@@ -23,6 +23,21 @@ config RPMON
Currently RPMON_QMI is available which uses QMI infrastructures
on Qualcomm SoC Platforms.
+config RPMON_QMI
+ tristate "RPMON QMI Driver Engine"
+ select RPMON_QMI_MSG_V1
+ depends on RPMON
+ depends on QCOM_QMI_HELPERS
+ help
+ RPMON_QMI is used by RPMON to communicate with remote processors
+ with QMI APIs if enabled. RPMON_QMI itself is designed as a modular
+ framework that would introduce different kinds of message sets
+ which may be updated for versions.
+
+ RPMON_QMI creates a device of rpmon_device type for each remote
+ processor endpoint. All the endpoint devices shares an unique set
+ of QMI suite.
+
config RPMON_QMI_MSG_V1
tristate "RPMON QMI Message Version 1.0"
depends on RPMON
@@ -1,2 +1,3 @@
obj-$(CONFIG_RPMON) += rpmon.o
+obj-$(CONFIG_RPMON_QMI) += rpmon_qmi.o
obj-$(CONFIG_RPMON_QMI_MSG_V1) += rpmon_qmi_msg_v1.o
new file mode 100644
@@ -0,0 +1,431 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (C) 2020 Vivo Communication Technology Co. Ltd.
+ * Copyright (C) 2020 Wang Wenhu <wenhu.wang@vivo.com>
+ * All rights reserved.
+ *
+ * RPMON: An implementation of remote processor monitor framework
+ * for modern SoCs that typically have heterogeneous remote processor
+ * devices in asymmetric multiprocessing configurations. It is
+ * implemented with chardev and sysfs class, which act as interfaces
+ * to communicate with user level. It supports different communication
+ * interfaces added modularly to communicate with remote processors.
+ *
+ * RPMON_QMI: Implements a kind of communication routine for RPMON
+ * to communicate with remote processors through QMI infrastructure.
+ * At least one set of RPMON_QMI_MSG should be available and RPMON_QMI
+ * initiates with the message set(s) to provide certain servicei(s)
+ * like stability checking of remote processors. Currently a set of
+ * messages, implemented by RPMON_QMI_MSG_V1 is available.
+ */
+
+#include <linux/module.h>
+#include <linux/rpmon.h>
+#include <linux/soc/qcom/qmi.h>
+#include <linux/of_platform.h>
+#include <linux/nospec.h>
+#include "rpmon_qmi.h"
+
+#define DRIVER_NAME "rpmon_qmi_drv"
+
+/* Remote processor registered. */
+#define RP_REGISTERED 0x0001
+
+/* work struct for message processing. */
+struct recv_work {
+ struct work_struct work;
+ struct sockaddr_qrtr sq;
+ void *msg;
+};
+
+/* Delayed work to take a reset action when a failure is detected. */
+struct exec_cb_work {
+ struct delayed_work dwk;
+ struct rpmon_qmi_device *rdev;
+ u32 checks;
+};
+
+struct rpmon_qmi_exec_fn {
+ int (*exec_call)(struct rpmon_qmi_device *rdev);
+};
+
+struct rpmon_qmi_cb_fn {
+ void (*callback)(struct work_struct *work);
+ u32 msg_len;
+};
+
+static DEFINE_MUTEX(rdev_list_lock);
+static LIST_HEAD(rdev_list);
+static struct rpmon_qmi *rpqmi;
+static struct workqueue_struct *rpqmi_wq;
+
+static void rpmon_qmi_register_req_cb(struct work_struct *work)
+{
+ struct rpmon_register_req *req;
+ struct rpmon_qmi_device *rdev;
+ struct recv_work *rwk = container_of(work, struct recv_work, work);
+
+ req = (struct rpmon_register_req *)rwk->msg;
+
+ mutex_lock(&rdev_list_lock);
+ list_for_each_entry(rdev, &rdev_list, list) {
+ if (strncmp(rdev->info->name, req->name, RP_NAME_LEN))
+ continue;
+
+ rdev->flag |= RP_REGISTERED;
+ memcpy(&rdev->addr, &rwk->sq, sizeof(rwk->sq));
+ if (req->timeout_valid)
+ rdev->timeout = req->timeout;
+ else
+ rdev->timeout = 5000;
+ rpmon_event_notify(rdev->info, RPMON_EVENT_REGISTER);
+ break;
+ }
+ mutex_unlock(&rdev_list_lock);
+
+ kfree(rwk->msg);
+ kfree(rwk);
+}
+
+void rpmon_qmi_conn_check_resp_cb(struct work_struct *work)
+{
+ struct rpmon_conn_check_resp *cc_resp;
+ struct rpmon_qmi_device *rdev;
+ struct sockaddr_qrtr *addr;
+ struct recv_work *rwk =
+ container_of(work, struct recv_work, work);
+
+ cc_resp = (struct rpmon_conn_check_resp *)rwk->msg;
+ mutex_lock(&rdev_list_lock);
+ list_for_each_entry(rdev, &rdev_list, list) {
+ addr = &rdev->addr;
+ if (addr->sq_node != rwk->sq.sq_node ||
+ addr->sq_port != rwk->sq.sq_port)
+ continue;
+
+ if (!cc_resp->result.error)
+ atomic_inc(&rdev->reports);
+ break;
+ }
+ mutex_unlock(&rdev_list_lock);
+
+ kfree(rwk->msg);
+ kfree(rwk);
+}
+
+/**
+ * rpmon_qmi_exec_cb_worker - callback worker for execution
+ * @work: work to been done
+ *
+ * Called as worker handler by the single worker thread of rpmon_wq.
