On Fri, Dec 10, 2010 at 04:57:24PM +0800, Ming Lei wrote:
2010/12/10 Felipe Balbi <balbi@xxxxxx>:
Ok, found it. It's a Makefile issue for sure. See that in
drives/Makefile we have:
obj-$(CONFIG_USB_MUSB_HDRC) += usb/musb/
since that evaluates to obj-m, looks like the Kbuild will only look for
.ko files in that directory and won't link drivers/usb/musb/built-in.o
to drivers/built-in.o and in turn it won't go into vmlinux.
Now there's a catch, because omap2430.c (and the other glue layers) use
a few symbols from musb_hdrc.ko so changing obj-$(CONFIG_USB_MUSB_HDRC)
to obj-y won't help. Some deeper thinking will be needed to decouple
glue layers even more from musb_hdrc.ko
I'm willing to leave this patch as bugfix during 2.6.38-rc cycle. Which
means current branch is ready to be pulled, I just have to add the omap4
patches from Hema on top. Another pull request will follow.
Agree, this patches should be merged into -next as soon as possible.
I have a couple of patches already to clean up glue layer:
- support each glue driver compiled as standalone module
- support all glue drivers compiled into one single image which
can be run on multiple machines
That's complicated for tusb6010, see that we have lots of ifdefs for
that one on struct musb and other parts of code. You'll also have to
take care of the different DMA engines and the easiest way is to use
drivers/dmaengine API. But if you have those patches, that's actually
great :-)
--
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
