Skip to contentSkip to conversation

tps65910: Work around silicon erratum SWCZ010

PATCH Superseded Arnout Vandecappelle (Essensium/Mind) 4 participants 11 messages Aug 22, 2014 – Aug 29, 2014

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 tps65910: Work around silicon erratum SWCZ010

Quoted 1 line

From http://www.ti.com/lit/pdf/SWCZ010 :

Glitch on SDA-SCL not managed correctly by the I2C IP

Impact:
The standard specifies that the I2C transfer should restart on a start event in all cases. The current design does not support two consecutive Start conditions. This can cause the first real access after such a glitch to be corrupted.

Description:
An unexpected glitch on SDA and SCL can generate a wrong start event. In the current design, the SCL line must toggle two times to detect a new start event and completely restart the I2C access; hence the real start event is not detected in the case of a single SCL toggle.

Workaround:
Repeat I2C access.

The first access to the tps65910 that we do is when loading the regmap in the probe function. If there has been a glitch during boot and the erratum is triggered, then the regmap loading will fail with -REMOTEIO since the I2C transfer is not ACKed by the tps65910.

A simple retry will work around it, because a subsequent I2C start condition is detected properly by the tps65910.

---
This patch is based on v3.17-rc1. Build-tested with omap2plus_defconfig. Runtime tested based on v3.4.97 with a custom config.

---
 drivers/mfd/tps65910.c | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)
drivers/mfd/tps65910.c +11 -1
index f243e75..cc1ca33 100644
@@ -487,8 +487,18 @@ static int tps65910_i2c_probe(struct i2c_client *i2c,
 	tps65910->id = chip_id;
 
 	tps65910->regmap = devm_regmap_init_i2c(i2c, &tps65910_regmap_config);
-	if (IS_ERR(tps65910->regmap)) {
+	if (IS_ERR(tps65910->regmap))
 		ret = PTR_ERR(tps65910->regmap);
+	if (ret == -EREMOTEIO) {
+		dev_warn(&i2c->dev, "regmap initialization failed, possibly due to erratum SWCZ010; trying again\n");
+		tps65910->regmap = regmap_init_i2c(i2c,
+			&tps65910_regmap_config);
+		if (IS_ERR(tps65910->regmap))
+			ret = PTR_ERR(tps65910->regmap);
+		else
+			ret = 0;
+	}
+	if (ret) {
 		dev_err(&i2c->dev, "regmap initialization failed: %d\n", ret);
 		return ret;
 	}
Signature · 1 line
2.1.0
#1 lore raw
Quoted 41 lines · Arnout Vandecappelle (Essensium/Mind)

From http://www.ti.com/lit/pdf/SWCZ010 :

Glitch on SDA-SCL not managed correctly by the I2C IP

Impact:
The standard specifies that the I2C transfer should restart on a start event in all cases. The current design does not support two consecutive Start conditions. This can cause the first real access after such a glitch to be corrupted.

Description:
An unexpected glitch on SDA and SCL can generate a wrong start event. In the current design, the SCL line must toggle two times to detect a new start event and completely restart the I2C access; hence the real start event is not detected in the case of a single SCL toggle.

Workaround:
Repeat I2C access.

The first access to the tps65910 that we do is when loading the regmap in the probe function. If there has been a glitch during boot and the erratum is triggered, then the regmap loading will fail with -REMOTEIO since the I2C transfer is not ACKed by the tps65910.

A simple retry will work around it, because a subsequent I2C start condition is detected properly by the tps65910.

---
This patch is based on v3.17-rc1. Build-tested with omap2plus_defconfig. Runtime tested based on v3.4.97 with a custom config.

---
  drivers/mfd/tps65910.c | 12 +++++++++++-
  1 file changed, 11 insertions(+), 1 deletion(-)
drivers/mfd/tps65910.c +0 -0
index f243e75..cc1ca33 100644

drivers/mfd/tps65910.c static int tps65910_i2c_probe(struct i2c_client *i2c, · line 488

Quoted 2 lines
drivers/mfd/tps65910.c +0 -0
@@ -487,8 +487,18 @@ static int tps65910_i2c_probe(struct i2c_client *i2c,
  	tps65910->id = chip_id;
 

I think one dummy read of any register unconditionally before regmap init will resolve this issue. Similar issue we face with Palmas TPS65913 and TI recommended to use dummy read before any valid transfer.

