Hi,
On Wed, Dec 15, 2010 at 02:57:41PM +0300, Sergei Shtylyov wrote:
glue layer has to probe before musb_core, it sets up the platform_device
for that.
Understood. I think we just can register the platfrom device
earlier on DaVinci...
Now that's too late to cancel the pull request. We will fix
during -rc. Sorry.
OK. I think I'll try to fix it in the DaVinci tree...
I was looking into it, maybe moving glue layer to subsysinitcall_sync()
will fix it (not maybe, it actually will) but I started to think that
it's becoming harder and harder to keep track of the initcall levels,
specially when a driver has lots of different dependencies.
MUSB has to probe before gadget drivers and USB host device drivers, but
glue layer has to probe before it and (in your case) gpiolib has to
probe before musb glue layer, which (in some cases) depend on I2C and so
on.
Maybe we are starting to need some smarter setup which would (maybe, I'm
only speculating now) link things differently based on the symbols they
use ?!? I don't know, just drifting here :-p At the end of the day,
module inter-dependency chain (heh, that's a nice name for it) is
growing larger. So, either we start having stuff like:
usb_transceiver_initcall();
usb_link_initcall();
gpio_initcall();
i2c_initcall();
etc etc, or we have something smarter to handle that. Don't know what
still..
--
balbi
--
To unsubscribe from this list: send the line "unsubscribe linux-usb" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
