capsules: ensure driver existence check #3613
Conversation
|
It's great that this is a comprehensive change because we do want cmd 0 to be a way to check if the driver exists. We don't want people starting from capsules that do not follow this convention so it's great to remove this broadly. We can't change the cmd==0 for the drivers with a check in the "2.0" column here: https://github.com/tock/tock/blob/master/doc/syscalls/README.md As the comments indicate we can't change those until Tock 3.0. For the others we can change, and I think we should.
I think generally using "1" would make sense, but I don't think it's worth changing existing drivers to match. Just adding them as the next available number makes sense to me for existing drivers. |
|
@Samir-Rashid anything that is just a change in comments seems good. Anything that adapts the behavior of non-stable driver (as @bradjc suggests) is fine. In lieu of finality on #3375, I suggest removing any functional changes to the stable drivers from this PR and then we can basically merge almost immediately IMO. As I noted in #3375 just now, it might be the best of all worlds to adapt TRD104 to allow these existing behaviors in the stable drivers, so those implementations may stay the same anyway (of course, that's just my stated opinion for now, not a conclusion). |
|
As I reverted stable drivers, I noticed there is no PWM driver listed. It should be there as it has an allocated driver number 10 (in hex?). |
Good catch, feel free to add that in this PR or another standalone one! |
|
Apology for delay. Here is a diff of what I changed in case GitHub gets confused when I squash again. I made accompanying PRs tock/libtock-c#349 and tock/libtock-rs#517 |
Add driver liveliness checks as defined in TRD104 4.3.1. Rename "driver check" to "driver existence check" as per tock#2908. "Driver check" falsely implies some sort of functionality checking. Note that stabilized capsules still need to be changed as noted in the TODOs for the Tock 3.0 release.
There was a problem hiding this comment.
I think this is looking pretty good. It's a lot of changes, but it's a nice consistency pass over a pretty old part of the kernel interface. It doesn't look like any stabilized interfaces are changed (though some of the documentation gets small updates in the name of consistency—which is good IMO).
| **Returns**: `Ok(())` if starting interrupts was succesful. | ||
|
|
||
| * ### Command number: `4` | ||
| * ### Command number: `3` |
There was a problem hiding this comment.
Just noting for myself: this is not a change per se, but a fix of a documentation bug
Pull Request Overview
The driver liveliness check is defined in TRD104 4.3.1. I have updated all drivers in advance of Tock v3 (#3375). Make code comments and documentation consistent with regards to "driver existence check" terminology -- change as per #2908. "Driver check" falsely implies some sort of functionality checking.
Testing Strategy
Unsure how to test. These are breaking changes.
TODO or Help Wanted
There seems to be an established convention that command code 0 should return the number of things the driver is controlling. I moved those commands to the next available command code. Should these syscalls have some standardized code number?
Documentation Updated
/docs, or no updates are required.Formatting
make prepush.