Skip to contentSkip to conversation

i3c: master: svc: some bug fixes

PATCH Superseded Clark Wang 2 participants 13 messages Jul 15, 2021 – Jul 16, 2021

Subsystems (2)

v1 of 5 Superseded Latest is v5 →

lore
Share & tools
lore

Clark Wang

author
#0 lore raw

cover i3c: master: svc: some bug fixes

Hi,

I am using SVC I3C module recently. I fix some problems and also have a question.

My question is:
Can I3C bus support pure I2C mode in kernel?
Or in other words, in mixed mode, must there be at least one I3C device on the I3C bus?

The pure I3C mode works fine. But when only have one I2C device on the I3C bus, the probe in function i3c_master_bus_init() will go error. Because there is no one on I3C bus can ACK the I3C message with I3C message speed. Then it will return error at function i3c_master_rstdaa_locked() because of no ACK for 0x7e start byte.
When I use the following dtb configuration, the above problem occurs.

&i3c2 {
	#address-cells = <3>;
	#size-cells = <0>;
	pinctrl-names = "default";
	pinctrl-0 = <&pinctrl_i3c2>;
	i2c-scl-hz = <400000>;
	status = "okay";

	lsm6dso_i2c: imu@6a {
		compatible = "st,lsm6dso";
		reg = <0x6a 0x0 0x50>;
	};
};

But I saw a similar configuration example in /home/nxf47749/work/kernel/i3c/Documentation/devicetree/bindings/i3c/snps,dw-i3c-master.txt. I wonder if that can work normally?

I know the definition in the specification is:
Mixed xxx Bus: I3C Bus topology with both I2C and I3C Devices present on the I3C Bus...
But I think it is feasible to use pure I2C mode with I3C module. I am not sure why the use of pure I2C mode is restricted in the software.

If there are errors in my ideas, please correct me in time. Thank you all.

Here are the fixes.

Clark Wang (4):
i3c: master: svc: move module reset behind clk enable
i3c: master: svc: fix atomic issue
i3c: master: svc: add support for slave to stop returning data
i3c: master: svc: set ODSTOP to let I2C device see the STOP signal

 drivers/i3c/master/svc-i3c-master.c | 45 +++++++++++++++++++----------
 1 file changed, 30 insertions(+), 15 deletions(-)
Signature · 1 line
2.25.1
author
#1 lore raw

PATCH 1/4 i3c: master: svc: move module reset behind clk enable

Reset I3C module will R/W its regs, so enable its clocks first.

---
 drivers/i3c/master/svc-i3c-master.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/i3c/master/svc-i3c-master.c +2 -2
index 879e5a64acaf..c25a372f6820 100644
@@ -1381,8 +1381,6 @@ static int svc_i3c_master_probe(struct platform_device *pdev)
 
 	master->dev = dev;
 
-	svc_i3c_master_reset(master);
-
 	ret = clk_prepare_enable(master->pclk);
 	if (ret)
 		return ret;
@@ -1419,6 +1417,8 @@ static int svc_i3c_master_probe(struct platform_device *pdev)
 
 	platform_set_drvdata(pdev, master);
 
+	svc_i3c_master_reset(master);
+
 	/* Register the master */
 	ret = i3c_master_register(&master->base, &pdev->dev,
 				  &svc_i3c_master_ops, false);
Signature · 1 line
2.25.1
author
#3 lore raw

PATCH 2/4 i3c: master: svc: fix atomic issue

In reply to Clark Wang

do_daa_locked() function is in a spin lock environment, use readl_poll_timeout_atomic() to replace the origin readl_poll_timeout().

---
 drivers/i3c/master/svc-i3c-master.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/i3c/master/svc-i3c-master.c +3 -3
index c25a372f6820..9d80435638ea 100644
@@ -656,7 +656,7 @@ static int svc_i3c_master_readb(struct svc_i3c_master *master, u8 *dst,
 	u32 reg;
 
 	for (i = 0; i < len; i++) {
-		ret = readl_poll_timeout(master->regs + SVC_I3C_MSTATUS, reg,
+		ret = readl_poll_timeout_atomic(master->regs + SVC_I3C_MSTATUS, reg,
 					 SVC_I3C_MSTATUS_RXPEND(reg), 0, 1000);
 		if (ret)
 			return ret;
@@ -687,7 +687,7 @@ static int svc_i3c_master_do_daa_locked(struct svc_i3c_master *master,
 		 * Either one slave will send its ID, or the assignment process
 		 * is done.
 		 */
-		ret = readl_poll_timeout(master->regs + SVC_I3C_MSTATUS, reg,
+		ret = readl_poll_timeout_atomic(master->regs + SVC_I3C_MSTATUS, reg,
 					 SVC_I3C_MSTATUS_RXPEND(reg) |
 					 SVC_I3C_MSTATUS_MCTRLDONE(reg),
 					 1, 1000);
@@ -744,7 +744,7 @@ static int svc_i3c_master_do_daa_locked(struct svc_i3c_master *master,
 		}
 
 		/* Wait for the slave to be ready to receive its address */
-		ret = readl_poll_timeout(master->regs + SVC_I3C_MSTATUS, reg,
+		ret = readl_poll_timeout_atomic(master->regs + SVC_I3C_MSTATUS, reg,
 					 SVC_I3C_MSTATUS_MCTRLDONE(reg) &&
 					 SVC_I3C_MSTATUS_STATE_DAA(reg) &&
 					 SVC_I3C_MSTATUS_BETWEEN(reg),
Signature · 1 line
2.25.1
maintainer Reviewed-by on v1 2/4
#4 lore raw

Re: [PATCH 2/4] i3c: master: svc: fix atomic issue

Hi Clark,

Clark Wang <[email protected]> wrote on Thu, 15 Jul 2021 16:24:11 +0800:

Quoted 14 lines

do_daa_locked() function is in a spin lock environment, use readl_poll_timeout_atomic() to replace the origin readl_poll_timeout().

