[RFC,06/12] Clk: qcom: support for dynamic updating the PLL
diff mbox

Message ID 1501153825-5181-7-git-send-email-absahu@codeaurora.org
State Superseded
Delegated to: Stephen Boyd
Headers show

Commit Message

Abhishek Sahu July 27, 2017, 11:10 a.m. UTC
Some of the Alpha PLL’s support dynamic update in which the
frequency can be changed dynamically without turning off the PLL.

This dynamic update requires the following sequence

1. Write the desired values to pll_l_val and pll_alpha_val.
2. Toggle pll_latch_input from low to high.
3. Wait for pll_ack_latch to transition from low to high.
   The new L and alpha values have been latched. It make
   take some time for the PLL to fully settle with these
   new values.
4. Pull pll_latch_input low.

Signed-off-by: Abhishek Sahu <absahu@codeaurora.org>
---
 drivers/clk/qcom/clk-alpha-pll.c | 62 +++++++++++++++++++++++++++++++++-------
 drivers/clk/qcom/clk-alpha-pll.h |  1 +
 2 files changed, 52 insertions(+), 11 deletions(-)

Comments

Stephen Boyd July 28, 2017, 6:34 p.m. UTC | #1
On 07/27, Abhishek Sahu wrote:
> Some of the Alpha PLL’s support dynamic update in which the
> frequency can be changed dynamically without turning off the PLL.
> 
> This dynamic update requires the following sequence
> 
> 1. Write the desired values to pll_l_val and pll_alpha_val.
> 2. Toggle pll_latch_input from low to high.
> 3. Wait for pll_ack_latch to transition from low to high.
>    The new L and alpha values have been latched. It make
>    take some time for the PLL to fully settle with these
>    new values.
> 4. Pull pll_latch_input low.
> 
> Signed-off-by: Abhishek Sahu <absahu@codeaurora.org>

I think Rajendra has a similar patch that was sent. Is this the
same? Can you please look on the list and find it and compare?
Abhishek Sahu July 30, 2017, 1:57 p.m. UTC | #2
On 2017-07-29 00:04, Stephen Boyd wrote:
> On 07/27, Abhishek Sahu wrote:
>> Some of the Alpha PLL’s support dynamic update in which the
>> frequency can be changed dynamically without turning off the PLL.
>> 
>> This dynamic update requires the following sequence
>> 
>> 1. Write the desired values to pll_l_val and pll_alpha_val.
>> 2. Toggle pll_latch_input from low to high.
>> 3. Wait for pll_ack_latch to transition from low to high.
>>    The new L and alpha values have been latched. It make
>>    take some time for the PLL to fully settle with these
>>    new values.
>> 4. Pull pll_latch_input low.
>> 
>> Signed-off-by: Abhishek Sahu <absahu@codeaurora.org>
> 
> I think Rajendra has a similar patch that was sent. Is this the
> same? Can you please look on the list and find it and compare?

  Checked the list. Rajendra has two patches

  1. https://www.spinics.net/lists/linux-arm-msm/msg23349.html

  Yes my patch does the same thing with minor diffs.
  My patch checks PLL_UPDATE_BYPASS and handles both
  the cases. We can merge both the patches. I will check
  with Rajendra and will work on this merge.

  2. Following patch fixes different issue although flag name
     is common.

  https://patchwork.kernel.org/patch/9662917/

  Shall I include this patch in my patch series but not
  sure we can directly turn off the PLL inside the PLL
  set rate operation since it will turn the PLL off for
  all its users.
Stephen Boyd Aug. 1, 2017, 9:12 p.m. UTC | #3
On 07/30, Abhishek Sahu wrote:
> On 2017-07-29 00:04, Stephen Boyd wrote:
> >On 07/27, Abhishek Sahu wrote:
> >>Some of the Alpha PLL’s support dynamic update in which the
> >>frequency can be changed dynamically without turning off the PLL.
> >>
> >>This dynamic update requires the following sequence
> >>
> >>1. Write the desired values to pll_l_val and pll_alpha_val.
> >>2. Toggle pll_latch_input from low to high.
> >>3. Wait for pll_ack_latch to transition from low to high.
> >>   The new L and alpha values have been latched. It make
> >>   take some time for the PLL to fully settle with these
> >>   new values.
> >>4. Pull pll_latch_input low.
> >>
> >>Signed-off-by: Abhishek Sahu <absahu@codeaurora.org>
> >
> >I think Rajendra has a similar patch that was sent. Is this the
> >same? Can you please look on the list and find it and compare?
> 
>  Checked the list. Rajendra has two patches
> 
>  1. https://www.spinics.net/lists/linux-arm-msm/msg23349.html
> 
>  Yes my patch does the same thing with minor diffs.
>  My patch checks PLL_UPDATE_BYPASS and handles both
>  the cases. We can merge both the patches. I will check
>  with Rajendra and will work on this merge.

