
+						word_quad);
+			tegra_se_write_32(se_dev,
+					SE_CTX_SAVE_CONFIG_REG_OFFSET, val);
+
+			/* SE context save operation */
+			ret = tegra_se_start_ctx_save_operation(se_dev,
+					TEGRA_SE_KEY_128_SIZE);
+			if (ret) {
+				ERROR("%s: ERR: pka1 keytable ctx save error\n",
+						__func__);
+				goto pka_keytable_save_err;
+			}
+
+			/* Update the pointer to the next word quad */
+			se_dev->dst_ll_buf->buffer[0].addr +=
+				TEGRA_SE_KEY_128_SIZE;
+		}
+	}
+
+pka_keytable_save_err:
+	return ret;
+}
+
+static int tegra_se_save_SRK(tegra_se_dev_t *se_dev)
+{
+	tegra_se_write_32(se_dev, SE_CTX_SAVE_CONFIG_REG_OFFSET,
+			SE_CTX_SAVE_SRC_SRK);
+
+	/* SE context save operation */
+	return tegra_se_start_ctx_save_operation(se_dev, 0);
+}
+
+/*
+ *  Lock both SE from non-TZ clients.
+ */
+static inline void tegra_se_lock(tegra_se_dev_t *se_dev)
+{
+	uint32_t val;
+
+	assert(se_dev);
+	val = tegra_se_read_32(se_dev, SE_SECURITY_REG_OFFSET);
+	val |= SE_SECURITY_TZ_LOCK_SOFT(SE_SECURE);
+	tegra_se_write_32(se_dev, SE_SECURITY_REG_OFFSET, val);
+}
+
+/*
+ * Use SRK to encrypt SE state and save to TZRAM carveout
+ */
+static int tegra_se_context_save_sw(tegra_se_dev_t *se_dev)
+{
+	int err = 0;
+
+	assert(se_dev);
+
+	/* Lock entire SE/SE2 as TZ protected */
+	tegra_se_lock(se_dev);
+
+	INFO("%s: generate SRK\n", __func__);
+	/* Generate SRK */
+	err = tegra_se_generate_srk(se_dev);
+	if (err) {
+		ERROR("%s: ERR: SRK generation failed\n", __func__);
+		return err;
+	}
+
+	INFO("%s: generate random data\n", __func__);
+	/* Generate random data */
+	err = tegra_se_lp_generate_random_data(se_dev);
+	if (err) {
+		ERROR("%s: ERR: LP random pattern generation failed\n", __func__);
+		return err;
+	}
+
+	INFO("%s: encrypt random data\n", __func__);
+	/* Encrypt the random data block */
+	err = tegra_se_lp_data_context_save(se_dev,
+		((uint64_t)(&(((tegra_se_context_t *)se_dev->
+					ctx_save_buf)->rand_data))),
+		((uint64_t)(&(((tegra_se_context_t *)se_dev->
+					ctx_save_buf)->rand_data))),
+		SE_CTX_SAVE_RANDOM_DATA_SIZE);
+	if (err) {
+		ERROR("%s: ERR: random pattern encryption failed\n", __func__);
+		return err;
+	}
+
+	INFO("%s: save SE sticky bits\n", __func__);
+	/* Save AES sticky bits context */
+	err = tegra_se_lp_sticky_bits_context_save(se_dev);
+	if (err) {
+		ERROR("%s: ERR: sticky bits context save failed\n", __func__);
+		return err;
+	}
+
+	INFO("%s: save AES keytables\n", __func__);
+	/* Save AES key table context */
+	err = tegra_se_aeskeytable_context_save(se_dev);
+	if (err) {
+		ERROR("%s: ERR: LP keytable save failed\n", __func__);
+		return err;
+	}
+
+	/* RSA key slot table context save */
+	INFO("%s: save RSA keytables\n", __func__);
+	err = tegra_se_lp_rsakeytable_context_save(se_dev);
+	if (err) {
+		ERROR("%s: ERR: rsa key table context save failed\n", __func__);
+		return err;
+	}
+
+	/* Only SE2 has an interface with PKA1; thus, PKA1's context is saved
+	 * via SE2.
+	 */
+	if (se_dev->se_num == 2) {
+		/* Encrypt PKA1 sticky bits on SE2 only */
+		INFO("%s: save PKA sticky bits\n", __func__);
+		err = tegra_se_pkakeytable_sticky_bits_save(se_dev);
+		if (err) {
+			ERROR("%s: ERR: PKA sticky bits context save failed\n", __func__);
+			return err;
+		}
+
+		/* Encrypt PKA1 keyslots on SE2 only */
+		INFO("%s: save PKA keytables\n", __func__);
+		err = tegra_se_pkakeytable_context_save(se_dev);
+		if (err) {
+			ERROR("%s: ERR: PKA key table context save failed\n", __func__);
+			return err;
+		}
+	}
+
+	/* Encrypt known pattern */
+	if (se_dev->se_num == 1) {
+		err = tegra_se_lp_data_context_save(se_dev,
+			((uint64_t)(&se_ctx_known_pattern_data)),
+			((uint64_t)(&(((tegra_se_context_blob_t *)se_dev->ctx_save_buf)->known_pattern))),
+			SE_CTX_KNOWN_PATTERN_SIZE);
+	} else if (se_dev->se_num == 2) {
+		err = tegra_se_lp_data_context_save(se_dev,
+			((uint64_t)(&se_ctx_known_pattern_data)),
+			((uint64_t)(&(((tegra_se2_context_blob_t *)se_dev->ctx_save_buf)->known_pattern))),
+			SE_CTX_KNOWN_PATTERN_SIZE);
+	}
+	if (err) {
+		ERROR("%s: ERR: save LP known pattern failure\n", __func__);
+		return err;
+	}
+
+	/* Write lp context buffer address into PMC scratch register */
+	if (se_dev->se_num == 1) {
+		/* SE context address */
+		mmio_write_32((uint64_t)TEGRA_PMC_BASE + PMC_SECURE_SCRATCH117_OFFSET,
+				((uint64_t)(se_dev->ctx_save_buf)));
+	} else if (se_dev->se_num == 2) {
+		/* SE2 & PKA1 context address */
+		mmio_write_32((uint64_t)TEGRA_PMC_BASE + PMC_SECURE_SCRATCH116_OFFSET,
+				((uint64_t)(se_dev->ctx_save_buf)));
+	}
+
+	/* Saves SRK to PMC secure scratch registers for BootROM, which
+	 * verifies and restores the security engine context on warm boot.
+	 */
+	err = tegra_se_save_SRK(se_dev);
+	if (err < 0) {
+		ERROR("%s: ERR: LP SRK save failure\n", __func__);
+		return err;
+	}
+
+	INFO("%s: SE context save done \n", __func__);
+
+	return err;
+}
+
 /*
  * Initialize the SE engine handle
  */
 void tegra_se_init(void)
 {
+	uint32_t val = 0;
 	INFO("%s: start SE init\n", __func__);
 
