davinci: da850: move input frequency to board specific files.

From: Bob Dunlop <[email protected]>

Currently the input frequency of the SoC is hardcoded in the SoC specific da850.c file to 24 MHz. Since the SoC accepts input frequencies in a wide range from 12 to 50 MHz, boards with different oscillator/crystal frequencies may be built.

This patch allows setting a different input frequency in the board specific files to support boards with oscillator/crystal frequencies other than 24 MHz.

Signed-off-by Bob Dunlop <[email protected]>

Signed-off-by Christian Riesch <[email protected]>

---

Hi,
in private email Bob Dunlop suggested to pass a pointer to a small structure instead of the frequency value to allow future expansions, e.g., for the OPP list. Therefore I submit his patch as V2. Christian

?arch/arm/mach-davinci/board-da850-evm.c ? ? | ? ?2 +-
?arch/arm/mach-davinci/board-mityomapl138.c ?| ? ?2 +-
?arch/arm/mach-davinci/board-omapl138-hawk.c | ? ?2 +-
?arch/arm/mach-davinci/da850.c ? ? ? ? ? ? ? | ? ?5 ++++-
?arch/arm/mach-davinci/include/mach/da8xx.h ?| ? ?6 +++++-

?5 files changed, 12 insertions(+), 5 deletions(-)

arch/arm/mach-davinci/board-da850-evm.c +1 -1

index a7b41bf..231ff87 100644
@@ -1252,7 +1252,7 @@ console_initcall(da850_evm_console_init);
 
 ?static void __init da850_evm_map_io(void)
 ?{
- ? ? ? da850_init();
+ ? ? ? da850_init(NULL);
 ?}
 
 ?MACHINE_START(DAVINCI_DA850_EVM, "DaVinci DA850/OMAP-L138/AM18x EVM")

arch/arm/mach-davinci/board-mityomapl138.c +1 -1

index 606a6f2..362770c 100644
@@ -561,7 +561,7 @@ console_initcall(mityomapl138_console_init);
 
 ?static void __init mityomapl138_map_io(void)
 ?{
- ? ? ? da850_init();
+ ? ? ? da850_init(NULL);
 ?}
 
 ?MACHINE_START(MITYOMAPL138, "MityDSP-L138/MityARM-1808")

arch/arm/mach-davinci/board-omapl138-hawk.c +1 -1

index 67c38d0..c43a6c3 100644
@@ -334,7 +334,7 @@ console_initcall(omapl138_hawk_console_init);
 
 ?static void __init omapl138_hawk_map_io(void)
 ?{
- ? ? ? da850_init();
+ ? ? ? da850_init(NULL);
 ?}
 
 ?MACHINE_START(OMAPL138_HAWKBOARD, "AM18x/OMAP-L138 Hawkboard")

arch/arm/mach-davinci/da850.c +4 -1

index 133aac4..ebd0603 100644
@@ -1104,10 +1104,13 @@ static struct davinci_soc_info davinci_soc_info_da850 = {
 ? ? ? ?.reset_device ? ? ? ? ? = &da8xx_wdt_device,
 ?};
 
-void __init da850_init(void)
+void __init da850_init(struct da850_init_board_info *board)
 ?{
 ? ? ? ?unsigned int v;
 
+ ? ? ? if (board && board->ref_clk_rate)
+ ? ? ? ? ? ? ? ref_clk.rate = board->ref_clk_rate;
+
 ? ? ? ?davinci_common_init(&davinci_soc_info_da850);
 
 ? ? ? ?da8xx_syscfg0_base = ioremap(DA8XX_SYSCFG0_BASE, SZ_4K);

arch/arm/mach-davinci/include/mach/da8xx.h +0 -0

index ad64da7..66efc5d 100644
@@ -69,8 +69,12 @@ extern unsigned int da850_max_speed;

?#define DA8XX_AEMIF_CTL_BASE ? 0x68000000
?#define DA8XX_ARM_RAM_BASE ? ? 0xffff0000

arch/arm/mach-davinci/include/mach/da8xx.h +4 -0

+struct da850_init_board_info {
+ ? ? ? unsigned long ref_clk_rate;
+};
+

?void __init da830_init(void);

arch/arm/mach-davinci/include/mach/da8xx.h +1 -1

-void __init da850_init(void);
+void __init da850_init(struct da850_init_board_info *board);

?int da830_register_edma(struct edma_rsv_info *rsv); ?int da850_register_edma(struct edma_rsv_info *rsv[2]); --
1.7.0.4

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel at lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

From Menon, Nishanth [mailto:nm at ti.com]

Quoted 7 lines · Christian Riesch

From: Bob Dunlop <[email protected]>

Currently the input frequency of the SoC is hardcoded in the SoC specific da850.c file to 24 MHz. Since the SoC accepts input frequencies in a wide range from 12 to 50 MHz, boards with different oscillator/crystal frequencies may be built.