Ok.

> 
>  2. Following patch fixes different issue although flag name
>     is common.
> 
>  https://patchwork.kernel.org/patch/9662917/
> 
>  Shall I include this patch in my patch series but not
>  sure we can directly turn off the PLL inside the PLL
>  set rate operation since it will turn the PLL off for
>  all its users.
> 

Hopefully the users of a PLL that doesn't support dynamic rate
update can accept the fact that the clk will turn off while the
rate is reprogrammed. At least that seems to be true for Taniya
in that patch set. If it isn't true for your hardware, then don't
specify the flag? Or is the problem that you may not have the
flag set for certain PLLs that you're supporting?
Abhishek Sahu Aug. 2, 2017, 1:50 p.m. UTC | #4
On 2017-08-02 02:42, Stephen Boyd wrote:
> On 07/30, Abhishek Sahu wrote:
>> On 2017-07-29 00:04, Stephen Boyd wrote:
>> >On 07/27, Abhishek Sahu wrote:

>>  2. Following patch fixes different issue although flag name
>>     is common.
>> 
>>  https://patchwork.kernel.org/patch/9662917/
>> 
>>  Shall I include this patch in my patch series but not
>>  sure we can directly turn off the PLL inside the PLL
>>  set rate operation since it will turn the PLL off for
>>  all its users.
>> 
> 
> Hopefully the users of a PLL that doesn't support dynamic rate
> update can accept the fact that the clk will turn off while the
> rate is reprogrammed. At least that seems to be true for Taniya
> in that patch set. If it isn't true for your hardware, then don't
> specify the flag? Or is the problem that you may not have the
> flag set for certain PLLs that you're supporting?

  The turning off PLL will happen in case of flag is not set.
  The turning off PLL in set rate is unsafe. If this PLL
  is driving multiple RCG's and one of the RCG is changing the
  PLL frequency by its clk_set_rate with CLK_SET_RATE_PARENT,
  then all the RCG's clock will go off for some
  time and it may trigger crash/silent reboot.

  If the user is aware, then it can turn off the clock first,
  then do the set rate and then it can enable again. In
  PLL set_rate we can check if PLL is enabled and can
  return EBUSY error.
--
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

Patch
diff mbox

diff --git a/drivers/clk/qcom/clk-alpha-pll.c b/drivers/clk/qcom/clk-alpha-pll.c
index 3a7ec42..e38f4d2 100644
--- a/drivers/clk/qcom/clk-alpha-pll.c
+++ b/drivers/clk/qcom/clk-alpha-pll.c
@@ -31,7 +31,10 @@ 
 # define PLL_VOTE_FSM_ENA	BIT(20)
 # define PLL_FSM_ENA		BIT(20)
 # define PLL_VOTE_FSM_RESET	BIT(21)
+# define PLL_UPDATE		BIT(22)
+# define PLL_UPDATE_BYPASS	BIT(23)
 # define PLL_OFFLINE_ACK	BIT(28)
+# define ALPHA_PLL_ACK_LATCH	BIT(29)
 # define PLL_ACTIVE_FLAG	BIT(30)
 # define PLL_LOCK_DET		BIT(31)
 
@@ -122,6 +125,15 @@  static int wait_for_pll(struct clk_alpha_pll *pll, u32 mask, bool inverse,
 #define wait_for_pll_offline(pll) \
 	wait_for_pll(pll, PLL_OFFLINE_ACK, 0, "offline")
 
