Skip to contentSkip to conversation

pwm: stm32: Minor cleanups

PATCH Superseded Thierry Reding 3 participants 11 messages Oct 16, 2019

Subsystems (3)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

cover pwm: stm32: Minor cleanups

Hi,

Looking at Fabrice's STM32 patches I noticed that we're now passing the breakinput values (u32) into a function via int parameters. The easiest way to fix this inconsistency is by just passing a pointer to the break input structure. There's some preparatory work here that makes the code slightly more readable, in my opinion, but it's really marginal, so I'm not terribly thrilled by this series in retrospect.

If nobody else thinks this is a big improvement I'll just scrap it.

Thierry

Thierry Reding (3):
pwm: stm32: Remove clutter from ternary operator
pwm: stm32: Remove confusing bitmask
pwm: stm32: Pass breakinput instead of its values

 drivers/pwm/pwm-stm32.c | 37 ++++++++++++++++++++-----------------
 1 file changed, 20 insertions(+), 17 deletions(-)
Signature · 1 line
2.23.0
#1 lore raw

PATCH 1/3 pwm: stm32: Remove clutter from ternary operator

Remove usage of the ternary operator to assign values for register fields. This removes clutter and improves readability.

---
 drivers/pwm/pwm-stm32.c | 18 +++++++++++++-----
 1 file changed, 13 insertions(+), 5 deletions(-)
drivers/pwm/pwm-stm32.c +13 -5
index 9430b4cd383f..b12fb11b7a55 100644
@@ -493,11 +493,19 @@ static const struct pwm_ops stm32pwm_ops = {
 static int stm32_pwm_set_breakinput(struct stm32_pwm *priv,
 				    int index, int level, int filter)
 {
-	u32 bke = (index == 0) ? TIM_BDTR_BKE : TIM_BDTR_BK2E;
-	int shift = (index == 0) ? TIM_BDTR_BKF_SHIFT : TIM_BDTR_BK2F_SHIFT;
-	u32 mask = (index == 0) ? TIM_BDTR_BKE | TIM_BDTR_BKP | TIM_BDTR_BKF
-				: TIM_BDTR_BK2E | TIM_BDTR_BK2P | TIM_BDTR_BK2F;
-	u32 bdtr = bke;
+	u32 bke, shift, mask, bdtr;
+
+	if (index == 0) {
+		bke = TIM_BDTR_BKE;
+		shift = TIM_BDTR_BKF_SHIFT;
+		mask = TIM_BDTR_BKE | TIM_BDTR_BKP | TIM_BDTR_BKF;
+	} else {
+		bke = TIM_BDTR_BK2E;
+		shift = TIM_BDTR_BK2F_SHIFT;
+		mask = TIM_BDTR_BK2E | TIM_BDTR_BK2P | TIM_BDTR_BK2F;
+	}
+
+	bdtr = bke;
 
 	/*
 	 * The both bits could be set since only one will be wrote
Signature · 1 line
2.23.0
Reviewed-by on v1 1/3
#2 lore raw

Re: [PATCH 1/3] pwm: stm32: Remove clutter from ternary operator

Quoted 24 lines · Thierry Reding

Remove usage of the ternary operator to assign values for register fields. This removes clutter and improves readability.

---
 drivers/pwm/pwm-stm32.c | 18 +++++++++++++-----
 1 file changed, 13 insertions(+), 5 deletions(-)
drivers/pwm/pwm-stm32.c +3 -5
index 9430b4cd383f..b12fb11b7a55 100644
@@ -493,11 +493,19 @@ static const struct pwm_ops stm32pwm_ops = {
 static int stm32_pwm_set_breakinput(struct stm32_pwm *priv,
 				    int index, int level, int filter)
 {
-	u32 bke = (index == 0) ? TIM_BDTR_BKE : TIM_BDTR_BK2E;
-	int shift = (index == 0) ? TIM_BDTR_BKF_SHIFT : TIM_BDTR_BK2F_SHIFT;
-	u32 mask = (index == 0) ? TIM_BDTR_BKE | TIM_BDTR_BKP | TIM_BDTR_BKF
-				: TIM_BDTR_BK2E | TIM_BDTR_BK2P | TIM_BDTR_BK2F;
-	u32 bdtr = bke;
+	u32 bke, shift, mask, bdtr;
+
+	if (index == 0) {

