Skip to contentSkip to conversation

Renesas R-Car Gen3 CPG support V4

PATCH Superseded Magnus Damm 2 participants 14 messages Aug 29, 2015 – Sep 1, 2015

Subsystems (5)

v4 of 8 Superseded Latest is v8 →

lore
Share & tools
lore

Magnus Damm

author maintainer
#0 lore raw

v4 cover Renesas R-Car Gen3 CPG support V4

Renesas R-Car Gen3 CPG support V4

[PATCH v4 01/05] clk: shmobile: Rework CONFIG_ARCH_SHMOBILE_MULTI [PATCH v4 02/05] clk: shmobile: Add Renesas R-Car Gen3 CPG support [PATCH v4 03/05] clk: shmobile: rcar-gen3: Add CPG/MSTP Clock Domain support [PATCH v4 04/05] clk: shmobile: Add r8a7795 SoC to MSTP bindings [PATCH v4 05/05] clk: shmobile: Make MSTP clock-output-names optional

This series is my second attempt to provide R-Car Gen3 clock support through a single "easy to use" patch series for drivers/clk and DT documentation bits.

There are no special runtime or build time dependencies. SoC / board specific integration code (DTS) is provided separately from this series.

This version includes several minor updates for all the patches. For details please see each individual patch.

---

Written to be picked up independently from other R-Car Gen3 patches.

 Documentation/devicetree/bindings/clock/renesas,cpg-mstp-clocks.txt      |    1 
 Documentation/devicetree/bindings/clock/renesas,rcar-gen3-cpg-clocks.txt |   57 ++
 drivers/clk/Makefile                                                     |    1 
 drivers/clk/shmobile/Makefile                                            |   23 
 drivers/clk/shmobile/clk-mstp.c                                          |   28 -
 drivers/clk/shmobile/clk-rcar-gen3.c                                     |  251 ++++++++++
 6 files changed, 338 insertions(+), 23 deletions(-)

damm@little-apple ~/patches/20150829/gen3-clk-v4 $

author maintainer
#1 lore raw

PATCH v4 1/5 clk: shmobile: Rework CONFIG_ARCH_SHMOBILE_MULTI

From: Magnus Damm <[email protected]>

Shmobile is all multiplatform these days, so in get rid of the reference to CONFIG_ARCH_SHMOBILE_MULTI in drivers/clk/shmobile/.

Also instead of always enabling DIV6 and MSTP adjust the Makefile to enable DIV6 and MSTP depending on if they are included in the SoC or not.

---

Earlier known as "clk: shmobile: Get rid of CONFIG_ARCH_SHMOBILE_MULTI"

Changes since V3:
- Got rid of hunks that modified drivers/clk/Makefile
- This to prevent build issue with non-CCF enabled SH arch
- A nice side effect is that this patch now can be merged any time

Changes since V2:
- Fixed patch subject typo CONFIG_SHMOBILE_MULTI -> CONFIG_ARCH_SHMOBILE_MULTI

 drivers/clk/shmobile/Makefile |   22 ++++++++++------------
 1 file changed, 10 insertions(+), 12 deletions(-)
drivers/clk/shmobile/Makefile +10 -12
@@ -1,13 +1,11 @@
 obj-$(CONFIG_ARCH_EMEV2)		+= clk-emev2.o
-obj-$(CONFIG_ARCH_R7S72100)		+= clk-rz.o
-obj-$(CONFIG_ARCH_R8A73A4)		+= clk-r8a73a4.o
-obj-$(CONFIG_ARCH_R8A7740)		+= clk-r8a7740.o
-obj-$(CONFIG_ARCH_R8A7778)		+= clk-r8a7778.o
-obj-$(CONFIG_ARCH_R8A7779)		+= clk-r8a7779.o
-obj-$(CONFIG_ARCH_R8A7790)		+= clk-rcar-gen2.o
-obj-$(CONFIG_ARCH_R8A7791)		+= clk-rcar-gen2.o
-obj-$(CONFIG_ARCH_R8A7793)		+= clk-rcar-gen2.o
-obj-$(CONFIG_ARCH_R8A7794)		+= clk-rcar-gen2.o
-obj-$(CONFIG_ARCH_SH73A0)		+= clk-sh73a0.o
-obj-$(CONFIG_ARCH_SHMOBILE_MULTI)	+= clk-div6.o
-obj-$(CONFIG_ARCH_SHMOBILE_MULTI)	+= clk-mstp.o
+obj-$(CONFIG_ARCH_R7S72100)		+= clk-rz.o clk-mstp.o
+obj-$(CONFIG_ARCH_R8A73A4)		+= clk-r8a73a4.o clk-mstp.o clk-div6.o
+obj-$(CONFIG_ARCH_R8A7740)		+= clk-r8a7740.o clk-mstp.o clk-div6.o
+obj-$(CONFIG_ARCH_R8A7778)		+= clk-r8a7778.o clk-mstp.o
+obj-$(CONFIG_ARCH_R8A7779)		+= clk-r8a7779.o clk-mstp.o
+obj-$(CONFIG_ARCH_R8A7790)		+= clk-rcar-gen2.o clk-mstp.o clk-div6.o
+obj-$(CONFIG_ARCH_R8A7791)		+= clk-rcar-gen2.o clk-mstp.o clk-div6.o
+obj-$(CONFIG_ARCH_R8A7793)		+= clk-rcar-gen2.o clk-mstp.o clk-div6.o
+obj-$(CONFIG_ARCH_R8A7794)		+= clk-rcar-gen2.o clk-mstp.o clk-div6.o
+obj-$(CONFIG_ARCH_SH73A0)		+= clk-sh73a0.o clk-mstp.o clk-div6.o
author maintainer
#2 lore raw

PATCH v4 2/5 clk: shmobile: Add Renesas R-Car Gen3 CPG support

In reply to Magnus Damm

From: Gaku Inami <[email protected]>

This patch adds initial CPG support for R-Car Generation 3 and in particular the R8A7795 SoC.

