soc: qcom: smsm: Add driver for Qualcomm SMSM

This driver exposed the Qualcomm Shared Memory State Machine bits.
Signed-off-by: Bjorn Andersson <bjorn.andersson@sonymobile.com>
Signed-off-by: Andy Gross <agross@codeaurora.org>

drivers/soc/qcom/Kconfig

@@ -52,3 +52,11 @@ config QCOM_SMD_RPM
 
 config QCOM_SMEM_STATE
 	bool
+
+config QCOM_SMSM
+	bool "Qualcomm Shared Memory State Machine"
+	depends on QCOM_SMEM
+	select QCOM_SMEM_STATE
+	help
+	  Say yes here to support the Qualcomm Shared Memory State Machine.
+	  The state machine is represented by bits in shared memory.

drivers/soc/qcom/Makefile

@@ -4,3 +4,4 @@ obj-$(CONFIG_QCOM_SMD) +=	smd.o
 obj-$(CONFIG_QCOM_SMD_RPM)	+= smd-rpm.o
 obj-$(CONFIG_QCOM_SMEM) +=	smem.o
 obj-$(CONFIG_QCOM_SMEM_STATE) += smem_state.o
+obj-$(CONFIG_QCOM_SMSM)	+= smsm.o

drivers/soc/qcom/smsm.c

+/*
+ * Copyright (c) 2015, Sony Mobile Communications Inc.
+ * Copyright (c) 2012-2013, The Linux Foundation. All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 and
+ * only version 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+#include <linux/interrupt.h>
+#include <linux/mfd/syscon.h>
+#include <linux/module.h>
+#include <linux/of_irq.h>
+#include <linux/platform_device.h>
+#include <linux/spinlock.h>
+#include <linux/regmap.h>
+#include <linux/soc/qcom/smem.h>
+#include <linux/soc/qcom/smem_state.h>
+
+/*
+ * This driver implements the Qualcomm Shared Memory State Machine, a mechanism
+ * for communicating single bit state information to remote processors.
+ *
+ * The implementation is based on two sections of shared memory; the first
+ * holding the state bits and the second holding a matrix of subscription bits.
+ *
+ * The state bits are structured in entries of 32 bits, each belonging to one
+ * system in the SoC. The entry belonging to the local system is considered
+ * read-write, while the rest should be considered read-only.
+ *
+ * The subscription matrix consists of N bitmaps per entry, denoting interest
+ * in updates of the entry for each of the N hosts. Upon updating a state bit
+ * each host's subscription bitmap should be queried and the remote system
+ * should be interrupted if they request so.
+ *
+ * The subscription matrix is laid out in entry-major order:
+ * entry0: [host0 ... hostN]
+ *	.
+ *	.
+ * entryM: [host0 ... hostN]
+ *
+ * A third, optional, shared memory region might contain information regarding
+ * the number of entries in the state bitmap as well as number of columns in
+ * the subscription matrix.
+ */
+
+/*
+ * Shared memory identifiers, used to acquire handles to respective memory
+ * region.
+ */
+#define SMEM_SMSM_SHARED_STATE		85
+#define SMEM_SMSM_CPU_INTR_MASK		333
+#define SMEM_SMSM_SIZE_INFO		419
+
+/*
+ * Default sizes, in case SMEM_SMSM_SIZE_INFO is not found.
