Skip to contentSkip to conversation

clk: imx: Refactor entire sccg pll clk

PATCH Applied Abel Vesa 3 participants 9 messages Feb 22, 2019 – Feb 26, 2019

Subsystems (3)
lore
Share & tools
lore

Abel Vesa

author
#0 lore raw

PATCH v2 1/1 clk: imx: Refactor entire sccg pll clk

Make the entire combination of plls to be one single clock. The parents used for bypasses are specified each as an index in the parents list. The determine_rate does a lookup throughout all the possible combinations for all the divs and returns the best possible 'setup' which in turn is used by set_rate later to set up all the divs and bypasses.

---

Changes since v1:
* fixed all the warnings and errors for arm32
* removed the explicit reparenting from clk-imx8mq leaving that work to be done by each clock driver explicitly

 drivers/clk/imx/clk-imx8mq.c   |  53 +----
 drivers/clk/imx/clk-sccg-pll.c | 514 ++++++++++++++++++++++++++++++++---------
 drivers/clk/imx/clk.h          |   9 +-
 3 files changed, 417 insertions(+), 159 deletions(-)
drivers/clk/imx/clk-imx8mq.c +8 -45
index 01c771f..081dc2d 100644
@@ -34,15 +34,10 @@ static const char * const audio_pll1_bypass_sels[] = {"audio_pll1", "audio_pll1_
 static const char * const audio_pll2_bypass_sels[] = {"audio_pll2", "audio_pll2_ref_sel", };
 static const char * const video_pll1_bypass_sels[] = {"video_pll1", "video_pll1_ref_sel", };
 
-static const char * const sys1_pll1_out_sels[] = {"sys1_pll1", "sys1_pll1_ref_sel", };
-static const char * const sys2_pll1_out_sels[] = {"sys2_pll1", "sys1_pll1_ref_sel", };
-static const char * const sys3_pll1_out_sels[] = {"sys3_pll1", "sys3_pll1_ref_sel", };
-static const char * const dram_pll1_out_sels[] = {"dram_pll1", "dram_pll1_ref_sel", };
-
-static const char * const sys1_pll2_out_sels[] = {"sys1_pll2_div", "sys1_pll1_ref_sel", };
-static const char * const sys2_pll2_out_sels[] = {"sys2_pll2_div", "sys2_pll1_ref_sel", };
-static const char * const sys3_pll2_out_sels[] = {"sys3_pll2_div", "sys2_pll1_ref_sel", };
-static const char * const dram_pll2_out_sels[] = {"dram_pll2_div", "dram_pll1_ref_sel", };
+static const char * const sys1_pll_out_sels[] = {"sys1_pll1_ref_sel", };
+static const char * const sys2_pll_out_sels[] = {"sys1_pll1_ref_sel", "sys2_pll1_ref_sel", };
+static const char * const sys3_pll_out_sels[] = {"sys3_pll1_ref_sel", "sys2_pll1_ref_sel", };
+static const char * const dram_pll_out_sels[] = {"dram_pll1_ref_sel", };
 
 /* CCM ROOT */
 static const char * const imx8mq_a53_sels[] = {"osc_25m", "arm_pll_out", "sys2_pll_500m", "sys2_pll_1000m",
@@ -313,10 +308,6 @@ static int imx8mq_clocks_probe(struct platform_device *pdev)
 	clks[IMX8MQ_AUDIO_PLL1_REF_DIV] = imx_clk_divider("audio_pll1_ref_div", "audio_pll1_ref_sel", base + 0x0, 5, 6);
 	clks[IMX8MQ_AUDIO_PLL2_REF_DIV] = imx_clk_divider("audio_pll2_ref_div", "audio_pll2_ref_sel", base + 0x8, 5, 6);
 	clks[IMX8MQ_VIDEO_PLL1_REF_DIV] = imx_clk_divider("video_pll1_ref_div", "video_pll1_ref_sel", base + 0x10, 5, 6);
-	clks[IMX8MQ_SYS1_PLL1_REF_DIV]	= imx_clk_divider("sys1_pll1_ref_div", "sys1_pll1_ref_sel", base + 0x38, 25, 3);
-	clks[IMX8MQ_SYS2_PLL1_REF_DIV]	= imx_clk_divider("sys2_pll1_ref_div", "sys2_pll1_ref_sel", base + 0x44, 25, 3);
-	clks[IMX8MQ_SYS3_PLL1_REF_DIV]	= imx_clk_divider("sys3_pll1_ref_div", "sys3_pll1_ref_sel", base + 0x50, 25, 3);
-	clks[IMX8MQ_DRAM_PLL1_REF_DIV]	= imx_clk_divider("dram_pll1_ref_div", "dram_pll1_ref_sel", base + 0x68, 25, 3);
 