---
 drivers/i3c/master/svc-i3c-master.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/i3c/master/svc-i3c-master.c +0 -0
index c25a372f6820..9d80435638ea 100644

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_readb(struct svc_i3c_master *master, u8 *dst, · line 660

Quoted 6 lines
drivers/i3c/master/svc-i3c-master.c +1 -1
@@ -656,7 +656,7 @@ static int svc_i3c_master_readb(struct svc_i3c_master *master, u8 *dst,
 	u32 reg;
 
 	for (i = 0; i < len; i++) {
-		ret = readl_poll_timeout(master->regs + SVC_I3C_MSTATUS, reg,
+		ret = readl_poll_timeout_atomic(master->regs + SVC_I3C_MSTATUS, reg,
 					 SVC_I3C_MSTATUS_RXPEND(reg), 0, 1000);

You forgot to align the parameters of the function here and below.

Otherwise,

Quoted 12 lines
drivers/i3c/master/svc-i3c-master.c +1 -1
 		if (ret)
 			return ret;
@@ -687,7 +687,7 @@ static int svc_i3c_master_do_daa_locked(struct svc_i3c_master *master,
 		 * Either one slave will send its ID, or the assignment process
 		 * is done.
 		 */
-		ret = readl_poll_timeout(master->regs + SVC_I3C_MSTATUS, reg,
+		ret = readl_poll_timeout_atomic(master->regs + SVC_I3C_MSTATUS, reg,
 					 SVC_I3C_MSTATUS_RXPEND(reg) |
 					 SVC_I3C_MSTATUS_MCTRLDONE(reg),
 					 1, 1000);

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_do_daa_locked(struct svc_i3c_master *master, · line 750

Quoted 8 lines
drivers/i3c/master/svc-i3c-master.c +1 -1
@@ -744,7 +744,7 @@ static int svc_i3c_master_do_daa_locked(struct svc_i3c_master *master,
 		}
 
 		/* Wait for the slave to be ready to receive its address */
-		ret = readl_poll_timeout(master->regs + SVC_I3C_MSTATUS, reg,
+		ret = readl_poll_timeout_atomic(master->regs + SVC_I3C_MSTATUS, reg,
 					 SVC_I3C_MSTATUS_MCTRLDONE(reg) &&
 					 SVC_I3C_MSTATUS_STATE_DAA(reg) &&
 					 SVC_I3C_MSTATUS_BETWEEN(reg),

Thanks,
Miquèl

author
#5 lore raw

RE: [PATCH 2/4] i3c: master: svc: fix atomic issue

Quoted 36 lines

-----Original Message-----
From: Miquel Raynal <[email protected]>
Sent: Friday, July 16, 2021 6:52
To: Clark Wang <[email protected]>
Cc: [email protected]; [email protected]; [email protected]; [email protected]; linux- [email protected]; [email protected]
Subject: Re: [PATCH 2/4] i3c: master: svc: fix atomic issue

Hi Clark,

Clark Wang <[email protected]> wrote on Thu, 15 Jul 2021 16:24:11 +0800:

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_readb(struct

Quoted 15 lines

do_daa_locked() function is in a spin lock environment, use readl_poll_timeout_atomic() to replace the origin readl_poll_timeout().

---
 drivers/i3c/master/svc-i3c-master.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/i3c/master/svc-i3c-master.c +0 -0
index c25a372f6820..9d80435638ea 100644
@@ -656,7 +656,7 @@ static int svc_i3c_master_readb(struct

svc_i3c_master *master, u8 *dst,

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_readb(struct svc_i3c_master *master, u8 *dst,

Quoted 4 lines
drivers/i3c/master/svc-i3c-master.c +0 -1
 	u32 reg;
 
 	for (i = 0; i < len; i++) {
-		ret = readl_poll_timeout(master->regs + SVC_I3C_MSTATUS,

reg,

Quoted 7 lines

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_readb(struct · line 659

Quoted 1 line
drivers/i3c/master/svc-i3c-master.c +1 -0
+		ret = readl_poll_timeout_atomic(master->regs +

SVC_I3C_MSTATUS,

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_readb(struct · line 661

Quoted 2 lines
drivers/i3c/master/svc-i3c-master.c +1 -0
+reg,
 					 SVC_I3C_MSTATUS_RXPEND(reg), 0,

1000);

You forgot to align the parameters of the function here and below.

Oh, sorry about this. I will send V2 to correct this.

By the way, can you help to check the question I mentioned in the cover letter?