Quoted 3 lines

This patch allows setting a different input frequency in the board specific files to support boards with oscillator/crystal frequencies other than 24 MHz.

Curious question: have you considered Documentation/power/opp.txt? Not entirely sure if that will help, but I am curious to know if there are reasons why it did not scale for you

From: Bob Dunlop <[email protected]>

Currently the input frequency of the SoC is hardcoded in the SoC specific da850.c file to 24 MHz. Since the SoC accepts input frequencies in a wide range from 12 to 50 MHz, boards with different oscillator/crystal frequencies may be built.

From: Bob Dunlop <[email protected]>

Currently the input frequency of the SoC is hardcoded in the SoC specific da850.c file to 24 MHz. Since the SoC accepts input frequencies in a wide range from 12 to 50 MHz, boards with different oscillator/crystal frequencies may be built.

This patch allows setting a different input frequency in the board specific files to support boards with oscillator/crystal frequencies other than 24 MHz.

Signed-off-by Bob Dunlop <[email protected]>

Signed-off-by Christian Riesch <[email protected]>

---

Hi,
in private email Bob Dunlop suggested to pass a pointer to a small structure instead of the frequency value to allow future expansions, e.g., for the OPP list. Therefore I submit his patch as V2. Christian

 arch/arm/mach-davinci/board-da850-evm.c     |    2 +-
 arch/arm/mach-davinci/board-mityomapl138.c  |    2 +-
 arch/arm/mach-davinci/board-omapl138-hawk.c |    2 +-
 arch/arm/mach-davinci/da850.c               |    5 ++++-
 arch/arm/mach-davinci/include/mach/da8xx.h  |    6 +++++-
 5 files changed, 12 insertions(+), 5 deletions(-)

arch/arm/mach-davinci/board-da850-evm.c +1 -1

index a7b41bf..231ff87 100644
@@ -1252,7 +1252,7 @@ console_initcall(da850_evm_console_init);
 
 static void __init da850_evm_map_io(void)
 {
-	da850_init();
+	da850_init(NULL);
 }
 
 MACHINE_START(DAVINCI_DA850_EVM, "DaVinci DA850/OMAP-L138/AM18x EVM")

arch/arm/mach-davinci/board-mityomapl138.c +1 -1

index 606a6f2..362770c 100644
@@ -561,7 +561,7 @@ console_initcall(mityomapl138_console_init);
 
 static void __init mityomapl138_map_io(void)
 {
-	da850_init();
+	da850_init(NULL);
 }
 
 MACHINE_START(MITYOMAPL138, "MityDSP-L138/MityARM-1808")

arch/arm/mach-davinci/board-omapl138-hawk.c +1 -1

index 67c38d0..c43a6c3 100644
@@ -334,7 +334,7 @@ console_initcall(omapl138_hawk_console_init);
 
 static void __init omapl138_hawk_map_io(void)
 {
-	da850_init();
+	da850_init(NULL);
 }
 
 MACHINE_START(OMAPL138_HAWKBOARD, "AM18x/OMAP-L138 Hawkboard")

arch/arm/mach-davinci/da850.c +4 -1

index 133aac4..ebd0603 100644
@@ -1104,10 +1104,13 @@ static struct davinci_soc_info davinci_soc_info_da850 = {
 	.reset_device		= &da8xx_wdt_device,
 };
 
-void __init da850_init(void)
+void __init da850_init(struct da850_init_board_info *board)
 {
 	unsigned int v;
 
