| Message ID | 1532010175-28364-3-git-send-email-pdeschrijver@nvidia.com (mailing list archive) |
|---|---|
| State | Changes Requested, archived |
| Headers | show |
On 19/07/18 15:22, Peter De Schrijver wrote: > The first 3 possible parents of clk_out_[1-3] are defined as clk_m, clk_m/2 > and clk_m/4. However They are actually osc, osc/2 and osc/4. In chips prior > to Tegra210 clk_m and osc had the same frequency, so this difference didn't > matter. Since Tegra210 however, clk_m is a divided version of osc. This > results in CCF reporting the rate as only half the actual rate on Tegra210. > To fix this, we add new clocks which have osc, osc/2, osc/4 and the > respective extern clock as their possible parents and use them for > Tegra210. Also 2 new DT clock defines are added to reflect the new > names of clk_m/2 and clk_m/4. The old defines are still kept for backwards > compatibility. > > Signed-off-by: Peter De Schrijver <pdeschrijver@nvidia.com> I gave this a whirl on the Tegra210 Smaug which uses the PMC CLK_OUT1 to clock the audio codec, but unfortunately this did not work. The good news is that I know why. Basically, for Tegra210 there is a slight change in the behaviour of the CLKx_ACCEPT_REQ bit. Essentially, this bit must always be set if you want to use any of the osc clocks to drive the CLK_OUTx. Setting this bit does not effect the CLK_OUTx if you are using the CAR to drive the pin. The simplest thing to do for Tegra210 is to always set the CLKx_ACCEPT_REQ. Cheers Jon
Quoting Jon Hunter (2018-07-19 08:06:50) > > On 19/07/18 15:22, Peter De Schrijver wrote: > > The first 3 possible parents of clk_out_[1-3] are defined as clk_m, clk_m/2 > > and clk_m/4. However They are actually osc, osc/2 and osc/4. In chips prior > > to Tegra210 clk_m and osc had the same frequency, so this difference didn't > > matter. Since Tegra210 however, clk_m is a divided version of osc. This > > results in CCF reporting the rate as only half the actual rate on Tegra210. > > To fix this, we add new clocks which have osc, osc/2, osc/4 and the > > respective extern clock as their possible parents and use them for > > Tegra210. Also 2 new DT clock defines are added to reflect the new > > names of clk_m/2 and clk_m/4. The old defines are still kept for backwards > > compatibility. > > > > Signed-off-by: Peter De Schrijver <pdeschrijver@nvidia.com> > I gave this a whirl on the Tegra210 Smaug which uses the PMC CLK_OUT1 to > clock the audio codec, but unfortunately this did not work. The good news > is that I know why. Basically, for Tegra210 there is a slight change in the > behaviour of the CLKx_ACCEPT_REQ bit. Essentially, this bit must always be > set if you want to use any of the osc clocks to drive the CLK_OUTx. Setting > this bit does not effect the CLK_OUTx if you are using the CAR to drive the > pin. The simplest thing to do for Tegra210 is to always set the > CLKx_ACCEPT_REQ. > So resend with that bit also set? -- To unsubscribe from this list: send the line "unsubscribe linux-clk" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Wed, Jul 25, 2018 at 03:44:55PM -0700, Stephen Boyd wrote: > Quoting Jon Hunter (2018-07-19 08:06:50) > > > > On 19/07/18 15:22, Peter De Schrijver wrote: > > > The first 3 possible parents of clk_out_[1-3] are defined as clk_m, clk_m/2 > > > and clk_m/4. However They are actually osc, osc/2 and osc/4. In chips prior > > > to Tegra210 clk_m and osc had the same frequency, so this difference didn't > > > matter. Since