On Tue, 14 Mar 2023 at 09:33, Marc Gonzalez <marc.w.gonzalez@xxxxxxx> wrote: > > On 14/03/2023 08:27, Ulf Hansson wrote: > > > On Mon, 13 Mar 2023 at 22:42, Marc Gonzalez wrote: > >> > >> On 13/03/2023 21:04, Martin Blumenstingl wrote: > >> > >>> On Mon, Mar 13, 2023 at 6:09 PM Marc Gonzalez wrote: > >>> [...] > >>>> QUESTION: > >>>> It's not clear to me why we sleep twice for host->ios.power_delay_ms? > >>> > >>> I'm not sure but I think host->ios.power_delay_ms is independent from > >>> the post-power-on-delay-ms property of "mmc-pwrseq-simple". > >> > >> They are distinct indeed (the props have the same name & similar purpose though). > > > > Correct. > > > >> > >> host->ios.power_delay_ms is set via: > >> > >> drivers/mmc/core/host.c: device_property_read_u32(dev, "post-power-on-delay-ms", > > > > This is a property that may be specified in the host's OF node. See > > Documentation/devicetree/bindings/mmc/mmc-controller.yaml. > > > >> drivers/mmc/core/host.c- &host->ios.power_delay_ms); > > > > Note that the default value is set to 10 ms, in mmc_alloc_host(). So, > > if you need a different value, that can be specified in DT. > > > >> > >> > >> pwrseq->post_power_on_delay_ms is set via: > >> > >> drivers/mmc/core/pwrseq_simple.c: device_property_read_u32(dev, "post-power-on-delay-ms", > > > > This is a property that may be specified in the pwrseq OF node. See > > Documentation/devicetree/bindings/mmc/mmc-pwrseq-simple.yaml. > > > >> drivers/mmc/core/pwrseq_simple.c- &pwrseq->post_power_on_delay_ms); > >> > >> > >> However, they are both used to delay mmc_power_up(), > >> and the host delay is used to sleep twice, which I > >> found confusing ;) > > > > It's certainly a bit hairy to follow, but I think you have understood > > things correctly. > > > >> > >> Regards > >> > > > > Looks like you are getting closer to the solution, at least for the > > first step to make sure the SDIO card is getting detected properly. > > Nice! > > What I'm confused about is: > > Why would I specify the delay in the host rather than in the pwrseq? If the delay is to manage vmmc and vqmmc, which is somewhat part of the generic specifications (SD/MMC), then it should be described in the host's node. A pwrseq is something special, which is also platform and device specific (the SDIO device). If the delays correspond to this, it should be part of the pwrseq node. > > I'll have to study mmc_power_up() a bit more ;) :-) Kind regards Uffe
