| Message ID | 20240123-gcc-ao-support-v1-2-6c18d5310874@quicinc.com (mailing list archive) |
|---|---|
| State | Changes Requested |
| Headers | show |
| Series | clk: qcom: gcc-sm8150: Add gcc_parents_0_ao support | expand |
On Tue, 23 Jan 2024 at 18:35, Satya Priya Kakitapalli <quic_skakitap@quicinc.com> wrote: > > Add active_only support for gcc_parents_0, this is needed because > some of the clocks under it are critical which would vote on xo > blocking the suspend. The kernel should be able to work with older DT files. Please add support for using bi_tcxo as a fallback. > > Signed-off-by: Satya Priya Kakitapalli <quic_skakitap@quicinc.com> > --- > drivers/clk/qcom/gcc-sm8150.c | 10 ++++++++-- > 1 file changed, 8 insertions(+), 2 deletions(-) > > diff --git a/drivers/clk/qcom/gcc-sm8150.c b/drivers/clk/qcom/gcc-sm8150.c > index 05d115c52dfe..2a0608c5a104 100644 > --- a/drivers/clk/qcom/gcc-sm8150.c > +++ b/drivers/clk/qcom/gcc-sm8150.c > @@ -123,6 +123,12 @@ static const struct clk_parent_data gcc_parents_0[] = { > { .hw = &gpll0_out_even.clkr.hw }, > }; > > +static const struct clk_parent_data gcc_parents_0_ao[] = { > + { .fw_name = "bi_tcxo_ao", .name = "bi_tcxo_ao" }, > + { .hw = &gpll0.clkr.hw }, > + { .hw = &gpll0_out_even.clkr.hw }, > +}; > + > static const struct parent_map gcc_parent_map_1[] = { > { P_BI_TCXO, 0 }, > { P_GPLL0_OUT_MAIN, 1 }, > @@ -222,8 +228,8 @@ static struct clk_rcg2 gcc_cpuss_ahb_clk_src = { > .freq_tbl = ftbl_gcc_cpuss_ahb_clk_src, > .clkr.hw.init = &(struct clk_init_data){ > .name = "gcc_cpuss_ahb_clk_src", > - .parent_data = gcc_parents_0, > - .num_parents = ARRAY_SIZE(gcc_parents_0), > + .parent_data = gcc_parents_0_ao, > + .num_parents = ARRAY_SIZE(gcc_parents_0_ao), > .flags = CLK_SET_RATE_PARENT, > .ops = &clk_rcg2_ops, > }, > > -- > 2.25.1 > >
On 1/23/24 17:34, Satya Priya Kakitapalli wrote: > Add active_only support for gcc_parents_0, this is needed because > some of the clocks under it are critical which would vote on xo > blocking the suspend. > > Signed-off-by: Satya Priya Kakitapalli <quic_skakitap@quicinc.com> > --- Is there a need to keep gcc_cpuss_ahb_clk_src around? Do we do any ratesetting on it? Should we ever turn it off? Konrad
On 1/23/2024 11:17 PM, Konrad Dybcio wrote: > > > On 1/23/24 17:34, Satya Priya Kakitapalli wrote: >> Add active_only support for gcc_parents_0, this is needed because >> some of the clocks under it are critical which would vote on xo >> blocking the suspend. >> >> Signed-off-by: Satya Priya Kakitapalli <quic_skakitap@quicinc.com> >> --- > > Is there a need to keep gcc_cpuss_ahb_clk_src around? Do we do any > ratesetting on it? Should we ever turn it off? > The branch clocks under gcc_cpuss_ahb_clk_src are critical clocks, which are running at 19.2Mhz causing vote on XO during suspend. As of now no rate setting is happening but this rcg is useful to get the exact rates from debugfs. Hence this change is needed to avoid XO shutdown issues.
