On Tue, 21 Apr 2009 10:54:40 -0700 (PDT) Trent Piepho <xyzzy@xxxxxxxxxxxxx> wrote: > On Tue, 21 Apr 2009, Uri Shkolnik wrote: > > --- On Tue, 4/21/09, Trent Piepho <xyzzy@xxxxxxxxxxxxx> wrote: > > > From: Trent Piepho <xyzzy@xxxxxxxxxxxxx> > > > Subject: Re: [PATCH] [0904_14] Siano: assemble all components to one kernel module > > > To: "Uri Shkolnik" <urishk@xxxxxxxxx> > > > Cc: "Mauro Carvalho Chehab" <mchehab@xxxxxxxxxxxxx>, "LinuxML" <linux-media@xxxxxxxxxxxxxxx> > > > Date: Tuesday, April 21, 2009, 6:17 AM > > > On Mon, 20 Apr 2009, Uri Shkolnik > > > wrote: > > > > > > > > "better to have the BUS configurable, e. g. just > > > because you have USB interface, it doesn't mean that you > > > want siano for USB, instead of using SDIO." > > > > > > > > Since the module is using dynamic registration, I > > > don't find it a problem. > > > > When the system has both USB and SDIO buses, both USB > > > and SDIO interface driver will be compiled and linked to the > > > module. When a Siano based device (or multiple Siano > > > devices) will be connected, they will be register internally > > > in the core and activated. Any combination is allow > > > (multiple SDIO, multiple USB and any mix). > > > > > > This is not the way linux drivers normally work. > > > Usually there are > > > multiple modules so that only the ones that need to be > > > loaded are loaded. > > > It sounds like you are designing this to be custom compiled > > > for each > > > system, but that's not usually they way things work. > > > > I think I didn't express myself clearly. > > > > Lets say that someone build a kernel X. > > This kernel has (from all buses) only USB. The Siano module will build with USB interface driver at this system. > > > > If the system includes SDIO and OMAP SPI/SPP, the module build will discard the USB interface driver, but the SDIO and the OMAP SPI will be built. The patch you've provided just merge everything. If you're proposing a newer model, you should send a patchset with the complete Kbuild refactor. For now, it is better to postpone this patch until we merge non-kbuild changes. > Can you name another driver that works this way? It is considered better > to build a new module for a different interface. That way one can decide > at run time if the interface is needed or not and only load the module if > needed. If everything is built into one module then one must decide at > compile time what interfaces to support. But it is often the case that > kernels are compiled on different systems than run them. This model also sounds different to me from what I've seen so far. Cheers, Mauro -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