drivers/pwm/pwm-stm32.c static const struct pwm_ops stm32pwm_ops = { · line 508

Quoted 10 lines
drivers/pwm/pwm-stm32.c +10 -0
+		bke = TIM_BDTR_BKE;
+		shift = TIM_BDTR_BKF_SHIFT;
+		mask = TIM_BDTR_BKE | TIM_BDTR_BKP | TIM_BDTR_BKF;
+	} else {
+		bke = TIM_BDTR_BK2E;
+		shift = TIM_BDTR_BK2F_SHIFT;
+		mask = TIM_BDTR_BK2E | TIM_BDTR_BK2P | TIM_BDTR_BK2F;
+	}
+
+	bdtr = bke;

Is index always in {0, 1}? Maybe a comment or a check about that would be helpful. (-> separate patch I think).

Best regards
Uwe

Signature · 2 lines
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
#3 lore raw

Re: [PATCH 1/3] pwm: stm32: Remove clutter from ternary operator

Quoted 38 lines · Uwe Kleine-König
Quoted 24 lines · Thierry Reding

Remove usage of the ternary operator to assign values for register fields. This removes clutter and improves readability.

---
 drivers/pwm/pwm-stm32.c | 18 +++++++++++++-----
 1 file changed, 13 insertions(+), 5 deletions(-)
drivers/pwm/pwm-stm32.c +3 -5
index 9430b4cd383f..b12fb11b7a55 100644
@@ -493,11 +493,19 @@ static const struct pwm_ops stm32pwm_ops = {
 static int stm32_pwm_set_breakinput(struct stm32_pwm *priv,
 				    int index, int level, int filter)
 {
-	u32 bke = (index == 0) ? TIM_BDTR_BKE : TIM_BDTR_BK2E;
-	int shift = (index == 0) ? TIM_BDTR_BKF_SHIFT : TIM_BDTR_BK2F_SHIFT;
-	u32 mask = (index == 0) ? TIM_BDTR_BKE | TIM_BDTR_BKP | TIM_BDTR_BKF
-				: TIM_BDTR_BK2E | TIM_BDTR_BK2P | TIM_BDTR_BK2F;
-	u32 bdtr = bke;
+	u32 bke, shift, mask, bdtr;
+
+	if (index == 0) {

drivers/pwm/pwm-stm32.c static const struct pwm_ops stm32pwm_ops = { · line 508

Quoted 10 lines
drivers/pwm/pwm-stm32.c +10 -0
+		bke = TIM_BDTR_BKE;
+		shift = TIM_BDTR_BKF_SHIFT;
+		mask = TIM_BDTR_BKE | TIM_BDTR_BKP | TIM_BDTR_BKF;
+	} else {
+		bke = TIM_BDTR_BK2E;
+		shift = TIM_BDTR_BK2F_SHIFT;
+		mask = TIM_BDTR_BK2E | TIM_BDTR_BK2P | TIM_BDTR_BK2F;
+	}
+
+	bdtr = bke;
Quoted 2 lines

Is index always in {0, 1}? Maybe a comment or a check about that would be helpful. (-> separate patch I think).

The bindings say that index can only be 0 or 1. I guess strictly it might actually depend on the number of break inputs, but given these register definitions, there's only ever two.

But yeah, it might be a good idea to sanitize the values upon probe. I'll add another patch to do that.

Thierry

#4 lore raw

PATCH 2/3 pwm: stm32: Remove confusing bitmask

In reply to Thierry Reding

Both BKP bits are set in the BDTR register and the code relies on the mask used during write to make sure only one of them is written. Since this isn't immediately obvious, a comment is needed to explain it. The same can be achieved by making explicit what happens, so add another temporary variable that contains only the one bit that is actually ORed into the register and get rid of the comment.