 	clks[IMX8MQ_ARM_PLL] = imx_clk_frac_pll("arm_pll", "arm_pll_ref_div", base + 0x28);
 	clks[IMX8MQ_GPU_PLL] = imx_clk_frac_pll("gpu_pll", "gpu_pll_ref_div", base + 0x18);
@@ -324,25 +315,6 @@ static int imx8mq_clocks_probe(struct platform_device *pdev)
 	clks[IMX8MQ_AUDIO_PLL1] = imx_clk_frac_pll("audio_pll1", "audio_pll1_ref_div", base + 0x0);
 	clks[IMX8MQ_AUDIO_PLL2] = imx_clk_frac_pll("audio_pll2", "audio_pll2_ref_div", base + 0x8);
 	clks[IMX8MQ_VIDEO_PLL1] = imx_clk_frac_pll("video_pll1", "video_pll1_ref_div", base + 0x10);
-	clks[IMX8MQ_SYS1_PLL1] = imx_clk_sccg_pll("sys1_pll1", "sys1_pll1_ref_div", base + 0x30, SCCG_PLL1);
-	clks[IMX8MQ_SYS2_PLL1] = imx_clk_sccg_pll("sys2_pll1", "sys2_pll1_ref_div", base + 0x3c, SCCG_PLL1);
-	clks[IMX8MQ_SYS3_PLL1] = imx_clk_sccg_pll("sys3_pll1", "sys3_pll1_ref_div", base + 0x48, SCCG_PLL1);
-	clks[IMX8MQ_DRAM_PLL1] = imx_clk_sccg_pll("dram_pll1", "dram_pll1_ref_div", base + 0x60, SCCG_PLL1);
-
-	clks[IMX8MQ_SYS1_PLL2] = imx_clk_sccg_pll("sys1_pll2", "sys1_pll1_out_div", base + 0x30, SCCG_PLL2);
-	clks[IMX8MQ_SYS2_PLL2] = imx_clk_sccg_pll("sys2_pll2", "sys2_pll1_out_div", base + 0x3c, SCCG_PLL2);
-	clks[IMX8MQ_SYS3_PLL2] = imx_clk_sccg_pll("sys3_pll2", "sys3_pll1_out_div", base + 0x48, SCCG_PLL2);
-	clks[IMX8MQ_DRAM_PLL2] = imx_clk_sccg_pll("dram_pll2", "dram_pll1_out_div", base + 0x60, SCCG_PLL2);
-
-	/* PLL divs */
-	clks[IMX8MQ_SYS1_PLL1_OUT_DIV] = imx_clk_divider("sys1_pll1_out_div", "sys1_pll1_out", base + 0x38, 19, 6);
-	clks[IMX8MQ_SYS2_PLL1_OUT_DIV] = imx_clk_divider("sys2_pll1_out_div", "sys2_pll1_out", base + 0x44, 19, 6);
-	clks[IMX8MQ_SYS3_PLL1_OUT_DIV] = imx_clk_divider("sys3_pll1_out_div", "sys3_pll1_out", base + 0x50, 19, 6);
-	clks[IMX8MQ_DRAM_PLL1_OUT_DIV] = imx_clk_divider("dram_pll1_out_div", "dram_pll1_out", base + 0x68, 19, 6);
-	clks[IMX8MQ_SYS1_PLL2_DIV] = imx_clk_divider("sys1_pll2_div", "sys1_pll2", base + 0x38, 1, 6);
-	clks[IMX8MQ_SYS2_PLL2_DIV] = imx_clk_divider("sys2_pll2_div", "sys2_pll2", base + 0x44, 1, 6);
-	clks[IMX8MQ_SYS3_PLL2_DIV] = imx_clk_divider("sys3_pll2_div", "sys3_pll2", base + 0x50, 1, 6);
-	clks[IMX8MQ_DRAM_PLL2_DIV] = imx_clk_divider("dram_pll2_div", "dram_pll2", base + 0x68, 1, 6);
 
 	/* PLL bypass out */
 	clks[IMX8MQ_ARM_PLL_BYPASS] = imx_clk_mux_flags("arm_pll_bypass", base + 0x28, 14, 1, arm_pll_bypass_sels, ARRAY_SIZE(arm_pll_bypass_sels), CLK_SET_RATE_PARENT);
@@ -352,15 +324,6 @@ static int imx8mq_clocks_probe(struct platform_device *pdev)
 	clks[IMX8MQ_AUDIO_PLL2_BYPASS] = imx_clk_mux("audio_pll2_bypass", base + 0x8, 14, 1, audio_pll2_bypass_sels, ARRAY_SIZE(audio_pll2_bypass_sels));
 	clks[IMX8MQ_VIDEO_PLL1_BYPASS] = imx_clk_mux("video_pll1_bypass", base + 0x10, 14, 1, video_pll1_bypass_sels, ARRAY_SIZE(video_pll1_bypass_sels));
 