The R-Car Gen3 clock hardware has a register write protection feature that needs to be shared between the CPG function needs to be shared between the CPG and MSTP hardware somehow. So far this feature is simply ignored.

---

Changes since V3: (Magnus Damm <[email protected]>)
- Reworked driver to incorporate most feedback from Stephen Boyd - thanks!!
- Major things like syscon and driver model require more discussion.
- Added hunk to build drivers/clk/shmobile if ARCH_RENESAS is set.

Changes since V2: (Magnus Damm <[email protected]>)
- Reworked driver to rely on clock index instead of strings.
- Dropped use of "clock-output-names".

Earlier versions: Kuninori Morimoto <[email protected]> Initial version: Gaku Inami <[email protected]>

 Documentation/devicetree/bindings/clock/renesas,rcar-gen3-cpg-clocks.txt |   31 +
 drivers/clk/Makefile                                                     |    1 
 drivers/clk/shmobile/Makefile                                            |    1 
 drivers/clk/shmobile/clk-rcar-gen3.c                                     |  249 ++++++++++
 4 files changed, 282 insertions(+)
Documentation/devicetree/bindings/clock/renesas,rcar-gen3-cpg-clocks.txt new file +31 -0
@@ -0,0 +1,31 @@
+* Renesas R-Car Gen3 Clock Pulse Generator (CPG)
+
+The CPG generates core clocks for the R-Car Gen3 SoCs. It includes three PLLs
+and several fixed ratio dividers.
+
+Required Properties:
+
+  - compatible: Must be one of
+    - "renesas,r8a7795-cpg-clocks" for the r8a7795 CPG
+    - "renesas,rcar-gen3-cpg-clocks" for the generic R-Car Gen3 CPG
+
+  - reg: Base address and length of the memory resource used by the CPG
+
+  - clocks: References to the parent clocks: first to the EXTAL clock
+  - #clock-cells: Must be 1
+  - clock-output-names: The names of the clocks. Supported clocks are
+    "main", "pll0", "pll1", "pll2", "pll3", "pll4"
+
+
+Example
+-------
+
+	cpg_clocks: cpg_clocks@e6150000 {
+		compatible = "renesas,r8a7795-cpg-clocks",
+			     "renesas,rcar-gen3-cpg-clocks";
+		reg = <0 0xe6150000 0 0x1000>;
+		clocks = <&extal_clk>;
+		#clock-cells = <1>;
+		clock-output-names = "main", "pll0", "pll1","pll2",
+				     "pll3", "pll4";
+	};
drivers/clk/Makefile +1 -0
@@ -67,6 +67,7 @@ obj-$(CONFIG_COMMON_CLK_QCOM)		+= qcom/
 obj-$(CONFIG_ARCH_ROCKCHIP)		+= rockchip/
 obj-$(CONFIG_COMMON_CLK_SAMSUNG)	+= samsung/
 obj-$(CONFIG_ARCH_SHMOBILE_MULTI)	+= shmobile/
+obj-$(CONFIG_ARCH_RENESAS)		+= shmobile/
 obj-$(CONFIG_ARCH_SIRF)			+= sirf/
 obj-$(CONFIG_ARCH_SOCFPGA)		+= socfpga/
 obj-$(CONFIG_PLAT_SPEAR)		+= spear/
drivers/clk/shmobile/Makefile +1 -0
@@ -8,4 +8,5 @@ obj-$(CONFIG_ARCH_R8A7790)		+= clk-rcar-
 obj-$(CONFIG_ARCH_R8A7791)		+= clk-rcar-gen2.o clk-mstp.o clk-div6.o
 obj-$(CONFIG_ARCH_R8A7793)		+= clk-rcar-gen2.o clk-mstp.o clk-div6.o
 obj-$(CONFIG_ARCH_R8A7794)		+= clk-rcar-gen2.o clk-mstp.o clk-div6.o
+obj-$(CONFIG_ARCH_R8A7795)		+= clk-rcar-gen3.o clk-mstp.o clk-div6.o
 obj-$(CONFIG_ARCH_SH73A0)		+= clk-sh73a0.o clk-mstp.o clk-div6.o
drivers/clk/shmobile/clk-rcar-gen3.c new file +249 -0

249 lines not on this page · raw message

author maintainer
#3 lore raw

Re: [PATCH v4 02/05] clk: shmobile: Add Renesas R-Car Gen3 CPG support

Quoted 56 lines · Magnus Damm

From: Gaku Inami <[email protected]>

This patch adds initial CPG support for R-Car Generation 3 and in particular the R8A7795 SoC.

The R-Car Gen3 clock hardware has a register write protection feature that needs to be shared between the CPG function needs to be shared between the CPG and MSTP hardware somehow. So far this feature is simply ignored.

