* Roger Quadros <rog...@ti.com> [151125 02:52]: > > On 03/09/15 10:39, Roger Quadros wrote: > > On 28/07/15 14:34, Roger Quadros wrote: > >> Paul, > >> > >> On 16/07/15 16:56, Roger Quadros wrote: > >>> On 16/07/15 04:25, Paul Walmsley wrote: > >>>> Hi > >>>> > >>>> On Tue, 23 Jun 2015, Roger Quadros wrote: > >>>> > >>>>> For some hwmods (e.g. DCAN on DRA7) we need the possibility to > >>>>> disable HW_AUTO for the clockdomain while the module is active. > >>>>> > >>>>> To achieve this there needs to be a refcounting mechanism to > >>>>> indicate whether any module in the clockdomain has requested > >>>>> to disable HW_AUTO. We keep track of this in 'noidlecount'. > >>>>> > >>>>> Hwmod code must use clkdm_hwmod_prevent_hwauto() to prevent > >>>>> HW_AUTO of the clockdomain in the future clkdm_hwmod_hwauto() calls. > >>>>> > >>>>> It must use clkdm_hwmod_allow_hwauto() to allow HW_AUTO in > >>>>> the future clkdm_hwmod_hwauto() calls. > >>>>> > >>>>> Hwmod code must use clkdm_hwmod_allow_hwauto() whenever it needs > >>>>> to request HW_AUTO of any clockdomain. (Typically after it has > >>>>> enabled the module). > >>>> > >>>> How about just modifying clkdm_{allow,deny}_idle*() to do the > >>>> idle-block-counting? The default state would be to allow idle, assuming > >>>> that the clockdomain flags support that state, and then > >>>> clkdm_deny_idle*() > >>>> would increment the idle-blocking count, clkdm_allow_idle*() would > >>>> decrement it. Then on the 0 -> 1 or 1 -> 0 transitions, the hardware > >>>> would be reprogrammed appropiately. > >>> > >>> That is not possible with current hwmod code as clkdm_allow_idle() and > >>> clkdm_deny_idle() > >>> are not symmetrically placed. > >>> > >>> The usual flow is > >>> clkdm_enable_hwmod(); > >>> if (hwsup) > >>> clkdm_allow_idle(); > >>> > >>> This is mainly because it is redundant to disable auto idle before > >>> enableing > >>> (SW_WKUP) the clockdomain. > >>> > >>> If we take your proposal above then we have to modify all users like so. > >>> > >>> if (hwsup) > >>> clkdm_deny_idle(); > >>> clkdm_enable_hwmod(); > >>> if (hwsup) > >>> clkdm_allow_idle(); > >>> > >>> Is this really what we want? > >> > >> Any comments on this? > > > > Paul. I'm waiting on your input to rework this series if needed. > > Early input would be much appreciated. Thanks. > > Not sure if Paul is receiving my e-mails but I'd like > others to give their opinion on how we can proceed with this. Thanks.
Well in the long run we want to have a proper bus driver for each interconnect instance and use genpd. I'm afraid that solution is not going to help immediately though. Regards, Tony -- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html