Re: pca953x: Probing too early

["Russell King (Oracle)" <linux@xxxxxxxxxxxxxxx>]

On Tue, Aug 20, 2024 at 11:29:07PM +0200, Andrew Lunn wrote:
> On Tue, Aug 20, 2024 at 05:47:27PM -0300, Fabio Estevam wrote:
> > Adding the i2c-folks on Cc.
> > 
> > On Tue, Aug 20, 2024 at 5:02 PM Fabio Estevam <festevam@xxxxxxxxx> wrote:
> > >
> > > Hi,
> > >
> > > I am seeing an issue with the PCA935X driver in 6.6.41 and
> > > 6.11.0-rc4-next-20240820.
> > >
> > > The pca953x is getting probed before its I2C parent (i2c-2):
> > >
> > > [    1.872917] pca953x 2-0020: supply vcc not found, using dummy regulator
> > > [    1.889195] pca953x 2-0020: using no AI
> > > [    1.893260] pca953x 2-0020: failed writing register
> > > [    1.898258] pca953x 2-0020: probe with driver pca953x failed with error -11
> 
> -11 is EAGAIN, which is a bit odd. Given your description, i would of
> expected ENODEV. My guess is, it needs another resource, a GPIO,
> regulator, or interrupt controller. That resources might not of probed
> yet. If that is true, you want the pca953x_probe() to return
> -EPROBE_DEFER. The driver core will then try the probe again sometime
> later, hopefully when all the needed resources are available.
> 
> Track down where the EAGAIN is coming from.

This is where:

        ret = regmap_bulk_write(chip->regmap, regaddr, value, NBANK(chip));
        if (ret < 0) {
                dev_err(&chip->client->dev, "failed writing register\n");

printing the error code in error messages would really help debugging.
Sadly, people don't do this. I don't know why we don't bulk replace
all error messages with just "Error!\n" to make them even more cryptic
and undebuggable!

It's likely that EAGAIN is coming from this - the probe function calls
one of the init functions, and propagates the error up, and as that
message is being printed... Tracing down, the I2C transfer function
returns -EAGAIN if it fails the transfer, and __i2c_smbus_xfer() will
itself retry it a number of times before propagating that -EAGAIN up.

EAGAIN is supposed to only be generated on arbitration loss - I'm
guessing that the I2C bus is in some kind of locked state, meaning
that devices on this bus are not accessible. Maybe the I2C bus
pull-ups aren't powered? Maybe there's a bad device on the bus
pulling the bus down?

Someone mentioned i2c-imx, maybe try enabling debug in that driver
to see why it's failing to access the device?

-- 
*** please note that I probably will only be occasionally responsive
*** for an unknown period of time due to recent eye surgery making
*** reading quite difficult.

RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!