---

Changes since V3: (Magnus Damm <[email protected]>)
- Reworked driver to incorporate most feedback from Stephen Boyd - thanks!!
- Major things like syscon and driver model require more discussion.
- Added hunk to build drivers/clk/shmobile if ARCH_RENESAS is set.

Changes since V2: (Magnus Damm <[email protected]>)
- Reworked driver to rely on clock index instead of strings.
- Dropped use of "clock-output-names".

Earlier versions: Kuninori Morimoto <[email protected]> Initial version: Gaku Inami <[email protected]>

 Documentation/devicetree/bindings/clock/renesas,rcar-gen3-cpg-clocks.txt |   31 +
 drivers/clk/Makefile                                                     |    1
 drivers/clk/shmobile/Makefile                                            |    1
 drivers/clk/shmobile/clk-rcar-gen3.c                                     |  249 ++++++++++
 4 files changed, 282 insertions(+)
Documentation/devicetree/bindings/clock/renesas,rcar-gen3-cpg-clocks.txt 2015-08-29 16:23:54.982366518 +0900 new file +21 -0
@@ -0,0 +1,31 @@
+* Renesas R-Car Gen3 Clock Pulse Generator (CPG)
+
+The CPG generates core clocks for the R-Car Gen3 SoCs. It includes three PLLs
+and several fixed ratio dividers.
+
+Required Properties:
+
+  - compatible: Must be one of
+    - "renesas,r8a7795-cpg-clocks" for the r8a7795 CPG
+    - "renesas,rcar-gen3-cpg-clocks" for the generic R-Car Gen3 CPG
+
+  - reg: Base address and length of the memory resource used by the CPG
+
+  - clocks: References to the parent clocks: first to the EXTAL clock
+  - #clock-cells: Must be 1
+  - clock-output-names: The names of the clocks. Supported clocks are
+    "main", "pll0", "pll1", "pll2", "pll3", "pll4"
+
+
+Example
+-------

Documentation/devicetree/bindings/clock/renesas,rcar-gen3-cpg-clocks.txt 2015-08-29 16:23:54.982366518 +0900 line 31

Quoted 10 lines
Documentation/devicetree/bindings/clock/renesas,rcar-gen3-cpg-clocks.txt 2015-08-29 16:23:54.982366518 +0900 new file +10 -0
+
+       cpg_clocks: cpg_clocks@e6150000 {
+               compatible = "renesas,r8a7795-cpg-clocks",
+                            "renesas,rcar-gen3-cpg-clocks";
+               reg = <0 0xe6150000 0 0x1000>;
+               clocks = <&extal_clk>;
+               #clock-cells = <1>;
+               clock-output-names = "main", "pll0", "pll1","pll2",
+                                    "pll3", "pll4";
+       };

Seems that I forgot update the DT binding bits. Will fix in next verison.

Thanks,

/ magnus

#4 lore raw

Re: [PATCH v4 02/05] clk: shmobile: Add Renesas R-Car Gen3 CPG support

In reply to Magnus Damm

Hi Magnus,

Quoted 12 lines · Magnus Damm

From: Gaku Inami <[email protected]>

This patch adds initial CPG support for R-Car Generation 3 and in particular the R8A7795 SoC.

The R-Car Gen3 clock hardware has a register write protection feature that needs to be shared between the CPG function needs to be shared between the CPG and MSTP hardware somehow. So far this feature is simply ignored.

Quoted 6 lines

---

Changes since V3: (Magnus Damm <[email protected]>)
- Reworked driver to incorporate most feedback from Stephen Boyd - thanks!!
- Major things like syscon and driver model require more discussion.
- Added hunk to build drivers/clk/shmobile if ARCH_RENESAS is set.

Thanks for the updates!

Quoted 2 lines

Changes since V2: (Magnus Damm <[email protected]>)
- Reworked driver to rely on clock index instead of strings.

I had completely missed this part in the previous version...

drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900

Quoted 13 lines

- Dropped use of "clock-output-names".

drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900 new file +6 -0
@@ -0,0 +1,249 @@
 
+#define RCAR_GEN3_CLK_MAIN     0
+#define RCAR_GEN3_CLK_PLL0     1
+#define RCAR_GEN3_CLK_PLL1     2
+#define RCAR_GEN3_CLK_PLL2     3
+#define RCAR_GEN3_CLK_PLL3     4
+#define RCAR_GEN3_CLK_PLL4     5

These values are intimately tied to R8A7795_CLK_* in include/dt-bindings/clock/r8a7795-clock.h. Perhaps we can use these instead?

If we ever have a Gen3 SoC with more clocks, we have to update this anyway.

drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900

Quoted 4 lines
drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900 new file +4 -0
+struct rcar_gen3_cpg {
+       struct clk_onecell_data data;
+       spinlock_t lock;
+       void __iomem *reg;

If you add the clocks here, you don't have to allocate them separately later:

        struct clk clocks[...];
Quoted 9 lines
drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900 new file +9 -0
+static void __init rcar_gen3_cpg_clocks_init(struct device_node *np)
+{
+       const struct cpg_pll_config *config;
+       struct rcar_gen3_cpg *cpg;
+       struct clk **clks;
+       u32 cpg_mode;
+       unsigned int i;
+       int num_clks;
+

drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900

Quoted 10 lines
drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900 new file +10 -0
+       cpg_mode = rcar_gen3_read_mode_pins();
+
+       num_clks = of_property_count_strings(np, "clock-indices");
+       if (num_clks < 0) {
+               pr_err("%s: failed to count clocks\n", __func__);
+               return;
+       }
+
+       cpg = kzalloc(sizeof(*cpg), GFP_KERNEL);
+       clks = kcalloc(num_clks, sizeof(*clks), GFP_KERNEL);

I.e. these 2 allocation can be combined by embedding the struct clks in struct rcar_gen3_cpg...

drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900

Quoted 3 lines
drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900 new file +3 -0
+       if (cpg == NULL || clks == NULL) {
+               kfree(cpg);
+               kfree(clks);

... and there's no longer a need to kfree() anything.

