[dm][hwspinlock] support hwspinlock

Hardware spinlock modules provide hardware assistance for
synchronization and mutual exclusion between heterogeneous processors
and those not operating under a single, shared operating system.

Signed-off-by: GuEe-GUI <[email protected]>

Author: GuEe-GUI

components/drivers/Kconfig

@@ -23,6 +23,7 @@ rsource "hwcrypto/Kconfig"
 rsource "wlan/Kconfig"
 rsource "led/Kconfig"
 rsource "mailbox/Kconfig"
+rsource "hwspinlock/Kconfig"
 rsource "phye/Kconfig"
 rsource "ata/Kconfig"
 rsource "nvme/Kconfig"

components/drivers/hwspinlock/Kconfig

@@ -0,0 +1,15 @@
+menuconfig RT_USING_HWSPINLOCK
+    bool "Using Hardware Spinlock device drivers"
+    depends on RT_USING_DM
+    depends on RT_USING_OFW
+    select RT_USING_ADT
+    select RT_USING_ADT_REF
+    default n
+    help
+      Hardware spinlock modules provide hardware assistance for
+      synchronization and mutual exclusion between heterogeneous processors
+      and those not operating under a single, shared operating system.
+
+if RT_USING_HWSPINLOCK
+    osource "$(SOC_DM_HWSPINLOCK_DIR)/Kconfig"
+endif

components/drivers/hwspinlock/SConscript

@@ -0,0 +1,15 @@
+from building import *
+
+group = []
+
+if not GetDepend(['RT_USING_HWSPINLOCK']):
+    Return('group')
+
+cwd = GetCurrentDir()
+CPPPATH = [cwd + '/../include']
+
+src = ['hwspinlock.c']
+
+group = DefineGroup('DeviceDrivers', src, depend = [''], CPPPATH = CPPPATH)
+
+Return('group')

