Skip to contentSkip to conversation

i2c: i801: Call i2c_register_spd for muxed child segments

RFC PATCH Superseded Heiner Kallweit 2 participants 11 messages Feb 22, 2024 – Mar 26, 2024

Subsystems (3)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

cover i2c: i801: Call i2c_register_spd for muxed child segments

Once the gpio mux driver binds to the "i2c-mux-gpio" platform device, this creates the i2c adapters for the muxed child segments. We can use the bus notifier mechanism to check for creation of the child i2d adapters, and call i2c_register_spd() for them. This allows to detect all DIMM's on systems with more than 8 memory slots.

I tested that the events are properly recognized. However I don't have hw with a muxed SMBUS, so I can't test the actual functionality.

Heiner Kallweit (2):
i2c: smbus: Prepare i2c_register_spd for usage on muxed segments
i2c: i801: Call i2c_register_spd for muxed child segments

 drivers/i2c/busses/i2c-i801.c | 23 +++++++++++++++++++++++
 drivers/i2c/i2c-smbus.c       | 18 +++++++++++-------
 2 files changed, 34 insertions(+), 7 deletions(-)
Signature · 1 line
2.43.2
#1 lore raw

RFC PATCH 1/2 i2c: smbus: Prepare i2c_register_spd for usage on muxed segments

If this is an adapter on a muxed bus segment, assume that each segment is connected to a subset of the (> 8) overall memory slots. In this case let's probe the maximum of 8 slots, however stop if the number of overall populated slots is reached.

If we're not on a muxed segment and the total number of slots is > 8, report an error, because then not all SPD eeproms can be addressed. Presumably the bus is muxed, but the mux config is missing.

---
 drivers/i2c/i2c-smbus.c | 18 +++++++++++-------
 1 file changed, 11 insertions(+), 7 deletions(-)
drivers/i2c/i2c-smbus.c +11 -7
index 74807c6db..ad7ea0215 100644
@@ -351,13 +351,17 @@ void i2c_register_spd(struct i2c_adapter *adap)
 	if (!dimm_count)
 		return;
 
-	dev_info(&adap->dev, "%d/%d memory slots populated (from DMI)\n",
-		 dimm_count, slot_count);
-
-	if (slot_count > 8) {
-		dev_warn(&adap->dev,
-			 "Systems with more than 8 memory slots not supported yet, not instantiating SPD\n");
-		return;
+	/* Check whether we're a child adapter on a muxed segment */
+	if (i2c_parent_is_i2c_adapter(adap)) {
+		slot_count = 8;
+	} else {
+		if (slot_count > 8) {
+			dev_err(&adap->dev,
+				"More than 8 memory slots on a single bus, mux config missing?\n");
+			return;
+		}
+		dev_info(&adap->dev, "%d/%d memory slots populated (from DMI)\n",
+			 dimm_count, slot_count);
 	}
 
 	/*
Signature · 1 line
2.43.2
#2 lore raw

Re: [PATCH RFC 1/2] i2c: smbus: Prepare i2c_register_spd for usage on muxed segments

Hi Heiner,

drivers/i2c/i2c-smbus.c void i2c_register_spd(struct i2c_adapter *adap) · line 354

Quoted 1 line
drivers/i2c/i2c-smbus.c +1 -0
+	/* Check whether we're a child adapter on a muxed segment */

The comment describes the 'if' but not 'then'. How about sth like "If we are a child on a muxed segment then limit slots to..."?

drivers/i2c/i2c-smbus.c void i2c_register_spd(struct i2c_adapter *adap) · line 356

Quoted 2 lines
drivers/i2c/i2c-smbus.c +2 -0
+	if (i2c_parent_is_i2c_adapter(adap)) {
+		slot_count = 8;

I don't know much about DMI. I just noticed that there are no printouts in this code path. Will there be one for the parent?

drivers/i2c/i2c-smbus.c void i2c_register_spd(struct i2c_adapter *adap) · line 360

Quoted 4 lines
drivers/i2c/i2c-smbus.c +4 -0
+	} else {
+		if (slot_count > 8) {
+			dev_err(&adap->dev,
+				"More than 8 memory slots on a single bus, mux config missing?\n");

With this error message, I as a user would think I need to setup a mux config somewhere. But it is missing from DMI, or? Then, we should probably use even FW_BUG in the message?