Quoted 1 line

+ pr_err("%s: failed to allocate cpg\n", __func__);

There's no need to print an error on allocation failures. The system will scream anyway (if there's already a console ;-).

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds

author maintainer
#5 lore raw

Re: [PATCH v4 02/05] clk: shmobile: Add Renesas R-Car Gen3 CPG support

Hi Geert,

Quoted 22 lines · Geert Uytterhoeven

Hi Magnus,

Quoted 12 lines · Magnus Damm

From: Gaku Inami <[email protected]>

This patch adds initial CPG support for R-Car Generation 3 and in particular the R8A7795 SoC.

The R-Car Gen3 clock hardware has a register write protection feature that needs to be shared between the CPG function needs to be shared between the CPG and MSTP hardware somehow. So far this feature is simply ignored.

Quoted 6 lines

---

Changes since V3: (Magnus Damm <[email protected]>)
- Reworked driver to incorporate most feedback from Stephen Boyd - thanks!!
- Major things like syscon and driver model require more discussion.
- Added hunk to build drivers/clk/shmobile if ARCH_RENESAS is set.

Quoted 6 lines

Thanks for the updates!

Quoted 2 lines

Changes since V2: (Magnus Damm <[email protected]>)
- Reworked driver to rely on clock index instead of strings.

I had completely missed this part in the previous version...

No worries, perhaps not so well documented. And there are many changes floating around...

Quoted 7 lines
Quoted 6 lines

- Dropped use of "clock-output-names".

drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900 new file +0 -0
@@ -0,0 +1,249 @@
Quoted 10 lines

drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900 line 6

Quoted 6 lines
drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900 new file +6 -0
+#define RCAR_GEN3_CLK_MAIN     0
+#define RCAR_GEN3_CLK_PLL0     1
+#define RCAR_GEN3_CLK_PLL1     2
+#define RCAR_GEN3_CLK_PLL2     3
+#define RCAR_GEN3_CLK_PLL3     4
+#define RCAR_GEN3_CLK_PLL4     5

These values are intimately tied to R8A7795_CLK_* in include/dt-bindings/clock/r8a7795-clock.h. Perhaps we can use these instead?

You are right that they are connected together, but I'm not sure if I prefer to include that file. If we start sharing files between the CPG and the integration code then a dependency hell breaks loose when it comes to merge order. If we want to share the #defines then I propose that we handle that incrementally once all pieces are merged.

On R-Car Gen2 we have a single CPG driver that is potentially shared across multiple SoCs in the same generation / family. In that case we use strings and the clock-output-names as part of the ABI if I'm not mistaken. On this driver we use clock index numbers instead, so I'd like to see these defines as part of a shared DT binding ABI where new SoCs can simply add new clocks at the end.

Quoted 2 lines

If we ever have a Gen3 SoC with more clocks, we have to update this anyway.

That's right!

Quoted 33 lines

drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900 line 10

Quoted 4 lines
drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900 new file +4 -0
+struct rcar_gen3_cpg {
+       struct clk_onecell_data data;
+       spinlock_t lock;
+       void __iomem *reg;

If you add the clocks here, you don't have to allocate them separately later:

        struct clk clocks[...];
Quoted 9 lines
drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900 new file +9 -0
+static void __init rcar_gen3_cpg_clocks_init(struct device_node *np)
+{
+       const struct cpg_pll_config *config;
+       struct rcar_gen3_cpg *cpg;
+       struct clk **clks;
+       u32 cpg_mode;
+       unsigned int i;
+       int num_clks;
+

drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900 line 29

Quoted 10 lines
drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900 new file +10 -0
+       cpg_mode = rcar_gen3_read_mode_pins();
+
+       num_clks = of_property_count_strings(np, "clock-indices");
+       if (num_clks < 0) {
+               pr_err("%s: failed to count clocks\n", __func__);
+               return;
+       }
+
+       cpg = kzalloc(sizeof(*cpg), GFP_KERNEL);
+       clks = kcalloc(num_clks, sizeof(*clks), GFP_KERNEL);

I.e. these 2 allocation can be combined by embedding the struct clks in struct rcar_gen3_cpg...

Quoted 5 lines

drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900 line 32

Quoted 3 lines
drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900 new file +3 -0
+       if (cpg == NULL || clks == NULL) {
+               kfree(cpg);
+               kfree(clks);

... and there's no longer a need to kfree() anything.

Good idea. I'll give it a shot.

Quoted 4 lines

drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900 line 33

Quoted 1 line
drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900 new file +1 -0
+               pr_err("%s: failed to allocate cpg\n", __func__);

There's no need to print an error on allocation failures. The system will scream anyway (if there's already a console ;-).