Tegra210 however, clk_m is a divided version of osc. This > > > results in CCF reporting the rate as only half the actual rate on Tegra210. > > > To fix this, we add new clocks which have osc, osc/2, osc/4 and the > > > respective extern clock as their possible parents and use them for > > > Tegra210. Also 2 new DT clock defines are added to reflect the new > > > names of clk_m/2 and clk_m/4. The old defines are still kept for backwards > > > compatibility. > > > > > > Signed-off-by: Peter De Schrijver <pdeschrijver@nvidia.com> > > I gave this a whirl on the Tegra210 Smaug which uses the PMC CLK_OUT1 to > > clock the audio codec, but unfortunately this did not work. The good news > > is that I know why. Basically, for Tegra210 there is a slight change in the > > behaviour of the CLKx_ACCEPT_REQ bit. Essentially, this bit must always be > > set if you want to use any of the osc clocks to drive the CLK_OUTx. Setting > > this bit does not effect the CLK_OUTx if you are using the CAR to drive the > > pin. The simplest thing to do for Tegra210 is to always set the > > CLKx_ACCEPT_REQ. > > > > So resend with that bit also set? > We haven't figured out who should set this bit as it has some other side-effects like overriding the pinmux settings. Also not all CLK_OUT clocks appear to behave the same way. Peter. -- To unsubscribe from this list: send the line "unsubscribe linux-clk" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Quoting Peter De Schrijver (2018-07-26 01:19:20) > On Wed, Jul 25, 2018 at 03:44:55PM -0700, Stephen Boyd wrote: > > Quoting Jon Hunter (2018-07-19 08:06:50) > > > > > > On 19/07/18 15:22, Peter De Schrijver wrote: > > > > The first 3 possible parents of clk_out_[1-3] are defined as clk_m, clk_m/2 > > > > and clk_m/4. However They are actually osc, osc/2 and osc/4. In chips prior > > > > to Tegra210 clk_m and osc had the same frequency, so this difference didn't > > > > matter. Since Tegra210 however, clk_m is a divided version of osc. This > > > > results in CCF reporting the rate as only half the actual rate on Tegra210. > > > > To fix this, we add new clocks which have osc, osc/2, osc/4 and the > > > > respective extern clock as their possible parents and use them for > > > > Tegra210. Also 2 new DT clock defines are added to reflect the new > > > > names of clk_m/2 and clk_m/4. The old defines are still kept for backwards > > > > compatibility. > > > > > > > > Signed-off-by: Peter De Schrijver <pdeschrijver@nvidia.com> > > > I gave this a whirl on the Tegra210 Smaug which uses the PMC CLK_OUT1 to > > > clock the audio codec, but unfortunately this did not work. The good news > > > is that I know why. Basically, for Tegra210 there is a slight change in the > > > behaviour of the CLKx_ACCEPT_REQ bit. Essentially, this bit must always be > > > set if you want to use any of the osc clocks to drive the CLK_OUTx. Setting > > > this bit does not effect the CLK_OUTx if you are using the CAR to drive the > > > pin. The simplest thing to do for Tegra210 is to always set the > > > CLKx_ACCEPT_REQ. > > > > > > > So resend with that bit also set? > > > > We haven't figured out who should set this bit as it has some other > side-effects like overriding the pinmux settings. Also not all CLK_OUT > clocks appear to behave the same way. > Ok! Thanks for letting me know. -- To unsubscribe from this list: send the line "unsubscribe linux-clk" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/clk/tegra/clk-id.h b/drivers/clk/tegra/clk-id.h