SPI LAN crash

[joyhope]

Bug description

using eth_spi_async code on ESP32S3 board. Crashed when lan driver dropped.

I (1493) esp_idf_svc::eth: Driver initialized
I (1493) esp_idf_svc::eth: Attached MAC address: [2, 0, 0, 18, 52, 86]
I (1493) esp_idf_svc::eth: Initialization complete
I (1493) esp_idf_svc::eth: Stopping
E (1493) esp_eth: esp_eth_stop(309): driver not started yet
I (1501) esp_idf_svc::eth: Stop requested
I (1509) esp_eth.netif.netif_glue: 02:00:00:12:34:56
I (1509) esp_eth.netif.netif_glue: ethernet attached to netif
I (1517) esp32test::eth_spi_async: Starting eth...
I (1525) esp_idf_svc::eth: Start requested
I (1525) esp32test::eth_spi_async: Waiting for DHCP lease...
I (4541) esp_netif_handlers: eth ip: 192.168.9.186, mask: 255.255.255.0, gw: 192.168.9.1
I (4541) esp32test::eth_spi_async: Eth DHCP info: IpInfo { ip: 192.168.9.186, subnet: Subnet { gateway: 192.168.9.1, mask: Mask(24) }, dns: Some(192.168.9.1), secondary_dns: Some(0.0.0.0) }
I (4549) esp32test::eth_spi_async: About to do some pings for 192.168.9.1
I (4557) esp_idf_svc::ping: About to run a summary ping 192.168.9.1 with configuration Configuration { count: 5, interval: 1s, timeout: 1s, data_size: 56, tos: 0 }
I (4573) esp_idf_svc::ping: Ping session established, got handle 0x3fcba7f8
I (4581) esp_idf_svc::ping: Ping session started
I (4581) esp_idf_svc::ping: Ping success callback invoked
I (4589) esp_idf_svc::ping: Waiting for the ping session to complete
I (4597) esp_idf_svc::ping: From 1.9.168.192 icmp_seq=1 ttl=64 time=3ms bytes=56
I (5581) esp_idf_svc::ping: Ping success callback invoked
I (5581) esp_idf_svc::ping: From 1.9.168.192 icmp_seq=2 ttl=64 time=1ms bytes=56
I (6581) esp_idf_svc::ping: Ping success callback invoked
I (6581) esp_idf_svc::ping: From 1.9.168.192 icmp_seq=3 ttl=64 time=1ms bytes=56
I (7581) esp_idf_svc::ping: Ping success callback invoked
I (7581) esp_idf_svc::ping: From 1.9.168.192 icmp_seq=4 ttl=64 time=1ms bytes=56
I (8581) esp_idf_svc::ping: Ping success callback invoked
I (8581) esp_idf_svc::ping: From 1.9.168.192 icmp_seq=5 ttl=64 time=1ms bytes=56
I (9581) esp_idf_svc::ping: Ping end callback invoked
I (9581) esp_idf_svc::ping: 5 packets transmitted, 5 received, time 7ms
I (9581) esp_idf_svc::ping: Ping session stopped
I (9581) esp_idf_svc::ping: Ping session 0x3fcba7f8 removed
I (9589) esp32test::eth_spi_async: Pinging done
I (9589) esp_idf_svc::eth: Stopping
I (9597) esp_idf_svc::eth: Stop requested
I (9597) esp_idf_svc::eth: EspEth dropped
I (9605) esp_idf_svc::netif: Dropped
I (9605) esp_idf_svc::eth: Stopping
E (9605) esp_eth: esp_eth_stop(309): driver not started yet
I (9613) esp_idf_svc::eth: Stop requested
I (9621) gpio: GPIO[15]| InputEn: 0| OutputEn: 0| OpenDrain: 0| Pullup: 1| Pulldown: 0| Intr:0 
I (9629) esp_idf_svc::eth: Driver deinitialized
I (9629) esp_idf_svc::eth: EthDriver dropped
E (9637) spi_master: spi_master_deinit_driver(355): not all CSses freed

thread 'main' panicked at /home/redux/.cargo/registry/src/index.crates.io-1949cf8c6b5b557f/esp-idf-hal-0.45.2/src/spi.rs:616:52:
called `Result::unwrap()` on an `Err` value: ESP_ERR_INVALID_STATE (error code 259)
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

• Would you like to work on a fix? no

To Reproduce

Just run the sample code.

Expected behavior

Do not crash.

The init will display error, i think this state is not correct, so it causes a crash.

I (1493) esp_idf_svc::eth: Initialization complete
I (1493) esp_idf_svc::eth: Stopping
E (1493) esp_eth: esp_eth_stop(309): driver not started yet
I (1501) esp_idf_svc::eth: Stop requested
I (1509) esp_eth.netif.netif_glue: 02:00:00:12:34:56

Environment

I use idf-esp 5.4.2

   Compiling embuild v0.33.1
   Compiling esp-idf-sys v0.36.1
   Compiling esp-idf-hal v0.45.2
   Compiling esp-idf-svc v0.51.0

running on ubunutu 24.04

[Vollbrecht]

The log seams to indicate that the stop() method was called twice in a row? And that consequently may got things out of order.

[Vollbrecht]

But the above may be unrelated overall.

The double stop happens because here in the EspEth drop handler through detach_netif stop is called first, than a bit later in the clear_all method on the EthDriver a second time stop is called on Drop of the EthDriver.

We can see that one gpio at least was reconfigured again before the error happesn through this line gpio: GPIO[15]| InputEn: 0| OutputEn: 0| OpenDrain: 0| Pullup: 1| Pulldown: 0| Intr:0

What gpio number do you use for the CS line ?

[joyhope]

My board

SCS   10
INT   15

This 15 belongs to INT. I confirmed my board is working correctly in C environment.

I could not understand this code

    fn internal_set_rx_callback<F>(&mut self, callback: F) -> Result<(), EspError>
    where
        F: FnMut(EthFrame) + Send + 'd,
    {
        let _ = self.stop();

Why stop is called?

[ivmarkov]

Why stop is called?

Why do you care? You are not calling set_callback* anyway.