Skip to contentSkip to conversation

ARM: S5P: Fix USB and 48M clock enable procedure

PATCH New Paulius Zaleckas 2 participants 5 messages Oct 12, 2010 – Oct 15, 2010

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 ARM: S5P: Fix USB and 48M clock enable procedure

---

 arch/arm/mach-s5p6440/clock.c              |   23 -------
 arch/arm/plat-s5p/clock.c                  |   86 ++++++++++++++++++++++++++++
 arch/arm/plat-s5p/include/plat/s5p-clock.h |    1 
 3 files changed, 86 insertions(+), 24 deletions(-)
arch/arm/mach-s5p6440/clock.c +0 -23
index ca6e48d..5a5b245 100644
@@ -295,27 +295,6 @@ static struct clksrc_clk clk_pclk_low = {
 	.reg_div = { .reg = S5P_CLK_DIV3, .shift = 12, .size = 4 },
 };
 
-int s5p6440_clk48m_ctrl(struct clk *clk, int enable)
-{
-	unsigned long flags;
-	u32 val;
-
-	/* can't rely on clock lock, this register has other usages */
-	local_irq_save(flags);
-
-	val = __raw_readl(S5P_OTHERS);
-	if (enable)
-		val |= S5P_OTHERS_USB_SIG_MASK;
-	else
-		val &= ~S5P_OTHERS_USB_SIG_MASK;
-
-	__raw_writel(val, S5P_OTHERS);
-
-	local_irq_restore(flags);
-
-	return 0;
-}
-
 static int s5p6440_pclk_ctrl(struct clk *clk, int enable)
 {
 	return s5p_gatectrl(S5P_CLK_GATE_PCLK, clk, enable);
@@ -769,8 +748,6 @@ void __init_or_cpufreq s5p6440_setup_clocks(void)
 	clk_fout_epll.enable = s5p6440_epll_enable;
 	clk_fout_epll.ops = &s5p6440_epll_ops;
 
-	clk_48m.enable = s5p6440_clk48m_ctrl;
-
 	xtal_clk = clk_get(NULL, "ext_xtal");
 	BUG_ON(IS_ERR(xtal_clk));
 
arch/arm/plat-s5p/clock.c +86 -0

112 lines not on this page · raw message

arch/arm/plat-s5p/include/plat/s5p-clock.h +0 -1
index 09418b1..1c9266d 100644
@@ -38,7 +38,6 @@ extern struct clksrc_sources clk_src_apll;
 extern struct clksrc_sources clk_src_mpll;
 extern struct clksrc_sources clk_src_epll;
 
-extern int s5p6440_clk48m_ctrl(struct clk *clk, int enable);
 extern int s5p_gatectrl(void __iomem *reg, struct clk *clk, int enable);
 
 #endif /* __ASM_PLAT_S5P_CLOCK_H */
#1 lore raw

PATCH 1/1 ARM: S5P: Fix USB and 48M clock enable procedure

Quoted 10 lines · Paulius Zaleckas

---

 arch/arm/mach-s5p6440/clock.c              |   23 -------
 arch/arm/plat-s5p/clock.c                  |   86
++++++++++++++++++++++++++++
 arch/arm/plat-s5p/include/plat/s5p-clock.h |    1
 3 files changed, 86 insertions(+), 24 deletions(-)

Hi Paulius,

There are some comments about your patches which includes previous S3C64XX patches.

Basically your approach looks good trial and structure...but I'm not sure whether your approach can be used commonly on Samsung's all SoCs or not. Need to do more test on boards and I already informed your patches to USB engineers in my team, actually need to discuss about this.

As a note, I know, 'xusbxti' clock is structure for external xtal which is used for generating USB clock on board... it depends on board condition, because can be used 12/24/48Mhz on board. The clk_48m means generated actual USB clock, 48Mhz. So should be implemented enable function by using clk_48m...

Anyway let you know about the result of internal discussion soon, then let's talk.

Thanks.

Best regards,
Kgene.
--
Kukjin Kim <[email protected]>, Senior Engineer, SW Solution Development Team, Samsung Electronics Co., Ltd.

#2 lore raw

PATCH 1/1 ARM: S5P: Fix USB and 48M clock enable procedure

Quoted 11 lines · Kukjin Kim
Quoted 10 lines · Paulius Zaleckas

---

?arch/arm/mach-s5p6440/clock.c ? ? ? ? ? ? ?| ? 23 ------- ?arch/arm/plat-s5p/clock.c ? ? ? ? ? ? ? ? ?| ? 86
++++++++++++++++++++++++++++
?arch/arm/plat-s5p/include/plat/s5p-clock.h | ? ?1 ?3 files changed, 86 insertions(+), 24 deletions(-)

Quoted 8 lines

Hi Paulius,

There are some comments about your patches which includes previous S3C64XX patches.

