Why TxTimestampStatus is alway 0, for esp32's IEEE 1588

[xien551]

Dear everyone, I would like to develop the PTP IEEE 1588.
But found some problems:

1) At first, I use netif and got the IP (just as \examples\ethernet\basic ), then broadcast a buffer just using eth (I refer to \components\esp_eth\test_apps\main\esp_eth_test.c), the program works, and I have gotten the packet from esp32 in wireshark.

2) then, I modify the "esp_eth_mac_new_esp32()".

I just let "emac_esp32_t *emac = NULL" defined outside the function, let it be a gloable var and I could access "hal.tx_desc->TimeStampLow" and "hal.tx_desc->TimeStampHigh"

I make a loop waiting for "emac->hal.tx_desc->TDES0.FirstSegment" and "TDES0.LastSegment" becomes 1, and this is also realized .

3) I found "emac->hal.tx_desc->TDES0.TxTimestampStatus" will never become 1. and the value of timestamp is always 0. And I have check that "TDES0.TransmitTimestampEnable" is 1.

so, what should I do. below is the main program.


void app_main(void)
{
// Initialize TCP/IP network interface (should be called only once in application)
ESP_ERROR_CHECK(esp_netif_init());
// Create default event loop that running in background
ESP_ERROR_CHECK(esp_event_loop_create_default());


// Create new default instance of esp-netif for Ethernet
esp_netif_config_t cfg = ESP_NETIF_DEFAULT_ETH();
esp_netif_t *eth_netif = esp_netif_new(&cfg);

// Init MAC and PHY configs to default
eth_mac_config_t mac_config = ETH_MAC_DEFAULT_CONFIG();
eth_phy_config_t phy_config = ETH_PHY_DEFAULT_CONFIG();

phy_config.phy_addr = CONFIG_EXAMPLE_ETH_PHY_ADDR;
phy_config.reset_gpio_num = CONFIG_EXAMPLE_ETH_PHY_RST_GPIO;
mac_config.smi_mdc_gpio_num = CONFIG_EXAMPLE_ETH_MDC_GPIO;
mac_config.smi_mdio_gpio_num = CONFIG_EXAMPLE_ETH_MDIO_GPIO;

esp_eth_mac_t *mac = esp_eth_mac_new_esp32_2(&mac_config); // using modified function
esp_eth_phy_t *phy = esp_eth_phy_new_lan87xx(&phy_config);

esp_eth_config_t config = ETH_DEFAULT_CONFIG(mac, phy);
esp_eth_handle_t eth_handle = NULL;
ESP_ERROR_CHECK(esp_eth_driver_install(&config, &eth_handle));
//--------------Initialize netif---------------------------------

// attach Ethernet driver to TCP/IP stack
ESP_ERROR_CHECK(esp_netif_attach(eth_netif, esp_eth_new_netif_glue(eth_handle)));

//TDES0.TransmitTimestampEnable
ESP_LOGI(TAG, "\n\n TransmitTimestampEnable:%d",emac->hal.tx_desc->TDES0.TransmitTimestampEnable);
//------------broadcast-------------------
SemaphoreHandle_t mutex = xSemaphoreCreateBinary();

esp_event_loop_create_default();
esp_event_handler_register(ETH_EVENT, ESP_EVENT_ANY_ID, &eth_event_handler, mutex);
ESP_ERROR_CHECK(esp_event_handler_register(IP_EVENT, IP_EVENT_ETH_GOT_IP, &got_ip_event_handler, NULL));

emac->hal.tx_desc->TDES0.TxTimestampStatus = 1;
ESP_LOGI(TAG, "\n\n TxTimestampStatus:%d",emac->hal.tx_desc->TDES0.TxTimestampStatus);
ESP_ERROR_CHECK(esp_eth_start(eth_handle));

ESP_LOGI(TAG, "\n\n TxTimestampStatus:%d",emac->hal.tx_desc->TDES0.TxTimestampStatus);


xSemaphoreTake(mutex, pdMS_TO_TICKS(3000));
// even if PHY (IP101) indicates autonegotiation done and link up, it sometimes may miss few packets after atonego reset, hence wait a bit
vTaskDelay(pdMS_TO_TICKS(100));

ESP_LOGI(TAG, "\n\n TxTimestampStatus:%d",emac->hal.tx_desc->TDES0.TxTimestampStatus);

emac_frame_t *pkt = malloc(1024);
pkt->proto = 0x2222;
memset(pkt->dest, 0xff, 6); // broadcast addr

printf("packet len:%d\n", 40 - ETH_HEADER_LEN);

for (int i = 0; i < (40 - ETH_HEADER_LEN); ++i){
pkt->data = i & 0xff;
}

esp_eth_transmit(eth_handle, pkt, 40);
vTaskDelay(pdMS_TO_TICKS(100));
free(pkt);
/*
ESP_ERROR_CHECK(esp_eth_stop(eth_handle));
ESP_ERROR_CHECK(esp_event_loop_delete_default());
ESP_ERROR_CHECK(esp_eth_driver_uninstall(eth_handle));
phy->del(phy);
mac->del(mac);
vSemaphoreDelete(mutex);
*/
printf("waiting for timestamp\n");
while(emac->hal.tx_desc->TDES0.FirstSegment == 0) {
printf("still waiting for timestamp\n");
vTaskDelay(pdMS_TO_TICKS(10));
}