Ok, will make it silent!

Thanks for your comments!!

Cheers,

/ magnus

#6 lore raw

Re: [PATCH v4 02/05] clk: shmobile: Add Renesas R-Car Gen3 CPG support

Hi Magnus,

Quoted 15 lines · Magnus Damm
Quoted 4 lines
Quoted 3 lines
drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900 new file +0 -0
@@ -0,0 +1,249 @@
Quoted 10 lines

drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900 line 6

Quoted 6 lines
drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900 new file +6 -0
+#define RCAR_GEN3_CLK_MAIN     0
+#define RCAR_GEN3_CLK_PLL0     1
+#define RCAR_GEN3_CLK_PLL1     2
+#define RCAR_GEN3_CLK_PLL2     3
+#define RCAR_GEN3_CLK_PLL3     4
+#define RCAR_GEN3_CLK_PLL4     5

These values are intimately tied to R8A7795_CLK_* in include/dt-bindings/clock/r8a7795-clock.h. Perhaps we can use these instead?

Quoted 5 lines

You are right that they are connected together, but I'm not sure if I prefer to include that file. If we start sharing files between the CPG and the integration code then a dependency hell breaks loose when it comes to merge order. If we want to share the #defines then I propose that we handle that incrementally once all pieces are merged.

OK, we can move them to include/dt-bindings/clock/rcar-gen3-clock.h later.

Quoted 6 lines

On R-Car Gen2 we have a single CPG driver that is potentially shared across multiple SoCs in the same generation / family. In that case we use strings and the clock-output-names as part of the ABI if I'm not mistaken. On this driver we use clock index numbers instead, so I'd like to see these defines as part of a shared DT binding ABI where new SoCs can simply add new clocks at the end.

OK, so rcar_gen3_cpg.clks[] is really intended to be variable-sized.

Hence I suggest to write it that way:

    struct rcar_gen3_cpg {
            struct clk_onecell_data data;
            spinlock_t lock;
            void __iomem *reg;
            struct clk *clks[0];
    };

Then you can allocate it with

kzalloc(sizeof(*cpg) + num_clks * sizeof(rcar_gen3_cpg.clks[0], ...)

and kill RCAR_GEN3_CLK_NR (one less definition that can go out of sync).

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds

author maintainer
#7 lore raw

Re: [PATCH v4 02/05] clk: shmobile: Add Renesas R-Car Gen3 CPG support

Hi Geert,

Quoted 18 lines · Geert Uytterhoeven

Hi Magnus,

Quoted 15 lines · Magnus Damm
Quoted 14 lines

drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900

Quoted 10 lines
drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900 new file +6 -0
@@ -0,0 +1,249 @@
 
+#define RCAR_GEN3_CLK_MAIN     0
+#define RCAR_GEN3_CLK_PLL0     1
+#define RCAR_GEN3_CLK_PLL1     2
+#define RCAR_GEN3_CLK_PLL2     3
+#define RCAR_GEN3_CLK_PLL3     4
+#define RCAR_GEN3_CLK_PLL4     5

These values are intimately tied to R8A7795_CLK_* in include/dt-bindings/clock/r8a7795-clock.h. Perhaps we can use these instead?

Quoted 8 lines
Quoted 5 lines

You are right that they are connected together, but I'm not sure if I prefer to include that file. If we start sharing files between the CPG and the integration code then a dependency hell breaks loose when it comes to merge order. If we want to share the #defines then I propose that we handle that incrementally once all pieces are merged.

OK, we can move them to include/dt-bindings/clock/rcar-gen3-clock.h later.

So in the future we may then want to change r8a7795-clock.h into:

#define R8A7795_CLK_MAIN RCAR_GEN3_CLK_MAIN

?

Quoted 18 lines
Quoted 6 lines

On R-Car Gen2 we have a single CPG driver that is potentially shared across multiple SoCs in the same generation / family. In that case we use strings and the clock-output-names as part of the ABI if I'm not mistaken. On this driver we use clock index numbers instead, so I'd like to see these defines as part of a shared DT binding ABI where new SoCs can simply add new clocks at the end.

OK, so rcar_gen3_cpg.clks[] is really intended to be variable-sized.

Hence I suggest to write it that way:

    struct rcar_gen3_cpg {
            struct clk_onecell_data data;
            spinlock_t lock;
            void __iomem *reg;
            struct clk *clks[0];
    };
Quoted 5 lines

Then you can allocate it with

   kzalloc(sizeof(*cpg) + num_clks * sizeof(rcar_gen3_cpg.clks[0], ...)

and kill RCAR_GEN3_CLK_NR (one less definition that can go out of sync).

I see RCAR_GEN3_CLK_NR as the number of different clocks that the Gen3 CPG driver can support. Exactly which one that the SoC specific DTS wants to use depends on each SoC. So for future SoCs we may want to keep on adding new clocks at the end. And the clock-indices list may be sparsely populated in DTS.

I am not 100% sure about how to handle error cases. Say a new DTS is booted on an old kernel then some clocks may not be supported by the CPG driver. For those cases the switch/case statement in the driver will simply return errors for the unsupported clocks. And the rest of the kernel has to figure out how to deal with that. That seems sane, right?

