feat(spmc): add FF-A secure partition manager core
This patch introduces the core support for enabling an SPMC in EL3
as per the FF-A spec.
The current implemented functionality is targeted to enable
initialization of the SPMC itself and initial support for
bringing up a single S-EL1 SP.
This includes initialization of the SPMC's internal state,
parsing of an SP's manifest, preparing the cpu contexts and
appropriate system registers for the Secure Partition.
The spmc_smc_handler is the main handler for all incoming SMCs
to the SPMC, FF-A ABI handlers and functionality will
be implemented in subsequent patches.
Signed-off-by: Marc Bonnici <marc.bonnici@arm.com>
Change-Id: Ib33c240b91e54cbd018a69fec880d02adfbe12b9
diff --git a/services/std_svc/spm/el3_spmc/spmc.h b/services/std_svc/spm/el3_spmc/spmc.h
new file mode 100644
index 0000000..e891516
--- /dev/null
+++ b/services/std_svc/spm/el3_spmc/spmc.h
@@ -0,0 +1,185 @@
+/*
+ * Copyright (c) 2022, ARM Limited and Contributors. All rights reserved.
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ */
+
+#ifndef SPMC_H
+#define SPMC_H
+
+#include <stdint.h>
+
+#include <lib/psci/psci.h>
+#include <lib/spinlock.h>
+#include "spm_common.h"
+
+/*
+ * Ranges of FF-A IDs for Normal world and Secure world components. The
+ * convention matches that used by other SPMCs i.e. Hafnium and OP-TEE.
+ */
+#define FFA_NWD_ID_BASE 0x0
+#define FFA_NWD_ID_LIMIT 0x7FFF
+#define FFA_SWD_ID_BASE 0x8000
+#define FFA_SWD_ID_LIMIT SPMD_DIRECT_MSG_ENDPOINT_ID - 1
+#define FFA_SWD_ID_MASK 0x8000
+
+/* First ID is reserved for the SPMC */
+#define FFA_SPMC_ID U(FFA_SWD_ID_BASE)
+/* SP IDs are allocated after the SPMC ID */
+#define FFA_SP_ID_BASE (FFA_SPMC_ID + 1)
+/* Align with Hafnium implementation */
+#define INV_SP_ID 0x7FFF
+
+/* FF-A warm boot types. */
+#define FFA_WB_TYPE_S2RAM 0
+#define FFA_WB_TYPE_NOTS2RAM 1
+
+/*
+ * Runtime states of an execution context as per the FF-A v1.1 specification.
+ */
+enum sp_runtime_states {
+ RT_STATE_WAITING,
+ RT_STATE_RUNNING,
+ RT_STATE_PREEMPTED,
+ RT_STATE_BLOCKED
+};
+
+/*
+ * Runtime model of an execution context as per the FF-A v1.1 specification. Its
+ * value is valid only if the execution context is not in the waiting state.
+ */
+enum sp_runtime_model {
+ RT_MODEL_DIR_REQ,
+ RT_MODEL_RUN,
+ RT_MODEL_INIT,
+ RT_MODEL_INTR
+};
+
+enum sp_runtime_el {
+ EL1 = 0,
+ S_EL0,
+ S_EL1
+};
+
+enum sp_execution_state {
+ SP_STATE_AARCH64 = 0,
+ SP_STATE_AARCH32
+};
+
+/*
+ * Execution context members for an SP. This is a bit like struct
+ * vcpu in a hypervisor.
+ */
+struct sp_exec_ctx {
+ /*
+ * Store the stack address to restore C runtime context from after
+ * returning from a synchronous entry into the SP.
+ */
+ uint64_t c_rt_ctx;
+
+ /* Space to maintain the architectural state of an SP. */
+ cpu_context_t cpu_ctx;
+
+ /* Track the current runtime state of the SP. */
+ enum sp_runtime_states rt_state;
+
+ /* Track the current runtime model of the SP. */
+ enum sp_runtime_model rt_model;
+};
+
+/*
+ * Structure to describe the cumulative properties of an SP.
+ */
+struct secure_partition_desc {
+ /*
+ * Execution contexts allocated to this endpoint. Ideally,
+ * we need as many contexts as there are physical cpus only
+ * for a S-EL1 SP which is MP-pinned.
+ */
+ struct sp_exec_ctx ec[PLATFORM_CORE_COUNT];
+
+ /* ID of the Secure Partition. */
+ uint16_t sp_id;
+
+ /* Runtime EL. */
+ enum sp_runtime_el runtime_el;
+
+ /* Partition UUID. */
+ uint32_t uuid[4];
+
+ /* Partition Properties. */
+ uint32_t properties;
+
+ /* Supported FF-A Version. */
+ uint32_t ffa_version;
+
+ /* Execution State. */
+ enum sp_execution_state execution_state;
+
+ /* Secondary entrypoint. Only valid for a S-EL1 SP. */
+ uintptr_t secondary_ep;
+};
+
+/*
+ * This define identifies the only SP that will be initialised and participate
+ * in FF-A communication. The implementation leaves the door open for more SPs
+ * to be managed in future but for now it is reasonable to assume that either a
+ * single S-EL0 or a single S-EL1 SP will be supported. This define will be used
+ * to identify which SP descriptor to initialise and manage during SP runtime.
+ */
+#define ACTIVE_SP_DESC_INDEX 0
+
+/*
+ * Structure to describe the cumulative properties of the Hypervisor and
+ * NS-Endpoints.
+ */
+struct ns_endpoint_desc {
+ /*
+ * ID of the NS-Endpoint or Hypervisor.
+ */
+ uint16_t ns_ep_id;
+
+ /*
+ * Supported FF-A Version.
+ */
+ uint32_t ffa_version;
+};
+
+/* Setup Function for different SP types. */
+void spmc_sp_common_setup(struct secure_partition_desc *sp,
+ entry_point_info_t *ep_info);
+void spmc_el1_sp_setup(struct secure_partition_desc *sp,
+ entry_point_info_t *ep_info);
+void spmc_sp_common_ep_commit(struct secure_partition_desc *sp,
+ entry_point_info_t *ep_info);
+
+/*
+ * Helper function to perform a synchronous entry into a SP.
+ */
+uint64_t spmc_sp_synchronous_entry(struct sp_exec_ctx *ec);
+
+/*
+ * Helper function to obtain the descriptor of the current SP on a physical cpu.
+ */
+struct secure_partition_desc *spmc_get_current_sp_ctx(void);
+
+/*
+ * Helper function to obtain the execution context of an SP on a
+ * physical cpu.
+ */
+struct sp_exec_ctx *spmc_get_sp_ec(struct secure_partition_desc *sp);
+
+/*
+ * Helper function to obtain the index of the execution context of an SP on a
+ * physical cpu.
+ */
+unsigned int get_ec_index(struct secure_partition_desc *sp);
+
+uint64_t spmc_ffa_error_return(void *handle, int error_code);
+
+/*
+ * Ensure a partition ID does not clash and follows the secure world convention.
+ */
+bool is_ffa_secure_id_valid(uint16_t partition_id);
+
+#endif /* SPMC_H */