On 1/25/24 06:49, Satya Priya Kakitapalli (Temp) wrote: > > On 1/23/2024 11:17 PM, Konrad Dybcio wrote: >> >> >> On 1/23/24 17:34, Satya Priya Kakitapalli wrote: >>> Add active_only support for gcc_parents_0, this is needed because >>> some of the clocks under it are critical which would vote on xo >>> blocking the suspend. >>> >>> Signed-off-by: Satya Priya Kakitapalli <quic_skakitap@quicinc.com> >>> --- >> >> Is there a need to keep gcc_cpuss_ahb_clk_src around? Do we do any >> ratesetting on it? Should we ever turn it off? >> > > The branch clocks under gcc_cpuss_ahb_clk_src are critical clocks, which are running at 19.2Mhz causing vote on XO during suspend. As of now no rate setting is happening but this rcg is useful to get the exact rates from debugfs. Hence this change is needed to avoid XO shutdown issues. So, if I underderstood you correctly, this clock serves no purpose other than getting rate? In this case, I'd say we should de-register it from the clock driver and use debugcc [1] (contributions welcome!) for precise measurements. Konrad [1] https://github.com/linux-msm/debugcc
On 1/25/2024 3:25 PM, Konrad Dybcio wrote: > > > On 1/25/24 06:49, Satya Priya Kakitapalli (Temp) wrote: >> >> On 1/23/2024 11:17 PM, Konrad Dybcio wrote: >>> >>> >>> On 1/23/24 17:34, Satya Priya Kakitapalli wrote: >>>> Add active_only support for gcc_parents_0, this is needed because >>>> some of the clocks under it are critical which would vote on xo >>>> blocking the suspend. >>>> >>>> Signed-off-by: Satya Priya Kakitapalli <quic_skakitap@quicinc.com> >>>> --- >>> >>> Is there a need to keep gcc_cpuss_ahb_clk_src around? Do we do any >>> ratesetting on it? Should we ever turn it off? >>> >> >> The branch clocks under gcc_cpuss_ahb_clk_src are critical clocks, >> which are running at 19.2Mhz causing vote on XO during suspend. As of >> now no rate setting is happening but this rcg is useful to get the >> exact rates from debugfs. Hence this change is needed to avoid XO >> shutdown issues. > > So, if I underderstood you correctly, this clock serves no purpose other > than getting rate? > > In this case, I'd say we should de-register it from the clock driver and > use debugcc [1] (contributions welcome!) for precise measurements. > Although currently there is no rate-setting happening now, its better to keep the rcg modelling as is, considering that it might be needed if some use case arises in future.
On 5.02.2024 09:37, Satya Priya Kakitapalli (Temp) wrote: > > On 1/25/2024 3:25 PM, Konrad Dybcio wrote: >> >> >> On 1/25/24 06:49, Satya Priya Kakitapalli (Temp) wrote: >>> >>> On 1/23/2024 11:17 PM, Konrad Dybcio wrote: >>>> >>>> >>>> On 1/23/24 17:34, Satya Priya Kakitapalli wrote: >>>>> Add active_only support for gcc_parents_0, this is needed because >>>>> some of the clocks under it are critical which would vote on xo >>>>> blocking the suspend. >>>>> >>>>> Signed-off-by: Satya Priya Kakitapalli <quic_skakitap@quicinc.com> >>>>> --- >>>> >>>> Is there a need to keep gcc_cpuss_ahb_clk_src around? Do we do any >>>> ratesetting on it? Should we ever turn it off? >>>> >>> >>> The branch clocks under gcc_cpuss_ahb_clk_src are critical clocks, which are running at 19.2Mhz causing vote on XO during suspend. As of now no rate setting is happening but this rcg is useful to get the exact rates from debugfs. Hence this change is needed to avoid XO shutdown issues. >> >> So, if I underderstood you correctly, this