On 10.12.2024 17:03:06, Stefano Offredi wrote:
> Il giorno mar 10 dic 2024 alle ore 16:28 Marc Kleine-Budde
> <mkl@xxxxxxxxxxxxxx> ha scritto:
> >
> > On 10.12.2024 16:18:07, Stefano Offredi wrote:
> > > > > Even with those lines of code, If I print spi->irq before the call
> > > > > to request_threaded_irq() it keeps different error codes
> > > > > values (-19, -22, depending on whether I set the gpio controller
> > > > > values in ACPI table).
> > > >
> > > > Just to clarify: from the Linux point of view "rx-int" is a GPIO, not an
> > > > interrupt. It's connected to the "nINT1/GPIO1" pin of the mcp251xfd and
> > > > this is optional. The IRQ which is available under spi->irq is connected
> > > > to the "nINT" pin of the mcp251xfd.
> > > >
> > > > > Even If I remove in ACPI table all the gpio references listed
> > > > > here below:
> > > > >
> > > > > - GpioInt (Edge, ActiveLow, ExclusiveAndWake, PullDefault...
> > > > > - Package () {"rx-int-gpios", Package () { ^CAN0, 0, 0, 0 } }
> > > > >
> > > > > the spi->irq keeps error value -2.
> > > >
> > > > I suggest to first remove all "rx-int" related stuff from your ACPI
> > > > table and concentrate that "spi->irq" has a proper value.
> > > >
> > > > The relevant code is in acpi_status acpi_register_spi_device():
> > > >
> > > > | https://elixir.bootlin.com/linux/v5.15/source/drivers/spi/spi.c#L2294
> > > >
> > >
> > > Ok thanks I will investigate on it.
> > >
> > > Just to be sure, will the driver, if rx-int and interrupt logic is not
> >
> > "rx-int" is not used as an interrupt source. To keep things simple
> > forget about "rx-int" for now.
> >
> > > used, perform a polling on the spi bus to check for messages incoming
> > > on the can controller?
> >
> > No - polling is not implemented and makes IMHO no sense, as it's quite
> > some overhead.
> >
>
> Ok, now it's clear, rx-int (INT0) is not necessary,
If you use "rx-int" it would be connected to "nINT1/GPIO1" not "INT0"
> but nINT is necessary.
ACK
> I will investigate starting from acpi_register_spi_device.
> I think that the problem depends on the fact that my gpio controller is not
> described in the ACPI table, (is loaded as an external module after the boot),
> and the reference from the ACPI mcp descriptor fails.
> Thanks again for your support. I'll be back when I have solved it.
If the GPIO controller is loaded late, you might run into the problem,
which this patch fixes:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=d24cfee7f63d6b44d45a67c5662bd1cc48e8b3ca
regards,
Marc
--
Pengutronix e.K. | Marc Kleine-Budde |
Embedded Linux | https://www.pengutronix.de |
Vertretung Nürnberg | Phone: +49-5121-206917-129 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-9 |
Attachment:
signature.asc
Description: PGP signature