+ */
+#define SMSM_DEFAULT_NUM_ENTRIES	8
+#define SMSM_DEFAULT_NUM_HOSTS		3
+
+struct smsm_entry;
+struct smsm_host;
+
+/**
+ * struct qcom_smsm - smsm driver context
+ * @dev:	smsm device pointer
+ * @local_host:	column in the subscription matrix representing this system
+ * @num_hosts:	number of columns in the subscription matrix
+ * @num_entries: number of entries in the state map and rows in the subscription
+ *		matrix
+ * @local_state: pointer to the local processor's state bits
+ * @subscription: pointer to local processor's row in subscription matrix
+ * @state:	smem state handle
+ * @lock:	spinlock for read-modify-write of the outgoing state
+ * @entries:	context for each of the entries
+ * @hosts:	context for each of the hosts
+ */
+struct qcom_smsm {
+	struct device *dev;
+
+	u32 local_host;
+
+	u32 num_hosts;
+	u32 num_entries;
+
+	u32 *local_state;
+	u32 *subscription;
+	struct qcom_smem_state *state;
+
+	spinlock_t lock;
+
+	struct smsm_entry *entries;
+	struct smsm_host *hosts;
+};
+
+/**
+ * struct smsm_entry - per remote processor entry context
+ * @smsm:	back-reference to driver context
+ * @domain:	IRQ domain for this entry, if representing a remote system
+ * @irq_enabled: bitmap of which state bits IRQs are enabled
+ * @irq_rising:	bitmap tracking if rising bits should be propagated
+ * @irq_falling: bitmap tracking if falling bits should be propagated
+ * @last_value:	snapshot of state bits last time the interrupts where propagated
+ * @remote_state: pointer to this entry's state bits
+ * @subscription: pointer to a row in the subscription matrix representing this
+ *		entry
+ */
+struct smsm_entry {
+	struct qcom_smsm *smsm;
+
+	struct irq_domain *domain;
+	DECLARE_BITMAP(irq_enabled, 32);
+	DECLARE_BITMAP(irq_rising, 32);
+	DECLARE_BITMAP(irq_falling, 32);
+	u32 last_value;
+
+	u32 *remote_state;
+	u32 *subscription;
+};
+
+/**
+ * struct smsm_host - representation of a remote host
+ * @ipc_regmap:	regmap for outgoing interrupt
+ * @ipc_offset:	offset in @ipc_regmap for outgoing interrupt
+ * @ipc_bit:	bit in @ipc_regmap + @ipc_offset for outgoing interrupt
+ */
+struct smsm_host {
+	struct regmap *ipc_regmap;
+	int ipc_offset;
+	int ipc_bit;
+};
+
+/**
+ * smsm_update_bits() - change bit in outgoing entry and inform subscribers
+ * @data:	smsm context pointer
+ * @offset:	bit in the entry
+ * @value:	new value
+ *
+ * Used to set and clear the bits in the outgoing/local entry and inform
+ * subscribers about the change.
+ */
+static int smsm_update_bits(void *data, u32 mask, u32 value)
+{
+	struct qcom_smsm *smsm = data;
+	struct smsm_host *hostp;
+	unsigned long flags;
+	u32 changes;
+	u32 host;
+	u32 orig;
+	u32 val;
+
+	spin_lock_irqsave(&smsm->lock, flags);
+
+	/* Update the entry */
+	val = orig = readl(smsm->local_state);
+	val &= ~mask;
+	val |= value;
+
+	/* Don't signal if we didn't change the value */
+	changes = val ^ orig;
+	if (!changes) {
+		spin_unlock_irqrestore(&smsm->lock, flags);
+		goto done;
+	}
+
+	/* Write out the new value */
+	writel(val, smsm->local_state);
+	spin_unlock_irqrestore(&smsm->lock, flags);
+
+	/* Make sure the value update is ordered before any kicks */
+	wmb();
+
+	/* Iterate over all hosts to check whom wants a kick */
+	for (host = 0; host < smsm->num_hosts; host++) {
+		hostp = &smsm->hosts[host];
+
+		val = readl(smsm->subscription + host);
+		if (val & changes && hostp->ipc_regmap) {
+			regmap_write(hostp->ipc_regmap,
+				     hostp->ipc_offset,
+				     BIT(hostp->ipc_bit));
+		}
+	}
+
+done:
+	return 0;
+}
+
+static const struct qcom_smem_state_ops smsm_state_ops = {
+	.update_bits = smsm_update_bits,
+};
+
+/**
+ * smsm_intr() - cascading IRQ handler for SMSM
+ * @irq:	unused
+ * @data:	entry related to this IRQ
+ *
+ * This function cascades an incoming interrupt from a remote system, based on
+ * the state bits and configuration.