+#define wait_for_pll_update(pll) \
+	wait_for_pll(pll, PLL_UPDATE, 1, "update")
+
+#define wait_for_pll_update_ack_set(pll) \
+	wait_for_pll(pll, ALPHA_PLL_ACK_LATCH, 0, "update_ack_set")
+
+#define wait_for_pll_update_ack_clear(pll) \
+	wait_for_pll(pll, ALPHA_PLL_ACK_LATCH, 1, "update_ack_clear")
+
 void clk_alpha_pll_configure(struct clk_alpha_pll *pll, struct regmap *regmap,
 			     const struct alpha_pll_config *config)
 {
@@ -398,7 +410,8 @@  static int clk_alpha_pll_set_rate(struct clk_hw *hw, unsigned long rate,
 {
 	struct clk_alpha_pll *pll = to_clk_alpha_pll(hw);
 	const struct pll_vco *vco;
-	u32 l, alpha_width = pll_alpha_width(pll);
+	u32 l, mode, alpha_width = pll_alpha_width(pll);
+	int ret;
 	u64 a;
 
 	rate = alpha_pll_round_rate(rate, prate, &l, &a, alpha_width);
@@ -410,22 +423,49 @@  static int clk_alpha_pll_set_rate(struct clk_hw *hw, unsigned long rate,
 
 	regmap_write(pll->clkr.regmap, pll_l(pll), l);
 
-	if (alpha_width > ALPHA_BITWIDTH) {
-		a <<= (alpha_width - ALPHA_BITWIDTH);
-		regmap_update_bits(pll->clkr.regmap, pll_alpha_u(pll),
-				   GENMASK(0, alpha_width - ALPHA_BITWIDTH - 1),
-				   a >> ALPHA_BITWIDTH);
-	}
+	a <<= ALPHA_REG_BITWIDTH - ALPHA_BITWIDTH;
+
+	regmap_write(pll->clkr.regmap, pll_alpha(pll), a);
+	regmap_write(pll->clkr.regmap, pll_alpha_u(pll), a >> 32);
 
 	regmap_update_bits(pll->clkr.regmap, pll_alpha(pll),
 			   GENMASK(0, alpha_width - 1), a);
 
 	regmap_update_bits(pll->clkr.regmap, pll_user_ctl(pll),
-			   PLL_VCO_MASK << PLL_VCO_SHIFT,
-			   vco->val << PLL_VCO_SHIFT);
+			   PLL_ALPHA_EN, PLL_ALPHA_EN);
+
+	if (!clk_hw_is_enabled(hw) || !(pll->flags & SUPPORTS_DYNAMIC_UPDATE))
+		return 0;
+
+	regmap_read(pll->clkr.regmap, pll_mode(pll), &mode);
+	regmap_update_bits(pll->clkr.regmap, pll_mode(pll), PLL_UPDATE,
+			   PLL_UPDATE);
+
+	/* Make sure PLL_UPDATE request goes through*/
+	mb();
 
-	regmap_update_bits(pll->clkr.regmap, pll_user_ctl(pll), PLL_ALPHA_EN,
-			   PLL_ALPHA_EN);
+	/*
+	 * PLL will latch the new L, Alpha and freq control word.
+	 * PLL will respond by raising PLL_ACK_LATCH output when new programming
+	 * has been latched in and PLL is being updated. When
+	 * UPDATE_LOGIC_BYPASS bit is not set, PLL_UPDATE will be cleared
+	 * automatically by hardware when PLL_ACK_LATCH is asserted by PLL.
+	 */
+	if (!(mode & PLL_UPDATE_BYPASS))
+		return wait_for_pll_update(pll);
+
+	ret = wait_for_pll_update_ack_set(pll);
+	if (ret)
+		return ret;
+
+	regmap_update_bits(pll->clkr.regmap, pll_mode(pll), PLL_UPDATE, 0);
+
+	/* Make sure PLL_UPDATE request goes through*/
+	mb();
+
+	ret = wait_for_pll_update_ack_clear(pll);
+	if (ret)
+		return ret;
 
 	return 0;
 }
diff --git a/drivers/clk/qcom/clk-alpha-pll.h b/drivers/clk/qcom/clk-alpha-pll.h
index 51a61a0..6e40e09 100644
--- a/drivers/clk/qcom/clk-alpha-pll.h
+++ b/drivers/clk/qcom/clk-alpha-pll.h
@@ -54,6 +54,7 @@  struct clk_alpha_pll {
 #define SUPPORTS_16BIT_ALPHA		BIT(1)
 #define SUPPORTS_FSM_MODE		BIT(2)
 #define SUPPORTS_64BIT_CONFIG_CTL	BIT(3)
+#define SUPPORTS_DYNAMIC_UPDATE		BIT(4)
 	u8 flags;
 
 	struct clk_regmap clkr;