-
Notifications
You must be signed in to change notification settings - Fork 60
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Make DMA-based AES API more ergonomic #19
Conversation
@arkorobotics Yes. #16 is fully ready. This PR (#19) still includes the commit from #16, but I'm marking it "Ready for review", as the one shared commit shouldn't be too much of a hindrance in reviewing. |
Sounds good, how would you like to resolve these merge conflicts? |
They shouldn't be there :-) I'll take a look later today! |
I'm working on some changes that require `Transfer::new` to be opened up to the rest of the crate, so I think it makes sense to not require weird PAC types when calling it.
The problem with these constructors is that they're not flexible enough. In general, peripheral APIs might want to present DMA transfers to the user in a different way than they work in hardware. For example, the AES peripheral supports only transfers of 32-bit words, but might want to present an API that works with byte slices (I'm currently working on this). Another example would be SPI, which might want to hide two simultaneous DMA transfers (for read and write) behind a single transfer struct. This commit doesn't make these use cases possible yet, as some modifications to the unsafe constructor are still required. But it paves the way, by exposing that constructor to the peripheral APIs. Please note that none of this exposes any unsafety to the users. They will still use the same (safe) peripheral APIs to access DMA.
The new trait, `Buffer`, provides everything the DMA code needs from the buffer, without requiring implementers to be able to return a slice. This makes the trait more flexible, as it can be implemented for a larger variety of types, without requiring any complex transmuting.
The type parameter names were wrong, as they were in a different order than was used everywhere else in the code. Please note that this didn't cause any incorrect behavior, as just the names were wrong and nothing else.
This should be more natural than operating with `u32` words.
I don't know what GitHub's problem is. This rebased flawlessly, no conflicts in sight. Anyways, all should be well now. |
This pull request changes the AES API to work with
u8
buffers, instead ofu32
buffers, which should make it easier to use in most scenarios.I've also snuck in a few commits that fix and improve some other things related to AES and DMA. It didn't seem worth it to create a separate pull request for those, especially since I've already got a bit a bit of a traffic jam going :-)
This pull request is based on #16. I recommend reviewing/merging that one first.
cc @lthiery