-	clks[IMX8MQ_SYS1_PLL1_OUT] = imx_clk_mux("sys1_pll1_out", base + 0x30, 5, 1, sys1_pll1_out_sels, ARRAY_SIZE(sys1_pll1_out_sels));
-	clks[IMX8MQ_SYS2_PLL1_OUT] = imx_clk_mux("sys2_pll1_out", base + 0x3c, 5, 1, sys2_pll1_out_sels, ARRAY_SIZE(sys2_pll1_out_sels));
-	clks[IMX8MQ_SYS3_PLL1_OUT] = imx_clk_mux("sys3_pll1_out", base + 0x48, 5, 1, sys3_pll1_out_sels, ARRAY_SIZE(sys3_pll1_out_sels));
-	clks[IMX8MQ_DRAM_PLL1_OUT] = imx_clk_mux("dram_pll1_out", base + 0x60, 5, 1, dram_pll1_out_sels, ARRAY_SIZE(dram_pll1_out_sels));
-	clks[IMX8MQ_SYS1_PLL2_OUT] = imx_clk_mux("sys1_pll2_out", base + 0x30, 4, 1, sys1_pll2_out_sels, ARRAY_SIZE(sys1_pll2_out_sels));
-	clks[IMX8MQ_SYS2_PLL2_OUT] = imx_clk_mux("sys2_pll2_out", base + 0x3c, 4, 1, sys2_pll2_out_sels, ARRAY_SIZE(sys2_pll2_out_sels));
-	clks[IMX8MQ_SYS3_PLL2_OUT] = imx_clk_mux("sys3_pll2_out", base + 0x48, 4, 1, sys3_pll2_out_sels, ARRAY_SIZE(sys3_pll2_out_sels));
-	clks[IMX8MQ_DRAM_PLL2_OUT] = imx_clk_mux("dram_pll2_out", base + 0x60, 4, 1, dram_pll2_out_sels, ARRAY_SIZE(dram_pll2_out_sels));
-
 	/* PLL OUT GATE */
 	clks[IMX8MQ_ARM_PLL_OUT] = imx_clk_gate("arm_pll_out", "arm_pll_bypass", base + 0x28, 21);
 	clks[IMX8MQ_GPU_PLL_OUT] = imx_clk_gate("gpu_pll_out", "gpu_pll_bypass", base + 0x18, 21);
@@ -368,11 +331,11 @@ static int imx8mq_clocks_probe(struct platform_device *pdev)
 	clks[IMX8MQ_AUDIO_PLL1_OUT] = imx_clk_gate("audio_pll1_out", "audio_pll1_bypass", base + 0x0, 21);
 	clks[IMX8MQ_AUDIO_PLL2_OUT] = imx_clk_gate("audio_pll2_out", "audio_pll2_bypass", base + 0x8, 21);
 	clks[IMX8MQ_VIDEO_PLL1_OUT] = imx_clk_gate("video_pll1_out", "video_pll1_bypass", base + 0x10, 21);
-	clks[IMX8MQ_SYS1_PLL_OUT] = imx_clk_gate("sys1_pll_out", "sys1_pll2_out", base + 0x30, 9);
-	clks[IMX8MQ_SYS2_PLL_OUT] = imx_clk_gate("sys2_pll_out", "sys2_pll2_out", base + 0x3c, 9);
-	clks[IMX8MQ_SYS3_PLL_OUT] = imx_clk_gate("sys3_pll_out", "sys3_pll2_out", base + 0x48, 9);
-	clks[IMX8MQ_DRAM_PLL_OUT] = imx_clk_gate("dram_pll_out", "dram_pll2_out", base + 0x60, 9);
 
