nanocoap/sock: re-try if wrong block was received

[benpicco]

Contribution description

Our go-coap server sometimes responds with block 0 to a block2 request. This breaks SUIT updates.

An easy fix is to just try again when an unexpected block was received, we then get the expected block.

Testing procedure

master

2024-11-05 17:26:32,695 # nanocoap: fetching block 29
2024-11-05 17:26:32,696 # nanocoap: send 20 bytes (4 tries left)
2024-11-05 17:26:32,712 # nanocoap: waiting for response (timeout: 1283696 µs)
2024-11-05 17:26:32,776 # nanocoap: response code=203
2024-11-05 17:26:32,777 # nanocoap got block 29 (offset 29696)
2024-11-05 17:26:32,777 # nanocoap: fetching block 30
2024-11-05 17:26:32,792 # nanocoap: send 20 bytes (4 tries left)
2024-11-05 17:26:32,795 # nanocoap: waiting for response (timeout: 1466309 µs)
2024-11-05 17:26:34,263 # nanocoap: timeout waiting for response
2024-11-05 17:26:34,264 # nanocoap: send 20 bytes (3 tries left)
2024-11-05 17:26:34,279 # nanocoap: waiting for response (timeout: 2937618 µs)
2024-11-05 17:26:34,343 # nanocoap: response code=203
2024-11-05 17:26:34,343 # nanocoap got block 30 (ffset 30720)
2024-11-05 17:26:34,349 # nanocoap: fetching block 31
2024-11-05 17:26:34,359 # nanocoap: send 20 bytes (4 tries left)
2024-11-05 17:26:34,360 # nanocoap: waiting for response (timeout: 1441263 µs)
2024-11-05 17:26:34,423 # nanocoap: response code=203
2024-11-05 17:26:34,424 # nanocoap got block 0 (offset 0)
2024-11-05 17:26:34,424 # Unexpected offset: 4 - expected: 31744
2024-11-05 17:26:34,439 # nanocoap: error fetching block 31: -50

this PR

2024-11-05 17:48:35,054 # nanocoap: fetching block 19
2024-11-05 17:48:35,054 # nanocoap: send 20 bytes (4 tries left)
2024-11-05 17:48:35,070 # nanocoap: waiting for response (timeout: 1168273 µs)
2024-11-05 17:48:35,135 # nanocoap: response code=203
2024-11-05 17:48:35,137 # nanocoap: got block 19 (offset 19456)
2024-11-05 17:48:35,140 # nanocoap: fetching block 20
2024-11-05 17:48:35,150 # nanocoap: send 20 bytes (4 tries left)
2024-11-05 17:48:35,153 # nanocoap: waiting for response (timeout: 1198003 µs)
2024-11-05 17:48:36,351 # nanocoap: timeout waiting for response
2024-11-05 17:48:36,353 # nanocoap: send 20 bytes (3 tries left)
2024-11-05 17:48:36,369 # nanocoap: waiting for response (timeout: 2402006 µs)
2024-11-05 17:48:36,432 # nanocoap: response code=203
2024-11-05 17:48:36,434 # nanocoap: got block 20 (offset 20480)
2024-11-05 17:48:36,436 # nanocoap: fetching block 21
2024-11-05 17:48:36,446 # nanocoap: send 20 bytes (4 tries left)
2024-11-05 17:48:36,449 # nanocoap: waiting for response (timeout: 1430197 µs)
2024-11-05 17:48:36,511 # nanocoap: response code=203
2024-11-05 17:48:36,513 # nanocoap: got block 0 (offset 0)
2024-11-05 17:48:36,514 # nanocoap: fetching block 21
2024-11-05 17:48:36,526 # nanocoap: send 20 bytes (4 tries left)
2024-11-05 17:48:36,527 # nanocoap: waiting for response (timeout: 1429462 µs)
2024-11-05 17:48:37,967 # nanocoap: timeout waiting for response
2024-11-05 17:48:37,968 # nanocoap: send 20 bytes (3 tries left)
2024-11-05 17:48:37,982 # nanocoap: waiting for response (timeout: 2862924 µs)
2024-11-05 17:48:38,031 # nanocoap: response code=203
2024-11-05 17:48:38,047 # nanocoap: got block 21 (offset 21504)
2024-11-05 17:48:38,049 # nanocoap: fetching block 22
2024-11-05 17:48:38,057 # nanocoap: send 20 bytes (4 tries left)
2024-11-05 17:48:38,066 # nanocoap: waiting for response (timeout: 1365115 µs)
2024-11-05 17:48:38,128 # nanocoap: response code=203
2024-11-05 17:48:38,137 # nanocoap: got block 22 (offset 22528)
2024-11-05 17:48:38,139 # nanocoap: fetching block 23
2024-11-05 17:48:38,142 # nanocoap: send 20 bytes (4 tries left)
2024-11-05 17:48:38,148 # nanocoap: waiting for response (timeout: 1033951 µs)
2024-11-05 17:48:38,206 # nanocoap: response code=203
2024-11-05 17:48:38,209 # nanocoap: got block 23 (offset 23552)

Issues/PRs references

[riot-ci]

Murdock results

✔️ PASSED

9b25c03 nanocoap/sock: re-try if wrong block was received

Success	Failures	Total	Runtime
10214	0	10215	17m:31s

Artifacts

• Documentation preview

[maribu]

Our go-coap server sometimes responds with block 0 to a block2 request.

That's not the best motivation.

But networks are not always order preserving and consistent in their latency. I think we are supposed to be prepared that an no longer expected block can arrive out of order even with a correct CoAP server, so this makes sense even for non go-coap users.

[benpicco]

networks are not always order preserving and consistent in their latency. I think we are supposed to be prepared that an no longer expected block can arrive out of order

Usually that would be caught by _id_or_token_missmatch() since we use a new ID for each block that we request.
Here we get a response with the right ID but the wrong block.

[maribu]

Ah, indeed, we do implement duplicate detection on client side.