On 04/10/2018 10:32, Grant Likely wrote:

On Fri, Sep 28, 2018 at 10:01 PM Li Yang [off-list ref] wrote:

quoted

On Fri, Sep 28, 2018 at 3:07 PM Rob Herring [off-list ref] wrote:

quoted

On Thu, Sep 27, 2018 at 5:25 PM Li Yang [off-list ref] wrote:

quoted

Hi Rob and Grant,

Various device tree specs are recommending to include all the
potential compatible strings in the device node, with the order from
most specific to most general.  But it looks like Linux kernel doesn't
provide a way to bind the device to the most specific driver, however,
the first registered compatible driver will be bound.

As more and more generic drivers are added to the Linux kernel, they
are competing with the more specific vendor drivers and causes problem
when both are built into the kernel.  I'm wondering if there is a
generic solution (or in plan) to make the most specific driver bound
to the device.   Or we have to disable the more general driver or
remove the more general compatible string from the device tree?

It's been a known limitation for a long time. However, in practice it
doesn't seem to be a common problem. Perhaps folks just remove the
less specific compatible from their DT (though that's not ideal). For
most modern bindings, there's so many other resources beyond
compatible (clocks, resets, pinctrl, etc.) that there are few generic
drivers that can work.

I guess if we want to fix this, we'd need to have weighted matching in
the driver core and unbind drivers when we get a better match. Though
it could get messy if the better driver probe fails. Then we've got to
rebind to the original driver.

Probably we can populate the platform devices from device tree after
the device_init phase?  So that all built-in drivers are already
registered when the devices are created and we can try find the best
match in one go?  For more specific loadable modules we probably need
to unbind from the old driver and bind to the new one.  But I agree
with you that it could be messy.

It's a tradeoff.

Oops! Accidentally hit send too early.

It's a tradeoff. If the platform device population is deferred until 
after all drivers are loaded, then there isn't any mechanism to ensure 
some devices get probed early in the init sequence.

As Rob said, while it is a problem in theory, there haven't been a lot 
of actual cases where it is a problem. The solution has been to either 
remove the generic match from the device, or we can blacklist particular 
devices from the generic driver.

g.

quoted

quoted

Do you have a specific case where you hit this?

Maybe not a new issue but "snps,dw-pcie" is competing with various
"fsl,<chip>-pcie" compatibles.  Also a specific PHY
"ethernet-phy-idAAAA.BBBB" with generic "ethernet-phy-ieee802.3-c45".

Regards,
Leo