+ */
+static irqreturn_t smsm_intr(int irq, void *data)
+{
+	struct smsm_entry *entry = data;
+	unsigned i;
+	int irq_pin;
+	u32 changed;
+	u32 val;
+
+	val = readl(entry->remote_state);
+	changed = val ^ entry->last_value;
+	entry->last_value = val;
+
+	for_each_set_bit(i, entry->irq_enabled, 32) {
+		if (!(changed & BIT(i)))
+			continue;
+
+		if (val & BIT(i)) {
+			if (test_bit(i, entry->irq_rising)) {
+				irq_pin = irq_find_mapping(entry->domain, i);
+				handle_nested_irq(irq_pin);
+			}
+		} else {
+			if (test_bit(i, entry->irq_falling)) {
+				irq_pin = irq_find_mapping(entry->domain, i);
+				handle_nested_irq(irq_pin);
+			}
+		}
+	}
+
+	return IRQ_HANDLED;
+}
+
+/**
+ * smsm_mask_irq() - un-subscribe from cascades of IRQs of a certain staus bit
+ * @irqd:	IRQ handle to be masked
+ *
+ * This un-subscribes the local CPU from interrupts upon changes to the defines
+ * status bit. The bit is also cleared from cascading.
+ */
+static void smsm_mask_irq(struct irq_data *irqd)
+{
+	struct smsm_entry *entry = irq_data_get_irq_chip_data(irqd);
+	irq_hw_number_t irq = irqd_to_hwirq(irqd);
+	struct qcom_smsm *smsm = entry->smsm;
+	u32 val;
+
+	if (entry->subscription) {
+		val = readl(entry->subscription + smsm->local_host);
+		val &= ~BIT(irq);
+		writel(val, entry->subscription + smsm->local_host);
+	}
+
+	clear_bit(irq, entry->irq_enabled);
+}
+
+/**
+ * smsm_unmask_irq() - subscribe to cascades of IRQs of a certain status bit
+ * @irqd:	IRQ handle to be unmasked
+ *
+
+ * This subscribes the local CPU to interrupts upon changes to the defined
+ * status bit. The bit is also marked for cascading.
+
+ */
+static void smsm_unmask_irq(struct irq_data *irqd)
+{
+	struct smsm_entry *entry = irq_data_get_irq_chip_data(irqd);
+	irq_hw_number_t irq = irqd_to_hwirq(irqd);
+	struct qcom_smsm *smsm = entry->smsm;
+	u32 val;
+
+	set_bit(irq, entry->irq_enabled);
+
+	if (entry->subscription) {
+		val = readl(entry->subscription + smsm->local_host);
+		val |= BIT(irq);
+		writel(val, entry->subscription + smsm->local_host);
+	}
+}
+
+/**
+ * smsm_set_irq_type() - updates the requested IRQ type for the cascading
+ * @irqd:	consumer interrupt handle
+ * @type:	requested flags
+ */
+static int smsm_set_irq_type(struct irq_data *irqd, unsigned int type)
+{
+	struct smsm_entry *entry = irq_data_get_irq_chip_data(irqd);
+	irq_hw_number_t irq = irqd_to_hwirq(irqd);
+
+	if (!(type & IRQ_TYPE_EDGE_BOTH))
+		return -EINVAL;
+
+	if (type & IRQ_TYPE_EDGE_RISING)
+		set_bit(irq, entry->irq_rising);
+	else
+		clear_bit(irq, entry->irq_rising);
+
+	if (type & IRQ_TYPE_EDGE_FALLING)
+		set_bit(irq, entry->irq_falling);
+	else
+		clear_bit(irq, entry->irq_falling);
+
+	return 0;
+}
+
+static struct irq_chip smsm_irq_chip = {
+	.name           = "smsm",
+	.irq_mask       = smsm_mask_irq,
+	.irq_unmask     = smsm_unmask_irq,
+	.irq_set_type	= smsm_set_irq_type,
+};
+
+/**
+ * smsm_irq_map() - sets up a mapping for a cascaded IRQ