Thank you very much!

Best Regards,
Clark Wang

Quoted 42 lines

Otherwise,

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_do_daa_locked(struct

Quoted 3 lines
drivers/i3c/master/svc-i3c-master.c +0 -0
 		if (ret)
 			return ret;
@@ -687,7 +687,7 @@ static int svc_i3c_master_do_daa_locked(struct

svc_i3c_master *master,

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_do_daa_locked(struct svc_i3c_master *master,

Quoted 4 lines
drivers/i3c/master/svc-i3c-master.c +0 -1
 		 * Either one slave will send its ID, or the assignment process
 		 * is done.
 		 */
-		ret = readl_poll_timeout(master->regs + SVC_I3C_MSTATUS,

reg,

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_do_daa_locked(struct · line 690

Quoted 1 line
drivers/i3c/master/svc-i3c-master.c +1 -0
+		ret = readl_poll_timeout_atomic(master->regs +

SVC_I3C_MSTATUS,

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_do_daa_locked(struct · line 693

Quoted 3 lines
drivers/i3c/master/svc-i3c-master.c +1 -0
+reg,
 					 SVC_I3C_MSTATUS_RXPEND(reg) |
 

SVC_I3C_MSTATUS_MCTRLDONE(reg),

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_do_daa_locked(struct

Quoted 2 lines
drivers/i3c/master/svc-i3c-master.c +0 -0
 					 1, 1000);
@@ -744,7 +744,7 @@ static int svc_i3c_master_do_daa_locked(struct

svc_i3c_master *master,

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_do_daa_locked(struct svc_i3c_master *master,

Quoted 4 lines
drivers/i3c/master/svc-i3c-master.c +0 -1
 		}
 
 		/* Wait for the slave to be ready to receive its address */
-		ret = readl_poll_timeout(master->regs + SVC_I3C_MSTATUS,

reg,

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_do_daa_locked(struct · line 747

Quoted 1 line
drivers/i3c/master/svc-i3c-master.c +1 -0
+		ret = readl_poll_timeout_atomic(master->regs +

SVC_I3C_MSTATUS,

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_do_daa_locked(struct · line 749

Quoted 2 lines
drivers/i3c/master/svc-i3c-master.c +1 -0
+reg,
 

SVC_I3C_MSTATUS_MCTRLDONE(reg) &&

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_do_daa_locked(struct · line 750

Quoted 1 line
drivers/i3c/master/svc-i3c-master.c +0 -0
 					 SVC_I3C_MSTATUS_STATE_DAA(reg)

&&

Quoted 1 line

SVC_I3C_MSTATUS_BETWEEN(reg),

Thanks,
Miquèl

author
#6 lore raw

PATCH 3/4 i3c: master: svc: add support for slave to stop returning data

In reply to Clark Wang

When i3c controller reads data from slave device, slave device can stop returning data with an ACK after any byte.
Add this support for svc i3c controller. Otherwise, it will go TIMEOUT error path when the slave device ends the read operation early.

---
 drivers/i3c/master/svc-i3c-master.c | 28 ++++++++++++++++++++--------
 1 file changed, 20 insertions(+), 8 deletions(-)
drivers/i3c/master/svc-i3c-master.c +20 -8
index 9d80435638ea..892e57fec4b0 100644
@@ -865,7 +865,7 @@ static int svc_i3c_master_read(struct svc_i3c_master *master,
 			       u8 *in, unsigned int len)
 {
 	int offset = 0, i, ret;
-	u32 mdctrl;
+	u32 mdctrl, mstatus;
 
 	while (offset < len) {
 		unsigned int count;
@@ -874,8 +874,15 @@ static int svc_i3c_master_read(struct svc_i3c_master *master,
 					 mdctrl,
 					 !(mdctrl & SVC_I3C_MDATACTRL_RXEMPTY),
 					 0, 1000);
-		if (ret)
-			return ret;
+		if (ret) {
+			ret = readl_poll_timeout(master->regs + SVC_I3C_MSTATUS,
+				 mstatus, SVC_I3C_MSTATUS_COMPLETE(mstatus),
+				 0, 1000);
+			if (ret)
+				return ret;
+			else
+				return offset;
+		}
 
 		count = SVC_I3C_MDATACTRL_RXCOUNT(mdctrl);
 		for (i = 0; i < count; i++)
@@ -884,7 +891,7 @@ static int svc_i3c_master_read(struct svc_i3c_master *master,
 		offset += count;
 	}
 