+	clks[IMX8MQ_SYS1_PLL_OUT] = imx_clk_sccg_pll("sys1_pll_out", sys1_pll_out_sels, ARRAY_SIZE(sys1_pll_out_sels), 0, 0, 0, base + 0x30, CLK_IS_CRITICAL);
+	clks[IMX8MQ_SYS2_PLL_OUT] = imx_clk_sccg_pll("sys2_pll_out", sys2_pll_out_sels, ARRAY_SIZE(sys2_pll_out_sels), 0, 0, 1, base + 0x3c, CLK_IS_CRITICAL);
+	clks[IMX8MQ_SYS3_PLL_OUT] = imx_clk_sccg_pll("sys3_pll_out", sys3_pll_out_sels, ARRAY_SIZE(sys3_pll_out_sels), 0, 0, 1, base + 0x48, CLK_IS_CRITICAL);
+	clks[IMX8MQ_DRAM_PLL_OUT] = imx_clk_sccg_pll("dram_pll_out", dram_pll_out_sels, ARRAY_SIZE(dram_pll_out_sels), 0, 0, 0, base + 0x60, CLK_IS_CRITICAL);
 	/* SYS PLL fixed output */
 	clks[IMX8MQ_SYS1_PLL_40M] = imx_clk_fixed_factor("sys1_pll_40m", "sys1_pll_out", 1, 20);
 	clks[IMX8MQ_SYS1_PLL_80M] = imx_clk_fixed_factor("sys1_pll_80m", "sys1_pll_out", 1, 10);
drivers/clk/imx/clk-sccg-pll.c +403 -111

621 lines not on this page · raw message

drivers/clk/imx/clk.h +6 -3
index 0aa48d3..5748ec8 100644
@@ -60,9 +60,12 @@ struct clk *imx_clk_pllv2(const char *name, const char *parent,
 struct clk *imx_clk_frac_pll(const char *name, const char *parent_name,
 			     void __iomem *base);
 
-struct clk *imx_clk_sccg_pll(const char *name, const char *parent_name,
-			     void __iomem *base,
-			     enum imx_sccg_pll_type pll_type);
+struct clk *imx_clk_sccg_pll(const char *name,
+				const char * const *parent_names,
+				u8 num_parents,
+				u8 parent, u8 bypass1, u8 bypass2,
+				void __iomem *base,
+				unsigned long flags);
 
 enum imx_pllv3_type {
 	IMX_PLLV3_GENERIC,
Signature · 1 line
2.7.4
maintainer
#1 lore raw

Quoting Abel Vesa (2019-02-22 09:07:32)

Quoted 9 lines

Make the entire combination of plls to be one single clock. The parents used for bypasses are specified each as an index in the parents list. The determine_rate does a lookup throughout all the possible combinations for all the divs and returns the best possible 'setup' which in turn is used by set_rate later to set up all the divs and bypasses.

I suspect these tested by and acked tags should have been dropped, unless you discussed and tested off-list?

author
#2 lore raw
Quoted 7 lines · Stephen Boyd

Quoting Abel Vesa (2019-02-22 09:07:32)

Quoted 6 lines

Make the entire combination of plls to be one single clock. The parents used for bypasses are specified each as an index in the parents list. The determine_rate does a lookup throughout all the possible combinations for all the divs and returns the best possible 'setup' which in turn is used by set_rate later to set up all the divs and bypasses.

Quoted 7 lines
Quoted 3 lines

I suspect these tested by and acked tags should have been dropped, unless you discussed and tested off-list?

Oups, I forgot to drop them.

maintainer
#3 lore raw

Quoting Abel Vesa (2019-02-23 02:58:14)

Quoted 12 lines · Stephen Boyd
Quoted 11 lines

Quoting Abel Vesa (2019-02-22 09:07:32)

Quoted 9 lines

Make the entire combination of plls to be one single clock. The parents used for bypasses are specified each as an index in the parents list. The determine_rate does a lookup throughout all the possible combinations for all the divs and returns the best possible 'setup' which in turn is used by set_rate later to set up all the divs and bypasses.

Quoted 5 lines
Quoted 3 lines

I suspect these tested by and acked tags should have been dropped, unless you discussed and tested off-list?

Oups, I forgot to drop them.

Ok. Can Lucas re-test?

Tested-by
#4 lore raw
Quoted 4 lines

From: Stephen Boyd [mailto:[email protected]]
Sent: Tuesday, February 26, 2019 1:23 AM>
Quoting Abel Vesa (2019-02-23 02:58:14)

Quoted 1 line · Stephen Boyd
Quoted 10 lines
Quoted 3 lines
Quoted 3 lines

Quoting Abel Vesa (2019-02-22 09:07:32)

Quoted 2 lines

Make the entire combination of plls to be one single clock. The parents used for bypasses are specified each as an index in the parents

list.

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

The determine_rate does a lookup throughout all the possible combinations for all the divs and returns the best possible 'setup' which in turn is used by set_rate later to set up all the divs and

bypasses.

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

This significantly simply the clock tree output.
Thanks for the work.

