Skip to contentSkip to conversation

mfd: madera: Allow more time for hardware reset

PATCH Applied Charles Keepax 2 participants 10 messages Jan 6, 2020 – Jan 16, 2020

Subsystems (2)

v1 of 2 Applied Latest is v2 →

lore
Share & tools
lore

Charles Keepax

author maintainer
#0 lore raw

PATCH 1/2 mfd: madera: Allow more time for hardware reset

Both manual and power on resets have a brief period where the chip will not be accessible immediately afterwards. Extend the time allowed for this from a minimum of 1mS to 2mS based on newer evaluation of the hardware and ensure this reset happens in all reset conditions. Whilst making the change also remove the redundant NULL checks in the reset functions as the GPIO functions already check for this.

---
 drivers/mfd/madera-core.c | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)
drivers/mfd/madera-core.c +10 -8
index a8cfadc1fc01e..f41ce408259fb 100644
@@ -238,6 +238,11 @@ static int madera_wait_for_boot(struct madera *madera)
 	return ret;
 }
 
+static inline void madera_reset_delay(void)
+{
+	usleep_range(2000, 3000);
+}
+
 static int madera_soft_reset(struct madera *madera)
 {
 	int ret;
@@ -249,16 +254,13 @@ static int madera_soft_reset(struct madera *madera)
 	}
 
 	/* Allow time for internal clocks to startup after reset */
-	usleep_range(1000, 2000);
+	madera_reset_delay();
 
 	return 0;
 }
 
 static void madera_enable_hard_reset(struct madera *madera)
 {
-	if (!madera->pdata.reset)
-		return;
-
 	/*
 	 * There are many existing out-of-tree users of these codecs that we
 	 * can't break so preserve the expected behaviour of setting the line
@@ -269,11 +271,9 @@ static void madera_enable_hard_reset(struct madera *madera)
 
 static void madera_disable_hard_reset(struct madera *madera)
 {
-	if (!madera->pdata.reset)
-		return;
-
 	gpiod_set_raw_value_cansleep(madera->pdata.reset, 1);
-	usleep_range(1000, 2000);
+
+	madera_reset_delay();
 }
 
 static int __maybe_unused madera_runtime_resume(struct device *dev)
@@ -292,6 +292,8 @@ static int __maybe_unused madera_runtime_resume(struct device *dev)
 	regcache_cache_only(madera->regmap, false);
 	regcache_cache_only(madera->regmap_32bit, false);
 
+	madera_reset_delay();
+
 	ret = madera_wait_for_boot(madera);
 	if (ret)
 		goto err;
Signature · 1 line
2.11.0
author maintainer
#1 lore raw

PATCH 2/2 mfd: madera: Wait for boot done before accessing any other registers

It is advised to wait for the boot done bit to be set before reading any other register, update the driver to respect this.

---
 drivers/mfd/madera-core.c | 17 +++++++++++++++--
 1 file changed, 15 insertions(+), 2 deletions(-)
drivers/mfd/madera-core.c +15 -2
index f41ce408259fb..edf3d22dd82f9 100644
@@ -199,7 +199,7 @@ EXPORT_SYMBOL_GPL(madera_name_from_type);
 #define MADERA_BOOT_POLL_INTERVAL_USEC		5000
 #define MADERA_BOOT_POLL_TIMEOUT_USEC		25000
 
-static int madera_wait_for_boot(struct madera *madera)
+static int madera_wait_for_boot_noack(struct madera *madera)
 {
 	ktime_t timeout;
 	unsigned int val = 0;
@@ -226,6 +226,13 @@ static int madera_wait_for_boot(struct madera *madera)
 		ret = -ETIMEDOUT;
 	}
 
