mirror of
https://github.com/espressif/esp-idf.git
synced 2024-10-05 20:47:46 -04:00
391 lines
16 KiB
C
391 lines
16 KiB
C
/*
|
|
* SPDX-FileCopyrightText: 2020-2021 Espressif Systems (Shanghai) CO LTD
|
|
*
|
|
* SPDX-License-Identifier: Apache-2.0
|
|
*/
|
|
|
|
// #define LOG_LOCAL_LEVEL ESP_LOG_DEBUG
|
|
|
|
#include <stdlib.h>
|
|
#include <string.h>
|
|
#include <sys/lock.h>
|
|
#include "sdkconfig.h"
|
|
#include "esp_compiler.h"
|
|
#include "esp_heap_caps.h"
|
|
#include "esp_intr_alloc.h"
|
|
#include "esp_log.h"
|
|
#include "esp_check.h"
|
|
#include "soc/soc_caps.h"
|
|
#include "soc/gpio_periph.h"
|
|
#include "soc/io_mux_reg.h"
|
|
#include "hal/cpu_hal.h"
|
|
#include "hal/cpu_ll.h"
|
|
#include "hal/gpio_hal.h"
|
|
#include "driver/periph_ctrl.h"
|
|
#include "esp_rom_gpio.h"
|
|
#include "freertos/FreeRTOS.h"
|
|
#include "driver/dedic_gpio.h"
|
|
#include "soc/dedic_gpio_periph.h"
|
|
#if SOC_DEDIC_GPIO_ALLOW_REG_ACCESS
|
|
#include "soc/dedic_gpio_struct.h"
|
|
#include "hal/dedic_gpio_ll.h"
|
|
#endif
|
|
|
|
|
|
static const char *TAG = "dedic_gpio";
|
|
|
|
typedef struct dedic_gpio_platform_t dedic_gpio_platform_t;
|
|
typedef struct dedic_gpio_bundle_t dedic_gpio_bundle_t;
|
|
|
|
// Dedicated GPIO driver platform, GPIO bundles will be installed onto it
|
|
static dedic_gpio_platform_t *s_platform[SOC_CPU_CORES_NUM];
|
|
// platform level mutex lock
|
|
static _lock_t s_platform_mutexlock[SOC_CPU_CORES_NUM];
|
|
|
|
struct dedic_gpio_platform_t {
|
|
portMUX_TYPE spinlock; // Spinlock, stop GPIO channels from accessing common resource concurrently
|
|
uint32_t out_occupied_mask; // mask of output channels that already occupied
|
|
uint32_t in_occupied_mask; // mask of input channels that already occupied
|
|
#if SOC_DEDIC_GPIO_HAS_INTERRUPT
|
|
intr_handle_t intr_hdl; // interrupt handle
|
|
dedic_gpio_isr_callback_t cbs[SOC_DEDIC_GPIO_IN_CHANNELS_NUM]; // array of callback function for input channel
|
|
void *cb_args[SOC_DEDIC_GPIO_IN_CHANNELS_NUM]; // array of callback arguments for input channel
|
|
dedic_gpio_bundle_t *in_bundles[SOC_DEDIC_GPIO_IN_CHANNELS_NUM]; // which bundle belongs to for input channel
|
|
#endif
|
|
#if SOC_DEDIC_GPIO_ALLOW_REG_ACCESS
|
|
dedic_dev_t *dev;
|
|
#endif
|
|
};
|
|
|
|
struct dedic_gpio_bundle_t {
|
|
uint32_t core_id; // CPU core ID, a GPIO bundle must be installed to a specific CPU core
|
|
uint32_t out_mask; // mask of output channels in the bank
|
|
uint32_t in_mask; // mask of input channels in the bank
|
|
uint32_t out_offset; // offset in the bank (seen from output channel)
|
|
uint32_t in_offset; // offset in the bank (seen from input channel)
|
|
