No, they were not switch timeouts. They were immediate failures due to touchiness about init sequence. The card firmware was corrected and the new samples don't have the issue. I don't think there are any of these buggy ones in the wild, but I expect to see buggy firmware coming from other vendors. Russ W. Knize Mobile Devices Android Platform Desk: 847-523-9381 Mobile: 224-622-1391 On Sun, Apr 3, 2011 at 11:11 PM, Andrei Warkentin <andreiw@xxxxxxxxxxxx> wrote: > On Fri, Apr 1, 2011 at 7:28 PM, Andrei Warkentin <andreiw@xxxxxxxxxxxx> wrote: >> Hi Russ, >> >> On Fri, Apr 1, 2011 at 10:58 AM, Russ Knize <Russ.Knize@xxxxxxxxxxxx> wrote: >>> + /* >>> + * Ignore switch errors from buggy cards that actually >>> + * do switch successfully. >>> + */ >>> + err = mmc_set_power_class(host, max_dtr, bus_width); >>> + if (err && err != -EBADMSG) >>> + goto free_card; >>> + >> >> 1) host->card is not yet set at this moment, you probably want to pass >> card to mmc_set_power_class as well . >> 2) It seems like you will call mmc_set_power_class even if bus_width >> == MMC_BUS_WIDTH_1. Is that supposed to work? 7.6.3 seems to imply >> power classes only apply to 4-bit and 8-bit bus widths. As does the >> paragraph on top of Page 138 of eMMC 4.41 spec, which says that for >> 1-bit bus configuration the default applies (class 0). >> 3) Since the current code supports DDR modes, you probably want to >> extend support for PWR_CL_DDR_52_(360|195) as well. >> >> Just curious, for the cards reporting switch errors but nonetheless >> switching to the power class, do they update the EXT_CSD field? >> > > There was just a patch posted by Chuanxiao Dong ([PATCH v4 0/3]mmc: > enable TRIM/ERASE caps for SDHCI host) against SDHCI core to enable > setting proper timeout for R1B response-type commands, one of which is > CMD6 (SWITCH). The patch is applicable to the kernel sources you've > tested your patch on. Were the failures you've seen for cards erroring > on MMC_SWITCH timeouts? Could you test with that change for the parts > affected, and see if your problem in programming power level goes > away? > > 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
