WiFi stack can freeze if CONFIG_ESP32_WIFI_RX_BA_WIN > CONFIG_ESP32_WIFI_STATIC_RX_BUFFER_NUM (IDFGH-1383)

[ssilverman]

Suggest you change the issue to this:

WiFi stack can freeze if CONFIG_ESP32_WIFI_RX_BA_WIN > CONFIG_ESP32_WIFI_STATIC_RX_BUFFER_NUM.

Tested on esp-idf v3.2

This configuration should be prevented or checked and an error thrown.

Originally posted by @negativekelvin in #3646 (comment)

[ssilverman]

See also: espressif/arduino-esp32#2899

[r1dd1ck]

Well, are you sure it's this?

Because the iperf example uses:
WIFI_STATIC_RX_BUFFER_NUM=16
WIFI_RX_BA_WIN=32

and it does not "freeze" 😐

[negativekelvin]

Well it solves the problem reported in espressif/arduino-esp32#2899 but maybe only happens "under certain conditions" or certain values so needs more constraints from wifi stack developers

[r1dd1ck]

Yes, I certainly agree that changing the values "fixes" the reported issue 👍

But what I was pointing at is, whether the supplied conditional really is the underlying cause of the freezing stack, because it is in direct conflict with the iperf example defaults (which doesn't seem to show the symptoms).

Other than that, all is fine 🙂

[ssilverman]

This issue still happens in the v1.0.3 Arduino core: espressif/arduino-esp32#3287
There, CONFIG_ESP32_WIFI_RX_BA_WIN = CONFIG_ESP32_WIFI_STATIC_RX_BUFFER_NUM = 16.

[ssilverman]

For those following along, the stack now fails to receive more packets than defined by CONFIG_ESP32_WIFI_DYNAMIC_RX_BUFFER_NUM. Doing some spelunking for the word dynamic_rx_buf_num (the consumer of this define, as found with a simple grep) led me to tools/sdk/lib/libesp32.a, that's why I think the issue is in esp-idf as opposed to just in the Arduino core. (Forgive me if that cursory assumption is incorrect.)

[espxiehang]

@ssilverman hi,I use your script to send upd packets, but I didn't reproduce the issue.
Can you give me an easy to reproduce method and your log?

[ssilverman]

When I filed this and related issues, this was the behaviour. However, the later Arduino releases seem to fix the problem. I haven’t dived in, but if you use settings from 1.0.2 (I think? See the issues in that repo that link to this one.), then you’ll see this behaviour. I realize this repo isn’t Arduino, and I haven’t explored much in the non-Arduino space of the esp-idf repo, but I think you’ll find the behaviour by using the settings mentioned in those Arduino versions.

There’s a bunch of related issues, but see this one (including the links there): espressif/arduino-esp32#2899
Also see here, where the comment author suggested I file this issue: #3646 (comment)

[negativekelvin]

I think this can be closed with dd26caf#diff-fdc67a5904b65c3885aad20459aa1cef

[ssilverman]

@negativekelvin was the source of this problem found?

[negativekelvin]

Yes, enforce restrictions on the value of CONFIG_ESP32_WIFI_RX_BA_WIN per the commit. The other issues are all outside the scope of esp-idf code so I think you can close this. Unless you want to request backport to another version.

[ssilverman]

Right, but I meant: why does not meeting those conditions cause the problem? Is there a deeper problem somewhere? Should a comment in the code that uses these values make note of a fault potential? I’m just suggesting another commit that addresses that problem or at least makes a note of it somewhere. Other than that, I can certainly close this if you’re saying it’s fixed.

@r1dd1ck what do you think?

Edit: I re-read what you wrote and you stated that the real problem is outside esp-idf. I’ll close. But what do you think of @r1dd1ck’s comment that this disagrees with iperf?

[negativekelvin]

The actual conditions in the commit are

(CONFIG_ESP32_WIFI_RX_BA_WIN ≤ CONFIG_ESP32_WIFI_DYNAMIC_RX_BUFFER_NUM) && (CONFIG_ESP32_WIFI_RX_BA_WIN ≤ 2 * CONFIG_ESP32_WIFI_STATIC_RX_BUFFER_NUM)

so iperf is not in violation

[espxiehang]

Yes, its rules are defined in “components/esp_wifi/src/wifi_init.c”