Re: [PATCH] of: Turn of_match_node into a static inline when CONFIG_OF isn't set

[Josh Cartwright <joshc-sgV2jX0FEOL9JmXXK+q4OQ@public.gmane.org>]

On Tue, Feb 11, 2014 at 03:30:33PM -0600, Rob Herring wrote:
> On Tue, Feb 11, 2014 at 12:29 PM, Geert Uytterhoeven
> <geert-Td1EMuHUCqxL1ZNQvxDV9g@public.gmane.org> wrote:
> > On Tue, Feb 11, 2014 at 7:08 PM, Josh Cartwright <joshc-sgV2jX0FEOL9JmXXK+q4OQ@public.gmane.org> wrote:
> >> It sure would be convenient if platform_device had a 'const struct
> >> of_device_id *of_id_entry' member similar to the existing struct
> >> platform_device_id one, that was set up during platform device matching.
> >> Most platform_driver users of of_match_node() would simply go away.
> >
> > Can't the entry be shared for both platform_device_id and of_device_id?
> > Only one of them can be valid at the same time, right?
> >
[..]
> 
> I believe this is the reason drivers have to call of_match_device:
> 
> commit b1608d69cb804e414d0887140ba08a9398e4e638
> Author: Grant Likely <grant.likely-s3s/WqlpOiPyB63q8FvJNQ@public.gmane.org>
> Date:   Wed May 18 11:19:24 2011 -0600
>
>     drivercore: revert addition of of_match to struct device
>
>     Commit b826291c, "drivercore/dt: add a match table pointer to struct
>     device" added an of_match pointer to struct device to cache the
>     of_match_table entry discovered at driver match time.  This was unsafe
>     because matching is not an atomic operation with probing a driver.  If
>     two or more drivers are attempted to be matched to a driver at the
>     same time, then the cached matching entry pointer could get
>     overwritten.
>
>     This patch reverts the of_match cache pointer and reworks all users to
>     call of_match_device() directly instead.
>
>     Signed-off-by: Grant Likely <grant.likely-s3s/WqlpOiPyB63q8FvJNQ@public.gmane.org>

Interesting, thanks for the history!  I'm wondering if this same problem
exists for the existing platform_device_id cached pointer as well.

Okay, so maybe caching a pointer in the device isn't the best option,
what if we considered extending the platform_driver callbacks to include
a set of per-method (?) probe callbacks which do provide a handle to
matched identifiers.

In the case of a totally contrived platform_driver supporting ACPI, OF,
and !OF configurations, it might look something like:

	static const struct of_device_id acme_of_table[] = {
		/* ... */
		{ },
	};
	MODULE_DEVICE_TABLE(of, acme_of_table);

	static int acme_probe_of(struct platform_device *pdev,
				 const struct of_device_id *id)
	{
		/* ... */
		return 0;
	}

	static const struct acpi_device_id acme_acpi_table[] = {
		/* ... */
		{ },
	};
	MODULE_DEVICE_TABLE(acpi, acme_acpi_table);

	static int acme_probe_acpi(struct platform_device *pdev,
				   const struct acpi_device_id *id)
	{
		/* ... */
		return 0;
	}

	static const struct platform_device_id acme_platform_table[] = {
		/* ... */
		{ },
	};
	MODULE_DEVICE_TABLE(platform, acme_platform_table);

	static int acme_probe_acpi(struct platform_device *pdev,
				   const struct platform_device_id *id)
	{
		/* ... */
		return 0;
	}

	static int acme_probe_name(struct platform_device *pdev)
	{
		/* ... */
		return 0;
	}

	static struct platform_driver acme_driver = {
		.probe_of	= acme_probe_of,
		.probe_acpi	= acme_probe_acpi,
		.probe_platform	= acme_probe_platform,
		.probe_name	= acme_probe_name,
		.remove		= acme_remove,
		.driver		= {
			.name			= "acme",
			.of_match_table		= of_match_ptr(acme_of_table),
			.acpi_match_table	= ACPI_PTR(acme_acpi_table),
		},
	};
	module_platform_driver(acme_driver);

-- 
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
hosted by The Linux Foundation
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html