size_t nr_gpio; // number of GPIOs in the gpio_array
|
|
int gpio_array[]; // array of GPIO numbers (configured by user)
|
|
};
|
|
|
|
static esp_err_t dedic_gpio_build_platform(uint32_t core_id)
|
|
{
|
|
esp_err_t ret = ESP_OK;
|
|
if (!s_platform[core_id]) {
|
|
// prevent building platform concurrently
|
|
_lock_acquire(&s_platform_mutexlock[core_id]);
|
|
if (!s_platform[core_id]) {
|
|
s_platform[core_id] = calloc(1, sizeof(dedic_gpio_platform_t));
|
|
if (s_platform[core_id]) {
|
|
// initialize platfrom members
|
|
s_platform[core_id]->spinlock = (portMUX_TYPE)portMUX_INITIALIZER_UNLOCKED;
|
|
#if SOC_DEDIC_GPIO_ALLOW_REG_ACCESS
|
|
s_platform[core_id]->dev = &DEDIC_GPIO;
|
|
#endif
|
|
periph_module_enable(dedic_gpio_periph_signals.module); // enable APB clock to peripheral
|
|
}
|
|
}
|
|
_lock_release(&s_platform_mutexlock[core_id]);
|
|
|
|
ESP_GOTO_ON_FALSE(s_platform[core_id], ESP_ERR_NO_MEM, err, TAG, "no mem for s_platform[%d]", core_id);
|
|
ESP_LOGD(TAG, "build platform on core[%d] at %p", core_id, s_platform);
|
|
}
|
|
|
|
err:
|
|
return ret;
|
|
}
|
|
|
|
static void dedic_gpio_break_platform(uint32_t core_id)
|
|
{
|
|
if (s_platform[core_id]) {
|
|
// prevent breaking platform concurrently
|
|
_lock_acquire(&s_platform_mutexlock[core_id]);
|
|
if (s_platform[core_id]) {
|
|
free(s_platform[core_id]);
|
|
s_platform[core_id] = NULL;
|
|
periph_module_disable(dedic_gpio_periph_signals.module); // disable module if no GPIO channel is being used
|
|
}
|
|
_lock_release(&s_platform_mutexlock[core_id]);
|
|
}
|
|
}
|
|
|
|
#if SOC_DEDIC_GPIO_HAS_INTERRUPT
|
|
static void dedic_gpio_default_isr(void *arg)
|
|
{
|
|
bool need_yield = false;
|
|
dedic_gpio_platform_t *platform = (dedic_gpio_platform_t *)arg;
|
|
|
|
// get and clear interrupt status
|
|
portENTER_CRITICAL_ISR(&platform->spinlock);
|
|
uint32_t status = dedic_gpio_ll_get_interrupt_status(platform->dev);
|
|
dedic_gpio_ll_clear_interrupt_status(platform->dev, status);
|
|
portEXIT_CRITICAL_ISR(&platform->spinlock);
|
|
|
|
// handle dedicated channel one by one
|
|
while (status) {
|
|
uint32_t channel = __builtin_ffs(status) - 1; // get dedicated channel number which triggered the interrupt
|
|
if (platform->cbs[channel]) {
|
|
if (platform->cbs[channel](platform->in_bundles[channel], channel - platform->in_bundles[channel]->in_offset, platform->cb_args[channel])) {
|
|
need_yield = true; // note that we need to yield at the end of isr
|
|
}
|
|
}
|
|
status = status & (status - 1); // clear the right most bit '1'
|
|
}
|
|
|
|
if (need_yield) {
|
|
portYIELD_FROM_ISR();
|
|
}
|
|
}
|
|
|
|
static esp_err_t dedic_gpio_install_interrupt(uint32_t core_id)
|
|
{
|
|
esp_err_t ret = ESP_OK;
