Hi Arnd, On Thu, Apr 7, 2011 at 10:47 PM, Arnd Bergmann <arnd@xxxxxxxx> wrote: > On Thursday 07 April 2011, Chris Ball wrote: >> On Wed, Apr 06 2011, Andrei Warkentin wrote: >> >> This is the latest version of the MMC device partition support. >> >> It relies on a few other changes that were discussed recently. >> >> >> >> Thanks, >> >> A >> >> >> >> TOC: >> >> [PATCH 1/4] MMC: Rename erase_timeout to cmd_timeout. >> >> [PATCH 2/4] MMC: SDHCI R1B command handling + MMC_CAP_ERASE. >> >> [PATCH 3/4] MMC: Allow setting CMD timeout for CMD6 (SWITCH). >> >> [PATCH 4/4] MMC: MMC boot partitions support. >> > >> > Any problems, comments? >> >> Looks good to me -- Arnd, did you want to add your Reviewed/Acked-by >> to this? >> > > I haven't done an in-depth review, but please add my Acked-by to the last > patch. For the other three, I have no idea what they do, but they all > seem harmless to me ;-) > > Thanks for the Ack. Patch 1 paves the way for using erase timeout (which is otherwise dead code). Patch 2 enables proper handling of R1B-response type commands in SDHCI host. R1B commands use the DAT line as busy, hence DAT timeout applies. Previously, DAT timeout is only calculated for data-bearing commands. Now for non-data commands with a cmd_timeout, the timeout is honored. This allows proper handling of CMD38 (erase). This also allows proper handling of CMD6 (SWITCH) writes to EXT_CSD register offsets, as certain operations invoked via this command (such as partition switch times) have different timeouts associated with them. Since erase timeout is now honored, the erase cap is turned on. Please note that all the work in Patch 2 applies only to SDHCI. I'm not familiar with other hosts, but a cursory glance revealed that they all have similar issues (no handling of R1B timeouts). I could roll patches, but I'd have no way of verifying (unless someone wants to volunteer and help out). If you (as a Linaro person) are interested, this is something you could add under (Write performance drop - https://wiki.linaro.org/WorkingGroups/Kernel/Specs/StoragePerfEMMC). Patch 3 exposes mmc_switch to block driver, and lets a timeout parameter be passed, as writes to different fields of EXT_CSD can have differerent timeouts. Thanks, A -- 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