* Nishanth Menon <nm@xxxxxx> [091016 08:04]: > Cory Maccarrone had written, on 10/16/2009 09:15 AM, the following: >> On Thu, Oct 15, 2009 at 9:53 PM, Nishanth Menon <nm@xxxxxx >> <mailto:nm@xxxxxx>> wrote: >> >> Pais, Allen had written, on 10/15/2009 11:53 PM, the following: >> >> a) A simple comment to all my comments: why cant we have these >> in bootloader and just simply leave the mux file alone? >> [Allen] Yes Nishanth, this would be a much cleaner approach. >> Even Santosh had suggested >> The same, if we can conclude on a approach here, I can go ahead >> and do the Mux Change it accordingly. >> >> Then lets please fix the bootloader and drop this patch. >> >> > [...] >> Maybe I'm missing something, but why is it more desirable to add the >> mux code to the bootloader instead of the kernel? Wouldn't adding it >> to the kernel guarantee it works regardless of the bootloader? > adding mux in kernel guarentees that it is independent of the bootloader > - yes - no questions on that. There are two strategies that can be taken: Muxing in the bootloader only leads to impossible to debug bugs of type "my device does not work" where the kernel has no clue what's going on. In addition, at least the GPIO pins need to be dynamically remuxed during the off-while-idle. Also, if CONFIG_MUX is not set, the code is optimized out. See the earlier discussion on the muxing. Basically we are going to move to init time muxing in board-*.c files hopefully this coming merge window. I'll post some patches probably next week on this. Meanwhile, the current mux system works just fine, but is not suitable for what we need for the dynamic GPIO pin muxing. > > A) Ensure that kernel is independent from bootloader variances > - this is desirable and would adhere to the rules of an entity uses > the resources(muxes) ONLY what it is required for itself to work -> e.g. > bootloader would use the minimal set that is required for itself to > function. > > B) Divide the load -> always assume that kernel and bootloaders work > together. > - Do all the static muxes in u-boot -> since the u-boot dies off once > the kernel starts up, the memory is completely freed up. > - Only the dynamic muxes (e.g. pins being used in two modes by two > drivers) are handled by the kernel. > > The current strategy that is implemented for OMAP3 is (B), but once the > kernel infrastructure for mux handling improves we should be able to > arrive at a compromise between A and B as far as execution latencies Vs > memory usage Vs independence issues are concerned. we are not there yet > unfortunately. Once we have all the mux stuff reworked, we should move to model where we mux all the pins during boot. This is with CONFIG_MUX set, without that it will be all optimized out. Regards, Tony -- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
