* Mike Frysinger wrote:
> On Tuesday 20 December 2011 03:52:39 Thierry Reding wrote:
> > This series of patches implements central card-detection within the MMC
> > framework. Before this patch series, many boards actually did implement
> > the board_mmc_getcd() function, but it wasn't used except by one driver
> > (fsl_esdhc). Unfortunately, implementations interpreted the meaning of
> > the cd parameter differently, some taking it to signal card absence and
> > others using it to detect card presence. Furthermore, the signature of
> > the board_mmc_getcd() function was not at all consistent with other MMC
> > related functions.
>
> as you've seen, GPIO pin detection is not novel or unique to any driver/SoC.
> i'd suggest instead finding a way to move this into "struct mmc". perhaps
> something like:
>
> include/mmc.h:
> struct mmc {
> ...
> #ifdef CONFIG_MMC_GPIO_CD
> unsigned gpio_cd;
> #endif
> ...
> };
>
> and then the default mmc_register() func can do:
> #ifdef CONFIG_MMC_GPIO_CD
> if (!mmc->getcd)
> mmc->getcd = mmc_getcd_gpio;
> #endif
>
> and we can have the drivers/mmc/mmc.c:
> #ifdef CONFIG_MMC_GPIO_CD
> int mmc_getcd_gpio(struct mmc *mmc)
> {
> return !gpio_get_value(mmc->getcd);
> }
> #endif
>
> then mmc drivers can extend their register function to take a GPIO number (if
> they don't already). and then they get everything else for free.Yes, there is some potential for refactoring drivers, though only a very limited number of boards actually implement card-detection. I guess something like what you are proposing could be added on top of the series. That would hard-code the fact that card-detect GPIOs are always low-active, though. Can we live with that? If a board handles that differently they can always override the behaviour by specifying their own board_mmc_getcd(). Thierry
pgpgY7c9w0hJ1.pgp
Description: PGP signature
_______________________________________________ U-Boot mailing list U-Boot@lists.denx.de http://lists.denx.de/mailman/listinfo/u-boot
