Implement support for 32-bit memory access scheme #39
Conversation
|
@usbalbin If you could test this on STM32H5 and confirm if this makes things work on H5 as well then that would be great. (Full disclosure: I have not looked at the STM32H5 refrence manual and have no idea if H5 USB implementation is in any way compatible with this driver. I just know from your issue that it uses a 32-bit memory access scheme.) |
|
I will try! |
|
Works great from what I can tell. Tested with this on a nucleo-h533 |
src/endpoint_memory.rs
Outdated
| } else { | ||
| self.mem[index * 2].set(value); | ||
| } | ||
| fn get_mem_slice<T>(&self) -> &mut [VolatileCell<T>] { |
There was a problem hiding this comment.
I believe this function is very unsafe because calling it twice within the same block is unsound.
There was a problem hiding this comment.
That's fair. I realized that I can return &[VolatileCell<T>], which relaxes the safety requirements quite a bit. I did also mark it as unsafe for posterity.
|
Looks good to me overall. I'm happy to approve the PR after changes to |
|
Thank you! |
|
Ok, CI is broken (fixed in #40), so merging the PR after running corresponding checks locally. |
Closes #37 and unblocks USB support for C0 and possibly some other STM32 families. Tested on STM32C071 devboard.
Note: I changed the UsbPeripheral trait a little, which unfortunately makes this a breaking change. In practice this should not be a big issue, since existing HAL-s can keep using the previous version pretty much indefinitely and the migration is very simple.