#2 lore raw
Quoted 37 lines · Laxman Dewangan
Quoted 36 lines · Arnout Vandecappelle (Essensium/Mind)

From http://www.ti.com/lit/pdf/SWCZ010 :

Glitch on SDA-SCL not managed correctly by the I2C IP

Impact:
The standard specifies that the I2C transfer should restart on a start event in all cases. The current design does not support two consecutive Start conditions. This can cause the first real access after such a glitch to be corrupted.

Description:
An unexpected glitch on SDA and SCL can generate a wrong start event. In the current design, the SCL line must toggle two times to detect a new start event and completely restart the I2C access; hence the real start event is not detected in the case of a single SCL toggle.

Workaround:
Repeat I2C access.

The first access to the tps65910 that we do is when loading the regmap in the probe function. If there has been a glitch during boot and the erratum is triggered, then the regmap loading will fail with -REMOTEIO since the I2C transfer is not ACKed by the tps65910.

A simple retry will work around it, because a subsequent I2C start condition is detected properly by the tps65910.

---
This patch is based on v3.17-rc1. Build-tested with omap2plus_defconfig. Runtime tested based on v3.4.97 with a custom config.

---
  drivers/mfd/tps65910.c | 12 +++++++++++-
  1 file changed, 11 insertions(+), 1 deletion(-)
Quoted 13 lines

drivers/mfd/tps65910.c static int tps65910_i2c_probe(struct i2c_client *i2c, · line 488

Quoted 7 lines
drivers/mfd/tps65910.c +0 -0
index f243e75..cc1ca33 100644
@@ -487,8 +487,18 @@ static int tps65910_i2c_probe(struct i2c_client *i2c,
      tps65910->id = chip_id;
 

I think one dummy read of any register unconditionally before regmap init will resolve this issue.
Similar issue we face with Palmas TPS65913 and TI recommended to use dummy read before any valid transfer.

Yeah, good idea.

So something like

i2c_master_send(i2c, "", 1);

just before the regmap_init.

Is abusing the empty string like that OK or should I declare a dummy buffer?

Regards,
Arnout

Signature · 6 lines
Arnout Vandecappelle arnout at mind be
Senior Embedded Software Architect +32-16-286500
Essensium/Mind http://www.mind.be
G.Geenslaan 9, 3001 Leuven, Belgium BE 872 984 063 RPR Leuven
LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle
GPG fingerprint: 7CB5 E4CC 6C2E EFD4 6E3D A754 F963 ECAB 2450 2F1F
#3 lore raw

PATCH v2 1/1 tps65910: Work around silicon erratum SWCZ010

In reply to Arnout Vandecappelle (Essensium/Mind)
Quoted 1 line

From http://www.ti.com/lit/pdf/SWCZ010 :

Glitch on SDA-SCL not managed correctly by the I2C IP

Impact:
The standard specifies that the I2C transfer should restart on a start event in all cases. The current design does not support two consecutive Start conditions. This can cause the first real access after such a glitch to be corrupted.

Description:
An unexpected glitch on SDA and SCL can generate a wrong start event. In the current design, the SCL line must toggle two times to detect a new start event and completely restart the I2C access; hence the real start event is not detected in the case of a single SCL toggle.

Workaround:
Repeat I2C access.

A simpler workaround is to make a dummy transfer just before the first access to the tps65910 chip. This can be done unconditionally.

