qca_spi: SPI_REG_SIGNATURE sometimes not recognized

qca_spi: SPI_REG_SIGNATURE sometimes not recognized

[Christoph Fritz]

Hello Stefan,

 working on a board with QCA7005, on probe() SPI_REG_SIGNATURE
sometimes fails (~1 out of 5 times) to be recognized correctly, even
after multiple reads and retries.

Any suggestions?

Thanks
 -- Christoph

Re: qca_spi: SPI_REG_SIGNATURE sometimes not recognized

[Stefan Wahren]

Hello Christoph,

Am 04.02.24 um 22:23 schrieb Christoph Fritz:
> Hello Stefan,
>
>   working on a board with QCA7005, on probe() SPI_REG_SIGNATURE
> sometimes fails (~1 out of 5 times) to be recognized correctly, even
> after multiple reads and retries.
at the time of writing the driver, i assumed the QCA7000 is always
powered up (takes ~ 1 second) during probe of the driver and the driver
could do the signature check during probe. But this doesn't work in all
cases. So qca_spi driver has the module parameter qcaspi_pluggable,
which is disabled per default. If you enable this, this will skip the
signature check during probe and let the spi thread handle the
synchronization.

Does this fix your problem?

Regardless of the results, i think qcaspi_pluggable should be enabled
per default.

Regards

>
> Any suggestions?
>
> Thanks
>   -- Christoph

Re: qca_spi: SPI_REG_SIGNATURE sometimes not recognized

[Christoph Fritz]

> >   working on a board with QCA7005, on probe() SPI_REG_SIGNATURE
> > sometimes fails (~1 out of 5 times) to be recognized correctly, even
> > after multiple reads and retries.

> at the time of writing the driver, i assumed the QCA7000 is always
> powered up (takes ~ 1 second) during probe of the driver and the driver
> could do the signature check during probe. But this doesn't work in all
> cases. So qca_spi driver has the module parameter qcaspi_pluggable,
> which is disabled per default. If you enable this, this will skip the
> signature check during probe and let the spi thread handle the
> synchronization.
> 
> Does this fix your problem?

Yes, thanks - this seems to fix the probe issue.

> Regardless of the results, i think qcaspi_pluggable should be enabled
> per default.

I think so too. Because the current default (off) is only appropriate
when the chip already gets powered up long before probing. This seems
unusual and at the same time understandable because the driver lacks
regulator and gpio-reset handling.

What about removing qcaspi_pluggable completely while adding a warning
when QCASPI_SYNC_UNKNOWN hits way too many times?

Thanks
  -- Christoph