ci: Enable github CI for developers forks #21
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Currently, if lowpan_adaptation_queue_size() returns > 2, the tun interface is not read. However, poll() still wake up the main loop. This behavior lead to a useless wake up of the main loop and a waste of CPU resources. This issue typically happens during throughput tests when the Linux stack send frames faster than the RCP is able to send. See WI_SUN-3494. Signed-off-by: Jérôme Pouiller <[email protected]>
Only the local copy of the variable was incremented.
tcflush() does not work for this purpose as it discards unsent bytes.
This prevents frame cancelation, which creates annoyances regarding framing detection on the RCP.
APIv2 has an additional field to allow for a different payload size between the request and the confirmation. This oversight went unnoticed because the way the payload is contructed starts with 00 01 02..., which was interpreted as "a 1 byte buffer with contents 0x02". Testing the TX mode makes the issue apparent as the RCP expects to read payload length of 0 but instead finds that there are not enough bytes.
With --no-daemon, dnsmasq outputs logs to both stderr and syslog, which results in line duplication in journalctl. Using --keep-in-foreground fixes this issue.
Since PAN Advertisements transmission can be canceled by trickle, it is best to also include JM-IEs in data frames.
WI_SUN-3496
Aparently this is how the API is supposed to work, but it makes the DELETE endpoint very confusing and/or useless.
Aparently this is how the API is supposed to work, but it makes the DELETE endpoint very confusing and/or useless.
Reviewed-by: Kylian Balan <[email protected]>
Reviewed-by: Kylian Balan <[email protected]>
The hop limit selected at a tunnel entry is a local configuration of the
ingress, and has no relation with the received hop limit of the packet
to encapsulate. RFC 2473 section 6.3 describes the use of hop limit in
the case of tunnels in more details.
IP tunnels are normally invisible to tools like traceroute since an
increment of the probe hop limit covers the whole tunnel. However with
this quirk, traceroute needs to increase to hop limit for every
intermediate Wi-SUN router in the tunnel, but it also does not receive
ICMP error messages since they are destined to the ingress.
Here is an example to demonstrate the difference in behavior:
+ 2001:db8::1
Backhaul--| | Ethernet link
+ 2001:db8::2 (eth0)
| Linux + wsbrd
+ fd12:3456::200:0:80:3 (tun0)
| | Radio
| fd12:3456::200:0:80:0
Wi-SUN----| | Radio
| fd12:3456::200:0:80:1
| | Radio
+ fd12:3456::200:0:80:2
Executing traceroute from the backhaul device:
Before:
$ traceroute fd12:3456::200:0:80:2
traceroute to fd12:3456::200:0:80:2 (fd12:3456::200:0:80:2), 30 hops max, 80 byte packets
1 2001:db8::1 (2001:db8::1) 0.323 ms 0.368 ms 0.280 ms
2 * * *
3 * * *
4 fd12:3456::200:0:80:2 (fd12:3456::200:0:80:2) 40.184 ms 43.054 ms 45.506 ms
After:
$ traceroute fd12:3456::200:0:80:2
traceroute to fd12:3456::200:0:80:2 (fd12:3456::200:0:80:2), 30 hops max, 80 byte packets
1 2001:db8::1 (2001:db8::1) 0.353 ms 0.393 ms 0.318 ms
2 fd12:3456::200:0:80:2 (fd12:3456::200:0:80:2) 19.228 ms 25.906 ms *
WI_SUN-3547
Unfortunately, sigdescr_np() is only available since glibc v2.32, which is not provided by older distros such as Raspbian bullseye which packages glibc v2.31.
After receiving an EAP reply from a target, this reply is forwarded to the radius server and the state machine transitions to the EAP_TLS_STATE_EAP_REQUEST state. If an incorrect frame is received from the same target while in this state, we would wrongly try to apply an offset to skip the socket and EAPOL headers. This would result in a segmentation fault. WI_SUN-3691
Exposing them would not make sense considering such FFNs are not operational yet. See the updated CHANGES.md. WI_SUN-3747
For most functions, CPC documents:
On error, a negative value of errno is returned.
Co-authored-by: Noé Masson <[email protected]> WI_SUN-3761
WI_SUN-3744
Note that with the current RCP API (2.4.1), there is no way to disable the destination address filtering, and the filter starts enabled with the hardware specific address. WI_SUN-3744
WI_SUN-3765
WI_SUN-3765
This new link has restricted access, but it turns out that the original swagger page was unofficial.
Keeping a "fragmenter_tx_entry_t" with field "buf" as NULL basically results in a corrupted entry in activeUnicastList, causing a SegFault in functions that try to process the buffer. Dropping the fragment altogether seems like the better thing to do. WI_SUN-3769
The max size is given as a STR_MAX_LEN_EUI64, but it really should be STR_MAX_LEN_EUI48 Origin: SiliconLabs#18
Valgrind reported:
==14914== 948,792 (4,896 direct, 943,896 indirect) bytes in 136 blocks are definitely lost in loss record 85 of 85
==14914== at 0x486AC6C: calloc (vg_replace_malloc.c:1328)
==14914== by 0x4908055: nl_cache_alloc (in /usr/lib/libnl-3.so.200.26.0)
WI_SUN-3772
Fixes: 279ba86193 ("ws: update regulation with values from Wi-SUN TPS 2v01")
The struct returned by rtnl_neigh_get() gets its reference count incremented so it must be explicitly unreferenced by calling rtnl_neigh_put(). WI_SUN-3772
Enable to setup a reference system for building, it can be used for various automation (eg: github actions) Most of changes are isolated in this simple makefile that more or less script what has been described in readme file Signed-off-by: Philippe Coval <[email protected]>
Just run the native helper script in docker, it is not optimized for bandwidth but minimize duplication of changes. Signed-off-by: Philippe Coval <[email protected]>
Enable github actions that build any branches using docker along helper script. To reduces the dependency on github and avoid bringing more ambiguity or complexity most tasks are isolated at lowerlevel (in helper for cmake). More checks to come next, unit tests, lint etc. Relate-to: https://github.com/rzr/wisun-br-linux/actions/runs/11687474586/job/32545660605 Signed-off-by: Philippe Coval <[email protected]>
|
Very nice! However, I wonder if it make sense to involve docker and a makefile for a such simple CI. I assume we could do the same with only |
|
well it is developer preference,. IHMO Depending on docker is not ideal, but supporting it is nice (again it can be a reference that can help troubleshooting), I think there are benefits to have minimal (like this one) docker file in tree, something nice for DX one can build directly with a oneliner: |
|
Hello, I have re-worked your proposal to not use Docker nor a bash script. I believe it is more suited to include everything in the workflow file in our case. I have also added a final step showing |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Check status in source project:
https://github.com/rzr/wisun-br-linux/tree/phcoval/ci/review/main
https://github.com/rzr/wisun-br-linux/actions/runs/11687571871/job/32545978157