On Sat, Nov 6, 2010 at 7:38 PM, Philip Rakity <prakity@xxxxxxxxxxx> wrote: > When clock gating is enabled with SDIO cards we have seen > a) Missed Interrupts > b) Transactions not completed As David suggested, these issues are probably card-specific. The 128x wlan card, e.g., supports both synchronous and asynchronous SDIO interrupts (and also external non-SDIO interrupts). So the decision whether to enable clock gating for SDIO should not be static (eventually. I'm not saying this is something Linus should take care of in his patch, it's up to us to make this work). By default, SDIO clock gating should be disabled, but we need to allow drivers to explicitly enable it if supported. I'm tempted to suggest that drivers should control this too using runtime PM API, but for that we need PM core to support several different levels of runtime suspended states (and not just active/suspended). It might be a good use case to push runtime PM core for that. We would then have three power states: * Active (powered on; clocks ungated) * Gated (powered on; clocks gated) * Suspended (powered off) > > We think this is because some SDIO cards need n clocks after the xfer completes from the host point of view to be in a state to handle clocks off. n clocks should be 8 for SD 3.0 but in sometimes the cards need more. We have seen SDIO cards work with no clocks needed and some needing more then 12 clocks. This is completely understandable considering that clock gating is a new feature. > > > b) hardware doing clock gating with SDIO is BAD -- we see the same issues that s/w clock gating patch see's .. The sdio cards do not work correctly. > > c) The quirk is needed to tell the core/ layer that h/w support is available. > > On Nov 6, 2010, at 10:24 AM, Ohad Ben-Cohen wrote: > >> Hi Phillip, >> >> On Sat, Nov 6, 2010 at 1:38 AM, Philip Rakity <prakity@xxxxxxxxxxx> wrote: >>> d) keep sdio clock gating OFF -- need this from our experience. >> >> I'm really interested to hear more about that one - what issues did you see ? >> Can you please elaborate ? >> >> Thanks, >> Ohad. > > -- 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