|
|
if (!s_platform[core_id]->intr_hdl) {
|
|
// prevent install interrupt concurrently
|
|
_lock_acquire(&s_platform_mutexlock[core_id]);
|
|
if (!s_platform[core_id]->intr_hdl) {
|
|
int isr_flags = 0;
|
|
ret = esp_intr_alloc(dedic_gpio_periph_signals.irq, isr_flags, dedic_gpio_default_isr, s_platform[core_id], &s_platform[core_id]->intr_hdl);
|
|
// clear pending interrupt
|
|
uint32_t status = dedic_gpio_ll_get_interrupt_status(s_platform[core_id]->dev);
|
|
dedic_gpio_ll_clear_interrupt_status(s_platform[core_id]->dev, status);
|
|
}
|
|
_lock_release(&s_platform_mutexlock[core_id]);
|
|
ESP_GOTO_ON_ERROR(ret, err, TAG, "alloc interrupt failed");
|
|
}
|
|
|
|
err:
|
|
return ret;
|
|
}
|
|
|
|
static void dedic_gpio_uninstall_interrupt(uint32_t core_id)
|
|
{
|
|
if (s_platform[core_id]->intr_hdl) {
|
|
// prevent uninstall interrupt concurrently
|
|
_lock_acquire(&s_platform_mutexlock[core_id]);
|
|
if (s_platform[core_id]->intr_hdl) {
|
|
esp_intr_free(s_platform[core_id]->intr_hdl);
|
|
s_platform[core_id]->intr_hdl = NULL;
|
|
// disable all interrupt
|
|
dedic_gpio_ll_enable_interrupt(s_platform[core_id]->dev, ~0UL, false);
|
|
}
|
|
_lock_release(&s_platform_mutexlock[core_id]);
|
|
}
|
|
}
|
|
|
|
static void dedic_gpio_set_interrupt(uint32_t core_id, uint32_t channel, dedic_gpio_intr_type_t type)
|
|
{
|
|
dedic_gpio_ll_set_interrupt_type(s_platform[core_id]->dev, channel, type);
|
|
if (type != DEDIC_GPIO_INTR_NONE) {
|
|
dedic_gpio_ll_enable_interrupt(s_platform[core_id]->dev, 1 << channel, true);
|
|
} else {
|
|
dedic_gpio_ll_enable_interrupt(s_platform[core_id]->dev, 1 << channel, false);
|
|
}
|
|
}
|
|
#endif // SOC_DEDIC_GPIO_HAS_INTERRUPT
|
|
|
|
esp_err_t dedic_gpio_new_bundle(const dedic_gpio_bundle_config_t *config, dedic_gpio_bundle_handle_t *ret_bundle)
|
|
{
|
|
esp_err_t ret = ESP_OK;
|
|
dedic_gpio_bundle_t *bundle = NULL;
|
|
uint32_t out_mask = 0;
|
|
uint32_t in_mask = 0;
|
|
uint32_t core_id = cpu_hal_get_core_id(); // dedicated GPIO will be binded to the CPU who invokes this API
|
|
|
|
ESP_GOTO_ON_FALSE(config && ret_bundle, ESP_ERR_INVALID_ARG, err, TAG, "invalid argument");
|
|
ESP_GOTO_ON_FALSE(config->gpio_array && config->array_size > 0, ESP_ERR_INVALID_ARG, err, TAG, "invalid GPIO array or size");
|
|
ESP_GOTO_ON_FALSE(config->flags.in_en || config->flags.out_en, ESP_ERR_INVALID_ARG, err, TAG, "no input/output mode specified");
|
|
// lazy install s_platform[core_id]
|
|
ESP_GOTO_ON_ERROR(dedic_gpio_build_platform(core_id), err, TAG, "build platform %d failed", core_id);
|
|
|
|
size_t bundle_size = sizeof(dedic_gpio_bundle_t) + config->array_size * sizeof(config->gpio_array[0]);
|
|