+	if (board && board->ref_clk_rate)
+		ref_clk.rate = board->ref_clk_rate;
+
 	davinci_common_init(&davinci_soc_info_da850);
 
 	da8xx_syscfg0_base = ioremap(DA8XX_SYSCFG0_BASE, SZ_4K);

arch/arm/mach-davinci/include/mach/da8xx.h +5 -1

index ad64da7..66efc5d 100644
@@ -69,8 +69,12 @@ extern unsigned int da850_max_speed;
 #define DA8XX_AEMIF_CTL_BASE	0x68000000
 #define DA8XX_ARM_RAM_BASE	0xffff0000
 
+struct da850_init_board_info {
+	unsigned long ref_clk_rate;
+};
+
 void __init da830_init(void);
-void __init da850_init(void);
+void __init da850_init(struct da850_init_board_info *board);
 
 int da830_register_edma(struct edma_rsv_info *rsv);
 int da850_register_edma(struct edma_rsv_info *rsv[2]);

Quoted 7 lines · Christian Riesch

From: Bob Dunlop <[email protected]>

Currently the input frequency of the SoC is hardcoded in the SoC specific da850.c file to 24 MHz. Since the SoC accepts input frequencies in a wide range from 12 to 50 MHz, boards with different oscillator/crystal frequencies may be built.

Quoted 6 lines

This patch allows setting a different input frequency in the board specific files to support boards with oscillator/crystal frequencies other than 24 MHz.

Signed-off-by Bob Dunlop <[email protected]>

Signed-off-by Christian Riesch <[email protected]>

Why not allow board code to just do a clk_set_rate()?

Currently the ref_clk struct clk does not have a .set_rate method implemented, but that should be easy enough to add.

Then the default ref_clk.rate would stay the 24MHz, but any boards that want to override that simply use clk_get(), clk_set_rate(), clk_put()

Hi Kevin,

Quoted 21 lines · Kevin

Quoted 7 lines · Christian Riesch

From: Bob Dunlop <[email protected]>

Currently the input frequency of the SoC is hardcoded in the SoC specific da850.c file to 24 MHz. Since the SoC accepts input frequencies in a wide range from 12 to 50 MHz, boards with different oscillator/crystal frequencies may be built.

Quoted 6 lines

This patch allows setting a different input frequency in the board specific files to support boards with oscillator/crystal frequencies other than 24 MHz.

Signed-off-by Bob Dunlop <[email protected]>

Signed-off-by Christian Riesch <[email protected]>

Why not allow board code to just do a clk_set_rate()?

Currently the ref_clk struct clk does not have a .set_rate method implemented, but that should be easy enough to add.

Quoted 2 lines

Then the default ref_clk.rate would stay the 24MHz, but any boards that want to override that simply use clk_get(), clk_set_rate(), clk_put()

That's certainly much more elegant, but this would mean the whole clock tree is traversed again on each boot.

I am doing some measurements to see if there is any big difference in boot-time if this is done. Will get back with results.

I don't expect this to be a big boot-time impact.

However, some of the clock.c assumptions might need to be updated as it currently is written from the perspective that the PLL clocks are the "root" clocks.

Setting (and propagating) clock rates is what the clock framework is for, so adding a new interface to set a custom clock rate just doesn't seem right. I understand that the reference oscillator might be considered a special case, but if this can be done with the clock framework, it is much preferred.

Quoted 1 line · Kevin

I don't expect this to be a big boot-time impact.

You are right. Using PRINTK_TIME on DM365 I saw no noticeable boot-time change. When I profiled the code using do_gettimeofday(), I saw it was taking ~95 usecs to complete the propagation. I was mainly worried about all the recursion and reading of PLL and sysclk registers. Seems its not so bad.

Quoted 3 lines

However, some of the clock.c assumptions might need to be updated as it currently is written from the perspective that the PLL clocks are the "root" clocks.

Hmm, just calling clk_set_rate() on refclk propagated the rate nicely across the tree. It seems DaVinci clock code is not in such a bad shape :) Or did I miss the concern?

Quoted 5 lines

Setting (and propagating) clock rates is what the clock framework is for, so adding a new interface to set a custom clock rate just doesn't seem right. I understand that the reference oscillator might be considered a special case, but if this can be done with the clock framework, it is much preferred.

Okay. Will modify the DM6467/T EVM code to use this method instead.

This patch allows setting the input clock frequency of the SoC from the board specific using the davinci_set_refclk_rate function.

