Re: [PATCH v3 2/2] spi: atmel-quadspi: Add support for sama7g5 QSPI

[Csókás Bence]

Hi,

On 2025. 01. 10. 11:40, Alexander Dahl wrote:
> > This function atmel_qspi_sama7g5_transfer() seems to be called from
> > atmel_qspi_exec_op() through ops->transfer() only.  I think the two
> > lines in the error handling of atmel_qspi_reg_sync() lead to
> > unbalanced calls of pm_runtime_xxx.  Compare with
> > atmel_qspi_transfer() which has no calls to pm_runtime, everything is
> > covered by atmel_qspi_exec_op() in this case where the pm_runtime
> > calls surround ->set_cfg() and ->transfer().  Right?
>
> This problem has been addressed in downstream kernel (linux4sam) by
> Claudiu Beznea back in 2023 already:
>
> https://github.com/linux4sam/linux-at91/commit/e59f646f516088fdab6d8213d8acda0c1063ec21

You are correct. I have picked it and will submit it on the ML shortly.

> > The whole call tree from atmel_qspi_sama7g5_setup() downwards is not
> > covered by pm_runtime get and put calls, although heavily doing i/o.
> > Further down in atmel_qspi_setup() there's a write to QSPI_SCR which
> > seems to be handled correctly.
>
> Same for this:
>
> https://github.com/linux4sam/linux-at91/commit/5ff0e74c1d548599fe85113e2f1817cb8a052b15
>
> Some hunks of that seem to have made it to upstream, not sure?

Yes, it seems to be an amalgam of patches from mainline, and original work.

> Maybe Microchip should upstream those fixes, now that SAMA7G5 support
> was ported to mainline?

I would prefer that as well.

> Greets
> Alex