-	return 0;
+	return offset;
 }
 
 static int svc_i3c_master_write(struct svc_i3c_master *master,
@@ -917,7 +924,7 @@ static int svc_i3c_master_write(struct svc_i3c_master *master,
 static int svc_i3c_master_xfer(struct svc_i3c_master *master,
 			       bool rnw, unsigned int xfer_type, u8 addr,
 			       u8 *in, const u8 *out, unsigned int xfer_len,
-			       unsigned int read_len, bool continued)
+			       unsigned int *read_len, bool continued)
 {
 	u32 reg;
 	int ret;
@@ -927,7 +934,7 @@ static int svc_i3c_master_xfer(struct svc_i3c_master *master,
 	       SVC_I3C_MCTRL_IBIRESP_NACK |
 	       SVC_I3C_MCTRL_DIR(rnw) |
 	       SVC_I3C_MCTRL_ADDR(addr) |
-	       SVC_I3C_MCTRL_RDTERM(read_len),
+	       SVC_I3C_MCTRL_RDTERM(*read_len),
 	       master->regs + SVC_I3C_MCTRL);
 
 	ret = readl_poll_timeout(master->regs + SVC_I3C_MSTATUS, reg,
@@ -939,8 +946,10 @@ static int svc_i3c_master_xfer(struct svc_i3c_master *master,
 		ret = svc_i3c_master_read(master, in, xfer_len);
 	else
 		ret = svc_i3c_master_write(master, out, xfer_len);
-	if (ret)
+	if (ret < 0)
 		goto emit_stop;
+	if (rnw)
+		*read_len = ret;
 
 	ret = readl_poll_timeout(master->regs + SVC_I3C_MSTATUS, reg,
 				 SVC_I3C_MSTATUS_COMPLETE(reg), 0, 1000);
@@ -1012,7 +1021,7 @@ static void svc_i3c_master_start_xfer_locked(struct svc_i3c_master *master)
 
 		ret = svc_i3c_master_xfer(master, cmd->rnw, xfer->type,
 					  cmd->addr, cmd->in, cmd->out,
-					  cmd->len, cmd->read_len,
+					  cmd->len, &cmd->read_len,
 					  cmd->continued);
 		if (ret)
 			break;
@@ -1141,6 +1150,9 @@ static int svc_i3c_master_send_direct_ccc_cmd(struct svc_i3c_master *master,
 	if (!wait_for_completion_timeout(&xfer->comp, msecs_to_jiffies(1000)))
 		svc_i3c_master_dequeue_xfer(master, xfer);
 
+	if (cmd->read_len != xfer_len)
+		ccc->dests[0].payload.len = cmd->read_len;
+
 	ret = xfer->ret;
 	svc_i3c_master_free_xfer(xfer);
 
Signature · 1 line
2.25.1
maintainer
#7 lore raw

Re: [PATCH 3/4] i3c: master: svc: add support for slave to stop returning data

Hi Clark,

Clark Wang <[email protected]> wrote on Thu, 15 Jul 2021 16:24:12 +0800:

Quoted 4 lines

When i3c controller reads data from slave device, slave device can stop returning data with an ACK after any byte.
Add this support for svc i3c controller. Otherwise, it will go TIMEOUT error path when the slave device ends the read operation early.

Is this part of the I3C specification? I am not aware about it.

Quoted 85 lines
---
 drivers/i3c/master/svc-i3c-master.c | 28 ++++++++++++++++++++--------
 1 file changed, 20 insertions(+), 8 deletions(-)
drivers/i3c/master/svc-i3c-master.c +17 -8
index 9d80435638ea..892e57fec4b0 100644
@@ -865,7 +865,7 @@ static int svc_i3c_master_read(struct svc_i3c_master *master,
 			       u8 *in, unsigned int len)
 {
 	int offset = 0, i, ret;
-	u32 mdctrl;
+	u32 mdctrl, mstatus;
 
 	while (offset < len) {
 		unsigned int count;
@@ -874,8 +874,15 @@ static int svc_i3c_master_read(struct svc_i3c_master *master,
 					 mdctrl,
 					 !(mdctrl & SVC_I3C_MDATACTRL_RXEMPTY),
 					 0, 1000);
-		if (ret)
-			return ret;
+		if (ret) {
+			ret = readl_poll_timeout(master->regs + SVC_I3C_MSTATUS,
+				 mstatus, SVC_I3C_MSTATUS_COMPLETE(mstatus),
+				 0, 1000);
+			if (ret)
+				return ret;
+			else
+				return offset;
+		}
 
 		count = SVC_I3C_MDATACTRL_RXCOUNT(mdctrl);
 		for (i = 0; i < count; i++)
@@ -884,7 +891,7 @@ static int svc_i3c_master_read(struct svc_i3c_master *master,
 		offset += count;
 	}
 
-	return 0;
+	return offset;
 }
 
 static int svc_i3c_master_write(struct svc_i3c_master *master,
@@ -917,7 +924,7 @@ static int svc_i3c_master_write(struct svc_i3c_master *master,
 static int svc_i3c_master_xfer(struct svc_i3c_master *master,
 			       bool rnw, unsigned int xfer_type, u8 addr,
 			       u8 *in, const u8 *out, unsigned int xfer_len,
-			       unsigned int read_len, bool continued)
+			       unsigned int *read_len, bool continued)
 {
 	u32 reg;
 	int ret;
@@ -927,7 +934,7 @@ static int svc_i3c_master_xfer(struct svc_i3c_master *master,
 	       SVC_I3C_MCTRL_IBIRESP_NACK |
 	       SVC_I3C_MCTRL_DIR(rnw) |
 	       SVC_I3C_MCTRL_ADDR(addr) |
-	       SVC_I3C_MCTRL_RDTERM(read_len),
+	       SVC_I3C_MCTRL_RDTERM(*read_len),
 	       master->regs + SVC_I3C_MCTRL);
 