 	/* Generate random SRK to initialize DRBG */
 	tegra_se_generate_srk(&se_dev_1);
 	tegra_se_generate_srk(&se_dev_2);
 
+	/* determine if ECID is valid */
+	val = mmio_read_32(TEGRA_FUSE_BASE + FUSE_JTAG_SECUREID_VALID);
+	ecid_valid = (val == ECID_VALID);
+
 	INFO("%s: SE init done\n", __func__);
 }
 
+static void tegra_se_enable_clocks(void)
+{
+	uint32_t val = 0;
+
+	/* Enable entropy clock */
+	val = mmio_read_32(TEGRA_CAR_RESET_BASE + TEGRA_CLK_OUT_ENB_W);
+	val |= ENTROPY_CLK_ENB_BIT;
+	mmio_write_32(TEGRA_CAR_RESET_BASE + TEGRA_CLK_OUT_ENB_W, val);
+
+	/* De-Assert Entropy Reset */
+	val = mmio_read_32(TEGRA_CAR_RESET_BASE + TEGRA_RST_DEVICES_W);
+	val &= ~ENTROPY_RESET_BIT;
+	mmio_write_32(TEGRA_CAR_RESET_BASE + TEGRA_RST_DEVICES_W, val);
+
+	/* Enable SE clock */
+	val = mmio_read_32(TEGRA_CAR_RESET_BASE + TEGRA_CLK_OUT_ENB_V);
+	val |= SE_CLK_ENB_BIT;
+	mmio_write_32(TEGRA_CAR_RESET_BASE + TEGRA_CLK_OUT_ENB_V, val);
+
+	/* De-Assert SE Reset */
+	val = mmio_read_32(TEGRA_CAR_RESET_BASE + TEGRA_RST_DEVICES_V);
+	val &= ~SE_RESET_BIT;
+	mmio_write_32(TEGRA_CAR_RESET_BASE + TEGRA_RST_DEVICES_V, val);
+}
+
+static void tegra_se_disable_clocks(void)
+{
+	uint32_t val = 0;
+
+	/* Disable entropy clock */
+	val = mmio_read_32(TEGRA_CAR_RESET_BASE + TEGRA_CLK_OUT_ENB_W);
+	val &= ~ENTROPY_CLK_ENB_BIT;
+	mmio_write_32(TEGRA_CAR_RESET_BASE + TEGRA_CLK_OUT_ENB_W, val);
+
+	/* Disable SE clock */
+	val = mmio_read_32(TEGRA_CAR_RESET_BASE + TEGRA_CLK_OUT_ENB_V);
+	val &= ~SE_CLK_ENB_BIT;
+	mmio_write_32(TEGRA_CAR_RESET_BASE + TEGRA_CLK_OUT_ENB_V, val);
+}
+
 /*
  * Security engine power suspend entry point.
  * This function is invoked from PSCI power domain suspend handler.
@@ -417,21 +965,57 @@ void tegra_se_init(void)
 int32_t tegra_se_suspend(void)
 {
 	int32_t ret = 0;
+	uint32_t val = 0;
 