+ * The worker is scheduled after timeout ms duration since the execution.
+ */
+static void rpmon_qmi_exec_cb_worker(struct work_struct *work)
+{
+ struct delayed_work *dwk = to_delayed_work(work);
+ struct exec_cb_work *ewk =
+ container_of(dwk, struct exec_cb_work, dwk);
+ struct rpmon_qmi_device *rdev = ewk->rdev;
+
+ mutex_lock(&rdev_list_lock);
+ if (ewk->checks <= atomic_read(&rdev->reports)) {
+ pr_debug("%s health check success", rdev->info->name);
+ goto out;
+ }
+
+ pr_err("subsystem %s failed to respond in time", rdev->info->name);
+
+ rpmon_event_notify(rdev->info, RPMON_EVENT_CHKCONN_FAIL);
+
+out:
+ mutex_unlock(&rdev_list_lock);
+ kfree(ewk);
+}
+
+static struct rpmon_qmi_cb_fn rpmon_qmi_event_callbacks[] = {
+ {
+ .callback = rpmon_qmi_register_req_cb,
+ .msg_len = sizeof(struct rpmon_register_req),
+ },
+ {
+ .callback = rpmon_qmi_conn_check_resp_cb,
+ .msg_len = sizeof(struct rpmon_conn_check_resp),
+ },
+};
+
+/**
+ * rpmon_qmi_conn_check - send indication, initiate and queue callback work
+ * @rdev: device interface of specific remote processor to be checked
+ */
+static int rpmon_qmi_conn_check(struct rpmon_qmi_device *rdev)
+{
+ struct exec_cb_work *ewk;
+
+ mutex_lock(&rdev_list_lock);
+ if (!(rdev->flag & RP_REGISTERED)) {
+ pr_err("%s has not registered", rdev->info->name);
+ return -ENONET;
+ }
+
+ if (!__ratelimit(&rdev->ratelimit)) {
+ pr_err("%s rate-limited", rdev->info->name);
+ return 0;
+ }
+ mutex_unlock(&rdev_list_lock);
+
+ rdev->rqmi->sendmsg(rdev, NULL, 0);
+
+ ewk = kzalloc(sizeof(*ewk), GFP_KERNEL);
+ if (!ewk)
+ return -ENOMEM;
+
+ ewk->rdev = rdev;
+ ewk->checks = atomic_inc_return(&rdev->checks);
+ INIT_DELAYED_WORK(&ewk->dwk, rpmon_qmi_exec_cb_worker);
+ queue_delayed_work(rpqmi_wq,
+ &ewk->dwk, msecs_to_jiffies(rdev->timeout));
+
+ return 0;
+}
+
+static struct rpmon_qmi_exec_fn rpmon_qmi_exec_calls[] = {
+ {.exec_call = rpmon_qmi_conn_check},
+};
+
+static int rpmon_qmi_monitor(struct rpmon_info *info, u32 event)
+{
+ struct rpmon_qmi_device *rdev = (struct rpmon_qmi_device *)info->priv;
+ int i, idx;
+
+ for (i = 0; i < RPMON_EXEC_MAX; i++) {
+ if (event & RPMON_ACTION(i)) {
+ if (i < ARRAY_SIZE(rpmon_qmi_exec_calls)) {
+ idx = array_index_nospec(i, ARRAY_SIZE(rpmon_qmi_exec_calls));
+ if (rpmon_qmi_exec_calls[idx].exec_call)
+ return rpmon_qmi_exec_calls[idx].exec_call(rdev);
+ else
+ return -ENOTSUPP;
+ } else
+ return -ENOPARAM;
+ }
+ }
+
+ return -EINVAL;
+}
+
+static int rpmon_qmi_drv_probe(struct platform_device *pdev)
+{
+ struct rpmon_info *info = pdev->dev.platform_data;
+ struct rpmon_qmi_device *rdev;
+ struct device_node *node = pdev->dev.of_node;
+ const char *name;
+ int ret = -ENODEV;
+
+ if (node) {
+ /* Allocate info for one device */
+ info = devm_kzalloc(&pdev->dev, sizeof(*info), GFP_KERNEL);
+ if (!info) {
+ ret = -ENOMEM;
+ goto out;
+ }
+
