On Wed, Mar 16, 2011 at 2:16 PM, Simon Horman <horms@xxxxxxxxxxxx> wrote: > On Wed, Mar 16, 2011 at 11:20:46AM +0900, Magnus Damm wrote: >> On Wed, Mar 16, 2011 at 10:14 AM, Simon Horman <horms@xxxxxxxxxxxx> wrote: >> > How do you envisage that the hardware block would be selected? >> > At compile time through Kconfig? If so the current #define mechanism >> > might be sufficient. >> >> Not through Kconfig. I think you should use Kconfig to enable the SDHI >> loader, but you should be able to select the SDHI base address during >> run-time. Similar to how we enable platform device drivers with >> Kconfig but put the instance information in the platform device >> resource and data outside the driver. >> >> So for instance, on some board we may want to read a GPIO pin at boot >> up time to select if we should boot from SDHI0 or SDHI1. I would like >> the SDHI loader to be designed so we can have support for multiple >> instances complied-in. Because of that I'd like to see the fixed >> SDHI_BASE disappear from the header, and letting the mmc_loader() >> function take the base address as an argument, or simply move the >> mmc_loader() function out of the SDHI loader code to give CPU specific >> and/or board specific code freedom to select which ever SDHI hardware >> block instance(s) they want to load from. >> >> Perhaps this would require some serious refactoring? > > Either making mmc_loader() CPU specific or allowing it > to take an argument should be pretty straight forward. Good! > However, it is entirely unclear to me how the argument to mmc_loader() > would be supplied or alternatively the variant of mmc_loader() be > selected at run-time. At this point, the sh7372 specific could would simply pass the same SDHI base address that your code is using. No special selection needed. Just compile it in. In the future we may want to read out the boot mode pins and select base address accordingly. It's too early to tell exactly what to do with the CPU specific bits for future processors, but at least we can make sure that the SDHI loader isn't tied to a single SDHI instance by design. > This code runs in very early boot. And as such I think that the two major > options are to either compile code in or pull it out of a register somehow. > We really don't have a whole lot of code that runs before mmc_loader() that > could do any kind of setup. Compiling in the base address is fine for sh7372, but please put that base address in the CPU specific code and feed that to the more generic SDHI loader using a function argument. Do you understand what I'm trying to do, or am I just going in circles? =) Thanks, / magnus -- To unsubscribe from this list: send the line "unsubscribe linux-mmc" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
