soc: nordic: common: Add mram latency manager

Add module for managing multiple requests for MRAM latency.

Signed-off-by: Krzysztof Chruściński <krzysztof.chruscinski@nordicsemi.no>

soc/nordic/common/CMakeLists.txt

@@ -34,3 +34,4 @@ if(CONFIG_TFM_PARTITION_PLATFORM)
 endif()
 
 zephyr_library_sources_ifdef(CONFIG_NRF_SYS_EVENT nrf_sys_event.c)
+zephyr_library_sources_ifdef(CONFIG_MRAM_LATENCY mram_latency.c)

soc/nordic/common/Kconfig

@@ -8,4 +8,24 @@ config NRF_SYS_EVENT
 	bool "nRF system event support"
 	select NRFX_POWER if !NRF_PLATFORM_HALTIUM
 
+config MRAM_LATENCY
+	bool "MRAM latency manager"
+	depends on NRFS_HAS_MRAM_SERVICE
+	select ONOFF
+	select NRFS_MRAM_SERVICE_ENABLED
+
+if MRAM_LATENCY
+
+config MRAM_LATENCY_SYNC_TIMEOUT
+	int "Timeout in synchronous request"
+	default 1000
+	help
+	  Timeout is given in milliseconds.
+
+module = MRAM_LATENCY
+module-str = mram_latency
+source "subsys/logging/Kconfig.template.log_config"
+
+endif # MRAM_LATENCY
+
 rsource "vpr/Kconfig"

soc/nordic/common/mram_latency.c

@@ -0,0 +1,175 @@
+/*
+ * Copyright (c) 2024 Nordic Semiconductor ASA
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+#include <mram_latency.h>
+#include <zephyr/kernel.h>
+#include <nrfs_mram.h>
+#include <nrfs_backend_ipc_service.h>
+#include <zephyr/logging/log.h>
+LOG_MODULE_REGISTER(mram_latency, CONFIG_MRAM_LATENCY_LOG_LEVEL);
+
+enum mram_latency_state {
+	MRAM_LATENCY_OFF = 0,
+	MRAM_LATENCY_OFF_PENDING,
+	MRAM_LATENCY_ON,
+};
+
+static struct k_work work;
+static bool no_latency;
+static enum mram_latency_state state;
+
+static onoff_notify_fn onoff_notify;
+struct onoff_manager mram_latency_mgr;
+
+struct sync_latency_req {
+	struct onoff_client cli;
+	struct k_sem sem;
+	int res;
+};
+
+static void latency_change_req(bool latency_not_allowed)
+{
+	nrfs_err_t err;
+
+	if (latency_not_allowed) {
+		err = nrfs_mram_set_latency(MRAM_LATENCY_NOT_ALLOWED, NULL);
+		if (err != NRFS_SUCCESS) {
+			onoff_notify(&mram_latency_mgr, -EIO);
+		}
+	} else {
+		/* There is no event for that setting so we can notify onoff manager
+		 * immediately.
+		 */
+		err = nrfs_mram_set_latency(MRAM_LATENCY_ALLOWED, NULL);
+		onoff_notify(&mram_latency_mgr, err != NRFS_SUCCESS ? -EIO : 0);
+	}
+}
+
+static void latency_change(bool latency_not_allowed)
+{
+	LOG_DBG("Request: latency %s allowed", latency_not_allowed ? "not " : "");
+	if (state == MRAM_LATENCY_OFF) {
+		state = MRAM_LATENCY_OFF_PENDING;
+	} else if (k_is_in_isr()) {
+		/* nrfs cannot be called from interrupt context so defer to the work
+		 * queue context and execute from there.
+		 */
+		no_latency = latency_not_allowed;
+		k_work_submit(&work);
+	} else {
+		latency_change_req(latency_not_allowed);
+	}
+}
+
+static void no_latency_start(struct onoff_manager *mgr, onoff_notify_fn notify)
+{
+	onoff_notify = notify;
+	latency_change(true);
+}
+
+static void no_latency_stop(struct onoff_manager *mgr, onoff_notify_fn notify)
+{
+	latency_change(false);
+}
+
+static void evt_handler(nrfs_mram_latency_evt_t const *p_evt, void *context)
+{
+	int res = p_evt->type == NRFS_MRAM_LATENCY_REQ_APPLIED ? 0 : -EIO;
+
+	LOG_DBG("Latency not allowed - applied");
+	onoff_notify(&mram_latency_mgr, res);
+}
+
+static void work_handler(struct k_work *work)
+{
+	latency_change_req(no_latency);
+}
+
+int mram_no_latency_cancel_or_release(struct onoff_client *cli)
+{
+	return onoff_cancel_or_release(&mram_latency_mgr, cli);
+}
+
+int mram_no_latency_request(struct onoff_client *cli)
+{
+	return onoff_request(&mram_latency_mgr, cli);
+}
+
+static void sync_req_cb(struct onoff_manager *mgr, struct onoff_client *cli, uint32_t state,
+			int res)
+{
+	struct sync_latency_req *req = CONTAINER_OF(cli, struct sync_latency_req, cli);
+
+	req->res = res;
+	k_sem_give(&req->sem);
+}
+
+int mram_no_latency_sync_request(void)
+{
+	struct sync_latency_req req;
+	int rv;
+
+	if (k_is_in_isr() || (state != MRAM_LATENCY_ON)) {
+		return -ENOTSUP;
+	}
+
+	k_sem_init(&req.sem, 0, 1);
+	sys_notify_init_callback(&req.cli.notify, sync_req_cb);
+	rv = onoff_request(&mram_latency_mgr, &req.cli);
+	if (rv < 0) {
+		return rv;
+	}
+
+	rv = k_sem_take(&req.sem, K_MSEC(CONFIG_MRAM_LATENCY_SYNC_TIMEOUT));
+	if (rv < 0) {
+		return rv;
+	}
+
+	return req.res;
+}
+
+int mram_no_latency_sync_release(void)
+{
+	return onoff_release(&mram_latency_mgr) >= 0 ? 0 : -EIO;
+}
+
+/* First initialize onoff manager to be able to accept requests. */
+static int init_manager(void)
+{
+	static const struct onoff_transitions transitions =
+		ONOFF_TRANSITIONS_INITIALIZER(no_latency_start, no_latency_stop, NULL);
+
+	return onoff_manager_init(&mram_latency_mgr, &transitions);
+}
+
+/* When kernel and IPC is running initialize nrfs. Optionally, execute pending request. */
+static int init_nrfs(void)
+{
+	nrfs_err_t err;
+	int rv;
+
+	err = nrfs_backend_wait_for_connection(K_FOREVER);
+	if (err != NRFS_SUCCESS) {
+		return -EIO;
+	}
+
+	err = nrfs_mram_init(evt_handler);
+	if (err != NRFS_SUCCESS) {
+		return -EIO;
+	}
+
+	k_work_init(&work, work_handler);
+
+	if (state == MRAM_LATENCY_OFF_PENDING) {
+		latency_change(true);
+	}
+
+	state = MRAM_LATENCY_ON;
+
+	return rv;
+}
+
+SYS_INIT(init_manager, PRE_KERNEL_1, 0);
+SYS_INIT(init_nrfs, APPLICATION, CONFIG_APPLICATION_INIT_PRIORITY);