index b616e33..d9a0a87 100644 --- a/drivers/clk/tegra/clk-id.h +++ b/drivers/clk/tegra/clk-id.h @@ -326,6 +326,11 @@ enum clk_id { tegra_clk_pll_a_out0_out_adsp, tegra_clk_adsp, tegra_clk_adsp_neon, + tegra_clk_osc_div2, + tegra_clk_osc_div4, + tegra_clk_clk_out_1_osc_mux, + tegra_clk_clk_out_2_osc_mux, + tegra_clk_clk_out_3_osc_mux, tegra_clk_max, }; diff --git a/drivers/clk/tegra/clk-tegra-fixed.c b/drivers/clk/tegra/clk-tegra-fixed.c index 91c38f1..e15b67b 100644 --- a/drivers/clk/tegra/clk-tegra-fixed.c +++ b/drivers/clk/tegra/clk-tegra-fixed.c @@ -106,4 +106,20 @@ void __init tegra_fixed_clk_init(struct tegra_clk *tegra_clks) CLK_SET_RATE_PARENT, 1, 4); *dt_clk = clk; } + + /* clk_m_div2 */ + dt_clk = tegra_lookup_dt_id(tegra_clk_osc_div2, tegra_clks); + if (dt_clk) { + clk = clk_register_fixed_factor(NULL, "osc_div2", "osc", + CLK_SET_RATE_PARENT, 1, 2); + *dt_clk = clk; + } + + /* clk_m_div4 */ + dt_clk = tegra_lookup_dt_id(tegra_clk_osc_div4, tegra_clks); + if (dt_clk) { + clk = clk_register_fixed_factor(NULL, "osc_div4", "osc", + CLK_SET_RATE_PARENT, 1, 4); + *dt_clk = clk; + } } diff --git a/drivers/clk/tegra/clk-tegra-pmc.c b/drivers/clk/tegra/clk-tegra-pmc.c index 90a353a..9ddaf1b 100644 --- a/drivers/clk/tegra/clk-tegra-pmc.c +++ b/drivers/clk/tegra/clk-tegra-pmc.c @@ -58,6 +58,19 @@ struct pmc_clk_init_data { .gate_shift = _gate_shift,\ } +#define PMC_OSC_CLK(_num, _mux_shift, _gate_shift)\ + {\ + .mux_name = "clk_out_" #_num "_mux",\ + .gate_name = "clk_out_" #_num,\ + .parents = clk_out ##_num ##_osc_parents,\ + .num_parents = ARRAY_SIZE(clk_out ##_num ##_osc_parents),\ + .mux_id = tegra_clk_clk_out_ ##_num ##_osc_mux,\ + .gate_id = tegra_clk_clk_out_ ##_num,\ + .dev_name = "extern" #_num,\ + .mux_shift = _mux_shift,\ + .gate_shift = _gate_shift,\ + } + static DEFINE_SPINLOCK(clk_out_lock); static const char * const clk_out1_parents[] = { "clk_m", "clk_m_div2", @@ -72,10 +85,25 @@ struct pmc_clk_init_data { "clk_m_div4", "extern3", }; +static const char * const clk_out1_osc_parents[] = { "osc", "osc_div2", + "osc_div4", "extern1", +}; + +static const char * const clk_out2_osc_parents[] = { "osc", "osc_div2", + "osc_div4", "extern2", +}; + +static const char * const clk_out3_osc_parents[] = { "osc", "osc_div2", + "osc_div4", "extern3", +}; + static struct pmc_clk_init_data pmc_clks[] = { PMC_CLK(1, 6, 2), PMC_CLK(2, 14, 10), PMC_CLK(3, 22, 18), + PMC_OSC_CLK(1, 6, 2), + PMC_OSC_CLK(2, 14, 10), + PMC_OSC_CLK(3, 22, 18), }; void __init tegra_pmc_clk_init(void __iomem *pmc_base, diff --git a/drivers/clk/tegra/clk-tegra210.c b/drivers/clk/tegra/clk-tegra210.c index 5435d01..d92f1d7 100644 --- a/drivers/clk/tegra/clk-tegra210.c +++ b/drivers/clk/tegra/clk-tegra210.c @@ -2370,8 +2370,8 @@ struct utmi_clk_param { [tegra_clk_fuse_burn] = { .dt_id = TEGRA210_CLK_FUSE_BURN, .present = true }, [tegra_clk_clk_32k] = { .dt_id = TEGRA210_CLK_CLK_32K, .present = true }, [tegra_clk_clk_m] = { .dt_id = TEGRA210_CLK_CLK_M, .present = true }, - [tegra_clk_clk_m_div2] = { .dt_id = TEGRA210_CLK_CLK_M_DIV2, .present = true }, - [tegra_clk_clk_m_div4] = { .dt_id = TEGRA210_CLK_CLK_M_DIV4, .present = true }, + [tegra_clk_osc_div2] = { .dt_id = TEGRA210_CLK_OSC_DIV2, .present = true }, + [tegra_clk_osc_div4] = { .dt_id = TEGRA210_CLK_OSC_DIV4, .present = true }, [tegra_clk_pll_ref] = { .dt_id = TEGRA210_CLK_PLL_REF, .present = true }, [tegra_clk_pll_c] = { .dt_id = TEGRA210_CLK_PLL_C, .present = true }, [tegra_clk_pll_c_out1] = { .dt_id = TEGRA210_CLK_PLL_C_OUT1, .present = true }, @@ -2451,9 +2451,9 @@ struct utmi_clk_param { [tegra_clk_audio3_mux] = { .dt_id = TEGRA210_CLK_AUDIO3_MUX, .present = true }, [tegra_clk_audio4_mux] = { .dt_id = TEGRA210_CLK_AUDIO4_MUX, .present = true }, [tegra_clk_spdif_mux] = { .dt_id = TEGRA210_CLK_SPDIF_MUX, .present = true }, - [tegra_clk_clk_out_1_mux] = { .dt_id = TEGRA210_CLK_CLK_OUT_1_MUX, .present = true }, - [tegra_clk_clk_out_2_mux] = { .dt_id = TEGRA210_CLK_CLK_OUT_2_MUX, .present = true }, - [tegra_clk_clk_out_3_mux] = { .dt_id = TEGRA210_CLK_CLK_OUT_3_MUX, .present = true }, + [tegra_clk_clk_out_1_osc_mux] = { .dt_id = TEGRA210_CLK_CLK_OUT_1_MUX, .present = true }, + [tegra_clk_clk_out_2_osc_mux] = { .dt_id = TEGRA210_CLK_CLK_OUT_2_MUX, .present = true }, + [tegra_clk_clk_out_3_osc_mux] = { .dt_id = TEGRA210_CLK_CLK_OUT_3_MUX, .present = true }, [tegra_clk_maud] = { .dt_id = TEGRA210_CLK_MAUD, .present = true }, [tegra_clk_mipibif] = { .dt_id = TEGRA210_CLK_MIPIBIF, .present = true }, [tegra_clk_qspi] = { .dt_id = TEGRA210_CLK_QSPI, .present = true }, @@ -2496,8 +2496,8 @@ struct utmi_clk_param { { .con_id = "clk_m", .dt_id = TEGRA210_CLK_CLK_M }, { .con_id = "pll_ref", .dt_id = TEGRA210_CLK_PLL_REF }, { .con_id = "clk_32k", .dt_id = TEGRA210_CLK_CLK_32K }, - { .con_id = "clk_m_div2", .dt_id = TEGRA210_CLK_CLK_M_DIV2 }, - { .con_id = "clk_m_div4", .dt_id = TEGRA210_CLK_CLK_M_DIV4 }, + { .con_id = "osc_div2", .dt_id = TEGRA210_CLK_OSC_DIV2 }, + { .con_id = "osc_div4", .dt_id = TEGRA210_CLK_OSC_DIV4 }, { .con_id = "pll_c", .dt_id = TEGRA210_CLK_PLL_C }, { .con_id = "pll_c_out1", .dt_id = TEGRA210_CLK_PLL_C_OUT1 }, { .con_id = "pll_c2", .dt_id = TEGRA210_CLK_PLL_C2 }, diff --git a/include/dt-bindings/clock/tegra210-car.h b/include/dt-bindings/clock/tegra210-car.h index 6b77e72..08ec981 100644 --- a/include/dt-bindings/clock/tegra210-car.h +++ b/include/dt-bindings/clock/tegra210-car.h @@ -260,8 +260,13 @@ #define TEGRA210_CLK_FUSE_BURN 231 #define TEGRA210_CLK_CLK_32K 232 #define TEGRA210_CLK_CLK_M 233 -#define TEGRA210_CLK_CLK_M_DIV2 234 -#define TEGRA210_CLK_CLK_M_DIV4 235 + +#define TEGRA210_CLK_OSC_DIV2 234 +#define TEGRA210_CLK_OSC_DIV4 235 +/* for backwards compatibility */ +#define TEGRA210_CLK_CLK_M_DIV2 TEGRA210_CLK_OSC_DIV2 +#define TEGRA210_CLK_CLK_M_DIV4 TEGRA210_CLK_OSC_DIV4 + #define TEGRA210_CLK_PLL_REF 236 #define TEGRA210_CLK_PLL_C 237 #define TEGRA210_CLK_PLL_C_OUT1 238
The first 3 possible parents of clk_out_[1-3] are defined as clk_m, clk_m/2 and clk_m/4. However They are actually osc, osc/2 and osc/4. In chips prior to Tegra210 clk_m and osc had the same frequency, so this difference didn't matter. Since Tegra210 however, clk_m is a divided version of osc. This results in CCF reporting the rate as only half the actual rate on Tegra210. To fix this, we add new clocks which have osc, osc/2, osc/4 and the respective extern clock as their possible parents and use them for Tegra210. Also 2 new DT clock defines are added to reflect the new names of clk_m/2 and clk_m/4. The old defines are still kept for backwards compatibility. Signed-off-by: Peter De Schrijver <pdeschrijver@nvidia.com> --- drivers/clk/tegra/clk-id.h | 5 +++++ drivers/clk/tegra/clk-tegra-fixed.c | 16 ++++++++++++++++ drivers/clk/tegra/clk-tegra-pmc.c | 28 ++++++++++++++++++++++++++++ drivers/clk/tegra/clk-tegra210.c | 14 +++++++------- include/dt-bindings/clock/tegra210-car.h | 9 +++++++-- 5 files changed, 63 insertions(+), 9 deletions(-)