Quoted 10 lines
Quoted 8 lines
Quoted 6 lines
Quoted 2 lines

I suspect these tested by and acked tags should have been dropped, unless you discussed and tested off-list?

Oups, I forgot to drop them.

Ok. Can Lucas re-test?

Tested on clk-next by cherry-pick a few defconfig patches from imx tree. So:

Regards
Dong Aisheng
_______________________________________________
linux-arm-kernel mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

author
#5 lore raw
Quoted 25 lines · Aisheng Dong
Quoted 4 lines

From: Stephen Boyd [mailto:[email protected]]
Sent: Tuesday, February 26, 2019 1:23 AM>
Quoting Abel Vesa (2019-02-23 02:58:14)

Quoted 1 line · Stephen Boyd
Quoted 10 lines
Quoted 3 lines
Quoted 3 lines

Quoting Abel Vesa (2019-02-22 09:07:32)

Quoted 2 lines

Make the entire combination of plls to be one single clock. The parents used for bypasses are specified each as an index in the parents

list.

Quoted 3 lines
Quoted 3 lines
Quoted 3 lines

The determine_rate does a lookup throughout all the possible combinations for all the divs and returns the best possible 'setup' which in turn is used by set_rate later to set up all the divs and

bypasses.

Quoted 2 lines
Quoted 2 lines
Quoted 2 lines

This significantly simply the clock tree output.
Thanks for the work.

Quoted 7 lines
Quoted 7 lines
Quoted 6 lines
Quoted 2 lines

I suspect these tested by and acked tags should have been dropped, unless you discussed and tested off-list?

Quoted 8 lines
Quoted 3 lines
Quoted 1 line

Oups, I forgot to drop them.

Ok. Can Lucas re-test?

Tested on clk-next by cherry-pick a few defconfig patches from imx tree. So:

Thanks for helping with the testing.

I don't think there should be any defconfig changes required by this patch.

Can you please list here the CONFIGs added?

Just wanna make sure I'm not missing anything.

Quoted 2 lines

Regards
Dong Aisheng

#6 lore raw

[...]

Quoted 16 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 6 lines
Quoted 2 lines

I suspect these tested by and acked tags should have been dropped, unless you discussed and tested off-list?

Quoted 8 lines
Quoted 3 lines
Quoted 1 line

Oups, I forgot to drop them.

Ok. Can Lucas re-test?

Tested on clk-next by cherry-pick a few defconfig patches from imx tree. So:

Quoted 8 lines

Thanks for helping with the testing.

I don't think there should be any defconfig changes required by this patch.

Can you please list here the CONFIGs added?

Just wanna make sure I'm not missing anything.

Clk-next does not have MX8 defconfig patches.
So I manually cherry-picked them from imx-next. Below two needed:
dee305e arm64: defconfig: Add i.MX8MQ boot necessary configs a1624ef arm64: defconfig: add imx8qxp support

Regards
Dong Aisheng

Quoted 2 lines
Quoted 2 lines

Regards
Dong Aisheng

author
#7 lore raw
Quoted 24 lines · Aisheng Dong

[...]

Quoted 22 lines
Quoted 7 lines
Quoted 7 lines
Quoted 7 lines
Quoted 6 lines
Quoted 2 lines

I suspect these tested by and acked tags should have been dropped, unless you discussed and tested off-list?

Quoted 8 lines
Quoted 3 lines
Quoted 1 line

Oups, I forgot to drop them.

Ok. Can Lucas re-test?

Tested on clk-next by cherry-pick a few defconfig patches from imx tree. So:

Thanks for helping with the testing.

I don't think there should be any defconfig changes required by this patch.

Can you please list here the CONFIGs added?

Quoted 9 lines
Quoted 2 lines

Just wanna make sure I'm not missing anything.

Clk-next does not have MX8 defconfig patches.
So I manually cherry-picked them from imx-next. Below two needed:
dee305e arm64: defconfig: Add i.MX8MQ boot necessary configs a1624ef arm64: defconfig: add imx8qxp support

Ah, makes sense.

Thanks.

Quoted 5 lines

Regards
Dong Aisheng

Quoted 2 lines
Quoted 2 lines

Regards
Dong Aisheng

maintainer Applied
#8 lore raw
In reply to Abel Vesa

Quoting Abel Vesa (2019-02-22 09:07:32)

Quoted 10 lines

Make the entire combination of plls to be one single clock. The parents used for bypasses are specified each as an index in the parents list. The determine_rate does a lookup throughout all the possible combinations for all the divs and returns the best possible 'setup' which in turn is used by set_rate later to set up all the divs and bypasses.

---

Applied to clk-next