ping fails when going via a gateway(?)

[shish]

I have a simple app which connects to wifi, pings an IP address, and lights up an LED based on how fast the response is.

• In the wokwi simulator, the device gets a 10.x.x.x IP address and can ping 1.1.1.1 and get responses just fine
• On an esp32-c3-devkit-rust-1 (risc-v), as well as a esp-wroom-32 (xtensa), I can ping hosts on the local network (192.168.1.x), but I can't ping hosts on the internet (1.1.1.1)

Running tcpdump on my access point, I notice something weird.

Here's the tcpdump when my laptop pings 1.1.1.1 (all good):

22:06:15.882597 IP 192.168.1.96 > 1.1.1.1: ICMP echo request, id 5273, seq 1, length 64
22:06:15.922553 IP 1.1.1.1 > 192.168.1.96: ICMP echo reply, id 5273, seq 1, length 64
22:06:16.885267 IP 192.168.1.96 > 1.1.1.1: ICMP echo request, id 5273, seq 2, length 64
22:06:16.927101 IP 1.1.1.1 > 192.168.1.96: ICMP echo reply, id 5273, seq 2, length 64
22:06:17.886193 IP 192.168.1.96 > 1.1.1.1: ICMP echo request, id 5273, seq 3, length 64
22:06:18.037990 IP 1.1.1.1 > 192.168.1.96: ICMP echo reply, id 5273, seq 3, length 64

And here's the tcpdump when my esp32 pings 1.1.1.1:

22:23:05.438439 IP 192.168.1.23 > 1.1.1.1: ICMP echo request, id 19580, seq 256, length 64
22:23:05.438661 IP 192.168.1.1 > 192.168.1.23: ICMP time exceeded in-transit, length 92
22:23:07.946762 IP 192.168.1.23 > 1.1.1.1: ICMP echo request, id 19580, seq 512, length 64
22:23:07.947317 IP 192.168.1.1 > 192.168.1.23: ICMP time exceeded in-transit, length 92
22:23:10.446840 IP 192.168.1.23 > 1.1.1.1: ICMP echo request, id 19580, seq 768, length 64
22:23:10.447511 IP 192.168.1.1 > 192.168.1.23: ICMP time exceeded in-transit, length 92
22:23:12.946633 IP 192.168.1.23 > 1.1.1.1: ICMP echo request, id 19580, seq 1024, length 64
22:23:12.947167 IP 192.168.1.1 > 192.168.1.23: ICMP time exceeded in-transit, length 92
22:23:15.446674 IP 192.168.1.23 > 1.1.1.1: ICMP echo request, id 19580, seq 1280, length 64
22:23:15.447205 IP 192.168.1.1 > 192.168.1.23: ICMP time exceeded in-transit, length 92

I see two things: the sequence numbers appear to be shifted by a couple of bits, and the only response is TTL timeout. I wonder if something is shifting the bits in the packet causing the sequence number to be too high and the TTL to be too low?

Here is my code: https://github.com/shish/esp-ping-leds/blob/main/src/main.rs

[shish]

Actually looking at the code in fn run_ping - I notice that the esp_ping_config_t doesn't have ttl specified, so it uses the ..Default::default() value, which I guess is zero? Perhaps just setting the ttl to non-zero would fix it...

[shish]

Yup, setting the TTL to 64 works. I have no idea how ttl=0 works in the wokwi simulator o_O

(I also have no idea what's up with the sequence numbers, maybe that's unrelated and harmless?)

[shish]

Fixed by #319