ot_spi_device: always upload payload for SW flash commands

[jwnrt]

The SPI device design only considers the PAYLOAD_EN bit of a slot for the passthrough mode, not for uploaded flash commands which always upload their payloads.

[AlexJones0]

References for review:

The documentation of the upload field loosely suggests that uploading only really cares about the address, but it doesn't make that very clear:

If upload field in the command info entry is set, the cmdparse activates the upload submodule when the opcode is received. addr_en, addr_4B_affected, and addr_4b_forced (TBD) affect the upload functionality. The three address related configs defines the command address field size.

RTL:

• payload_en is loaded into the cmd_info from the regs: https://github.com/lowRISC/opentitan/blob/e4ce7cf160f508d5787ed2e272872326c9f9159c/hw/ip/spi_device/rtl/spi_device.sv#L641
• The cmd_info is passed through SPI command parsing (which only observes the opcode): https://github.com/lowRISC/opentitan/blob/e4ce7cf160f508d5787ed2e272872326c9f9159c/hw/ip/spi_device/rtl/spi_cmdparse.sv#L253
• spi_passthrough (unimplemented in QEMU) does use the payload_en: https://github.com/lowRISC/opentitan/blob/e4ce7cf160f508d5787ed2e272872326c9f9159c/hw/ip/spi_device/rtl/spi_passthrough.sv#L810 etc.
• SPI read commands use the payload_en to determine the number of bits read from the FIFO at a time: https://github.com/lowRISC/opentitan/blob/e4ce7cf160f508d5787ed2e272872326c9f9159c/hw/ip/spi_device/rtl/spi_readcmd.sv#L472-L486
• SPI upload basically uses none of the cmd_info fields, including ignoring payload_en. It only uses the addr_mode and busy fields: https://github.com/lowRISC/opentitan/blob/e4ce7cf160f508d5787ed2e272872326c9f9159c/hw/ip/spi_device/rtl/spid_upload.sv#L212-L225

[jwnrt]

Changed to check is_upload instead of has_input_payload

[rivos-eblot]

It seems that clang-format complains about the line length on the comments.
This is a known issue: GitLab CI stops whenever ot-format.sh generates a diff, but we have not integrated this is GitHub CI. The check would better be done in ot-format.sh itself I believe.

qemu/.gitlab-ci.d/opentitan/build.yml

Line 66 in 9bc668a

- test -z "$(git status -s)" || git diff

[jwnrt]

Ah my bad, I didn't re-run ot-format after copying this comment!

#226

[rivos-eblot]

Ah my bad, I didn't re-run ot-format after copying this comment!

I keep doing this mistake, I think we need to address it @ CI level.