---
 drivers/pwm/pwm-stm32.c | 10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)
drivers/pwm/pwm-stm32.c +4 -6
index b12fb11b7a55..8f1f3371e1dd 100644
@@ -493,26 +493,24 @@ static const struct pwm_ops stm32pwm_ops = {
 static int stm32_pwm_set_breakinput(struct stm32_pwm *priv,
 				    int index, int level, int filter)
 {
-	u32 bke, shift, mask, bdtr;
+	u32 bke, bkp, shift, mask, bdtr;
 
 	if (index == 0) {
 		bke = TIM_BDTR_BKE;
+		bkp = TIM_BDTR_BKP;
 		shift = TIM_BDTR_BKF_SHIFT;
 		mask = TIM_BDTR_BKE | TIM_BDTR_BKP | TIM_BDTR_BKF;
 	} else {
 		bke = TIM_BDTR_BK2E;
+		bkp = TIM_BDTR_BK2P;
 		shift = TIM_BDTR_BK2F_SHIFT;
 		mask = TIM_BDTR_BK2E | TIM_BDTR_BK2P | TIM_BDTR_BK2F;
 	}
 
 	bdtr = bke;
 
-	/*
-	 * The both bits could be set since only one will be wrote
-	 * due to mask value.
-	 */
 	if (level)
-		bdtr |= TIM_BDTR_BKP | TIM_BDTR_BK2P;
+		bdtr |= bkp;
 
 	bdtr |= (filter & TIM_BDTR_BKF_MASK) << shift;
 
Signature · 1 line
2.23.0
#5 lore raw

Re: [PATCH 2/3] pwm: stm32: Remove confusing bitmask

Quoted 23 lines · Thierry Reding

Both BKP bits are set in the BDTR register and the code relies on the mask used during write to make sure only one of them is written. Since this isn't immediately obvious, a comment is needed to explain it. The same can be achieved by making explicit what happens, so add another temporary variable that contains only the one bit that is actually ORed into the register and get rid of the comment.

---
 drivers/pwm/pwm-stm32.c | 10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)
drivers/pwm/pwm-stm32.c +1 -1
index b12fb11b7a55..8f1f3371e1dd 100644
@@ -493,26 +493,24 @@ static const struct pwm_ops stm32pwm_ops = {
 static int stm32_pwm_set_breakinput(struct stm32_pwm *priv,
 				    int index, int level, int filter)
 {
-	u32 bke, shift, mask, bdtr;
+	u32 bke, bkp, shift, mask, bdtr;
 

drivers/pwm/pwm-stm32.c static const struct pwm_ops stm32pwm_ops = { · line 507

Quoted 10 lines
drivers/pwm/pwm-stm32.c +2 -0
 	if (index == 0) {
 		bke = TIM_BDTR_BKE;
+		bkp = TIM_BDTR_BKP;
 		shift = TIM_BDTR_BKF_SHIFT;
 		mask = TIM_BDTR_BKE | TIM_BDTR_BKP | TIM_BDTR_BKF;
 	} else {
 		bke = TIM_BDTR_BK2E;
+		bkp = TIM_BDTR_BK2P;
 		shift = TIM_BDTR_BK2F_SHIFT;
 		mask = TIM_BDTR_BK2E | TIM_BDTR_BK2P | TIM_BDTR_BK2F;

Assuming in the else branch index is always 1, the following would be IMHO nicer:

#define TIM_BDTR_BKE(i) BIT(12 + 12 * (i))
#define TIM_BDTR_BKP(i) BIT(13 + 12 * (i))
#define TIM_BDTR_BKF_SHIFT(i) (16 + 4 * (i))

..

	bke = TIM_BDTR_BKE(index);
	bkp = TIM_BDTR_BKP(index);

Best regards
Uwe

Signature · 2 lines
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
#6 lore raw

Re: [PATCH 2/3] pwm: stm32: Remove confusing bitmask

Quoted 38 lines · Uwe Kleine-König
Quoted 23 lines · Thierry Reding

Both BKP bits are set in the BDTR register and the code relies on the mask used during write to make sure only one of them is written. Since this isn't immediately obvious, a comment is needed to explain it. The same can be achieved by making explicit what happens, so add another temporary variable that contains only the one bit that is actually ORed into the register and get rid of the comment.