Thanks,

/ magnus

#8 lore raw

Re: [PATCH v4 02/05] clk: shmobile: Add Renesas R-Car Gen3 CPG support

Hi Magnus,

Quoted 24 lines · Magnus Damm
Quoted 22 lines · Geert Uytterhoeven
Quoted 20 lines · Magnus Damm
Quoted 4 lines
Quoted 3 lines
drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900 new file +0 -0
@@ -0,0 +1,249 @@
Quoted 10 lines

drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900 line 6

Quoted 6 lines
drivers/clk/shmobile/clk-rcar-gen3.c 2015-08-29 16:30:59.032366518 +0900 new file +6 -0
+#define RCAR_GEN3_CLK_MAIN     0
+#define RCAR_GEN3_CLK_PLL0     1
+#define RCAR_GEN3_CLK_PLL1     2
+#define RCAR_GEN3_CLK_PLL2     3
+#define RCAR_GEN3_CLK_PLL3     4
+#define RCAR_GEN3_CLK_PLL4     5

These values are intimately tied to R8A7795_CLK_* in include/dt-bindings/clock/r8a7795-clock.h. Perhaps we can use these instead?

You are right that they are connected together, but I'm not sure if I prefer to include that file. If we start sharing files between the CPG and the integration code then a dependency hell breaks loose when it comes to merge order. If we want to share the #defines then I propose that we handle that incrementally once all pieces are merged.

Quoted 8 lines
Quoted 2 lines

OK, we can move them to include/dt-bindings/clock/rcar-gen3-clock.h later.

So in the future we may then want to change r8a7795-clock.h into:

#define R8A7795_CLK_MAIN RCAR_GEN3_CLK_MAIN

?

Exactly.

Quoted 22 lines
Quoted 22 lines
Quoted 6 lines

On R-Car Gen2 we have a single CPG driver that is potentially shared across multiple SoCs in the same generation / family. In that case we use strings and the clock-output-names as part of the ABI if I'm not mistaken. On this driver we use clock index numbers instead, so I'd like to see these defines as part of a shared DT binding ABI where new SoCs can simply add new clocks at the end.

OK, so rcar_gen3_cpg.clks[] is really intended to be variable-sized.

Hence I suggest to write it that way:

    struct rcar_gen3_cpg {
            struct clk_onecell_data data;
            spinlock_t lock;
            void __iomem *reg;
            struct clk *clks[0];
    };

Then you can allocate it with

   kzalloc(sizeof(*cpg) + num_clks * sizeof(rcar_gen3_cpg.clks[0], ...)
Quoted 7 lines
Quoted 1 line

and kill RCAR_GEN3_CLK_NR (one less definition that can go out of sync).

I see RCAR_GEN3_CLK_NR as the number of different clocks that the Gen3 CPG driver can support. Exactly which one that the SoC specific DTS wants to use depends on each SoC. So for future SoCs we may want to keep on adding new clocks at the end. And the clock-indices list may be sparsely populated in DTS.

OK. And it's just a few pointers (initially I thought it contained full clk structs).

Quoted 6 lines

I am not 100% sure about how to handle error cases. Say a new DTS is booted on an old kernel then some clocks may not be supported by the CPG driver. For those cases the switch/case statement in the driver will simply return errors for the unsupported clocks. And the rest of the kernel has to figure out how to deal with that. That seems sane, right?

Indeed. Those clocks will not exist, and any driver trying to use them will fail to clk_get() them.

Note that fixed-factor-clocks pointing to them will still be instantiated, with a zero parent rate.

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds

author maintainer
#9 lore raw

PATCH v4 3/5 clk: shmobile: rcar-gen3: Add CPG/MSTP Clock Domain support

In reply to Magnus Damm

From: Geert Uytterhoeven <[email protected]>

Add Clock Domain support to the R-Car Gen3 Clock Pulse Generator (CPG) driver using the generic PM Domain. This allows to power-manage the module clocks of SoC devices that are part of the CPG/MSTP Clock Domain using Runtime PM, or for system suspend/resume.

SoC devices that are part of the CPG/MSTP Clock Domain and can be power-managed through an MSTP clock should be tagged in DT with a proper "power-domains" property.

---

Changes since V3:
- Fixed so correct file is actually used. =)
- Took V2 from Geert but filtered out Kconfig.platforms bits

 Documentation/devicetree/bindings/clock/renesas,rcar-gen3-cpg-clocks.txt |   26 +++++++++-
 drivers/clk/shmobile/clk-rcar-gen3.c                                     |    2 
 2 files changed, 26 insertions(+), 2 deletions(-)
Documentation/devicetree/bindings/clock/renesas,rcar-gen3-cpg-clocks.txt +24 -2
@@ -2,6 +2,8 @@
 
 The CPG generates core clocks for the R-Car Gen3 SoCs. It includes three PLLs
 and several fixed ratio dividers.
+The CPG also provides a Clock Domain for SoC devices, in combination with the
+CPG Module Stop (MSTP) Clocks.
 
 Required Properties:
 
@@ -15,10 +17,18 @@ Required Properties:
   - #clock-cells: Must be 1
   - clock-output-names: The names of the clocks. Supported clocks are
     "main", "pll0", "pll1", "pll2", "pll3", "pll4"
+  - #power-domain-cells: Must be 0
 
+SoC devices that are part of the CPG/MSTP Clock Domain and can be power-managed
+through an MSTP clock should refer to the CPG device node in their
+"power-domains" property, as documented by the generic PM domain bindings in
+Documentation/devicetree/bindings/power/power_domain.txt.
 
