Prepare LPSPI driver for `embedded-hal 1.0` rework

Finomnis commented 6 months ago

Some fixes/modifications to the existing LPSPI driver.

This is in preparation for the embedded-hal 1.0 rework, which will introduce a new high level driver that wraps this low level one.

Changes:

Remove CS_PIN from Spi device
- Rationale: eh1 primarily uses software CS pins. There is an extension for hardware CS pin support, but I think we should add it later as a special case instead of by default. This will simplify API design a lot.
Split Spi into Spi + SpiCsPin in the board subcrate
Add initial state to gpio::Output constructor. This is important for cases where the state needs to be set before enabling the pin.
Fix/rework check for valid frame sizes (existing implementation did not cover all corner cases)
Rework SPI clock initialization
- Make sure we never get a baud rate faster than specified (slower is pretty much always ok for peripherals)
- Configure PCSSCK and SCKPCS properly for contiguous transfers
Some improvements in reset logic
Add soft_reset that cancels the current transfer without requiring a full re-initialization
Move more values into enqueue_transaction so that multiple transactions with different settings can be enqueued in series
Add flush function that waits for the current transaction to be finished whilst repeatedly checking for errors
Fix disabled() not waiting for the current transfer to be finished
Take set_watermark out of disabled, because it doesn't require it.

Finomnis commented 4 months ago

@mciantyre Sorry that I was gone for a while. Don't know how much time I will be able to spend on the SPI device, but I'll try to do a little :)

I finally finished this PR. I didn't update examples yet, but I think there's no point really in updating them until we have a properly reworked driver.

Any feedback?

Finomnis commented 4 months ago

@mciantyre Ping :)

teburd commented 3 months ago

I’d say squash this and I’ll merge, changes look good. Ian is presumably busy with other endeavors at the moment.

Finomnis commented 3 months ago

@teburd I don't think I agree time to continue this any time soon, so let's keep that one open because it's a breaking change.

mciantyre commented 2 months ago

There's useful non-breaking changes in this branch, changes that could help today's users. I extracted them into #156.

Finomnis commented 2 months ago

@mciantyre Welcome back! Glad to see your life has cooled down enough to dabble in some Rust again :)

imxrt-rs / imxrt-hal

Prepare LPSPI driver for `embedded-hal 1.0` rework #149