bundle = calloc(1, bundle_size);
|
|
ESP_GOTO_ON_FALSE(bundle, ESP_ERR_NO_MEM, err, TAG, "no mem for bundle");
|
|
|
|
// for performance reasons, we only search for continuous channels
|
|
uint32_t pattern = (1 << config->array_size) - 1;
|
|
// configure outwards channels
|
|
uint32_t out_offset = 0;
|
|
if (config->flags.out_en) {
|
|
ESP_GOTO_ON_FALSE(config->array_size <= SOC_DEDIC_GPIO_OUT_CHANNELS_NUM, ESP_ERR_INVALID_ARG, err, TAG,
|
|
"array size(%d) exceeds maximum supported out channels(%d)", config->array_size, SOC_DEDIC_GPIO_OUT_CHANNELS_NUM);
|
|
// prevent install bundle concurrently
|
|
portENTER_CRITICAL(&s_platform[core_id]->spinlock);
|
|
for (size_t i = 0; i <= SOC_DEDIC_GPIO_OUT_CHANNELS_NUM - config->array_size; i++) {
|
|
if ((s_platform[core_id]->out_occupied_mask & (pattern << i)) == 0) {
|
|
out_mask = pattern << i;
|
|
out_offset = i;
|
|
break;
|
|
}
|
|
}
|
|
if (out_mask) {
|
|
s_platform[core_id]->out_occupied_mask |= out_mask;
|
|
#if SOC_DEDIC_GPIO_ALLOW_REG_ACCESS
|
|
// always enable instruction to access output GPIO, which has better performance than register access
|
|
dedic_gpio_ll_enable_instruction_access_out(s_platform[core_id]->dev, out_mask, true);
|
|
#endif
|
|
}
|
|
portEXIT_CRITICAL(&s_platform[core_id]->spinlock);
|
|
ESP_GOTO_ON_FALSE(out_mask, ESP_ERR_NOT_FOUND, err, TAG, "no free outward channels on core[%d]", core_id);
|
|
ESP_LOGD(TAG, "new outward bundle(%p) on core[%d], offset=%d, mask(%x)", bundle, core_id, out_offset, out_mask);
|
|
}
|
|
|
|
// configure inwards channels
|
|
uint32_t in_offset = 0;
|
|
if (config->flags.in_en) {
|
|
ESP_GOTO_ON_FALSE(config->array_size <= SOC_DEDIC_GPIO_IN_CHANNELS_NUM, ESP_ERR_INVALID_ARG, err, TAG,
|
|
"array size(%d) exceeds maximum supported in channels(%d)", config->array_size, SOC_DEDIC_GPIO_IN_CHANNELS_NUM);
|
|
// prevent install bundle concurrently
|
|
portENTER_CRITICAL(&s_platform[core_id]->spinlock);
|
|
for (size_t i = 0; i <= SOC_DEDIC_GPIO_IN_CHANNELS_NUM - config->array_size; i++) {
|
|
if ((s_platform[core_id]->in_occupied_mask & (pattern << i)) == 0) {
|
|
in_mask = pattern << i;
|
|
in_offset = i;
|
|
break;
|
|
}
|
|
}
|
|
if (in_mask) {
|
|
s_platform[core_id]->in_occupied_mask |= in_mask;
|
|
}
|
|
portEXIT_CRITICAL(&s_platform[core_id]->spinlock);
|
|
ESP_GOTO_ON_FALSE(in_mask, ESP_ERR_NOT_FOUND, err, TAG, "no free inward channels on core[%d]", core_id);
|
|
ESP_LOGD(TAG, "new inward bundle(%p) on core[%d], offset=%d, mask(%x)", bundle, core_id, in_offset, in_mask);
|
|
}
|
|
|
|
// route dedicated GPIO channel signals to GPIO matrix
|
|
if (config->flags.in_en) {
|
|
for (size_t i = 0; i < config->array_size; i++) {
|
|