+ if (!of_property_read_string(node, "linux,subsys", &name))
+ info->name = devm_kstrdup(&pdev->dev, name, GFP_KERNEL);
+ else
+ info->name = devm_kasprintf(&pdev->dev, GFP_KERNEL,
+ "%pOFn", node);
+ info->version = "devicetree";
+ }
+
+ if (!info || !info->name || !info->version) {
+ dev_dbg(&pdev->dev, "%s: err_info\n", __func__);
+ return ret;
+ }
+
+ /* Allocate device for qmi specific reference */
+ rdev = devm_kzalloc(&pdev->dev, sizeof(*rdev), GFP_KERNEL);
+ if (!rdev) {
+ ret = -ENOMEM;
+ goto err_info_free;
+ }
+
+ rdev->rqmi = rpqmi;
+ rdev->info = info;
+ info->priv = rdev;
+ info->monitor = rpmon_qmi_monitor;
+ platform_set_drvdata(pdev, rdev);
+
+ ret = rpmon_register_device(&pdev->dev, info);
+ if (ret) {
+ dev_err(&pdev->dev, "unable to register rpmon_qmi_device\n");
+ goto err_rdev_free;
+ }
+
+ mutex_lock(&rdev_list_lock);
+ list_add_tail(&rdev->list, &rdev_list);
+ mutex_unlock(&rdev_list_lock);
+
+ return ret;
+
+err_rdev_free:
+ devm_kfree(&pdev->dev, rdev);
+err_info_free:
+ devm_kfree(&pdev->dev, info);
+out:
+ return ret;
+}
+
+static int rpmon_qmi_drv_remove(struct platform_device *pdev)
+{
+ struct rpmon_qmi_device *rdev = platform_get_drvdata(pdev);
+
+ rpmon_unregister_device(rdev->info);
+
+ return 0;
+}
+
+static void rpmon_qmi_msg_callback(enum rpmon_qmi_msg_type type,
+ struct sockaddr_qrtr *sq,
+ const void *msg)
+{
+ struct recv_work *rwk;
+
+ if (type >= ARRAY_SIZE(rpmon_qmi_event_callbacks)) {
+ pr_err("Error non-supported message type.\n");
+ return;
+ }
+
+ if (rpmon_qmi_event_callbacks[type].callback) {
+ rwk = kzalloc(sizeof(*rwk), GFP_KERNEL);
+ if (!rwk) {
+ pr_err("Error to alloc recv_work");
+ return;
+ }
+
+ INIT_WORK(&rwk->work, rpmon_qmi_event_callbacks[type].callback);
+ memcpy(&rwk->sq, sq, sizeof(*sq));
+
+ rwk->msg = kzalloc(rpmon_qmi_event_callbacks[type].msg_len,
+ GFP_KERNEL);
+ if (!rwk->msg) {
+ pr_err("Error to alloc message of recv_work");
+ kfree(rwk);
+ return;
+ }
+
+ memcpy(rwk->msg, msg, rpmon_qmi_event_callbacks[type].msg_len);
+ queue_work(rpqmi_wq, &rwk->work);
+ }
+}
+
+static const struct of_device_id rpmon_of_qmi_match[] = {
+ { .compatible = "rpmon-qmi" },
+ { /* Sentinel */ },
+};
+
+static struct platform_driver rpmon_qmi_drv = {
+ .probe = rpmon_qmi_drv_probe,
+ .remove = rpmon_qmi_drv_remove,
+ .driver = {
+ .name = DRIVER_NAME,
+ .owner = THIS_MODULE,
+ .of_match_table = of_match_ptr(rpmon_of_qmi_match),
+ },
+};
+
+static int __init rpmon_qmi_drv_init(void)
+{
+ int ret;
+
+ rpqmi_wq = create_singlethread_workqueue("rpmon_qmi_wq");
+ if (!rpqmi_wq) {
+ pr_err("Error creating workqueue\n");
+ ret = -EFAULT;
+ goto out;
+ }
+
+ rpqmi = kzalloc(sizeof(*rpqmi), GFP_KERNEL);
+ if (!rpqmi) {
+ ret = -ENOMEM;
+ goto err_wq_free;
+ }
+
+ ret = rpmon_qmi_handle_init(rpqmi, rpmon_qmi_msg_callback);
+ if (ret)
+ goto err_rpqmi_free;
+
+ ret = qmi_handle_init(&rpqmi->qmi,