-Example
--------
+
+Examples
+--------
+
+  - CPG device node:
 
 	cpg_clocks: cpg_clocks@e6150000 {
 		compatible = "renesas,r8a7795-cpg-clocks",
@@ -28,4 +38,16 @@ Example
 		#clock-cells = <1>;
 		clock-output-names = "main", "pll0", "pll1","pll2",
 				     "pll3", "pll4";
+		#power-domain-cells = <0>;
+	};
+
+  - CPG/MSTP Clock Domain member device node:
+
+	scif2: serial@e6e88000 {
+		compatible = "renesas,scif-r8a7795", "renesas,scif";
+		reg = <0 0xe6e88000 0 64>;
+		interrupts = <GIC_SPI 164 IRQ_TYPE_LEVEL_HIGH>;
+		clocks = <&mstp3_clks RCAR_R8A7795_CLK_SCIF2>;
+		clock-names = "sci_ick";
+		power-domains = <&cpg_clocks>;
 	};
drivers/clk/shmobile/clk-rcar-gen3.c +2 -0
@@ -244,6 +244,8 @@ static void __init rcar_gen3_cpg_clocks_
 	}
 
 	of_clk_add_provider(np, of_clk_src_onecell_get, &cpg->data);
+
+	cpg_mstp_add_clk_domain(np);
 }
 CLK_OF_DECLARE(rcar_gen3_cpg_clks, "renesas,rcar-gen3-cpg-clocks",
 	       rcar_gen3_cpg_clocks_init);
author maintainer
#10 lore raw

PATCH v4 4/5 clk: shmobile: Add r8a7795 SoC to MSTP bindings

In reply to Magnus Damm

From: Kuninori Morimoto <[email protected]>

Add r8a7795 to the MSTP DT binding documentation.

---

Changes since V3: (Magnus Damm <[email protected]>)
- Picked up patch floating around on ML, added change log

 Documentation/devicetree/bindings/clock/renesas,cpg-mstp-clocks.txt |    1 +
 1 file changed, 1 insertion(+)
Documentation/devicetree/bindings/clock/renesas,cpg-mstp-clocks.txt +1 -0
@@ -19,6 +19,7 @@ Required Properties:
     - "renesas,r8a7791-mstp-clocks" for R8A7791 (R-Car M2-W) MSTP gate clocks
     - "renesas,r8a7793-mstp-clocks" for R8A7793 (R-Car M2-N) MSTP gate clocks
     - "renesas,r8a7794-mstp-clocks" for R8A7794 (R-Car E2) MSTP gate clocks
+    - "renesas,r8a7795-mstp-clocks" for R8A7795 (R-Car H3) MSTP gate clocks
     - "renesas,sh73a0-mstp-clocks" for SH73A0 (SH-MobileAG5) MSTP gate clocks
     and "renesas,cpg-mstp-clocks" as a fallback.
   - reg: Base address and length of the I/O mapped registers used by the MSTP
Acked-by on v4 4/5
#11 lore raw

Re: [PATCH v4 04/05] clk: shmobile: Add r8a7795 SoC to MSTP bindings

Quoted 6 lines · Magnus Damm

From: Kuninori Morimoto <[email protected]>

Add r8a7795 to the MSTP DT binding documentation.

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds

author maintainer
#12 lore raw

PATCH v4 5/5 clk: shmobile: Make MSTP clock-output-names optional

In reply to Magnus Damm

From: Magnus Damm <[email protected]>

This patch updates the MSTP driver to make the "clock-output-names" DT property optional instead of mandatory. In case the DT property is omitted then the function clk_register_clkdev() is skipped.

The default for new SoCs is to not use "clock-output-names".

---

Changes since V3:
- Added support for multiple MSTP bits. =)

 drivers/clk/shmobile/clk-mstp.c |   28 +++++++++++++++++++---------
 1 file changed, 19 insertions(+), 9 deletions(-)
drivers/clk/shmobile/clk-mstp.c +19 -9
@@ -199,26 +199,29 @@ static void __init cpg_mstp_clocks_init(
 	for (i = 0; i < MSTP_MAX_CLOCKS; ++i) {
 		const char *parent_name;
 		const char *name;
+		bool allocated_name;
 		u32 clkidx;
 		int ret;
 
-		/* Skip clocks with no name. */
-		ret = of_property_read_string_index(np, "clock-output-names",
-						    i, &name);
-		if (ret < 0 || strlen(name) == 0)
-			continue;
-
 		parent_name = of_clk_get_parent_name(np, i);
 		ret = of_property_read_u32_index(np, idxname, i, &clkidx);
 		if (parent_name == NULL || ret < 0)
 			break;
 
 		if (clkidx >= MSTP_MAX_CLOCKS) {
-			pr_err("%s: invalid clock %s %s index %u)\n",
-			       __func__, np->name, name, clkidx);
+			pr_err("%s: invalid clock %s index %u)\n",
+			       __func__, np->name, clkidx);
 			continue;
 		}
 
