Re: [PATCH] spi/acpi: fix incorrect ACPI parent check

[Ard Biesheuvel <ard.biesheuvel@xxxxxxxxxx>]

On Wed, 19 Jun 2019 at 16:43, Mika Westerberg
<mika.westerberg@xxxxxxxxxxxxxxx> wrote:
>
> On Wed, Jun 19, 2019 at 05:17:59PM +0300, Jarkko Nikula wrote:
> > Hi
> >
> > On 6/19/19 4:58 PM, Ard Biesheuvel wrote:
> > > So can you explain how exactly the I2C GPIO expander is failing? I
> > > struggle to understand how the SPI slave probing could be related to
> > > that.
> > >
> > They don't show up in /sys/kernel/debug/gpio, are not present in
> > /sys/bus/i2c/devices/ but SPI core instead tries add them with a bogus Chip
> > Select number:
> >
> > [    5.727699][    T1] pxa2xx-spi pxa2xx-spi.5: cs56 >= max 4
> > [    5.733545][    T1] spi_master spi5: failed to add SPI device INT3491:00
> > from ACPI
>
> Just a guess but looking at the 4c3c59544f33 acpi_register_spi_device()
> does not seem to zero fill the whole struct acpi_spi_lookup structure so
> when it is supposed to bail out when SPI slave was not found:
>
>         if (!lookup.max_speed_hz)
>                 return AE_OK
>
> it instead continues to register SPI slave because lookup.max_speed_hz
> may contain whatever garbage there is in the stack at that address.

Good point.

Jarkko, does this help?

diff --git a/drivers/spi/spi.c b/drivers/spi/spi.c
index 50d230b33c42..d072efdd65ba 100644
--- a/drivers/spi/spi.c
+++ b/drivers/spi/spi.c
@@ -1914,6 +1914,7 @@ static acpi_status
acpi_register_spi_device(struct spi_controller *ctlr,
                return AE_OK;

        lookup.ctlr             = ctlr;
+       lookup.max_speed_hz     = 0;
        lookup.mode             = 0;
        lookup.bits_per_word    = 0;
        lookup.irq              = -1;