 	ret = readl_poll_timeout(master->regs + SVC_I3C_MSTATUS, reg,
@@ -939,8 +946,10 @@ static int svc_i3c_master_xfer(struct svc_i3c_master *master,
 		ret = svc_i3c_master_read(master, in, xfer_len);
 	else
 		ret = svc_i3c_master_write(master, out, xfer_len);
-	if (ret)
+	if (ret < 0)
 		goto emit_stop;
+	if (rnw)
+		*read_len = ret;
 
 	ret = readl_poll_timeout(master->regs + SVC_I3C_MSTATUS, reg,
 				 SVC_I3C_MSTATUS_COMPLETE(reg), 0, 1000);
@@ -1012,7 +1021,7 @@ static void svc_i3c_master_start_xfer_locked(struct svc_i3c_master *master)
 
 		ret = svc_i3c_master_xfer(master, cmd->rnw, xfer->type,
 					  cmd->addr, cmd->in, cmd->out,
-					  cmd->len, cmd->read_len,
+					  cmd->len, &cmd->read_len,
 					  cmd->continued);
 		if (ret)
 			break;

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_send_direct_ccc_cmd(struct svc_i3c_master *master, · line 1158

Quoted 9 lines
drivers/i3c/master/svc-i3c-master.c +3 -0
@@ -1141,6 +1150,9 @@ static int svc_i3c_master_send_direct_ccc_cmd(struct svc_i3c_master *master,
 	if (!wait_for_completion_timeout(&xfer->comp, msecs_to_jiffies(1000)))
 		svc_i3c_master_dequeue_xfer(master, xfer);
 
+	if (cmd->read_len != xfer_len)
+		ccc->dests[0].payload.len = cmd->read_len;
+
 	ret = xfer->ret;
 	svc_i3c_master_free_xfer(xfer);
 

Thanks,
Miquèl

author
#8 lore raw

RE: [PATCH 3/4] i3c: master: svc: add support for slave to stop returning data

Hi Miquel,

Thanks for your review!

Quoted 15 lines

-----Original Message-----
From: Miquel Raynal <[email protected]>
Sent: Friday, July 16, 2021 6:56
To: Clark Wang <[email protected]>
Cc: [email protected]; [email protected]; [email protected]; [email protected]; linux- [email protected]; [email protected]
Subject: Re: [PATCH 3/4] i3c: master: svc: add support for slave to stop returning data

Hi Clark,

Clark Wang <[email protected]> wrote on Thu, 15 Jul 2021 16:24:12 +0800:

Quoted 6 lines
Quoted 4 lines

When i3c controller reads data from slave device, slave device can stop returning data with an ACK after any byte.
Add this support for svc i3c controller. Otherwise, it will go TIMEOUT error path when the slave device ends the read operation early.

Is this part of the I3C specification? I am not aware about it.

Yes. I think the spec mentions about this. You can check section "5.1.2.3 I3C SDR Data Words" of <mipi_i3c_basic_specification>. [3. Ninth Bit of SDR Slave Returned (Read) Data as End-of-Data: In I2C, the ninth Data bit from
Slave to Master is an ACK by the Master. By contrast, in I3C this bit allows the Slave to end a
Read, and allows the Master to Abort a Read. In SDR terms, the ninth bit of Read data is referred
to as the T-Bit (for ‘Transition’) (see Section 5.1.2.3.4).]

Best Regards,
Clark Wang

Quoted 112 lines

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_read(struct

Quoted 12 lines
---
 drivers/i3c/master/svc-i3c-master.c | 28 ++++++++++++++++++++--------
 1 file changed, 20 insertions(+), 8 deletions(-)
drivers/i3c/master/svc-i3c-master.c +0 -0
index 9d80435638ea..892e57fec4b0 100644
@@ -865,7 +865,7 @@ static int svc_i3c_master_read(struct

svc_i3c_master *master,

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_read(struct

Quoted 9 lines
drivers/i3c/master/svc-i3c-master.c +1 -1
 			       u8 *in, unsigned int len)
 {
 	int offset = 0, i, ret;
-	u32 mdctrl;
+	u32 mdctrl, mstatus;
 
 	while (offset < len) {
 		unsigned int count;
@@ -874,8 +874,15 @@ static int svc_i3c_master_read(struct

svc_i3c_master *master,

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_read(struct · line 875

Quoted 2 lines
drivers/i3c/master/svc-i3c-master.c +0 -0
 					 mdctrl,
 					 !(mdctrl &

SVC_I3C_MDATACTRL_RXEMPTY),

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_read(struct svc_i3c_master *master,

Quoted 5 lines
drivers/i3c/master/svc-i3c-master.c +2 -2
 					 0, 1000);
-		if (ret)
-			return ret;
+		if (ret) {
+			ret = readl_poll_timeout(master->regs +

SVC_I3C_MSTATUS,

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_read(struct · line 879

Quoted 1 line
drivers/i3c/master/svc-i3c-master.c +1 -0
+				 mstatus,

SVC_I3C_MSTATUS_COMPLETE(mstatus),

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_read(struct

Quoted 10 lines
drivers/i3c/master/svc-i3c-master.c +6 -0
+				 0, 1000);
+			if (ret)
+				return ret;
+			else
+				return offset;
+		}
 
 		count = SVC_I3C_MDATACTRL_RXCOUNT(mdctrl);
 		for (i = 0; i < count; i++)
@@ -884,7 +891,7 @@ static int svc_i3c_master_read(struct

svc_i3c_master *master,

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_xfer(struct svc_i3c_master

Quoted 18 lines
drivers/i3c/master/svc-i3c-master.c +1 -2
 		offset += count;
 	}
 
