interface for changing bitrate

[cperkulator]

Quick implementation of changing the bitrate at runtime. I just want to get a feel if this is the interface we would like or if there is something better.

I just implemented the Vector interface for now but I can add socketcan eventually (would close #549 ).

This might be a start for implementing autobaud as well.

TODO

• need to add error handling

[codecov[bot]]

Codecov Report

Merging #1030 (48b8f8c) into develop (7389b23) will decrease coverage by 2.12%. The diff coverage is 39.21%.

@@             Coverage Diff             @@
##           develop    #1030      +/-   ##
===========================================
- Coverage    66.51%   64.38%   -2.13%     
===========================================
  Files           78       79       +1     
  Lines         7561     7651      +90     
===========================================
- Hits          5029     4926     -103     
- Misses        2532     2725     +193     

[felixdivo]

Wouldn't it be nice to implement a more generic change_parameters() method that takes any parameters (just like on bus init)? The method can then complain whether that specific setting is supported to be changed or not.

[felixdivo]

Your proposal for implementing it as a property looks really nice! However, as a change to the core class of the library this should get more feedback.

(Side note: This could also be used in other places since it modularizes the configuration of the bus. Sometimes, that might make things more readable.)

[zariiii9003]

Some questions:

• How would you deal with CAN FD?
• Regarding the Vector interface: what happens if you do not have Init Access?

[felixdivo]

• How would you deal with CAN FD?

I guess we would not have to implement it right away. Besides, wouldn't the bitrate be handled automatically?

• Regarding the Vector interface: what happens if you do not have Init Access?

I'd propose to simply throw one of the new exceptions, like a CanOperationError.

[cperkulator]

• How would you deal with CAN FD?

I guess we would not have to implement it right away. Besides, wouldn't the bitrate be handled automatically?

This will only set the arbitration bitrate. Would probably need a separate property for the data bitrate.

• Regarding the Vector interface: what happens if you do not have Init Access?

I'd propose to simply throw one of the new exceptions, like a CanOperationError.

Check out the latest commit. I think it should handle these issues. I still need to add unit tests but testing from home this all works fine. I'll need to test on my work setup too next week.

[zariiii9003]

@hartkopp and @lumagi : What do you think about the socketcan part of this PR? Is that a viable solution to implement bit timing configuration for socketcan?

[hartkopp]

@hartkopp and @lumagi : What do you think about the socketcan part of this PR? Is that a viable solution to implement bit timing configuration for socketcan?

The bitrate setting of 'real' CAN interfaces can only be performed as network-admin (aka root) on Linux. That's why the ip tool is mostly used with sudo ip ...

And the setting affects all users on the system (not only python-can users), as the interface has to be shut down to change the bitrates (and has to be set to "up" afterwards). Shutting down the interfaces leads to -ENETDOWN return values for all user space applications working on that interface. Additionally the python-can application needs to be run as root to manipulate the interface status and the bitrate.

So reading the interface properties via ip tool or the netlink socket is fine - but setting new bitrates from python-can is probably not as feasible as suggested by the patch.

[zariiii9003]

@hartkopp Thank you for your input. It's always neat to have the expert here 👍 It also seems like libsocketcan is not maintained anymore. The last commit is from 2020 and it does not support CAN FD. So it's probably not a good foundation to build upon.

[hartkopp]

It also seems like libsocketcan is not maintained anymore. The last commit is from 2020 and it does not support CAN FD. So it's probably not a good foundation to build upon.

@marckleinebudde : Is there any ongoing activity planned for libsocketcan ?

[lumagi]

I agree with @hartkopp. The executing process must be root or at least have the NET_ADMIN capability to change the interface. Neither one of those things is something you want to do to your Python interpreter on a regular basis.

[marckleinebudde]

There is no immediate activity planned for libsocketcan, but patches are welcome. However I'd rather have a proper (dbus?) interface for systemd-networkd to configure CAN (and networking in general).