Basically your approach looks good trial and structure...but I'm not sure whether your approach can be used commonly on Samsung's all SoCs or not. Need to do more test on boards and I already informed your patches to USB engineers in my team, actually need to discuss about this.

As a note, I know, 'xusbxti' clock is structure for external xtal which is used for generating USB clock on board... it depends on board condition, because can be used 12/24/48Mhz on board. The clk_48m means generated actual USB clock, 48Mhz. So should be implemented enable function by using clk_48m...

I don't agree that enable should be for clk_48m. The reason is that IMO it is possible
to enable 48m clock, but suspend the clock for USB device part (I am not sure about this yet...).

If that is true than I think we will need one more clk for USB device:

        /->clk_48m
xusbxti-|
        \->clk_usb_device
Quoted 9 lines

Anyway let you know about the result of internal discussion soon, then let's talk.

Thanks.

Best regards,
Kgene.
--
Kukjin Kim <[email protected]>, Senior Engineer, SW Solution Development Team, Samsung Electronics Co., Ltd.

#3 lore raw

PATCH 1/1 ARM: S5P: Fix USB and 48M clock enable procedure

Quoted 2 lines · Paulius Zaleckas

On Thu, Oct 14, 2010 at 12:03 PM, Kukjin Kim <[email protected]>

Quoted 11 lines · wrote:
Quoted 11 lines
Quoted 10 lines · Paulius Zaleckas

---

 arch/arm/mach-s5p6440/clock.c              |   23 -------
 arch/arm/plat-s5p/clock.c                  |   86
++++++++++++++++++++++++++++
 arch/arm/plat-s5p/include/plat/s5p-clock.h |    1
 3 files changed, 86 insertions(+), 24 deletions(-)
Quoted 3 lines
Quoted 3 lines

Hi Paulius,

There are some comments about your patches which includes previous

S3C64XX

Quoted 3 lines

patches.

Quoted 2 lines

Basically your approach looks good trial and structure...but I'm not

sure

Quoted 2 lines

whether your approach can be used commonly on Samsung's all SoCs or not.

Quoted 1 line

Need to do more test on boards and I already informed your patches to

USB

Quoted 3 lines

engineers in my team, actually need to discuss about this.

Quoted 2 lines

As a note, I know, 'xusbxti' clock is structure for external xtal which

is

Quoted 2 lines

used for generating USB clock on board... it depends on board condition, because can be used 12/24/48Mhz on board. The clk_48m means generated

actual

Quoted 1 line

USB clock, 48Mhz. So should be implemented enable function by using

clk_48m...

Quoted 5 lines

I don't agree that enable should be for clk_48m. The reason is that IMO it is possible
to enable 48m clock, but suspend the clock for USB device part (I am not sure about this yet...).

Umm...please see below.

          |\
xxti------|O|
          |M|-----..System..-- HCLK --> usb_device...
xusbxti---| |  |
        | |/   |
        |      |
       (1)    (2)  -------------
         ---------| USB OTG Phy | --> clk_48M
                   -------------

Firstly, 'xusbxti' can be used to system clock when selected by OM pin. As I said, the rate of xusbxti can be 12/24/48Mhz and there is it on board...usually is used 12Mhz or 24Mhz on SMDK board.
It means depends on board and always enabled, so no need following 'rate' and 'enable' in your following patch.

struct clk clk_xusbxti = {
.name = "xusbxti",
.id = -1,
+ .rate = 48000000,
+ .enable = clk_xusbxti_ctrl,
};

struct clk s5p_clk_27m = {
@@ -49,6 +134,7 @@ struct clk clk_48m = {
.name = "clk_48m",
.id = -1,
.rate = 48000000,
+ .parent = &clk_xusbxti,
};

According to above clock diagram, the parent of 'clk_48m' can be 'xxti' or 'xusbxti', it is decided by SoC and OM pin. For example, there is line (1) on S5P6440 so your patch is right, but there is line (2) on S5PC100 but line (1) so it's wrong..the parent of clk_48m can be xxti by OM Pin.

It means the parent of clk_48m depends on SoC, can't define it in the plat- s5p. And maybe as you know, in the S5PV210/S5PC110 case, need two clk_48m structures for Host and Device...

So its clock structure has SoC(ARCH) dependency and it's difficult to merge it into plat-s5p now.

We need to re-think to implement it...and I and my team will sort out it.

Quoted 7 lines

If that is true than I think we will need one more clk for USB device:

        /->clk_48m
xusbxti-|
        \->clk_usb_device

Actually, it's different on each SoCs...