+ RPQMI_BUF_SIZE, NULL, rpqmi->handlers);
+ if (ret < 0) {
+ pr_err("Error init qmi handle, %d", ret);
+ goto err_rpqmi_free;
+ }
+
+ ret = qmi_add_server(&rpqmi->qmi,
+ rpqmi->svc->service,
+ rpqmi->svc->version,
+ rpqmi->svc->instance);
+ if (ret < 0) {
+ pr_err("Error add qmi server, %d", ret);
+ goto err_rpqmi_free;
+ }
+ mutex_init(&rdev_list_lock);
+
+ return platform_driver_register(&rpmon_qmi_drv);
+
+err_rpqmi_free:
+ kfree(rpqmi);
+err_wq_free:
+ destroy_workqueue(rpqmi_wq);
+out:
+ return ret;
+}
+late_initcall_sync(rpmon_qmi_drv_init);
+
+static void rpmon_qmi_del_server(void)
+{
+ struct qrtr_ctrl_pkt pkt;
+ struct sockaddr_qrtr sq;
+ struct msghdr msg = { };
+ struct kvec iv = { &pkt, sizeof(pkt) };
+ struct qmi_service *svc = rpqmi->svc;
+ struct qmi_handle *qmi = &rpqmi->qmi;
+ int ret;
+
+ memset(&pkt, 0, sizeof(pkt));
+ pkt.cmd = cpu_to_le32(QRTR_TYPE_DEL_SERVER);
+ pkt.server.service = cpu_to_le32(svc->service);
+ pkt.server.instance = cpu_to_le32(svc->version | svc->instance << 8);
+ pkt.server.node = cpu_to_le32(qmi->sq.sq_node);
+ pkt.server.port = cpu_to_le32(qmi->sq.sq_port);
+
+ sq.sq_family = qmi->sq.sq_family;
+ sq.sq_node = qmi->sq.sq_node;
+ sq.sq_port = QRTR_PORT_CTRL;
+
+ msg.msg_name = &sq;
+ msg.msg_namelen = sizeof(sq);
+
+ mutex_lock(&qmi->sock_lock);
+ if (qmi->sock) {
+ ret = kernel_sendmsg(qmi->sock, &msg, &iv, 1, sizeof(pkt));
+ if (ret < 0)
+ pr_err("send service delete message failed: %d\n", ret);
+ }
+ mutex_unlock(&qmi->sock_lock);
+}
+
+static void __exit rpmon_qmi_drv_exit(void)
+{
+ rpmon_qmi_del_server();
+
+ qmi_handle_release(&rpqmi->qmi);
+
+ platform_driver_unregister(&rpmon_qmi_drv);
+}
+module_exit(rpmon_qmi_drv_exit);
+
+MODULE_AUTHOR("Wang Wenhu <wenhu.wang@vivo.com>");
+MODULE_DESCRIPTION("Subsystem Monitor via QMI platform driver");
+MODULE_ALIAS("platform:" DRIVER_NAME);
+MODULE_LICENSE("GPL v2");
Implements a kind of communication routine for RPMON to communicate with remote processors through QMI infrastructure. RPMON_QMI itself is designed as a modular framework that would introduce different kind of message sets binding to different services. RPMON_QMI creates a device of rpmon_device type for each remote processor endpoint. All the endpoint devices share an unique set of QMI suite. Cc: Andy Gross <agross@kernel.org> Cc: Bjorn Andersson <bjorn.andersson@linaro.org> Cc: Ohad Ben-Cohen <ohad@wizery.com> Cc: linux-remoteproc@vger.kernel.org Cc: linux-arm-msm@vger.kernel.org Signed-off-by: Wang Wenhu <wenhu.wang@vivo.com> --- Changes since v1: - Addressed review comments from Randy Changes since v2: - Added Cc list - Use ARRAY_SIZE instead of multiple sizeof calculation --- drivers/rpmon/Kconfig | 15 ++ drivers/rpmon/Makefile | 1 + drivers/rpmon/rpmon_qmi.c | 431 ++++++++++++++++++++++++++++++++++++++ 3 files changed, 447 insertions(+) create mode 100644 drivers/rpmon/rpmon_qmi.c