[spi_device] Disallow certain state changes when busy

[a-will]

Description

On a real SPI flash device, many commands are ignored when the device is busy. We might want to consider adding this behavior to our IP as well.

Currently, command uploads, WREN, WRDI, EN4B, and EX4B all are allowed while the busy bit is set.

[a-will]

I have created this primarily for future reference (and evaluation in the D2 sign-off meeting). I don't think this is necessary for the next earlgrey release--It primarily helps for achieving robust behavior when the host is doing something illegal, likely due to a loss of state from a reset or similar.

[a-will]

A similar discussion about this happened here: #12053 (comment)

[andreaskurth]

@a-will: This is currently in M4, but your proposal would be rather to not do this for this release (thus backlog), right?

[a-will]

In the D2 sign-off, I think we decided to just get it done.

[andreaskurth]

@moidx: I think we already have this feature, so this may only be a coverage gap. Needs to be taken into account for DV.

[antmarzam]

@moidx: I think we already have this feature, so this may only be a coverage gap. Needs to be taken into account for DV.

@andreaskurth - This needs doing in both RTL and TB (I just checked with @a-will)

[a-will]

Off the top of my head, I'd probably simply add the current "busy" value as an input to spi_cmdparse, then use it to block the commands that should be ignored. For example, if busy is 1, we could prevent decoding the opcode here:

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

Lines 194 to 195 in c84524e

	assign opcode_en4b = cmd_info_i[CmdInfoEn4B].valid
	&& (data_i == cmd_info_i[CmdInfoEn4B].opcode);

There are likely a number of equivalent means of doing it, though.

[andreaskurth]

Ah, in that case we were wrong about what needs to happen here during today's triage -- sorry! Thanks for clarifying 👍

@antmarzam, as current coordinator for spi_device, may I leave it to you to find someone to do the RTL change (or do it yourself)?