+ * @d:		IRQ domain representing an entry
+ * @irq:	IRQ to set up
+ * @hw:		unused
+ */
+static int smsm_irq_map(struct irq_domain *d,
+			unsigned int irq,
+			irq_hw_number_t hw)
+{
+	struct smsm_entry *entry = d->host_data;
+
+	irq_set_chip_and_handler(irq, &smsm_irq_chip, handle_level_irq);
+	irq_set_chip_data(irq, entry);
+	irq_set_nested_thread(irq, 1);
+
+	return 0;
+}
+
+static const struct irq_domain_ops smsm_irq_ops = {
+	.map = smsm_irq_map,
+	.xlate = irq_domain_xlate_twocell,
+};
+
+/**
+ * smsm_parse_ipc() - parses a qcom,ipc-%d device tree property
+ * @smsm:	smsm driver context
+ * @host_id:	index of the remote host to be resolved
+ *
+ * Parses device tree to acquire the information needed for sending the
+ * outgoing interrupts to a remote host - identified by @host_id.
+ */
+static int smsm_parse_ipc(struct qcom_smsm *smsm, unsigned host_id)
+{
+	struct device_node *syscon;
+	struct device_node *node = smsm->dev->of_node;
+	struct smsm_host *host = &smsm->hosts[host_id];
+	char key[16];
+	int ret;
+
+	snprintf(key, sizeof(key), "qcom,ipc-%d", host_id);
+	syscon = of_parse_phandle(node, key, 0);
+	if (!syscon)
+		return 0;
+
+	host->ipc_regmap = syscon_node_to_regmap(syscon);
+	if (IS_ERR(host->ipc_regmap))
+		return PTR_ERR(host->ipc_regmap);
+
+	ret = of_property_read_u32_index(node, key, 1, &host->ipc_offset);
+	if (ret < 0) {
+		dev_err(smsm->dev, "no offset in %s\n", key);
+		return -EINVAL;
+	}
+
+	ret = of_property_read_u32_index(node, key, 2, &host->ipc_bit);
+	if (ret < 0) {
+		dev_err(smsm->dev, "no bit in %s\n", key);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+/**
+ * smsm_inbound_entry() - parse DT and set up an entry representing a remote system
+ * @smsm:	smsm driver context
+ * @entry:	entry context to be set up
+ * @node:	dt node containing the entry's properties
+ */
+static int smsm_inbound_entry(struct qcom_smsm *smsm,
+			      struct smsm_entry *entry,
+			      struct device_node *node)
+{
+	int ret;
+	int irq;
+
+	irq = irq_of_parse_and_map(node, 0);
+	if (!irq) {
+		dev_err(smsm->dev, "failed to parse smsm interrupt\n");
+		return -EINVAL;
+	}
+
+	ret = devm_request_threaded_irq(smsm->dev, irq,
+					NULL, smsm_intr,
+					IRQF_ONESHOT,
+					"smsm", (void *)entry);
+	if (ret) {
+		dev_err(smsm->dev, "failed to request interrupt\n");
+		return ret;
+	}
+
+	entry->domain = irq_domain_add_linear(node, 32, &smsm_irq_ops, entry);
+	if (!entry->domain) {
+		dev_err(smsm->dev, "failed to add irq_domain\n");
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+
+/**
+ * smsm_get_size_info() - parse the optional memory segment for sizes
+ * @smsm:	smsm driver context
+ *
+ * Attempt to acquire the number of hosts and entries from the optional shared
+ * memory location. Not being able to find this segment should indicate that
+ * we're on a older system where these values was hard coded to
+ * SMSM_DEFAULT_NUM_ENTRIES and SMSM_DEFAULT_NUM_HOSTS.