gpio_hal_iomux_func_sel(GPIO_PIN_MUX_REG[config->gpio_array[i]], PIN_FUNC_GPIO);
|
|
esp_rom_gpio_connect_in_signal(config->gpio_array[i], dedic_gpio_periph_signals.cores[core_id].in_sig_per_channel[in_offset + i], config->flags.in_invert);
|
|
}
|
|
}
|
|
if (config->flags.out_en) {
|
|
for (size_t i = 0; i < config->array_size; i++) {
|
|
gpio_hal_iomux_func_sel(GPIO_PIN_MUX_REG[config->gpio_array[i]], PIN_FUNC_GPIO);
|
|
esp_rom_gpio_connect_out_signal(config->gpio_array[i], dedic_gpio_periph_signals.cores[core_id].out_sig_per_channel[out_offset + i], config->flags.out_invert, false);
|
|
}
|
|
}
|
|
|
|
// it's safe to initialize bundle members without locks here
|
|
bundle->core_id = core_id;
|
|
bundle->out_mask = out_mask;
|
|
bundle->in_mask = in_mask;
|
|
bundle->out_offset = out_offset;
|
|
bundle->in_offset = in_offset;
|
|
bundle->nr_gpio = config->array_size;
|
|
memcpy(bundle->gpio_array, config->gpio_array, config->array_size * sizeof(config->gpio_array[0]));
|
|
|
|
*ret_bundle = bundle; // return bundle instance
|
|
return ESP_OK;
|
|
|
|
err:
|
|
if (s_platform[core_id] && (out_mask || in_mask)) {
|
|
portENTER_CRITICAL(&s_platform[core_id]->spinlock);
|
|
s_platform[core_id]->out_occupied_mask &= ~out_mask;
|
|
s_platform[core_id]->in_occupied_mask &= ~in_mask;
|
|
portEXIT_CRITICAL(&s_platform[core_id]->spinlock);
|
|
}
|
|
if (bundle) {
|
|
free(bundle);
|
|
}
|
|
return ret;
|
|
}
|
|
|
|
esp_err_t dedic_gpio_del_bundle(dedic_gpio_bundle_handle_t bundle)
|
|
{
|
|
esp_err_t ret = ESP_OK;
|
|
bool recycle_all = false;
|
|
ESP_GOTO_ON_FALSE(bundle, ESP_ERR_INVALID_ARG, err, TAG, "invalid argument");
|
|
|
|
uint32_t core_id = cpu_hal_get_core_id();
|
|
ESP_GOTO_ON_FALSE(core_id == bundle->core_id, ESP_FAIL, err, TAG, "del bundle on wrong CPU");
|
|
|
|
portENTER_CRITICAL(&s_platform[core_id]->spinlock);
|
|
s_platform[core_id]->out_occupied_mask &= ~(bundle->out_mask);
|
|
s_platform[core_id]->in_occupied_mask &= ~(bundle->in_mask);
|
|
if (!s_platform[core_id]->in_occupied_mask && !s_platform[core_id]->out_occupied_mask) {
|
|
recycle_all = true;
|
|
}
|
|
portEXIT_CRITICAL(&s_platform[core_id]->spinlock);
|
|
|
|
free(bundle);
|
|
|
|
if (recycle_all) {
|
|
#if SOC_DEDIC_GPIO_HAS_INTERRUPT
|
|
dedic_gpio_uninstall_interrupt(core_id);
|
|
#endif
|
|
dedic_gpio_break_platform(core_id);
|
|
}
|
|
|
|
err:
|
|
return ret;
|
|
}
|
|
|
|
esp_err_t dedic_gpio_get_out_mask(dedic_gpio_bundle_handle_t bundle, uint32_t *mask)
|
|
{
|
|
esp_err_t ret = ESP_OK;
|
|
ESP_GOTO_ON_FALSE(bundle && mask, ESP_ERR_INVALID_ARG, err, TAG, "invalid argument");
|
|
*mask = bundle->out_mask;
|
|
err:
|
|
return ret;
|
|
}
|
|
|
|