---
This patch is based on v3.17-rc1. Build-tested with omap2plus_defconfig. Runtime tested based on v3.4.97 with a custom config.

v2: use dummy transfer instead of retrying (Laxman Dewangan)

---
 drivers/mfd/tps65910.c | 5 +++++
 1 file changed, 5 insertions(+)
drivers/mfd/tps65910.c +5 -0
index f243e75..7612d89 100644
@@ -486,6 +486,11 @@ static int tps65910_i2c_probe(struct i2c_client *i2c,
 	tps65910->i2c_client = i2c;
 	tps65910->id = chip_id;
 
+	/* Work around silicon erratum SWCZ010: the tps65910 may miss the
+	 * first I2C transfer. So issue a dummy transfer before the first
+	 * real transfer.
+	 */
+	i2c_master_send(i2c, "", 1);
 	tps65910->regmap = devm_regmap_init_i2c(i2c, &tps65910_regmap_config);
 	if (IS_ERR(tps65910->regmap)) {
 		ret = PTR_ERR(tps65910->regmap);
Signature · 1 line
2.1.0
#5 lore raw
Quoted 4 lines · Mark Brown
Quoted 3 lines · Arnout Vandecappelle (Essensium/Mind)

From http://www.ti.com/lit/pdf/SWCZ010 :

Glitch on SDA-SCL not managed correctly by the I2C IP

 drivers/mfd/tps65910.c | 5 +++++
 1 file changed, 5 insertions(+)
Quoted 1 line

Applied, thanks.

Eh?

Signature · 4 lines
Lee Jones
Linaro STMicroelectronics Landing Team Lead
Linaro.org │ Open source software for ARM SoCs
Follow Linaro: Facebook | Twitter | Blog
#6 lore raw
Quoted 5 lines · Lee Jones
Quoted 3 lines · Mark Brown
Quoted 2 lines · Arnout Vandecappelle (Essensium/Mind)

From http://www.ti.com/lit/pdf/SWCZ010 :

Quoted 8 lines
Quoted 1 line
Quoted 1 line

Glitch on SDA-SCL not managed correctly by the I2C IP

 drivers/mfd/tps65910.c | 5 +++++
 1 file changed, 5 insertions(+)
Quoted 1 line

Applied, thanks.

Eh?

So, this is an example of why it's important to both use subject lines matching the pattern for the subsystem and try to select CC lists sensibly. I've dropped the patch.

#7 lore raw
Quoted 11 lines · Mark Brown
Quoted 9 lines · Lee Jones
Quoted 4 lines · Mark Brown
Quoted 3 lines · Arnout Vandecappelle (Essensium/Mind)

From http://www.ti.com/lit/pdf/SWCZ010 :

Glitch on SDA-SCL not managed correctly by the I2C IP

 drivers/mfd/tps65910.c | 5 +++++
 1 file changed, 5 insertions(+)
Quoted 7 lines
Quoted 3 lines
Quoted 1 line

Applied, thanks.

Eh?

So, this is an example of why it's important to both use subject lines matching the pattern for the subsystem and try to select CC lists sensibly. I've dropped the patch.

So, what is the proper subsystem in this case? mfd?

Should I repost with the subsystem in the subject?

BTW, while I'm asking questions: is this something that should be sent to stable?

Regards,
Arnout

Signature · 6 lines
Arnout Vandecappelle                          arnout at mind be
Senior Embedded Software Architect            +32-16-286500
Essensium/Mind                                http://www.mind.be
G.Geenslaan 9, 3001 Leuven, Belgium           BE 872 984 063 RPR Leuven
LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle
GPG fingerprint:  7CB5 E4CC 6C2E EFD4 6E3D A754 F963 ECAB 2450 2F1F
#8 lore raw
In reply to Arnout Vandecappelle (Essensium/Mind)

Because this patch is dropped from Mark's tree, here is opportunity to revisit patch.

Quoted 10 lines · Arnout Vandecappelle (Essensium/Mind)

From http://www.ti.com/lit/pdf/SWCZ010 :

Workaround:
Repeat I2C access.

A simpler workaround is to make a dummy transfer just before the first access to the tps65910 chip. This can be done unconditionally.

Quoted 1 line

id = chip_id;

drivers/mfd/tps65910.c static int tps65910_i2c_probe(struct i2c_client *i2c, · line 493

Quoted 5 lines
drivers/mfd/tps65910.c +5 -0
+	/* Work around silicon erratum SWCZ010: the tps65910 may miss the
+	 * first I2C transfer. So issue a dummy transfer before the first
+	 * real transfer.
+	 */
+	i2c_master_send(i2c, "", 1);

I think dummy read is more safe operation than dummy write.
Dummy write can create the write on any register which can damage the critical settings or it may be possible that it will be incomplete calls. Datasheet has not been explained this clearly.