-	return 0;
+	return offset;
 }
 
 static int svc_i3c_master_write(struct svc_i3c_master *master, @@
-917,7 +924,7 @@ static int svc_i3c_master_write(struct svc_i3c_master

*master, static int svc_i3c_master_xfer(struct svc_i3c_master *master,

drivers/i3c/master/svc-i3c-master.c +1 -1
 			       bool rnw, unsigned int xfer_type, u8 addr,
 			       u8 *in, const u8 *out, unsigned int xfer_len,
-			       unsigned int read_len, bool continued)
+			       unsigned int *read_len, bool continued)
 {
 	u32 reg;
 	int ret;
@@ -927,7 +934,7 @@ static int svc_i3c_master_xfer(struct svc_i3c_master

*master,

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_xfer(struct svc_i3c_master *master, · line 940

Quoted 8 lines
drivers/i3c/master/svc-i3c-master.c +1 -1
 	       SVC_I3C_MCTRL_IBIRESP_NACK |
 	       SVC_I3C_MCTRL_DIR(rnw) |
 	       SVC_I3C_MCTRL_ADDR(addr) |
-	       SVC_I3C_MCTRL_RDTERM(read_len),
+	       SVC_I3C_MCTRL_RDTERM(*read_len),
 	       master->regs + SVC_I3C_MCTRL);
 
 	ret = readl_poll_timeout(master->regs + SVC_I3C_MSTATUS, reg,

@@

Quoted 1 line

-939,8 +946,10 @@ static int svc_i3c_master_xfer(struct svc_i3c_master

*master,

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_xfer(struct svc_i3c_master *master,

Quoted 11 lines
drivers/i3c/master/svc-i3c-master.c +3 -1
 		ret = svc_i3c_master_read(master, in, xfer_len);
 	else
 		ret = svc_i3c_master_write(master, out, xfer_len);
-	if (ret)
+	if (ret < 0)
 		goto emit_stop;
+	if (rnw)
+		*read_len = ret;
 
 	ret = readl_poll_timeout(master->regs + SVC_I3C_MSTATUS, reg,
 				 SVC_I3C_MSTATUS_COMPLETE(reg), 0,

1000); @@ -1012,7 +1021,7 @@

drivers/i3c/master/svc-i3c-master.c static int

Quoted 11 lines

static void svc_i3c_master_start_xfer_locked(struct svc_i3c_master *master)

drivers/i3c/master/svc-i3c-master.c +1 -1
 		ret = svc_i3c_master_xfer(master, cmd->rnw, xfer->type,
 					  cmd->addr, cmd->in, cmd->out,
-					  cmd->len, cmd->read_len,
+					  cmd->len, &cmd->read_len,
 					  cmd->continued);
 		if (ret)
 			break;
@@ -1141,6 +1150,9 @@ static int

svc_i3c_master_send_direct_ccc_cmd(struct svc_i3c_master *master,

drivers/i3c/master/svc-i3c-master.c static int · line 1150

Quoted 1 line
drivers/i3c/master/svc-i3c-master.c +0 -0
 	if (!wait_for_completion_timeout(&xfer->comp,

msecs_to_jiffies(1000)))

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_send_direct_ccc_cmd(struct svc_i3c_master *master, · line 1158

Quoted 8 lines
drivers/i3c/master/svc-i3c-master.c +3 -0
 		svc_i3c_master_dequeue_xfer(master, xfer);
 
+	if (cmd->read_len != xfer_len)
+		ccc->dests[0].payload.len = cmd->read_len;
+
 	ret = xfer->ret;
 	svc_i3c_master_free_xfer(xfer);
 

Thanks,
Miquèl

author
#9 lore raw

PATCH 4/4 i3c: master: svc: set ODSTOP to let I2C device see the STOP signal

In reply to Clark Wang

If using I2C/I3C mixed mode, need to set ODSTOP. Otherwise, the I2C devices cannot see the stop signal. It may cause message sending errors.

---
 drivers/i3c/master/svc-i3c-master.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)
drivers/i3c/master/svc-i3c-master.c +5 -2
index 892e57fec4b0..b05cc7f521e6 100644
@@ -449,7 +449,7 @@ static int svc_i3c_master_bus_init(struct i3c_master_controller *m)
 	struct i3c_device_info info = {};
 	unsigned long fclk_rate, fclk_period_ns;
 	unsigned int high_period_ns, od_low_period_ns;
-	u32 ppbaud, pplow, odhpp, odbaud, i2cbaud, reg;
+	u32 ppbaud, pplow, odhpp, odbaud, odstop, i2cbaud, reg;
 	int ret;
 