Happy hacking,

Wolfram

#3 lore raw

Re: [PATCH RFC 1/2] i2c: smbus: Prepare i2c_register_spd for usage on muxed segments

Quoted 7 lines · Wolfram Sang

Hi Heiner,

drivers/i2c/i2c-smbus.c void i2c_register_spd(struct i2c_adapter *adap) · line 354

Quoted 1 line
drivers/i2c/i2c-smbus.c +1 -0
+	/* Check whether we're a child adapter on a muxed segment */

The comment describes the 'if' but not 'then'. How about sth like "If we are a child on a muxed segment then limit slots to..."?

OK, this would be better.

Quoted 6 lines

drivers/i2c/i2c-smbus.c void i2c_register_spd(struct i2c_adapter *adap) · line 356

Quoted 2 lines
drivers/i2c/i2c-smbus.c +2 -0
+	if (i2c_parent_is_i2c_adapter(adap)) {
+		slot_count = 8;

I don't know much about DMI. I just noticed that there are no printouts in this code path. Will there be one for the parent?

With the patch as-is there's we omit printout for systems with > 8 memory slots. I'm not aware of any way to find out how many memory slots belong to a specific child bus segment. So all we could do is print per child segment how many slots are populated. But we have a printout per populated slot
already: "Successfully instantiated SPD at 0x%hx\n"
So IMO we don't loose any relevant info.

Quoted 9 lines

drivers/i2c/i2c-smbus.c void i2c_register_spd(struct i2c_adapter *adap) · line 360

Quoted 4 lines
drivers/i2c/i2c-smbus.c +4 -0
+	} else {
+		if (slot_count > 8) {
+			dev_err(&adap->dev,
+				"More than 8 memory slots on a single bus, mux config missing?\n");

With this error message, I as a user would think I need to setup a mux config somewhere. But it is missing from DMI, or? Then, we should probably use even FW_BUG in the message?

Actually a developer has to add the config to i801's mux_dmi_table[]. So yes, we should change the message to something like:
"More than 8 memory slots on a single bus, contact i801 maintainer to add the missing mux configuration"

Quoted 4 lines

Happy hacking,

Wolfram

Heiner

#4 lore raw

Re: [PATCH RFC 1/2] i2c: smbus: Prepare i2c_register_spd for usage on muxed segments

Quoted 4 lines

how many slots are populated. But we have a printout per populated slot
already: "Successfully instantiated SPD at 0x%hx\n"
So IMO we don't loose any relevant info.

Right. That makes me think we could even remove the dev_info() entirely. But I leave this up to you.

Quoted 4 lines

Actually a developer has to add the config to i801's mux_dmi_table[]. So yes, we should change the message to something like:
"More than 8 memory slots on a single bus, contact i801 maintainer to add the missing mux configuration"

Sounds good!

#5 lore raw

RFC PATCH 2/2 i2c: i801: Call i2c_register_spd for muxed child segments

In reply to Heiner Kallweit

Once the gpio mux driver binds to the "i2c-mux-gpio" platform device, this creates the i2c adapters for the muxed child segments. We can use the bus notifier mechanism to check for creation of the child i2d adapters, and call i2c_register_spd() for them. This allows to detect all DIMM's on systems with more than 8 memory slots.

---
 drivers/i2c/busses/i2c-i801.c | 23 +++++++++++++++++++++++
 1 file changed, 23 insertions(+)
drivers/i2c/busses/i2c-i801.c +23 -0
index 8f225cd7b..4b9d04f20 100644
@@ -105,6 +105,7 @@
 #include <linux/ioport.h>
 #include <linux/init.h>
 #include <linux/i2c.h>
+#include <linux/i2c-mux.h>
 #include <linux/i2c-smbus.h>
 #include <linux/acpi.h>
 #include <linux/io.h>
@@ -291,6 +292,7 @@ struct i801_priv {
 #if IS_ENABLED(CONFIG_I2C_MUX_GPIO) && defined CONFIG_DMI
 	struct platform_device *mux_pdev;
 	struct gpiod_lookup_table *lookup;
+	struct notifier_block mux_notifier_block;
 #endif
 	struct platform_device *tco_pdev;
 
@@ -1388,6 +1390,23 @@ static const struct dmi_system_id mux_dmi_table[] = {
 	{ }
 };
 