Quoted 1 line
#9 lore raw
Quoted 19 lines · Laxman Dewangan

Because this patch is dropped from Mark's tree, here is opportunity to revisit patch.

Quoted 10 lines · Arnout Vandecappelle (Essensium/Mind)

From http://www.ti.com/lit/pdf/SWCZ010 :

Workaround:
Repeat I2C access.

A simpler workaround is to make a dummy transfer just before the first access to the tps65910 chip. This can be done unconditionally.

Quoted 1 line

id = chip_id;

+ /* Work around silicon erratum SWCZ010: the tps65910 may miss the

drivers/mfd/tps65910.c static int tps65910_i2c_probe(struct i2c_client *i2c, · line 493

Quoted 4 lines
drivers/mfd/tps65910.c +4 -0
+     * first I2C transfer. So issue a dummy transfer before the first
+     * real transfer.
+     */
+    i2c_master_send(i2c, "", 1);
Quoted 4 lines

I think dummy read is more safe operation than dummy write.
Dummy write can create the write on any register which can damage the critical settings or it may be possible that it will be incomplete calls. Datasheet has not been explained this clearly.

We're just sending the register address 0 of the SMBus transfer, without an
actual read/write request. If we do an i2c_master_recv, it's not a valid SMBus transfer so we're equally unsure about how the chip will react to that.

But if you like, I can check how the chip reacts to it - most likely it'll just
ignore it, possibly it won't even ack it.

Regards,
Arnout

Signature · 6 lines
Arnout Vandecappelle                          arnout at mind be
Senior Embedded Software Architect            +32-16-286500
Essensium/Mind                                http://www.mind.be
G.Geenslaan 9, 3001 Leuven, Belgium           BE 872 984 063 RPR Leuven
LinkedIn profile: http://www.linkedin.com/in/arnoutvandecappelle
GPG fingerprint:  7CB5 E4CC 6C2E EFD4 6E3D A754 F963 ECAB 2450 2F1F
#10 lore raw
In reply to Arnout Vandecappelle (Essensium/Mind)
Quoted 28 lines · Arnout Vandecappelle (Essensium/Mind)

From http://www.ti.com/lit/pdf/SWCZ010 :

Glitch on SDA-SCL not managed correctly by the I2C IP

Impact:
The standard specifies that the I2C transfer should restart on a start event in all cases. The current design does not support two consecutive Start conditions. This can cause the first real access after such a glitch to be corrupted.

Description:
An unexpected glitch on SDA and SCL can generate a wrong start event. In the current design, the SCL line must toggle two times to detect a new start event and completely restart the I2C access; hence the real start event is not detected in the case of a single SCL toggle.

Workaround:
Repeat I2C access.

A simpler workaround is to make a dummy transfer just before the first access to the tps65910 chip. This can be done unconditionally.

---
This patch is based on v3.17-rc1. Build-tested with omap2plus_defconfig. Runtime tested based on v3.4.97 with a custom config.

Quoted 4 lines

v2: use dummy transfer instead of retrying (Laxman Dewangan)

---
 drivers/mfd/tps65910.c | 5 +++++
 1 file changed, 5 insertions(+)

Applied with Mark's Ack.

Quoted 16 lines
drivers/mfd/tps65910.c +5 -0
index f243e75..7612d89 100644
@@ -486,6 +486,11 @@ static int tps65910_i2c_probe(struct i2c_client *i2c,
 	tps65910->i2c_client = i2c;
 	tps65910->id = chip_id;
 
+	/* Work around silicon erratum SWCZ010: the tps65910 may miss the
+	 * first I2C transfer. So issue a dummy transfer before the first
+	 * real transfer.
+	 */
+	i2c_master_send(i2c, "", 1);
 	tps65910->regmap = devm_regmap_init_i2c(i2c, &tps65910_regmap_config);
 	if (IS_ERR(tps65910->regmap)) {
 		ret = PTR_ERR(tps65910->regmap);
Signature · 4 lines
Lee Jones
Linaro STMicroelectronics Landing Team Lead
Linaro.org │ Open source software for ARM SoCs
Follow Linaro: Facebook | Twitter | Blog