-	/* Atomic context save se2 and pka1 */
-	INFO("%s: SE2/PKA1 atomic context save\n", __func__);
-	ret = tegra_se_context_save_atomic(&se_dev_2);
+	/* SE does not use SMMU in EL3, disable SMMU.
+	 * This will be re-enabled by kernel on resume */
+	val = mmio_read_32(TEGRA_MC_BASE + MC_SMMU_PPCS_ASID_0);
+	val &= ~PPCS_SMMU_ENABLE;
+	mmio_write_32(TEGRA_MC_BASE + MC_SMMU_PPCS_ASID_0, val);
 
-	/* Atomic context save se */
-	if (ret == 0) {
-		INFO("%s: SE1 atomic context save\n", __func__);
-		ret = tegra_se_context_save_atomic(&se_dev_1);
-	}
+	tegra_se_enable_clocks();
 
-	if (ret == 0) {
-		INFO("%s: SE atomic context save done\n", __func__);
+	if (tegra_se_atomic_save_enabled(&se_dev_2) &&
+			tegra_se_atomic_save_enabled(&se_dev_1)) {
+		/* Atomic context save se2 and pka1 */
+		INFO("%s: SE2/PKA1 atomic context save\n", __func__);
+		if (ret == 0) {
+			ret = tegra_se_context_save_atomic(&se_dev_2);
+		}
+
+		/* Atomic context save se */
+		if (ret == 0) {
+			INFO("%s: SE1 atomic context save\n", __func__);
+			ret = tegra_se_context_save_atomic(&se_dev_1);
+		}
+
+		if (ret == 0) {
+			INFO("%s: SE atomic context save done\n", __func__);
+		}
+	} else if (!tegra_se_atomic_save_enabled(&se_dev_2) &&
+			!tegra_se_atomic_save_enabled(&se_dev_1)) {
+		/* SW context save se2 and pka1 */
+		INFO("%s: SE2/PKA1 legacy(SW) context save\n", __func__);
+		if (ret == 0) {
+			ret = tegra_se_context_save_sw(&se_dev_2);
+		}
+
+		/* SW context save se */
+		if (ret == 0) {
+			INFO("%s: SE1 legacy(SW) context save\n", __func__);
+			ret = tegra_se_context_save_sw(&se_dev_1);
+		}
+
+		if (ret == 0) {
+			INFO("%s: SE SW context save done\n", __func__);
+		}
+	} else {
+		ERROR("%s: One SE set for atomic CTX save, the other is not\n",
+			 __func__);
 	}
 
+	tegra_se_disable_clocks();
+
 	return ret;
 }
 
@@ -445,6 +1029,7 @@ int32_t tegra_se_save_tzram(void)
 	uint32_t timeout;
 
 	INFO("%s: SE TZRAM save start\n", __func__);
+	tegra_se_enable_clocks();
 
 	val = (SE_TZRAM_OP_REQ_INIT | SE_TZRAM_OP_MODE_SAVE);
 	tegra_se_write_32(&se_dev_1, SE_TZRAM_OPERATION, val);
@@ -465,6 +1050,8 @@ int32_t tegra_se_save_tzram(void)
 		INFO("%s: SE TZRAM save done!\n", __func__);
 	}
 
+	tegra_se_disable_clocks();
+
 	return ret;
 }
 
@@ -483,12 +1070,6 @@ static void tegra_se_warm_boot_resume(const tegra_se_dev_t *se_dev)
 		DRBG_RO_ENT_SRC_ENABLE;
 	tegra_se_write_32(se_dev, SE_RNG_SRC_CONFIG_REG_OFFSET, val);
 
-	/* Enable and lock the SE atomic context save setting */
-	if (tegra_se_ctx_save_auto_enable(se_dev) != 0) {
-		ERROR("%s: ERR: enable SE%d context save auto failed!\n",
-			__func__, se_dev->se_num);
-	}
-
 	/* Set a random value to SRK to initialize DRBG */
 	tegra_se_generate_srk(se_dev);
 }