soc/nordic/common/mram_latency.h

@@ -0,0 +1,91 @@
+/*
+ * Copyright (c) 2024 Nordic Semiconductor ASA
+ * SPDX-License-Identifier: Apache-2.0
+ */
+
+/**
+ * @file
+ * nRF SoC specific public APIs for MRAM latency management
+ * @brief Experimental. It will be replaced by the PM latency policy API in the future.
+ */
+
+#ifndef SOC_NORDIC_COMMON_MRAM_LATENCY_H_
+#define SOC_NORDIC_COMMON_MRAM_LATENCY_H_
+
+#include <zephyr/sys/onoff.h>
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/** @internal For test purposes only. */
+extern struct onoff_manager mram_latency_mgr;
+
+/** @brief Request MRAM operations without latency.
+ *
+ * The return value indicates the success or failure of an attempt to initiate
+ * an operation to request the MRAM low latency. If initiation of the
+ * operation succeeds, the result of the request operation is provided through
+ * the configured client notification method, possibly before this call returns.
+ *
+ * @param cli pointer to client state providing instructions on synchronous
+ *            expectations and how to notify the client when the request
+ *            completes. Behavior is undefined if client passes a pointer
+ *            object associated with an incomplete service operation.
+ *
+ * @retval non-negative the observed state of the on-off service associated
+ *                      with the MRAM latency service.
+ * @retval -EIO if MRAM latency service returned error.
+ * @retval -EINVAL if the parameters are invalid.
+ * @retval -EAGAIN if the reference count would overflow.
+ */
+int mram_no_latency_request(struct onoff_client *cli);
+
+/** @brief Request MRAM operations without latency.
+ *
+ * Request is synchronous and blocks until it is completed. It can be called only
+ * from the thread context and cannot be called in the pre kernel stage.
+ *
+ * @retval 0 on successful request.
+ * @retval -EIO if MRAM latency service returned error.
+ * @retval -EAGAIN if request was not completed on time.
+ */
+int mram_no_latency_sync_request(void);
+
+/**
+ * @brief Safely cancel a request for MRAM operations without latency.
+ *
+ * It may be that a client has issued a reservation request but needs to
+ * shut down before the request has completed. This function attempts to
+ * cancel the request and issues a release if cancellation fails because
+ * the request was completed. This synchronously ensures that ownership
+ * data reverts to the client so is available for a future request.
+ *
+ * @param cli a pointer to the same client state that was provided
+ *            when the operation to be cancelled was issued.
+ *
+ * @retval ONOFF_STATE_TO_ON if the cancellation occurred before the transition
+ *                           completed.
+ * @retval ONOFF_STATE_ON if the cancellation occurred after the transition
+ *                        completed.
+ * @retval -EINVAL if the parameters are invalid.
+ * @retval -EIO if MRAM latency service returned error.
+ * @retval negative other errors produced by onoff_release().
+ */
+int mram_no_latency_cancel_or_release(struct onoff_client *cli);
+
+/**
+ * @brief Release a request for MRAM operations without latency.
+ *
+ * It should match with a completed @ref mram_no_latency_sync_request call.
+ *
+ * @retval 0 on successful request.
+ * @retval -EIO if MRAM latency service returned error.
+ */
+int mram_no_latency_sync_release(void);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* SOC_NORDIC_COMMON_MRAM_LATENCY_H_ */