//--------------try capture Timestamps-------------------
ESP_LOGI(TAG, "\n\n TxTimestampStatus:%d",emac->hal.tx_desc->TDES0.TxTimestampStatus);
//TDES0.TransmitTimestampEnable
ESP_LOGI(TAG, "\n\n TransmitTimestampEnable:%d",emac->hal.tx_desc->TDES0.TransmitTimestampEnable);

ESP_LOGI(TAG, "if tx timestamp is available TDES0.FirstSegment %d TDES0.LastSegment %d",emac->hal.tx_desc->TDES0.FirstSegment,
emac->hal.tx_desc->TDES0.LastSegment );
ESP_LOGI(TAG, "tx timestmap: TimeStampLow:%d, TimeStampHigh:%d\n",emac->hal.tx_desc->TimeStampLow,
emac->hal.tx_desc->TimeStampHigh );
/*
ESP_LOGI(TAG, "if rx timestamp is available, RDES0.LastDescriptor: %d ",emac->hal.rx_desc->RDES0.LastDescriptor );
ESP_LOGI(TAG, "rx timestmap: TimeStampLow %d TimeStampHigh %d\n",emac->hal.rx_desc->TimeStampLow,
emac->hal.rx_desc->TimeStampHigh );
*/
ESP_LOGI(TAG, "Buffer1Addr:%d, Buffer2NextDescAddr %d\n",emac->hal.tx_desc->Buffer1Addr,
emac->hal.tx_desc->Buffer2NextDescAddr);
ESP_LOGI(TAG,"capture system time %lld", esp_timer_get_time() );
//---------------

ESP_LOGI(TAG, "\n\n TxTimestampStatus:%d",emac->hal.tx_desc->TDES0.TxTimestampStatus);
while(emac->hal.tx_desc->TDES0.TxTimestampStatus == 0) {
//printf("still waiting TTTS becaome 1\n");
emac->hal.tx_desc->TDES0.TxTimestampStatus = 1;
ESP_LOGI(TAG, "1 TxTimestampStatus:%d",emac->hal.tx_desc->TDES0.TxTimestampStatus);
esp_eth_transmit(eth_handle, pkt, 40);
ESP_LOGI(TAG, "2 TxTimestampStatus:%d",emac->hal.tx_desc->TDES0.TxTimestampStatus);
//vTaskDelay(pdMS_TO_TICKS(10));
ESP_LOGI(TAG, "if tx timestamp is available TDES0.FirstSegment %d TDES0.LastSegment %d",emac->hal.tx_desc->TDES0.FirstSegment,
emac->hal.tx_desc->TDES0.LastSegment );
ESP_LOGI(TAG, "tx timestmap: TimeStampLow:%d, TimeStampHigh:%d\n",emac->hal.tx_desc->TimeStampLow,
emac->hal.tx_desc->TimeStampHigh );
}
printf("Finished!\n");
//------------broadcast end-------------------

}

[ESP_ondrej]

Hi xien551,

thanks for the question. EMAC IP in ESP32 SoC is third party product for which Espressif pays to could use it in its chips. It supports multiple features. One of the advertised supported features is IEE1588 time stamping. Unfortunately, we have conducted only partial testing of the time stamping on ESP32 so far. Therefore I can provide you only very basic guidance. In addition, it seems there are some mistakes in ESP32 TRM which we need to fix. The EMAC used in the ESP32 supports IEEE1588 but only IEEE1588-2002 (i.e. PTPv1). In addition, description of registers to configure/enable EMAC's time stamp engine is missing. That is the reason why you did not get any time stamp - it was not properly configured.

The main concern though is to evaluate if it is sufficient to use PTPv1 for your application since most applications use IEEE1588-2008 (PTPv2) which is not backwards compatible.

In any case, I am planning to have look at it in more details during the next year. Thank you for understanding.

Ondrej

[xien551]

Hi, Ondrej!

Thanks for your reply. I think you are right, the second edtion is more suitable. I would like to consider STM32F4 or H7, since there are some worked 1588 examples in Github.

But if you have any progress, please tell me, thank you!

[Pe3ucTop]

Hi Ondrej ( @ESP_ondrej )

I also intereted in packet timestamping and PTPv1 could be enough for my needs, so any update on subject ?

@xien551 sorry, did you get/had RX timestaps ? Or you get none ?

Best regards,
Maksims M.

[ESP_ondrej]

I investigated and discussed internally and I have bad news... Hardware time-stamping cannot be used on ESP32 (neither PTPv1 nor PTPv2).

[xien551]

Thanks, I have changed my ideas.

Maybe FTM of c3 or s3 is a good alternative choice. And I have make some progress.