Something I didn't notice in v4, sorry.
On Sat, 2015-06-20 at 00:28 +0800, Yakir Yang wrote:
--- /dev/null +++ b/drivers/gpu/drm/bridge/dw_hdmi-i2s-audio.c
+#define DRIVER_NAME "dw-hdmi-i2s-audio"
+MODULE_ALIAS(PLATFORM_MODULE_PREFIX DRIVER_NAME);
0) Side note: this is the first time that PLATFORM_MODULE_PREFIX is used inside MODULE_ALIAS(). But none of the 1000+ other "platform:" aliases do that. And neither does 5/6 of this series! That suggests, I think, that this shouldn't be done.
You could consider adding something like #define MODULE_ALIAS_PLATFORM(NAME) MODULE_ALIAS(PLATFORM_MODULE_PREFIX NAME)
But then, I think, all the current 1000+ platform: aliases should be converted to that macro. Would that be worth it?
1) Now on to my remark: this alias seems to be only useful if there also is a struct platform_device with a "dw-hdmi-i2s-audio" name. Because that platform_device would, badly summarized, fire of a "MODALIAS=platform:dw-hdmi-i2s-audio" uevent when created. Which, in its turn, would trigger userspace to load this module, correct?
But I think there's no platform_device with a "dw-hdmi-i2s-audio" name. So I wonder whether this MODULE_ALIAS() is actually needed. What breaks if you leave it out?
(Likewise for 5/6, but there the platform_device should have a "rockchip-hdmi-audio" name.)
Thanks,
Paul Bolle