 	/* Timings derivation */
@@ -479,6 +479,7 @@ static int svc_i3c_master_bus_init(struct i3c_master_controller *m)
 	switch (bus->mode) {
 	case I3C_BUS_MODE_PURE:
 		i2cbaud = 0;
+		odstop = 0;
 		break;
 	case I3C_BUS_MODE_MIXED_FAST:
 	case I3C_BUS_MODE_MIXED_LIMITED:
@@ -487,6 +488,7 @@ static int svc_i3c_master_bus_init(struct i3c_master_controller *m)
 		 * between the high and low period does not really matter.
 		 */
 		i2cbaud = DIV_ROUND_UP(1000, od_low_period_ns) - 2;
+		odstop = 1;
 		break;
 	case I3C_BUS_MODE_MIXED_SLOW:
 		/*
@@ -494,6 +496,7 @@ static int svc_i3c_master_bus_init(struct i3c_master_controller *m)
 		 * constraints as the FM+ mode.
 		 */
 		i2cbaud = DIV_ROUND_UP(2500, od_low_period_ns) - 2;
+		odstop = 1;
 		break;
 	default:
 		return -EINVAL;
@@ -502,7 +505,7 @@ static int svc_i3c_master_bus_init(struct i3c_master_controller *m)
 	reg = SVC_I3C_MCONFIG_MASTER_EN |
 	      SVC_I3C_MCONFIG_DISTO(0) |
 	      SVC_I3C_MCONFIG_HKEEP(0) |
-	      SVC_I3C_MCONFIG_ODSTOP(0) |
+	      SVC_I3C_MCONFIG_ODSTOP(odstop) |
 	      SVC_I3C_MCONFIG_PPBAUD(ppbaud) |
 	      SVC_I3C_MCONFIG_PPLOW(pplow) |
 	      SVC_I3C_MCONFIG_ODBAUD(odbaud) |
Signature · 7 lines
2.25.1


-- 
linux-i3c mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-i3c
maintainer Reviewed-by on v1 4/4
#10 lore raw

Re: [PATCH 4/4] i3c: master: svc: set ODSTOP to let I2C device see the STOP signal

Hi Clark,

Clark Wang <[email protected]> wrote on Thu, 15 Jul 2021 16:24:13 +0800:

Quoted 2 lines

If using I2C/I3C mixed mode, need to set ODSTOP. Otherwise, the I2C devices cannot see the stop signal. It may cause message sending errors.

Quoted 44 lines
---
 drivers/i3c/master/svc-i3c-master.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)
drivers/i3c/master/svc-i3c-master.c +4 -1
index 892e57fec4b0..b05cc7f521e6 100644
@@ -449,7 +449,7 @@ static int svc_i3c_master_bus_init(struct i3c_master_controller *m)
 	struct i3c_device_info info = {};
 	unsigned long fclk_rate, fclk_period_ns;
 	unsigned int high_period_ns, od_low_period_ns;
-	u32 ppbaud, pplow, odhpp, odbaud, i2cbaud, reg;
+	u32 ppbaud, pplow, odhpp, odbaud, odstop, i2cbaud, reg;
 	int ret;
 
 	/* Timings derivation */
@@ -479,6 +479,7 @@ static int svc_i3c_master_bus_init(struct i3c_master_controller *m)
 	switch (bus->mode) {
 	case I3C_BUS_MODE_PURE:
 		i2cbaud = 0;
+		odstop = 0;
 		break;
 	case I3C_BUS_MODE_MIXED_FAST:
 	case I3C_BUS_MODE_MIXED_LIMITED:
@@ -487,6 +488,7 @@ static int svc_i3c_master_bus_init(struct i3c_master_controller *m)
 		 * between the high and low period does not really matter.
 		 */
 		i2cbaud = DIV_ROUND_UP(1000, od_low_period_ns) - 2;
+		odstop = 1;
 		break;
 	case I3C_BUS_MODE_MIXED_SLOW:
 		/*
@@ -494,6 +496,7 @@ static int svc_i3c_master_bus_init(struct i3c_master_controller *m)
 		 * constraints as the FM+ mode.
 		 */
 		i2cbaud = DIV_ROUND_UP(2500, od_low_period_ns) - 2;
+		odstop = 1;
 		break;
 	default:
 		return -EINVAL;

drivers/i3c/master/svc-i3c-master.c static int svc_i3c_master_bus_init(struct i3c_master_controller *m) · line 511

Quoted 8 lines
drivers/i3c/master/svc-i3c-master.c +1 -1
@@ -502,7 +505,7 @@ static int svc_i3c_master_bus_init(struct i3c_master_controller *m)
 	reg = SVC_I3C_MCONFIG_MASTER_EN |
 	      SVC_I3C_MCONFIG_DISTO(0) |
 	      SVC_I3C_MCONFIG_HKEEP(0) |
-	      SVC_I3C_MCONFIG_ODSTOP(0) |
+	      SVC_I3C_MCONFIG_ODSTOP(odstop) |
 	      SVC_I3C_MCONFIG_PPBAUD(ppbaud) |
 	      SVC_I3C_MCONFIG_PPLOW(pplow) |
 	      SVC_I3C_MCONFIG_ODBAUD(odbaud) |

Thanks,
Miquèl

maintainer
#11 lore raw
In reply to Clark Wang

Hi Clark,

Clark Wang <[email protected]> wrote on Thu, 15 Jul 2021 16:24:09 +0800:

Quoted 34 lines

Hi,

I am using SVC I3C module recently. I fix some problems and also have a question.