components/drivers/hwspinlock/hwspinlock.c

@@ -0,0 +1,313 @@
+/*
+ * Copyright (c) 2006-2023, RT-Thread Development Team
+ *
+ * SPDX-License-Identifier: Apache-2.0
+ *
+ * Change Logs:
+ * Date           Author       Notes
+ * 2023-09-23     GuEe-GUI     first version
+ */
+
+#include <rtthread.h>
+
+#include <cpuport.h>
+
+#define DBG_TAG "rtdm.hwspinlock"
+#define DBG_LVL DBG_INFO
+#include <rtdbg.h>
+
+#include "hwspinlock_dm.h"
+
+static RT_DEFINE_SPINLOCK(hwspinlock_ops_lock);
+static rt_list_t hwspinlock_bank_nodes = RT_LIST_OBJECT_INIT(hwspinlock_bank_nodes);
+
+rt_err_t rt_hwspinlock_bank_register(struct rt_hwspinlock_bank *bank)
+{
+    struct rt_hwspinlock *hwlock;
+
+    if (!bank || !bank->ops || bank->locks_nr <= 0 || !bank->dev)
+    {
+        return -RT_EINVAL;
+    }
+
+    rt_list_init(&bank->list);
+    rt_ref_init(&bank->ref);
+
+    hwlock = &bank->locks[0];
+
+    for (int i = 0; i < bank->locks_nr; ++i, ++hwlock)
+    {
+        rt_spin_lock_init(&hwlock->lock);
+        hwlock->used = RT_FALSE;
+    }
+
+    rt_spin_lock(&hwspinlock_ops_lock);
+    rt_list_insert_after(&hwspinlock_bank_nodes, &bank->list);
+    rt_spin_unlock(&hwspinlock_ops_lock);
+
+    rt_dm_dev_bind_fwdata(bank->dev, RT_NULL, bank);
+
+    return RT_EOK;
+}
+
+rt_err_t rt_hwspinlock_bank_unregister(struct rt_hwspinlock_bank *bank)
+{
+    rt_err_t err;
+
+    if (!bank)
+    {
+        return -RT_EINVAL;
+    }
+
+    rt_spin_lock(&hwspinlock_ops_lock);
+
+    if (rt_ref_read(&bank->ref) == 1)
+    {
+        rt_dm_dev_unbind_fwdata(bank->dev, RT_NULL);
+
+        err = RT_EOK;
+    }
+    else
+    {
+        err = -RT_EBUSY;
+    }
+
+    rt_spin_unlock(&hwspinlock_ops_lock);
+
+    return err;
+}
+
+rt_err_t rt_hwspin_trylock_raw(struct rt_hwspinlock *hwlock,
+        rt_ubase_t *out_irq_level)
+{
+    rt_err_t err;
+
+    if (!hwlock)
+    {
+        return -RT_EINVAL;
+    }
+
+    if (out_irq_level)
+    {
+        *out_irq_level = rt_spin_lock_irqsave(&hwlock->lock);
+    }
+    else
+    {
+        rt_spin_lock(&hwlock->lock);
+    }
+
+    err = hwlock->bank->ops->trylock(hwlock);
+
+    if (err)
+    {
+        if (out_irq_level)
+        {
+            rt_spin_unlock_irqrestore(&hwlock->lock, *out_irq_level);
+        }
+        else
+        {
+            rt_spin_unlock(&hwlock->lock);
+        }
+    }
+
+    rt_hw_dmb();
+
+    return err;
+}
+
+rt_err_t rt_hwspin_lock_timeout_raw(struct rt_hwspinlock *hwlock,
+        rt_uint32_t timeout_ms, rt_ubase_t *out_irq_level)
+{
+    rt_err_t err;
+    rt_tick_t timeout = rt_tick_get() + rt_tick_from_millisecond(timeout_ms);
+
+    for (;;)
+    {
+        err = rt_hwspin_trylock_raw(hwlock, out_irq_level);
+
+        if (err != -RT_EBUSY)
+        {
+            break;
+        }
+
+        if (timeout < rt_tick_get())
+        {
+            return -RT_ETIMEOUT;
+        }
+
+        if (hwlock->bank->ops->relax)
+        {
+            hwlock->bank->ops->relax(hwlock);
+        }
+    }
+
+    return err;
+}
+
+void rt_hwspin_unlock_raw(struct rt_hwspinlock *hwlock,
+        rt_ubase_t *out_irq_level)
+{
+    if (!hwlock)
+    {
+        return;
+    }
+
+    rt_hw_dmb();
+
+    hwlock->bank->ops->unlock(hwlock);
+
+    if (out_irq_level)
+    {
+        rt_spin_unlock_irqrestore(&hwlock->lock, *out_irq_level);
+    }
+    else
+    {
+        rt_spin_unlock(&hwlock->lock);
+    }
+}
+
+static struct rt_hwspinlock *hwspinlock_get(struct rt_hwspinlock_bank *bank, int id)
+{
+    struct rt_hwspinlock *hwlock = RT_NULL;
+
+    if (bank)
+    {
+        int offset = id - bank->base_id;
+
+        if (!bank->locks[offset].used)
+        {
+            hwlock = &bank->locks[offset];
+        }
+    }
+    else
+    {
+        rt_list_for_each_entry(bank, &hwspinlock_bank_nodes, list)
+        {
+            hwlock = rt_err_ptr(-RT_EBUSY);
+
+            for (int i = 0; i < bank->locks_nr; ++i)
+            {
+                if (!bank->locks[i].used)
+                {
+                    hwlock = &bank->locks[i];
+                    goto _found;
+                }
+            }
+        }
+    }
+
+_found:
+    if (!rt_is_err_or_null(hwlock))
+    {
+        hwlock->used = RT_TRUE;
+    }
+
+    return hwlock;
+}
+
+struct rt_hwspinlock *rt_hwspinlock_get(void)
+{
+    struct rt_hwspinlock *lock;
+
+    rt_spin_lock(&hwspinlock_ops_lock);
+
+    lock = hwspinlock_get(RT_NULL, -1);
+
+    rt_spin_unlock(&hwspinlock_ops_lock);
+
+    return lock;
+}
+
+struct rt_hwspinlock *rt_hwspinlock_get_by_index(struct rt_device *dev, int index)
+{
+    return rt_ofw_get_hwspinlock_by_index(dev->ofw_node, index);
+}
+
+struct rt_hwspinlock *rt_hwspinlock_get_by_name(struct rt_device *dev, const char *name)
+{
+    return rt_ofw_get_hwspinlock_by_name(dev->ofw_node, name);
+}
+
+static void hwspinlock_release(struct rt_ref *r)
+{
+    struct rt_hwspinlock_bank *bank = rt_container_of(r, struct rt_hwspinlock_bank, ref);
+
+    LOG_E("%s is release", rt_dm_dev_get_name(bank->dev));
+    (void)bank;
+
+    RT_ASSERT(0);
+}
+
+void rt_hwspinlock_put(struct rt_hwspinlock *hwlock)
+{
+    if (hwlock)
+    {
+        rt_ref_put(&hwlock->bank->ref, &hwspinlock_release);
+        hwlock->used = RT_TRUE;
+    }
+}
+
+struct rt_hwspinlock *rt_ofw_get_hwspinlock_by_index(struct rt_ofw_node *np, int index)
+{
+    rt_err_t err;
+    struct rt_ofw_node *bank_np;
+    struct rt_ofw_cell_args args;
+    struct rt_hwspinlock *lock;
+    struct rt_hwspinlock_bank *bank;
+
+    if (!np && index < 0)
+    {
+        return rt_err_ptr(-RT_EINVAL);
+    }
+
+    err = rt_ofw_parse_phandle_cells(np, "hwlocks", "#hwlock-cells", index, &args);
+
+    if (err)
+    {
+        return rt_err_ptr(err);
+    }
+
+    bank_np = args.data;
+
+    if (!rt_ofw_data(bank_np))
+    {
+        rt_platform_ofw_request(bank_np);
+    }
+
+    rt_spin_lock(&hwspinlock_ops_lock);
+
+    bank = rt_ofw_data(bank_np);
+    rt_ofw_node_put(bank_np);
+
+    if (!bank || args.args_count != 1)
+    {
+        lock = rt_err_ptr(-RT_ENOSYS);
+    }
+    else
+    {
+        lock = hwspinlock_get(bank, bank->base_id + args.args[0]);
+    }
+
+    rt_spin_unlock(&hwspinlock_ops_lock);
+
+    return lock;
+}
+
+struct rt_hwspinlock *rt_ofw_get_hwspinlock_by_name(struct rt_ofw_node *np, const char *name)
+{
+    int index;
+
+    if (!np || !name)
+    {
+        return rt_err_ptr(-RT_EINVAL);
+    }
+
+    index = rt_ofw_prop_index_of_string(np, "hwlock-names", name);
+
+    if (index < 0)
+    {
+        return rt_err_ptr(index);
+    }
+
+    return rt_ofw_get_hwspinlock_by_index(np, index);
+}