Suggested-by Kevin Hilman <[email protected]>

Cc Sekhar Nori <[email protected]>

Signed-off-by Christian Riesch <[email protected]>

---

This patch applies on top of Sekhar's patch.
Best regards, Christian

 arch/arm/mach-davinci/da850.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

arch/arm/mach-davinci/da850.c +0 -0

index 133aac4..27fe3ac 100644

This patch allows setting the input clock frequency of the SoC from the board specific code using the davinci_set_refclk_rate function.

Suggested-by Kevin Hilman <[email protected]>

Signed-off-by Christian Riesch <[email protected]>

---

v2: fixed the commit message and indentation.
Regards, Christian

 arch/arm/mach-davinci/da850.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

arch/arm/mach-davinci/da850.c +0 -0

index 133aac4..27fe3ac 100644

Quoted 18 lines · Sekhar

Quoted 1 line · Kevin

I don't expect this to be a big boot-time impact.

You are right. Using PRINTK_TIME on DM365 I saw no noticeable boot-time change. When I profiled the code using do_gettimeofday(), I saw it was taking ~95 usecs to complete the propagation. I was mainly worried about all the recursion and reading of PLL and sysclk registers. Seems its not so bad.

Quoted 3 lines

However, some of the clock.c assumptions might need to be updated as it currently is written from the perspective that the PLL clocks are the "root" clocks.

Hmm, just calling clk_set_rate() on refclk propagated the rate nicely across the tree. It seems DaVinci clock code is not in such a bad shape :) Or did I miss the concern?

Quoted 8 lines

Quoted 5 lines

Setting (and propagating) clock rates is what the clock framework is for, so adding a new interface to set a custom clock rate just doesn't seem right. I understand that the reference oscillator might be considered a special case, but if this can be done with the clock framework, it is much preferred.

Okay. Will modify the DM6467/T EVM code to use this method instead.

So, here is the patch. I suspect reference clock information should come from devicetree data when available. I hope it is OK to take this approach till that time?

Thanks,
Sekhar

-----8<------------
From: Sekhar Nori <[email protected]>
Date: Thu, 2 Jun 2011 14:10:50 +0530
Subject: [PATCH 1/1] davinci: dm6467/T EVM: fix setting up of reference clock rate

The DM6467 and DM6467T EVMs use different reference clock frequencies. This difference is currently supported by having the SoC code call a public board routine which sets up the reference clock frequency. This does not scale as more boards are added.

Instead, use the clk_set_rate() API to setup the reference clock frequency to a different value from the board file.

Suggested-by Kevin Hilman <[email protected]>

Signed-off-by Sekhar Nori <[email protected]>

Quoted 7 lines · Christian Riesch

From: Bob Dunlop <[email protected]>

Currently the input frequency of the SoC is hardcoded in the SoC specific da850.c file to 24 MHz. Since the SoC accepts input frequencies in a wide range from 12 to 50 MHz, boards with different oscillator/crystal frequencies may be built.

Quoted 6 lines

This patch allows setting a different input frequency in the board specific files to support boards with oscillator/crystal frequencies other than 24 MHz.

Signed-off-by Bob Dunlop <[email protected]>

Signed-off-by Christian Riesch <[email protected]>

Why not allow board code to just do a clk_set_rate()?

Hi Kevin,

Quoted 16 lines · Kevin Hilman

Quoted 13 lines · Christian Riesch

From: Bob Dunlop <[email protected]>

Currently the input frequency of the SoC is hardcoded in the SoC specific da850.c file to 24 MHz. Since the SoC accepts input frequencies in a wide range from 12 to 50 MHz, boards with different oscillator/crystal frequencies may be built.

This patch allows setting a different input frequency in the board specific files to support boards with oscillator/crystal frequencies other than 24 MHz.

Signed-off-by Bob Dunlop <[email protected]>

Signed-off-by Christian Riesch <[email protected]>

Quoted 1 line

Why not allow board code to just do a clk_set_rate()?

I'm fine with this method (In fact it was the first thing that I tried, I added a .set_rate method to ref_clk, it worked well for me). However I wonder whether first initializing the clock with the wrong value (24 MHz) and later correcting it via clk_set_rate() would break something. In the meantime, the data in the clock tree do not reflect the actual frequencies that are present on the SoC.