clock serves no purpose other >> than getting rate? >> >> In this case, I'd say we should de-register it from the clock driver and >> use debugcc [1] (contributions welcome!) for precise measurements. >> > > Although currently there is no rate-setting happening now, its better to keep the rcg modelling as is, considering that it might be needed if some use case arises in future. We don't tend to solve imaginary problems, please lay out what it would be used for, if at all. Then, we can assess whether it's necessary. Konrad
On 2/5/2024 5:03 PM, Konrad Dybcio wrote: > On 5.02.2024 09:37, Satya Priya Kakitapalli (Temp) wrote: >> On 1/25/2024 3:25 PM, Konrad Dybcio wrote: >>> >>> On 1/25/24 06:49, Satya Priya Kakitapalli (Temp) wrote: >>>> On 1/23/2024 11:17 PM, Konrad Dybcio wrote: >>>>> >>>>> On 1/23/24 17:34, Satya Priya Kakitapalli wrote: >>>>>> Add active_only support for gcc_parents_0, this is needed because >>>>>> some of the clocks under it are critical which would vote on xo >>>>>> blocking the suspend. >>>>>> >>>>>> Signed-off-by: Satya Priya Kakitapalli <quic_skakitap@quicinc.com> >>>>>> --- >>>>> Is there a need to keep gcc_cpuss_ahb_clk_src around? Do we do any >>>>> ratesetting on it? Should we ever turn it off? >>>>> >>>> The branch clocks under gcc_cpuss_ahb_clk_src are critical clocks, which are running at 19.2Mhz causing vote on XO during suspend. As of now no rate setting is happening but this rcg is useful to get the exact rates from debugfs. Hence this change is needed to avoid XO shutdown issues. >>> So, if I underderstood you correctly, this clock serves no purpose other >>> than getting rate? >>> >>> In this case, I'd say we should de-register it from the clock driver and >>> use debugcc [1] (contributions welcome!) for precise measurements. >>> >> Although currently there is no rate-setting happening now, its better to keep the rcg modelling as is, considering that it might be needed if some use case arises in future. > We don't tend to solve imaginary problems, please lay out what it would be > used for, if at all. Then, we can assess whether it's necessary. Okay, I'll de-register this RCG.
diff --git a/drivers/clk/qcom/gcc-sm8150.c b/drivers/clk/qcom/gcc-sm8150.c index 05d115c52dfe..2a0608c5a104 100644 --- a/drivers/clk/qcom/gcc-sm8150.c +++ b/drivers/clk/qcom/gcc-sm8150.c @@ -123,6 +123,12 @@ static const struct clk_parent_data gcc_parents_0[] = { { .hw = &gpll0_out_even.clkr.hw }, }; +static const struct clk_parent_data gcc_parents_0_ao[] = { + { .fw_name = "bi_tcxo_ao", .name = "bi_tcxo_ao" }, + { .hw = &gpll0.clkr.hw }, + { .hw = &gpll0_out_even.clkr.hw }, +}; + static const struct parent_map gcc_parent_map_1[] = { { P_BI_TCXO, 0 }, { P_GPLL0_OUT_MAIN, 1 }, @@ -222,8 +228,8 @@ static struct clk_rcg2 gcc_cpuss_ahb_clk_src = { .freq_tbl = ftbl_gcc_cpuss_ahb_clk_src, .clkr.hw.init = &(struct clk_init_data){ .name = "gcc_cpuss_ahb_clk_src", - .parent_data = gcc_parents_0, - .num_parents = ARRAY_SIZE(gcc_parents_0), + .parent_data = gcc_parents_0_ao, + .num_parents = ARRAY_SIZE(gcc_parents_0_ao), .flags = CLK_SET_RATE_PARENT, .ops = &clk_rcg2_ops, },
Add active_only support for gcc_parents_0, this is needed because some of the clocks under it are critical which would vote on xo blocking the suspend. Signed-off-by: Satya Priya Kakitapalli <quic_skakitap@quicinc.com> --- drivers/clk/qcom/gcc-sm8150.c | 10 ++++++++-- 1 file changed, 8 insertions(+), 2 deletions(-)