+	return ret;
+}
+
+static int madera_wait_for_boot(struct madera *madera)
+{
+	int ret = madera_wait_for_boot_noack(madera);
+
 	/*
 	 * BOOT_DONE defaults to unmasked on boot so we must ack it.
 	 * Do this even after a timeout to avoid interrupt storms.
@@ -547,6 +554,12 @@ int madera_dev_init(struct madera *madera)
 	regcache_cache_only(madera->regmap, false);
 	regcache_cache_only(madera->regmap_32bit, false);
 
+	ret = madera_wait_for_boot_noack(madera);
+	if (ret) {
+		dev_err(madera->dev, "Device failed initial boot: %d\n", ret);
+		goto err_reset;
+	}
+
 	/*
 	 * Now we can power up and verify that this is a chip we know about
 	 * before we start doing any writes to its registers.
@@ -652,7 +665,7 @@ int madera_dev_init(struct madera *madera)
 
 	ret = madera_wait_for_boot(madera);
 	if (ret) {
-		dev_err(madera->dev, "Device failed initial boot: %d\n", ret);
+		dev_err(madera->dev, "Failed to clear boot done: %d\n", ret);
 		goto err_reset;
 	}
 
Signature · 1 line
2.11.0
#2 lore raw

Re: [PATCH 2/2] mfd: madera: Wait for boot done before accessing any other registers

Quoted 7 lines · Charles Keepax

It is advised to wait for the boot done bit to be set before reading any other register, update the driver to respect this.

---
 drivers/mfd/madera-core.c | 17 +++++++++++++++--
 1 file changed, 15 insertions(+), 2 deletions(-)

I'm assuming this patch is orthogonal to the last?

Can I take it on its own?

For my own reference:

Signature · 4 lines
Lee Jones [李琼斯]
Linaro Services Technical Lead
Linaro.org │ Open source software for ARM SoCs
Follow Linaro: Facebook | Twitter | Blog
author maintainer
#3 lore raw

Re: [PATCH 2/2] mfd: madera: Wait for boot done before accessing any other registers

Quoted 5 lines · Lee Jones
Quoted 3 lines · Charles Keepax

It is advised to wait for the boot done bit to be set before reading any other register, update the driver to respect this.

Quoted 9 lines
Quoted 4 lines
---
 drivers/mfd/madera-core.c | 17 +++++++++++++++--
 1 file changed, 15 insertions(+), 2 deletions(-)

I'm assuming this patch is orthogonal to the last?

Can I take it on its own?

Yeah these can be taken separately to the other series we are waiting on Mark to review.

Thanks,
Charles

Quoted 8 lines

For my own reference:

Signature · 4 lines
Lee Jones [李琼斯]
Linaro Services Technical Lead
Linaro.org │ Open source software for ARM SoCs
Follow Linaro: Facebook | Twitter | Blog
#4 lore raw

Re: [PATCH 2/2] mfd: madera: Wait for boot done before accessing any other registers

Quoted 11 lines · Charles Keepax
Quoted 10 lines · Lee Jones
Quoted 7 lines · Charles Keepax

It is advised to wait for the boot done bit to be set before reading any other register, update the driver to respect this.

---
 drivers/mfd/madera-core.c | 17 +++++++++++++++--
 1 file changed, 15 insertions(+), 2 deletions(-)
Quoted 7 lines
Quoted 4 lines

I'm assuming this patch is orthogonal to the last?

Can I take it on its own?

Yeah these can be taken separately to the other series we are waiting on Mark to review.

I mean the patch. Can I take 2/2 without 1/2?

Signature · 4 lines
Lee Jones [李琼斯]
Linaro Services Technical Lead
Linaro.org │ Open source software for ARM SoCs
Follow Linaro: Facebook | Twitter | Blog
author maintainer
#5 lore raw

Re: [PATCH 2/2] mfd: madera: Wait for boot done before accessing any other registers

Quoted 15 lines · Lee Jones
Quoted 13 lines · Charles Keepax
Quoted 12 lines · Lee Jones
Quoted 7 lines · Charles Keepax

It is advised to wait for the boot done bit to be set before reading any other register, update the driver to respect this.