---
 drivers/pwm/pwm-stm32.c | 10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)
drivers/pwm/pwm-stm32.c +1 -1
index b12fb11b7a55..8f1f3371e1dd 100644
@@ -493,26 +493,24 @@ static const struct pwm_ops stm32pwm_ops = {
 static int stm32_pwm_set_breakinput(struct stm32_pwm *priv,
 				    int index, int level, int filter)
 {
-	u32 bke, shift, mask, bdtr;
+	u32 bke, bkp, shift, mask, bdtr;
 

drivers/pwm/pwm-stm32.c static const struct pwm_ops stm32pwm_ops = { · line 507

Quoted 10 lines
drivers/pwm/pwm-stm32.c +2 -0
 	if (index == 0) {
 		bke = TIM_BDTR_BKE;
+		bkp = TIM_BDTR_BKP;
 		shift = TIM_BDTR_BKF_SHIFT;
 		mask = TIM_BDTR_BKE | TIM_BDTR_BKP | TIM_BDTR_BKF;
 	} else {
 		bke = TIM_BDTR_BK2E;
+		bkp = TIM_BDTR_BK2P;
 		shift = TIM_BDTR_BK2F_SHIFT;
 		mask = TIM_BDTR_BK2E | TIM_BDTR_BK2P | TIM_BDTR_BK2F;

Assuming in the else branch index is always 1, the following would be IMHO nicer:

Quoted 8 lines

#define TIM_BDTR_BKE(i) BIT(12 + 12 * (i))
#define TIM_BDTR_BKP(i) BIT(13 + 12 * (i))
#define TIM_BDTR_BKF_SHIFT(i) (16 + 4 * (i))

..

	bke = TIM_BDTR_BKE(index);
	bkp = TIM_BDTR_BKP(index);

I had thought about that, but ultimately decided against it because the original defines might match exactly what's in the datasheet, so there's some value to keep the originals.

I suppose one other alternative would be to let the macros be and do the computations in the driver instead, something like:

	bke = TIM_BDTR_BKE << (index * 12);
	bkp = TIM_BDTR_BKP << (index * 12);
	bkf = TIM_BDTR_BKF << (index *  4);

But yeah, I agree that having the parameters be part of the macros is even better.

Fabrice, any objection to redefining the macros as above?

Thierry

#7 lore raw

Re: [PATCH 2/3] pwm: stm32: Remove confusing bitmask

Quoted 55 lines · Thierry Reding
Quoted 38 lines · Uwe Kleine-König
Quoted 23 lines · Thierry Reding

Both BKP bits are set in the BDTR register and the code relies on the mask used during write to make sure only one of them is written. Since this isn't immediately obvious, a comment is needed to explain it. The same can be achieved by making explicit what happens, so add another temporary variable that contains only the one bit that is actually ORed into the register and get rid of the comment.

---
 drivers/pwm/pwm-stm32.c | 10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)
drivers/pwm/pwm-stm32.c +1 -1
index b12fb11b7a55..8f1f3371e1dd 100644
@@ -493,26 +493,24 @@ static const struct pwm_ops stm32pwm_ops = {
 static int stm32_pwm_set_breakinput(struct stm32_pwm *priv,
 				    int index, int level, int filter)
 {
-	u32 bke, shift, mask, bdtr;
+	u32 bke, bkp, shift, mask, bdtr;
 

drivers/pwm/pwm-stm32.c static const struct pwm_ops stm32pwm_ops = { · line 507

Quoted 10 lines
drivers/pwm/pwm-stm32.c +2 -0
 	if (index == 0) {
 		bke = TIM_BDTR_BKE;
+		bkp = TIM_BDTR_BKP;
 		shift = TIM_BDTR_BKF_SHIFT;
 		mask = TIM_BDTR_BKE | TIM_BDTR_BKP | TIM_BDTR_BKF;
 	} else {
 		bke = TIM_BDTR_BK2E;
+		bkp = TIM_BDTR_BK2P;
 		shift = TIM_BDTR_BK2F_SHIFT;
 		mask = TIM_BDTR_BK2E | TIM_BDTR_BK2P | TIM_BDTR_BK2F;

Assuming in the else branch index is always 1, the following would be IMHO nicer:

Quoted 8 lines

#define TIM_BDTR_BKE(i) BIT(12 + 12 * (i))
#define TIM_BDTR_BKP(i) BIT(13 + 12 * (i))
#define TIM_BDTR_BKF_SHIFT(i) (16 + 4 * (i))

..

	bke = TIM_BDTR_BKE(index);
	bkp = TIM_BDTR_BKP(index);

I had thought about that, but ultimately decided against it because the original defines might match exactly what's in the datasheet, so there's some value to keep the originals.

I suppose one other alternative would be to let the macros be and do the computations in the driver instead, something like:

Quoted 8 lines
	bke = TIM_BDTR_BKE << (index * 12);
	bkp = TIM_BDTR_BKP << (index * 12);
	bkf = TIM_BDTR_BKF << (index *  4);

But yeah, I agree that having the parameters be part of the macros is even better.

Fabrice, any objection to redefining the macros as above?

Hi Thierry,

No objection from me, it will probably improve readability.

I'd just suggest an alternative to this: maybe a simple struct array with two entries can improve readability ? E.g. keep the defines matching the datasheet, and get rid of the conditional code ?

Dirty proposal:

static const struct stm32_pwm_brk[] = {
	/* {bke, bkp, shift, mask} */
	{ TIM_BDTR_BKE, TIM_BDTR_BKP, ...},
	{ TIM_BDTR_BK2E, TIM_BDTR_BK2P, ...},
}

and use "index" to index it ?

But I'm fine with the macros as well: there's already similar things in this driver to deal with the channels for instance.

Thanks,
Fabrice

Quoted 3 lines

Thierry

#8 lore raw

Re: [PATCH 2/3] pwm: stm32: Remove confusing bitmask

Quoted 81 lines · Fabrice Gasnier
Quoted 55 lines · Thierry Reding
Quoted 38 lines · Uwe Kleine-König
Quoted 23 lines · Thierry Reding

Both BKP bits are set in the BDTR register and the code relies on the mask used during write to make sure only one of them is written. Since this isn't immediately obvious, a comment is needed to explain it. The same can be achieved by making explicit what happens, so add another temporary variable that contains only the one bit that is actually ORed into the register and get rid of the comment.

---
 drivers/pwm/pwm-stm32.c | 10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)
drivers/pwm/pwm-stm32.c +1 -1
index b12fb11b7a55..8f1f3371e1dd 100644
@@ -493,26 +493,24 @@ static const struct pwm_ops stm32pwm_ops = {
 static int stm32_pwm_set_breakinput(struct stm32_pwm *priv,
 				    int index, int level, int filter)
 {
-	u32 bke, shift, mask, bdtr;
+	u32 bke, bkp, shift, mask, bdtr;
 

drivers/pwm/pwm-stm32.c static const struct pwm_ops stm32pwm_ops = { · line 507

Quoted 10 lines
drivers/pwm/pwm-stm32.c +2 -0
 	if (index == 0) {
 		bke = TIM_BDTR_BKE;
+		bkp = TIM_BDTR_BKP;
 		shift = TIM_BDTR_BKF_SHIFT;
 		mask = TIM_BDTR_BKE | TIM_BDTR_BKP | TIM_BDTR_BKF;
 	} else {
 		bke = TIM_BDTR_BK2E;
+		bkp = TIM_BDTR_BK2P;
 		shift = TIM_BDTR_BK2F_SHIFT;
 		mask = TIM_BDTR_BK2E | TIM_BDTR_BK2P | TIM_BDTR_BK2F;

Assuming in the else branch index is always 1, the following would be IMHO nicer:

Quoted 8 lines

#define TIM_BDTR_BKE(i) BIT(12 + 12 * (i))
#define TIM_BDTR_BKP(i) BIT(13 + 12 * (i))
#define TIM_BDTR_BKF_SHIFT(i) (16 + 4 * (i))

..

	bke = TIM_BDTR_BKE(index);
	bkp = TIM_BDTR_BKP(index);

I had thought about that, but ultimately decided against it because the original defines might match exactly what's in the datasheet, so there's some value to keep the originals.