My question is:
Can I3C bus support pure I2C mode in kernel?
Or in other words, in mixed mode, must there be at least one I3C device on the I3C bus?

The pure I3C mode works fine. But when only have one I2C device on the I3C bus, the probe in function i3c_master_bus_init() will go error. Because there is no one on I3C bus can ACK the I3C message with I3C message speed. Then it will return error at function i3c_master_rstdaa_locked() because of no ACK for 0x7e start byte.
When I use the following dtb configuration, the above problem occurs.

&i3c2 {
	#address-cells = <3>;
	#size-cells = <0>;
	pinctrl-names = "default";
	pinctrl-0 = <&pinctrl_i3c2>;
	i2c-scl-hz = <400000>;
	status = "okay";

	lsm6dso_i2c: imu@6a {
		compatible = "st,lsm6dso";
		reg = <0x6a 0x0 0x50>;
	};
};

But I saw a similar configuration example in /home/nxf47749/work/kernel/i3c/Documentation/devicetree/bindings/i3c/snps,dw-i3c-master.txt. I wonder if that can work normally?

Quoted 7 lines

I know the definition in the specification is:
Mixed xxx Bus: I3C Bus topology with both I2C and I3C Devices present on the I3C Bus...
But I think it is feasible to use pure I2C mode with I3C module. I am not sure why the use of pure I2C mode is restricted in the software.

If there are errors in my ideas, please correct me in time. Thank you all.

As you pointed out, I am not aware of a specific I2C only bus setting but if you find a way to workaround the issue raised above by software in a rather clean way, then... why not?

Quoted 11 lines

Here are the fixes.

Clark Wang (4):
i3c: master: svc: move module reset behind clk enable
i3c: master: svc: fix atomic issue
i3c: master: svc: add support for slave to stop returning data
i3c: master: svc: set ODSTOP to let I2C device see the STOP signal

 drivers/i3c/master/svc-i3c-master.c | 45 +++++++++++++++++++----------
 1 file changed, 30 insertions(+), 15 deletions(-)

Thanks,
Miquèl

author
#12 lore raw
Quoted 59 lines

-----Original Message-----
From: Miquel Raynal <[email protected]>
Sent: Friday, July 16, 2021 15:52
To: Clark Wang <[email protected]>
Cc: [email protected]; [email protected]; [email protected]; [email protected]; linux- [email protected]; [email protected]
Subject: Re: [PATCH 0/4] i3c: master: svc: some bug fixes

Hi Clark,

Clark Wang <[email protected]> wrote on Thu, 15 Jul 2021 16:24:09 +0800:

Quoted 5 lines

Hi,

I am using SVC I3C module recently. I fix some problems and also have a question.

Quoted 8 lines

My question is:
Can I3C bus support pure I2C mode in kernel?
Or in other words, in mixed mode, must there be at least one I3C device on the I3C bus?

The pure I3C mode works fine. But when only have one I2C device on the I3C bus, the probe in function i3c_master_bus_init() will go error. Because there is no one on I3C bus can ACK the I3C message with I3C message

speed. Then

Quoted 1 line

it will return error at function i3c_master_rstdaa_locked() because of no

ACK

Quoted 16 lines

for 0x7e start byte.
When I use the following dtb configuration, the above problem occurs.

&i3c2 {
	#address-cells = <3>;
	#size-cells = <0>;
	pinctrl-names = "default";
	pinctrl-0 = <&pinctrl_i3c2>;
	i2c-scl-hz = <400000>;
	status = "okay";

	lsm6dso_i2c: imu@6a {
		compatible = "st,lsm6dso";
		reg = <0x6a 0x0 0x50>;
	};
};
Quoted 2 lines

But I saw a similar configuration example in

/home/nxf47749/work/kernel/i3c/Documentation/devicetree/bindings/i3c/ snps,dw-i3c-master.txt.

Quoted 8 lines

I wonder if that can work normally?

I know the definition in the specification is:
Mixed xxx Bus: I3C Bus topology with both I2C and I3C Devices present on the I3C Bus...
But I think it is feasible to use pure I2C mode with I3C module. I am not sure why the use of pure I2C mode is restricted in the software.

Quoted 5 lines
Quoted 1 line

If there are errors in my ideas, please correct me in time. Thank you all.

As you pointed out, I am not aware of a specific I2C only bus setting but if you find a way to workaround the issue raised above by software in a rather clean way, then... why not?

Hi Miquel,

Ok. At present, I just commented part of the code for registering i3c for pure I2C mode testing. I will try to make a clean revision afterwards. Thank you very much for your suggestion!

Best Regards,
Clark Wang

Quoted 15 lines
Quoted 12 lines

Here are the fixes.

Clark Wang (4):
i3c: master: svc: move module reset behind clk enable
i3c: master: svc: fix atomic issue
i3c: master: svc: add support for slave to stop returning data
i3c: master: svc: set ODSTOP to let I2C device see the STOP signal

 drivers/i3c/master/svc-i3c-master.c | 45 +++++++++++++++++++----------
 1 file changed, 30 insertions(+), 15 deletions(-)

Thanks,
Miquèl