+		if (of_property_read_string_index(np, "clock-output-names",
+						  i, &name) == 0) {
+			allocated_name = false;
+		} else {
+			name = kasprintf(GFP_KERNEL, "%s.%u", np->name, clkidx);
+			allocated_name = true;
+		}
+
 		clks[clkidx] = cpg_mstp_clock_register(name, parent_name,
 						       clkidx, group);
 		if (!IS_ERR(clks[clkidx])) {
@@ -231,12 +234,19 @@ static void __init cpg_mstp_clocks_init(
 			 *
 			 * FIXME: Remove this when all devices that require a
 			 * clock will be instantiated from DT.
+			 *
+			 * We currently register clkdev only when the DT
+			 * property clock-output-names is present.
 			 */
-			clk_register_clkdev(clks[clkidx], name, NULL);
+			if (!allocated_name)
+				clk_register_clkdev(clks[clkidx], name, NULL);
 		} else {
 			pr_err("%s: failed to register %s %s clock (%ld)\n",
 			       __func__, np->name, name, PTR_ERR(clks[clkidx]));
 		}
+
+		if (allocated_name)
+			kfree(name);
 	}
 
 	of_clk_add_provider(np, of_clk_src_onecell_get, &group->data);
#13 lore raw

Re: [PATCH v4 05/05] clk: shmobile: Make MSTP clock-output-names optional

drivers/clk/shmobile/clk-mstp.c 2015-08-29 18:01:00.662366518 +0900 static void __init cpg_mstp_clocks_init( · line 202

Quoted 7 lines · Magnus Damm
drivers/clk/shmobile/clk-mstp.c 2015-08-29 18:01:00.662366518 +0900 renamed from drivers/clk/shmobile/clk-mstp.c +1 -0
@@ -199,26 +199,29 @@ static void __init cpg_mstp_clocks_init(
        for (i = 0; i < MSTP_MAX_CLOCKS; ++i) {
                const char *parent_name;
                const char *name;
+               bool allocated_name;

If you make it "const char *allocated_name = NULL;" ...

Quoted 15 lines
drivers/clk/shmobile/clk-mstp.c 2015-08-29 18:01:00.662366518 +0900 renamed from drivers/clk/shmobile/clk-mstp.c +0 -6
                u32 clkidx;
                int ret;
 
-               /* Skip clocks with no name. */
-               ret = of_property_read_string_index(np, "clock-output-names",
-                                                   i, &name);
-               if (ret < 0 || strlen(name) == 0)
-                       continue;
-
                parent_name = of_clk_get_parent_name(np, i);
                ret = of_property_read_u32_index(np, idxname, i, &clkidx);
                if (parent_name == NULL || ret < 0)
                        break;
 
                if (clkidx >= MSTP_MAX_CLOCKS) {

drivers/clk/shmobile/clk-mstp.c 2015-08-29 18:01:00.662366518 +0900 static void __init cpg_mstp_clocks_init( · line 219

Quoted 10 lines
drivers/clk/shmobile/clk-mstp.c 2015-08-29 18:01:00.662366518 +0900 renamed from drivers/clk/shmobile/clk-mstp.c +5 -2
-                       pr_err("%s: invalid clock %s %s index %u)\n",
-                              __func__, np->name, name, clkidx);
+                       pr_err("%s: invalid clock %s index %u)\n",
+                              __func__, np->name, clkidx);
                        continue;
                }
 
+               if (of_property_read_string_index(np, "clock-output-names",
+                                                 i, &name) == 0) {
+                       allocated_name = false;

... you can drop this branch...

drivers/clk/shmobile/clk-mstp.c 2015-08-29 18:01:00.662366518 +0900 static void __init cpg_mstp_clocks_init( · line 222

Quoted 3 lines
drivers/clk/shmobile/clk-mstp.c 2015-08-29 18:01:00.662366518 +0900 renamed from drivers/clk/shmobile/clk-mstp.c +3 -0
+               } else {
+                       name = kasprintf(GFP_KERNEL, "%s.%u", np->name, clkidx);
+                       allocated_name = true;

... and use "allocated_name = name = kasprintf(...);"

Quoted 13 lines
drivers/clk/shmobile/clk-mstp.c 2015-08-29 18:01:00.662366518 +0900 renamed from drivers/clk/shmobile/clk-mstp.c +5 -0
+               }
+
                clks[clkidx] = cpg_mstp_clock_register(name, parent_name,
                                                       clkidx, group);
                if (!IS_ERR(clks[clkidx])) {
@@ -231,12 +234,19 @@ static void __init cpg_mstp_clocks_init(
                         *
                         * FIXME: Remove this when all devices that require a
                         * clock will be instantiated from DT.
+                        *
+                        * We currently register clkdev only when the DT
+                        * property clock-output-names is present.
                         */

drivers/clk/shmobile/clk-mstp.c 2015-08-29 18:01:00.662366518 +0900 static void __init cpg_mstp_clocks_init( · line 249

Quoted 10 lines
drivers/clk/shmobile/clk-mstp.c 2015-08-29 18:01:00.662366518 +0900 renamed from drivers/clk/shmobile/clk-mstp.c +5 -1
-                       clk_register_clkdev(clks[clkidx], name, NULL);
+                       if (!allocated_name)
+                               clk_register_clkdev(clks[clkidx], name, NULL);
                } else {
                        pr_err("%s: failed to register %s %s clock (%ld)\n",
                               __func__, np->name, name, PTR_ERR(clks[clkidx]));
                }
+
+               if (allocated_name)
+                       kfree(name);

... and kfree(allocated_name) unconditionally.

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- [email protected]

In personal conversations with technical people, I call myself a hacker. But when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds