RFC draft - patch to add IrDA SIR mode support to the usart code.

[tim-seoss]

A few queries about this:

• Should it be behind a feature flag?
• Should IrdaMode be replaced with SpecialMode or similar, and support all of IrDA, LIN and Smartcard?

Any comments welcome!

n.b. I'm seeing some weird behaviour in transmit mode (missing or truncated start bit pulse) - which I'm currently debugging - might be a clock setup issue outside of the serial driver code.

[burrbull]

• Should it be behind a feature flag?

In release mode it should be optimized out if not used. So I think No.

[tim-seoss]

Re f405 tests failing - the docs say that register should be present, so maybe an SVD error? Is that likely?

[burrbull]

Re f405 tests failing - the docs say that register should be present, so maybe an SVD error? Is that likely?

No. It present in usart1.
There is bigger problem. It is because of:

stm32f4xx-hal/src/serial.rs

Line 695 in 7dced36

use crate::pac::uart4 as uart_base;

[tim-seoss]

Re f405 tests failing - the docs say that register should be present, so maybe an SVD error? Is that likely?

No. It present in usart1. There is bigger problem. It is because of:

stm32f4xx-hal/src/serial.rs

Line 695 in 7dced36

use crate::pac::uart4 as uart_base;

OK, thanks, so this needs to be conditionally not compiled for uarts (as opposed for usarts) [edit: this was incorrect - see below - the SVDs were missing the IrDA fields for UARTs - see https://github.com/stm32-rs/stm32-rs/pull/713 ], I'll take a look at that once I've worked out what's going on with the mystery of the missing start bits.

[burrbull]

OK, thanks, so this needs to be conditionally not compiled for uarts (as opposed for usarts), I'll take a look at that once I've worked out what's going on with the mystery of the missing start bits.

I don't think it will be easy with current implementation.
Looks like it is time to split implementations of USARTs and UARTs.

[burrbull]

I think I found some workaround witch don't requires total rework.

Instead of adding this to new just create other constructor function (for example irda) with advanced restriction:

USART: Instance + core::ops::Deref<Target = crate::pac::usart1::RegisterBlock>

Something like ab28a0a

[burrbull]

Also in this case you don't need IrdaMode::None so you can optimize it.

[tim-seoss]

The plot thickens...

According to RM0090 30.5 USART mode configuration (page 1007 on rev 19), IrDA mode is supported on UART peripherals as well USART peripherals.

30.6.7 Guard time and prescaler register (USART_GTPR) (p. 1017) suggests that this register is unavailable on UARTs, which would (I think) make IrDA low power mode non-functional on UARTs (but there's no mention of that elsewhere in the documentation). I have an f446 here, so out of curiosity, I'll do some experimenting with low power mode to see if the register is really not there...

n.b. The GTPR register is also used for Smartcard mode - which is not present on UART peripherals.

[tim-seoss]

I noticed that the GTPR register is included in the UART4 peripheral of the stm32f413, but is missing on the other models which include UART4: https://stm32-rs.github.io/stm32-rs/stm32f/stm32f4/UART4_0x40004C00.html

I only have an stm32f446 here, so I tried writing to the GTPR register to see if it would have any effect. It works exactly as expected (writing to the PCS bits of the location where the GTPR reg "should" be does change the transmit pulse length in IrDA low-power mode, exactly as it does on USART peripherals on the 446, and on the 411), so I suspect that this omission is just an SVD bug, and I'll submit a PR to stm32-rs.

As to why the register is missing from the SVDs... The GTPR register is used in IrDA mode, and also in Smartcard mode. The UART peripheral includes IrDA support, but doesn't include Smartcard support, so I think it was accidentally omitted from the SVD because of this (the wording in the reference manual is ambiguous, so maybe this is the original source of the error).

[tim-seoss]

PAC issue:
blocked-by: stm32-rs/stm32-rs#711