Ambiguity in use of cs_wait field in SPI_CMD_SOT vs. wait_cyc from SPI_CMD_WAIT in udma_spim_ctrl

[cst-aditya]

As per the RTL implementation, The SPI_CMD_SOT command implementation in udma_spim_ctrl can consume a prgrammable delay defined by SPI_CMD_WAIT. This implementation seems confusing for the below reasons: -

a) SPI_CMD_SOT has a cs_wait field that is not used anywhere; rather, SPI_CMD_WAIT is used.

b) If it is correct to use SPI_CMD_WAIT instead of the cs_wait field of SPI_CMD_WAIT, then there should be code to check whether the cd_wait_cyc are for clock-based wait or event-based wait. In the case of the SPI_CMD_SOT command, the wait_cyc field of SPI_CMD_WAIT is used blindly.

c) Using the wait cycle from the SPI_CMD_WAIT command instead of the SPI_CMD_SOT command puts an additional constraint that the SPI_CMD_WAIT command should come before the SPI_CMD_SOT command. Using the wait field of SPI_CMD_SOT should have been sufficient here.

Kindly review and confirm the implementation. It is not clear whether this is intentional or a typo.

[MikeOpenHWGroup]

Good catch @cst-aditya. At first glance, I am inclined to agree with you. However, I'd like to spend more time with this before labelling it a bug. For now it is labelled a question.

[cst-aditya]

Hi @MikeOpenHWGroup , Thank you for the feedback. For now I have documented it as per the current RTL implementation. Once we have closure on the issue, we can plan to take the necessary action.

[MikeOpenHWGroup]

It seems that the SPI controller is actually a JEDEC SPI NOR Flash controller. We should make sure this is explicitly mentioned in the User Manual. I also think you are right that the wait_cyc field of SPI_CMD_WAIT should not be used blindly. However, that is the current implementation and it would require a significant effort to fully verify any change we make. My recommendation is to document this behavior in the User Manual in a "note". Something like this:

> [!NOTE]
> There should be code to check whether the cd_wait_cyc are for clock-based wait or event-based wait.
> However, in the case of the SPI_CMD_SOT command, the wait_cyc field of SPI_CMD_WAIT is used blindly.

(Hopefully you can think of better wording!)

[cst-aditya]

Thank you for providing the additional details. I will take care of document updates!!

Do you have the technical specs of the JEDEC SPI NOR Flash controller? If yes, could you please share it with us? If not, could you please arrange it?

Thank you for continuous support and suggestions.

[MikeOpenHWGroup]

I just emailed you a couple of specs.

[cst-aditya]

Thank you for providing the specifications; they are quite helpful.

I have added the below warning in the user manual:

In the current implementation, the WAIT_CYC field of the SPI_CMD_WAIT command is used in place of the CS_WAIT field of the SPI_CMD_SOT command. When applied in the context of SPI_CMD_SOT, the WAIT_CYC field must, at all times, represent the required wait cycles.