---
 drivers/mfd/madera-core.c | 17 +++++++++++++++--
 1 file changed, 15 insertions(+), 2 deletions(-)

I'm assuming this patch is orthogonal to the last?

Quoted 8 lines
Quoted 5 lines
Quoted 2 lines

Can I take it on its own?

Yeah these can be taken separately to the other series we are waiting on Mark to review.

I mean the patch. Can I take 2/2 without 1/2?

Sorry, yes that should work too. Just tested it on my system seems fine.

Thanks,
Charles

Applied
#6 lore raw

Re: [PATCH 2/2] mfd: madera: Wait for boot done before accessing any other registers

In reply to Charles Keepax
Quoted 7 lines · Charles Keepax

It is advised to wait for the boot done bit to be set before reading any other register, update the driver to respect this.

---
 drivers/mfd/madera-core.c | 17 +++++++++++++++--
 1 file changed, 15 insertions(+), 2 deletions(-)

Applied, thanks.

Signature · 4 lines
Lee Jones [李琼斯]
Linaro Services Technical Lead
Linaro.org │ Open source software for ARM SoCs
Follow Linaro: Facebook | Twitter | Blog
#7 lore raw
In reply to Charles Keepax
Quoted 17 lines · Charles Keepax

Both manual and power on resets have a brief period where the chip will not be accessible immediately afterwards. Extend the time allowed for this from a minimum of 1mS to 2mS based on newer evaluation of the hardware and ensure this reset happens in all reset conditions. Whilst making the change also remove the redundant NULL checks in the reset functions as the GPIO functions already check for this.

---
 drivers/mfd/madera-core.c | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)
drivers/mfd/madera-core.c +0 -0
index a8cfadc1fc01e..f41ce408259fb 100644

drivers/mfd/madera-core.c static int madera_wait_for_boot(struct madera *madera) · line 244

Quoted 7 lines
drivers/mfd/madera-core.c +4 -0
@@ -238,6 +238,11 @@ static int madera_wait_for_boot(struct madera *madera)
 	return ret;
 }
 
+static inline void madera_reset_delay(void)
+{
+	usleep_range(2000, 3000);
+}

Hmm ... We usually shy away from abstraction for the sake of abstraction. What's preventing you from using the preferred method of simply calling the abstracted function from each of the call-sites?

I could understand (a little) if you needed to frequently change these values, since changing them in once place is obviously simpler than changing them in 3, but even then it's marginal.

Quoted 44 lines
drivers/mfd/madera-core.c +5 -8
 static int madera_soft_reset(struct madera *madera)
 {
 	int ret;
@@ -249,16 +254,13 @@ static int madera_soft_reset(struct madera *madera)
 	}
 
 	/* Allow time for internal clocks to startup after reset */
-	usleep_range(1000, 2000);
+	madera_reset_delay();
 
 	return 0;
 }
 
 static void madera_enable_hard_reset(struct madera *madera)
 {
-	if (!madera->pdata.reset)
-		return;
-
 	/*
 	 * There are many existing out-of-tree users of these codecs that we
 	 * can't break so preserve the expected behaviour of setting the line
@@ -269,11 +271,9 @@ static void madera_enable_hard_reset(struct madera *madera)
 
 static void madera_disable_hard_reset(struct madera *madera)
 {
-	if (!madera->pdata.reset)
-		return;
-
 	gpiod_set_raw_value_cansleep(madera->pdata.reset, 1);
-	usleep_range(1000, 2000);
+
+	madera_reset_delay();
 }
 
 static int __maybe_unused madera_runtime_resume(struct device *dev)
@@ -292,6 +292,8 @@ static int __maybe_unused madera_runtime_resume(struct device *dev)
 	regcache_cache_only(madera->regmap, false);
 	regcache_cache_only(madera->regmap_32bit, false);
 