I suppose one other alternative would be to let the macros be and do the computations in the driver instead, something like:

Quoted 8 lines
	bke = TIM_BDTR_BKE << (index * 12);
	bkp = TIM_BDTR_BKP << (index * 12);
	bkf = TIM_BDTR_BKF << (index *  4);

But yeah, I agree that having the parameters be part of the macros is even better.

Fabrice, any objection to redefining the macros as above?

Hi Thierry,

No objection from me, it will probably improve readability.

I'd just suggest an alternative to this: maybe a simple struct array with two entries can improve readability ? E.g. keep the defines matching the datasheet, and get rid of the conditional code ?

Dirty proposal:

static const struct stm32_pwm_brk[] = {
	/* {bke, bkp, shift, mask} */
	{ TIM_BDTR_BKE, TIM_BDTR_BKP, ...},
	{ TIM_BDTR_BK2E, TIM_BDTR_BK2P, ...},
}
Quoted 4 lines

and use "index" to index it ?

But I'm fine with the macros as well: there's already similar things in this driver to deal with the channels for instance.

I didn't test but I wouldn't be surprised if the compiler could better optimize the solution I suggested. Might be worth a test however.

Best regards
Uwe

Signature · 2 lines
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
#9 lore raw

PATCH 3/3 pwm: stm32: Pass breakinput instead of its values

In reply to Thierry Reding

Instead of passing the individual values of the breakpoint, pass a pointer to the breakpoint.

---
 drivers/pwm/pwm-stm32.c | 13 +++++--------
 1 file changed, 5 insertions(+), 8 deletions(-)
drivers/pwm/pwm-stm32.c +5 -8
index 8f1f3371e1dd..9e500928c37d 100644
@@ -491,11 +491,11 @@ static const struct pwm_ops stm32pwm_ops = {
 };
 
 static int stm32_pwm_set_breakinput(struct stm32_pwm *priv,
-				    int index, int level, int filter)
+				    const struct stm32_breakinput *bi)
 {
 	u32 bke, bkp, shift, mask, bdtr;
 
-	if (index == 0) {
+	if (bi->index == 0) {
 		bke = TIM_BDTR_BKE;
 		bkp = TIM_BDTR_BKP;
 		shift = TIM_BDTR_BKF_SHIFT;
@@ -509,10 +509,10 @@ static int stm32_pwm_set_breakinput(struct stm32_pwm *priv,
 
 	bdtr = bke;
 
-	if (level)
+	if (bi->level)
 		bdtr |= bkp;
 
-	bdtr |= (filter & TIM_BDTR_BKF_MASK) << shift;
+	bdtr |= (bi->filter & TIM_BDTR_BKF_MASK) << shift;
 
 	regmap_update_bits(priv->regmap, TIM_BDTR, mask, bdtr);
 
@@ -527,10 +527,7 @@ static int stm32_pwm_apply_breakinputs(struct stm32_pwm *priv)
 	int ret;
 
 	for (i = 0; i < priv->num_breakinputs; i++) {
-		ret = stm32_pwm_set_breakinput(priv,
-					       priv->breakinputs[i].index,
-					       priv->breakinputs[i].level,
-					       priv->breakinputs[i].filter);
+		ret = stm32_pwm_set_breakinput(priv, &priv->breakinputs[i]);
 		if (ret < 0)
 			return ret;
 	}
Signature · 1 line
2.23.0
#10 lore raw
In reply to Thierry Reding
Quoted 10 lines · Thierry Reding

Hi,

Looking at Fabrice's STM32 patches I noticed that we're now passing the breakinput values (u32) into a function via int parameters. The easiest way to fix this inconsistency is by just passing a pointer to the break input structure. There's some preparatory work here that makes the code slightly more readable, in my opinion, but it's really marginal, so I'm not terribly thrilled by this series in retrospect.

If nobody else thinks this is a big improvement I'll just scrap it.

I like it. Together with my suggestion to add parameters to the register offsets I think it's worth the effort.

Best regards
Uwe

Signature · 2 lines
Pengutronix e.K.                           | Uwe Kleine-König            |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |