On Wed, May 04, 2011 at 10:06:20PM +0530, Prashanth Bhat wrote: > On Wednesday 04 May 2011 04:53 PM, Michał Mirosław wrote: > >On Tue, May 03, 2011 at 05:00:03PM +0530, Prashanth Bhat wrote: > >>To be more specific, I would think that the code change required in > >>include/linux/mmc/card.h is: > >> > >>#define mmc_card_mmc(c) ((c)->type == MMC_TYPE_MMC) > >>- #define mmc_card_sd(c) ((c)->type == MMC_TYPE_SD) > >>- #define mmc_card_sdio(c) ((c)->type == MMC_TYPE_SDIO) > >> > >>+ #define mmc_card_sd(c) ((c)->type == MMC_TYPE_SD || > >>(c)->type == MMC_TYPE_SD_COMBO) > >>+ #define mmc_card_sdio(c) ((c)->type == MMC_TYPE_SDIO || > >>(c)->type == MMC_TYPE_SD_COMBO) > >You can actually use (c->type& MMC_TYPE_SD) and (c->type& MMC_TYPE_SDIO). > >Unless there will be more types of SD cards (unlikely) this way > >will generate less code on average. > Interesting coincidence that the value of MMC_TYPE_SD and > MMC_TYPE_SDIO and MMC_TYPE_SD_COMBO have the relationship that > SD_COMBO is the bitwise OR of the other two. However, this seems to > be more of a coincidence than intentional. The #defines were clearly > meant to be numeric values rather than bit-masks. > > #define MMC_TYPE_MMC 0 /* MMC card */ > #define MMC_TYPE_SD 1 /* SD card */ > #define MMC_TYPE_SDIO 2 /* SDIO card */ > #define MMC_TYPE_SD_COMBO 3 /* SD combo (IO+mem) card */ > > > Using the bit-mask approach therefore doesn't feel like the natural > way to me. Perhaps the #defines could be changed to > > #define MMC_TYPE_MMC (1 << 0) /* MMC card */ > #define MMC_TYPE_SD (1 << 1) /* SD card */ > #define MMC_TYPE_SDIO (1 << 2) /* SDIO card */ > #define MMC_TYPE_SD_COMBO (MMC_TYPE_SD | MMC_TYPE_SDIO) /* > SD combo (IO+mem) card */ Since MMC == !SD && !SDIO, then MMC_TYPE_MMC bit would carry redundant information. Best Regards, Michał Mirosław PS. Please do not top-post. -- 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
