On Thu, Nov 12, 2015 at 9:53 AM, Uwe Kleine-König <u.kleine-koe...@pengutronix.de> wrote: > CC += devicet...@vger.kernel.org, gregkh
You added linux@pengutronix instead of devicetree. > > On Thu, Nov 12, 2015 at 09:36:55AM +0100, Frans Klaver wrote: >> On Thu, Nov 12, 2015 at 9:26 AM, Uwe Kleine-König >> <u.kleine-koe...@pengutronix.de> wrote: >> > On Thu, Nov 12, 2015 at 09:03:11AM +0100, Frans Klaver wrote: >> >> Hi, >> >> >> >> On Thu, Nov 12, 2015 at 8:46 AM, LABBE Corentin >> >> <clabbe.montj...@gmail.com> wrote: >> >> > of_match_device could return NULL, and so cause a NULL pointer >> >> > dereference later. >> >> >> >> Did you actually run into this? It seems to me that this driver is >> >> only probed if and only if we have a match and that therefore >> >> of_match_device will always return a valid pointer (it is using the >> >> same match table). Am I missing something? >> > >> > Yes, you're missing something. The driver would probe for a dt snippet >> > like: >> > >> > mxc_nand { >> > compatible = "foobar"; >> > } >> > >> > In this case dev->of_node is non-NULL but of_match_device(mxcnd_dt_ids, >> > dev) is. >> > >> > (I didn't actually test this, so there is a chance I'm wrong here. And >> > if not I wonder if it is sensible at all to match the device name on >> > driver name for of-created platform devices.) >> >> Yea, looks like you're right. platform devices check a number of >> things to determine a match, among which is driver name if all else >> fails (platform.c, platform_match()). > > Maybe something like this would help to reduce surprises: > > diff --git a/drivers/base/platform.c b/drivers/base/platform.c > index f80aaaf9f610..a9fc22c86552 100644 > --- a/drivers/base/platform.c > +++ b/drivers/base/platform.c > @@ -840,8 +840,8 @@ static int platform_match(struct device *dev, struct > device_driver *drv) > return !strcmp(pdev->driver_override, drv->name); > > /* Attempt an OF style match first */ > - if (of_driver_match_device(dev, drv)) > - return 1; > + if (pdev->dev.of_node) > + return of_driver_match_device(dev, drv); > > /* Then try ACPI style match */ > if (acpi_driver_match_device(dev, drv)) That looks sensible, yea. There is a chance that misbehaving DT nodes will fail after this change, of course. Thanks, Frans -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/