+	madera_reset_delay();
+
 	ret = madera_wait_for_boot(madera);
 	if (ret)
 		goto err;
Signature · 4 lines
Lee Jones [李琼斯]
Linaro Services Technical Lead
Linaro.org │ Open source software for ARM SoCs
Follow Linaro: Facebook | Twitter | Blog
author maintainer
#8 lore raw
Quoted 31 lines · Lee Jones
Quoted 17 lines · Charles Keepax

Both manual and power on resets have a brief period where the chip will not be accessible immediately afterwards. Extend the time allowed for this from a minimum of 1mS to 2mS based on newer evaluation of the hardware and ensure this reset happens in all reset conditions. Whilst making the change also remove the redundant NULL checks in the reset functions as the GPIO functions already check for this.

---
 drivers/mfd/madera-core.c | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)
drivers/mfd/madera-core.c +0 -0
index a8cfadc1fc01e..f41ce408259fb 100644

drivers/mfd/madera-core.c static int madera_wait_for_boot(struct madera *madera) · line 244

Quoted 7 lines
drivers/mfd/madera-core.c +4 -0
@@ -238,6 +238,11 @@ static int madera_wait_for_boot(struct madera *madera)
 	return ret;
 }
 
+static inline void madera_reset_delay(void)
+{
+	usleep_range(2000, 3000);
+}

Hmm ... We usually shy away from abstraction for the sake of abstraction. What's preventing you from using the preferred method of simply calling the abstracted function from each of the call-sites?

Quoted 4 lines

I could understand (a little) if you needed to frequently change these values, since changing them in once place is obviously simpler than changing them in 3, but even then it's marginal.

I don't mind manually inline it, we don't plan on changing the values very often certainly. It really was just to avoid future bugs if someone adds a new place that needs the delay or does indeed change the delay. Would you mind if I used a define for the time instead, if I am manually inlining? That keeps the same single place to update, but without the extra function.

Thanks,
Charles

#9 lore raw
Quoted 37 lines · Charles Keepax
Quoted 31 lines · Lee Jones
Quoted 17 lines · Charles Keepax

Both manual and power on resets have a brief period where the chip will not be accessible immediately afterwards. Extend the time allowed for this from a minimum of 1mS to 2mS based on newer evaluation of the hardware and ensure this reset happens in all reset conditions. Whilst making the change also remove the redundant NULL checks in the reset functions as the GPIO functions already check for this.

---
 drivers/mfd/madera-core.c | 18 ++++++++++--------
 1 file changed, 10 insertions(+), 8 deletions(-)
drivers/mfd/madera-core.c +0 -0
index a8cfadc1fc01e..f41ce408259fb 100644

drivers/mfd/madera-core.c static int madera_wait_for_boot(struct madera *madera) · line 244

Quoted 7 lines
drivers/mfd/madera-core.c +4 -0
@@ -238,6 +238,11 @@ static int madera_wait_for_boot(struct madera *madera)
 	return ret;
 }
 
+static inline void madera_reset_delay(void)
+{
+	usleep_range(2000, 3000);
+}

Hmm ... We usually shy away from abstraction for the sake of abstraction. What's preventing you from using the preferred method of simply calling the abstracted function from each of the call-sites?

Quoted 4 lines

I could understand (a little) if you needed to frequently change these values, since changing them in once place is obviously simpler than changing them in 3, but even then it's marginal.

Quoted 6 lines

I don't mind manually inline it, we don't plan on changing the values very often certainly. It really was just to avoid future bugs if someone adds a new place that needs the delay or does indeed change the delay. Would you mind if I used a define for the time instead, if I am manually inlining? That keeps the same single place to update, but without the extra function.

That would be my preference, yes.

Signature · 4 lines
Lee Jones [李琼斯]
Linaro Services Technical Lead
Linaro.org │ Open source software for ARM SoCs
Follow Linaro: Facebook | Twitter | Blog