2010/12/13 Felipe Balbi <balbi@xxxxxx>: > Hi, > > On Mon, Dec 13, 2010 at 09:15:22PM +0800, Ming Lei wrote: >>>> >>>> see size of glue modules built with this patch set in my beagle-xm: >>>> >>>> -rw-rw-r-- 1 500 500 132068 Dec 13 2010 am35x.ko >>>> -rw-rw-r-- 1 500 500 135279 Dec 13 2010 omap2430.ko >>>> -rw-rw-r-- 1 500 500 152696 Dec 13 2010 tusb6010.ko >>> >>> that's all on .init.text section, no ?!? then it'll be freed after we're >>> out of initcalls. >> >> No, only very few function/variables of glue driver is declared >> as __init, and is in .init section. Others will still stay in ram for >> ever. > > But what's the problem ? We want a single binary for several glue layers > of same ARCH anyway. When building products, device makers will > optimize what they think needs to be optimized, counting removing unused > drivers from kernel configuration. > > On a distro-like kernel we will have all of that configured to work. > Which is the same in both approaches (glue built-in or as module), the > difference is that there's not behavior change for users which have been > issuing 'modprobe musb_hdrc' since it has been added to mainline. OTOH, > if we change glue layers to modules, users will have to remember that > now they're running omap2plus boards and they have to type modprobe > musb-omap2430, instead of modprobe musb_hdrc. No, udev can do it automatically, and the default rule is enough intelligent. > See the problem ? We have a false abstraction as user has to know which > glue layer to load. See above, user does not need to care it. > > Users will never issue 'modprobe ehci-omap' they run 'modprobe > ehci-hcd'. In fact, user doesn't need to issue all modprobe things in theroy, :-) -- Lei Ming -- 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