Quoted 3 lines
Quoted 1 line

Anyway let you know about the result of internal discussion soon, then

let's talk.

Quoted 1 line

Thanks.

Best regards,
Kgene.
--
Kukjin Kim <[email protected]>, Senior Engineer, SW Solution Development Team, Samsung Electronics Co., Ltd.

#4 lore raw

PATCH 1/1 ARM: S5P: Fix USB and 48M clock enable procedure

Quoted 79 lines · Kukjin Kim
Quoted 2 lines · Paulius Zaleckas

On Thu, Oct 14, 2010 at 12:03 PM, Kukjin Kim<[email protected]>

Quoted 11 lines · wrote:
Quoted 11 lines
Quoted 10 lines · Paulius Zaleckas

---

  arch/arm/mach-s5p6440/clock.c              |   23 -------
  arch/arm/plat-s5p/clock.c                  |   86
++++++++++++++++++++++++++++
  arch/arm/plat-s5p/include/plat/s5p-clock.h |    1
  3 files changed, 86 insertions(+), 24 deletions(-)
Quoted 3 lines
Quoted 3 lines

Hi Paulius,

There are some comments about your patches which includes previous

S3C64XX

Quoted 3 lines

patches.

Quoted 2 lines

Basically your approach looks good trial and structure...but I'm not

sure

Quoted 2 lines

whether your approach can be used commonly on Samsung's all SoCs or not.

Quoted 1 line

Need to do more test on boards and I already informed your patches to

USB

Quoted 3 lines

engineers in my team, actually need to discuss about this.

Quoted 2 lines

As a note, I know, 'xusbxti' clock is structure for external xtal which

is

Quoted 2 lines

used for generating USB clock on board... it depends on board condition, because can be used 12/24/48Mhz on board. The clk_48m means generated

actual

Quoted 1 line

USB clock, 48Mhz. So should be implemented enable function by using

clk_48m...

Quoted 5 lines

I don't agree that enable should be for clk_48m. The reason is that IMO it is possible
to enable 48m clock, but suspend the clock for USB device part (I am not sure about this yet...).

Umm...please see below.

           |\
xxti------|O|
           |M|-----..System..-- HCLK -->  usb_device...
xusbxti---| |  |
         | |/   |
         |      |
        (1)    (2)  -------------
          ---------| USB OTG Phy | -->  clk_48M
                    -------------

Firstly, 'xusbxti' can be used to system clock when selected by OM pin. As I said, the rate of xusbxti can be 12/24/48Mhz and there is it on board...usually is used 12Mhz or 24Mhz on SMDK board.
It means depends on board and always enabled, so no need following 'rate' and 'enable' in your following patch.

  struct clk clk_xusbxti = {
  	.name		= "xusbxti",
  	.id		= -1,
arch/arm/plat-s5p/clock.c +3 -0
+	.rate		= 48000000,
+	.enable		= clk_xusbxti_ctrl,
  };
 
  struct clk s5p_clk_27m = {
@@ -49,6 +134,7 @@ struct clk clk_48m = {
  	.name		= "clk_48m",
  	.id		= -1,
  	.rate		= 48000000,
+	.parent		=&clk_xusbxti,
  };
 

According to above clock diagram, the parent of 'clk_48m' can be 'xxti' or 'xusbxti', it is decided by SoC and OM pin. For example, there is line (1) on S5P6440 so your patch is right, but there is line (2) on S5PC100 but line (1) so it's wrong..the parent of clk_48m can be xxti by OM Pin.

Quoted 6 lines

It means the parent of clk_48m depends on SoC, can't define it in the plat- s5p. And maybe as you know, in the S5PV210/S5PC110 case, need two clk_48m structures for Host and Device...

So its clock structure has SoC(ARCH) dependency and it's difficult to merge it into plat-s5p now.

I see... So maybe the best option is to define enable function not per platform, but per machine.

Originally I was trying to solve bug where OHCI was not working on S3C6410. I discovered that I need to select different clock source for it or fix enable procedure for 48M. For S5P platform I have datasheet only for S5PC100 and yes you are right about clock structure there.

Quoted 1 line

We need to re-think to implement it...and I and my team will sort out it.

Looking forward!

Quoted 22 lines
Quoted 7 lines

If that is true than I think we will need one more clk for USB device:

         /->clk_48m
xusbxti-|
         \->clk_usb_device

Actually, it's different on each SoCs...

Quoted 3 lines
Quoted 1 line

Anyway let you know about the result of internal discussion soon, then

let's talk.

Quoted 1 line

Thanks.

Best regards,
Kgene.
--
Kukjin Kim<[email protected]>, Senior Engineer, SW Solution Development Team, Samsung Electronics Co., Ltd.