esp_err_t dedic_gpio_get_in_mask(dedic_gpio_bundle_handle_t bundle, uint32_t *mask)
|
|
{
|
|
esp_err_t ret = ESP_OK;
|
|
ESP_GOTO_ON_FALSE(bundle && mask, ESP_ERR_INVALID_ARG, err, TAG, "invalid argument");
|
|
*mask = bundle->in_mask;
|
|
err:
|
|
return ret;
|
|
}
|
|
|
|
void dedic_gpio_bundle_write(dedic_gpio_bundle_handle_t bundle, uint32_t mask, uint32_t value)
|
|
{
|
|
// For performace reasons, we don't want to check the validation of parameters here
|
|
// Even didn't check if we're working on the correct CPU core (i.e. bundle->core_id == current core_id)
|
|
cpu_ll_write_dedic_gpio_mask(bundle->out_mask & (mask << bundle->out_offset), value << bundle->out_offset);
|
|
}
|
|
|
|
uint32_t dedic_gpio_bundle_read_out(dedic_gpio_bundle_handle_t bundle)
|
|
{
|
|
// For performace reasons, we don't want to check the validation of parameters here
|
|
// Even didn't check if we're working on the correct CPU core (i.e. bundle->core_id == current core_id)
|
|
uint32_t value = cpu_ll_read_dedic_gpio_out();
|
|
return (value & bundle->out_mask) >> (bundle->out_offset);
|
|
}
|
|
|
|
uint32_t dedic_gpio_bundle_read_in(dedic_gpio_bundle_handle_t bundle)
|
|
{
|
|
// For performace reasons, we don't want to check the validation of parameters here
|
|
// Even didn't check if we're working on the correct CPU core (i.e. bundle->core_id == current core_id)
|
|
uint32_t value = cpu_ll_read_dedic_gpio_in();
|
|
return (value & bundle->in_mask) >> (bundle->in_offset);
|
|
}
|
|
|
|
#if SOC_DEDIC_GPIO_HAS_INTERRUPT
|
|
esp_err_t dedic_gpio_bundle_set_interrupt_and_callback(dedic_gpio_bundle_handle_t bundle, uint32_t mask, dedic_gpio_intr_type_t intr_type, dedic_gpio_isr_callback_t cb_isr, void *cb_args)
|
|
{
|
|
esp_err_t ret = ESP_OK;
|
|
ESP_GOTO_ON_FALSE(bundle, ESP_ERR_INVALID_ARG, err, TAG, "invalid argument");
|
|
uint32_t core_id = cpu_hal_get_core_id();
|
|
// lazy alloc interrupt
|
|
ESP_GOTO_ON_ERROR(dedic_gpio_install_interrupt(core_id), err, TAG, "allocate interrupt on core %d failed", core_id);
|
|
|
|
uint32_t channel_mask = bundle->in_mask & (mask << bundle->in_offset);
|
|
uint32_t channel = 0;
|
|
while (channel_mask) {
|
|
channel = __builtin_ffs(channel_mask) - 1;
|
|
portENTER_CRITICAL(&s_platform[core_id]->spinlock);
|
|
dedic_gpio_set_interrupt(core_id, channel, intr_type);
|
|
portEXIT_CRITICAL(&s_platform[core_id]->spinlock);
|
|
|
|
s_platform[core_id]->cbs[channel] = cb_isr;
|
|
s_platform[core_id]->cb_args[channel] = cb_args;
|
|
s_platform[core_id]->in_bundles[channel] = bundle;
|
|
channel_mask = channel_mask & (channel_mask - 1); // clear the right most bit '1'
|
|
}
|
|
|
|
err:
|
|
return ret;
|
|
}
|
|
#endif // SOC_DEDIC_GPIO_HAS_INTERRUPT
|