+static int i801_notifier_call(struct notifier_block *nb, unsigned long action,
+			      void *data)
+{
+	struct i801_priv *priv = container_of(nb, struct i801_priv, mux_notifier_block);
+	struct device *dev = data;
+
+	if (action != BUS_NOTIFY_ADD_DEVICE ||
+	    dev->type != &i2c_adapter_type ||
+	    i2c_root_adapter(dev) != &priv->adapter)
+		return NOTIFY_DONE;
+
+	/* Call i2c_register_spd for muxed child segments */
+	i2c_register_spd(to_i2c_adapter(dev));
+
+	return NOTIFY_OK;
+}
+
 /* Setup multiplexing if needed */
 static void i801_add_mux(struct i801_priv *priv)
 {
@@ -1425,6 +1444,9 @@ static void i801_add_mux(struct i801_priv *priv)
 	gpiod_add_lookup_table(lookup);
 	priv->lookup = lookup;
 
+	priv->mux_notifier_block.notifier_call = i801_notifier_call;
+	if (bus_register_notifier(&i2c_bus_type, &priv->mux_notifier_block))
+		return;
 	/*
 	 * Register the mux device, we use PLATFORM_DEVID_NONE here
 	 * because since we are referring to the GPIO chip by name we are
@@ -1443,6 +1465,7 @@ static void i801_add_mux(struct i801_priv *priv)
 
 static void i801_del_mux(struct i801_priv *priv)
 {
+	bus_unregister_notifier(&i2c_bus_type, &priv->mux_notifier_block);
 	platform_device_unregister(priv->mux_pdev);
 	gpiod_remove_lookup_table(priv->lookup);
 }
Signature · 1 line
2.43.2
Reviewed-by on v1 2/2
#6 lore raw
Quoted 7 lines · Heiner Kallweit

Once the gpio mux driver binds to the "i2c-mux-gpio" platform device, this creates the i2c adapters for the muxed child segments. We can use the bus notifier mechanism to check for creation of the child i2d adapters, and call i2c_register_spd() for them. This allows to detect all DIMM's on systems with more than 8 memory slots.

Yay, this looks *much* better. Thanks for doing it! Sadly, I can also only review, not test. Let's hope someone can step up to do the real testing. Maybe resend it as RFT to lkml?

#7 lore raw
Quoted 7 lines · Wolfram Sang
Quoted 6 lines · Heiner Kallweit

Once the gpio mux driver binds to the "i2c-mux-gpio" platform device, this creates the i2c adapters for the muxed child segments. We can use the bus notifier mechanism to check for creation of the child i2d adapters, and call i2c_register_spd() for them. This allows to detect all DIMM's on systems with more than 8 memory slots.

Quoted 8 lines
Quoted 1 line

Yay, this looks *much* better. Thanks for doing it! Sadly, I can also only review, not test. Let's hope someone can step up to do the real testing. Maybe resend it as RFT to lkml?

Ah yes, I'll do this.

#8 lore raw
In reply to Wolfram Sang
Quoted 7 lines · Wolfram Sang
Quoted 6 lines · Heiner Kallweit

Once the gpio mux driver binds to the "i2c-mux-gpio" platform device, this creates the i2c adapters for the muxed child segments. We can use the bus notifier mechanism to check for creation of the child i2d adapters, and call i2c_register_spd() for them. This allows to detect all DIMM's on systems with more than 8 memory slots.

Quoted 6 lines
Quoted 1 line

Yay, this looks *much* better. Thanks for doing it! Sadly, I can also only review, not test. Let's hope someone can step up to do the real testing. Maybe resend it as RFT to lkml?

3 weeks ago I sent the change as RFT to lkml, as suggested. However I didn't receive a single response yet.
Now that 6.9-rc1 is out, we would have several weeks in linux-next before 6.10-rc1. Would it be ok for you to apply the patch and see whether we get any feedback from linux-next users?
If yes, can you apply it as-is, or shall I resubmit the patch?

Quoted 2 lines
#9 lore raw
Quoted 4 lines

Now that 6.9-rc1 is out, we would have several weeks in linux-next before 6.10-rc1. Would it be ok for you to apply the patch and see whether we get any feedback from linux-next users?

Yes, we can do that.

Quoted 1 line

If yes, can you apply it as-is, or shall I resubmit the patch?

This is a question for Andi who picks up driver patches meanwhile.

#10 lore raw
Quoted 11 lines · Wolfram Sang
Quoted 4 lines

Now that 6.9-rc1 is out, we would have several weeks in linux-next before 6.10-rc1. Would it be ok for you to apply the patch and see whether we get any feedback from linux-next users?

Yes, we can do that.

Quoted 1 line

If yes, can you apply it as-is, or shall I resubmit the patch?

This is a question for Andi who picks up driver patches meanwhile.

I have to rebase the patch anyway, so I'll resubmit.