[spi_device,rtl] Consider adding signalling mechanism to abort

[hcallahan-lowrisc]

Description

opentitan/hw/ip/spi_device/rtl/spi_device.sv

Line 204 in 8da2b0d

//logic abort_done; // TODO: Not implemented yet

[msfschaffner]

@a-will is this related to #20654?

[a-will]

@a-will is this related to #20654?

No, this issue appears to be about some kind of OT-triggered abort (which doesn't make sense to me). Notice that it's assigned by CSRs:

opentitan/hw/ip/spi_device/rtl/spi_device.sv

Line 424 in 8da2b0d

assign abort = reg2hw.control.abort.q;

The other issue is about being able to handle the host abandoning SPI TPM transactions before they have completed. The buffering architecture has a poor design for handling such sequences, since it adds ambiguity.

[msfschaffner]

Ah I see. It looks like this is allocated for generic mode, which we plan to deprecate:

opentitan/hw/ip/spi_device/data/spi_device.hjson

Lines 407 to 418 in 143098e

	{ bits: "0",
	name: "ABORT",
	desc: '''Abort pending TX data in Generic mode.
	If TX_FIFO (Asynchronous) is full, the TXF waits indefinitely to
	push the next byte into the asynchronous FIFO. SW may reset the
	Async FIFO along with aborting the current task. SW should update
	the write pointer of the TXF in order not to push the byte to
	Asynchronous FIFO again by TXF logic.
	'''
	resval: "0"
	},

Hence this can be resolved by removing it entirely.

[a-will]

Closing, since we plan to remove firmware mode.