+ *
+ * Returns 0 on success, negative errno on failure.
+ */
+static int smsm_get_size_info(struct qcom_smsm *smsm)
+{
+	size_t size;
+	struct {
+		u32 num_hosts;
+		u32 num_entries;
+		u32 reserved0;
+		u32 reserved1;
+	} *info;
+
+	info = qcom_smem_get(QCOM_SMEM_HOST_ANY, SMEM_SMSM_SIZE_INFO, &size);
+	if (PTR_ERR(info) == -ENOENT || size != sizeof(*info)) {
+		dev_warn(smsm->dev, "no smsm size info, using defaults\n");
+		smsm->num_entries = SMSM_DEFAULT_NUM_ENTRIES;
+		smsm->num_hosts = SMSM_DEFAULT_NUM_HOSTS;
+		return 0;
+	} else if (IS_ERR(info)) {
+		dev_err(smsm->dev, "unable to retrieve smsm size info\n");
+		return PTR_ERR(info);
+	}
+
+	smsm->num_entries = info->num_entries;
+	smsm->num_hosts = info->num_hosts;
+
+	dev_dbg(smsm->dev,
+		"found custom size of smsm: %d entries %d hosts\n",
+		smsm->num_entries, smsm->num_hosts);
+
+	return 0;
+}
+
+static int qcom_smsm_probe(struct platform_device *pdev)
+{
+	struct device_node *local_node;
+	struct device_node *node;
+	struct smsm_entry *entry;
+	struct qcom_smsm *smsm;
+	u32 *intr_mask;
+	size_t size;
+	u32 *states;
+	u32 id;
+	int ret;
+
+	smsm = devm_kzalloc(&pdev->dev, sizeof(*smsm), GFP_KERNEL);
+	if (!smsm)
+		return -ENOMEM;
+	smsm->dev = &pdev->dev;
+	spin_lock_init(&smsm->lock);
+
+	ret = smsm_get_size_info(smsm);
+	if (ret)
+		return ret;
+
+	smsm->entries = devm_kcalloc(&pdev->dev,
+				     smsm->num_entries,
+				     sizeof(struct smsm_entry),
+				     GFP_KERNEL);
+	if (!smsm->entries)
+		return -ENOMEM;
+
+	smsm->hosts = devm_kcalloc(&pdev->dev,
+				   smsm->num_hosts,
+				   sizeof(struct smsm_host),
+				   GFP_KERNEL);
+	if (!smsm->hosts)
+		return -ENOMEM;
+
+	local_node = of_find_node_with_property(pdev->dev.of_node, "#qcom,state-cells");
+	if (!local_node) {
+		dev_err(&pdev->dev, "no state entry\n");
+		return -EINVAL;
+	}
+
+	of_property_read_u32(pdev->dev.of_node,
+			     "qcom,local-host",
+			     &smsm->local_host);
+
+	/* Parse the host properties */
+	for (id = 0; id < smsm->num_hosts; id++) {
+		ret = smsm_parse_ipc(smsm, id);
+		if (ret < 0)
+			return ret;
+	}
+
+	/* Acquire the main SMSM state vector */
+	ret = qcom_smem_alloc(QCOM_SMEM_HOST_ANY, SMEM_SMSM_SHARED_STATE,
+			      smsm->num_entries * sizeof(u32));
+	if (ret < 0 && ret != -EEXIST) {
+		dev_err(&pdev->dev, "unable to allocate shared state entry\n");
+		return ret;
+	}
+
+	states = qcom_smem_get(QCOM_SMEM_HOST_ANY, SMEM_SMSM_SHARED_STATE, NULL);
+	if (IS_ERR(states)) {
+		dev_err(&pdev->dev, "Unable to acquire shared state entry\n");
+		return PTR_ERR(states);
+	}
+
+	/* Acquire the list of interrupt mask vectors */
+	size = smsm->num_entries * smsm->num_hosts * sizeof(u32);
+	ret = qcom_smem_alloc(QCOM_SMEM_HOST_ANY, SMEM_SMSM_CPU_INTR_MASK, size);
+	if (ret < 0 && ret != -EEXIST) {
+		dev_err(&pdev->dev, "unable to allocate smsm interrupt mask\n");
+		return ret;
+	}
+
+	intr_mask = qcom_smem_get(QCOM_SMEM_HOST_ANY, SMEM_SMSM_CPU_INTR_MASK, NULL);
+	if (IS_ERR(intr_mask)) {
+		dev_err(&pdev->dev, "unable to acquire shared memory interrupt mask\n");
+		return PTR_ERR(intr_mask);
+	}
+
+	/* Setup the reference to the local state bits */
+	smsm->local_state = states + smsm->local_host;
+	smsm->subscription = intr_mask + smsm->local_host * smsm->num_hosts;
+
+	/* Register the outgoing state */
+	smsm->state = qcom_smem_state_register(local_node, &smsm_state_ops, smsm);
+	if (IS_ERR(smsm->state)) {
+		dev_err(smsm->dev, "failed to register qcom_smem_state\n");
+		return PTR_ERR(smsm->state);
+	}
+
+	/* Register handlers for remote processor entries of interest. */
+	for_each_available_child_of_node(pdev->dev.of_node, node) {
+		if (!of_property_read_bool(node, "interrupt-controller"))
+			continue;
+
+		ret = of_property_read_u32(node, "reg", &id);
+		if (ret || id >= smsm->num_entries) {
+			dev_err(&pdev->dev, "invalid reg of entry\n");
+			if (!ret)
+				ret = -EINVAL;
+			goto unwind_interfaces;
+		}
+		entry = &smsm->entries[id];
+
+		entry->smsm = smsm;
+		entry->remote_state = states + id;
+
+		/* Setup subscription pointers and unsubscribe to any kicks */
+		entry->subscription = intr_mask + id * smsm->num_hosts;
+		writel(0, entry->subscription + smsm->local_host);
+
+		ret = smsm_inbound_entry(smsm, entry, node);
+		if (ret < 0)
+			goto unwind_interfaces;
+	}
+
+	platform_set_drvdata(pdev, smsm);
+
+	return 0;
+
+unwind_interfaces:
+	for (id = 0; id < smsm->num_entries; id++)
+		if (smsm->entries[id].domain)
+			irq_domain_remove(smsm->entries[id].domain);
+
+	qcom_smem_state_unregister(smsm->state);
+
+	return ret;
+}
+
+static int qcom_smsm_remove(struct platform_device *pdev)
+{
+	struct qcom_smsm *smsm = platform_get_drvdata(pdev);
+	unsigned id;
+
+	for (id = 0; id < smsm->num_entries; id++)
+		if (smsm->entries[id].domain)
+			irq_domain_remove(smsm->entries[id].domain);
+
+	qcom_smem_state_unregister(smsm->state);
+
+	return 0;
+}
+
+static const struct of_device_id qcom_smsm_of_match[] = {
+	{ .compatible = "qcom,smsm" },
+	{}
+};
+MODULE_DEVICE_TABLE(of, qcom_smsm_of_match);
+
+static struct platform_driver qcom_smsm_driver = {
+	.probe = qcom_smsm_probe,
+	.remove = qcom_smsm_remove,
+	.driver  = {
+		.name  = "qcom-smsm",
+		.of_match_table = qcom_smsm_of_match,
+	},
+};
+module_platform_driver(qcom_smsm_driver);
+
+MODULE_DESCRIPTION("Qualcomm Shared Memory State Machine driver");
+MODULE_LICENSE("GPL v2");