Skip to contentSkip to conversation

i2c: npcm: read/write operation, checkpatch

PATCH Superseded [email protected] 3 participants 18 messages Sep 20, 2024 – Sep 27, 2024

Subsystems (2)

v4 of 8 Superseded Latest is v8 →

lore
Share & tools
lore
#0 lore raw

v4 cover i2c: npcm: read/write operation, checkpatch

From: Tyrone Ting <[email protected]>

This patchset includes the following fixes:

- Enable the target functionality in the interrupt handling routine when the i2c transfer is about to finish.
- Correct the read/write operation procedure.
- Introduce a software flag to handle the bus error (BER) condition which is not caused by the i2c transfer.
- Modify timeout calculation.
- Assign the client address earlier logically.
- Use an i2c frequency table for the frequency parameters assignment.
- Coding style fix.

The NPCM I2C driver is tested on NPCM750 and NPCM845 evaluation boards.

Addressed comments from:
- Andy Shevchenko : https://lore.kernel.org/linux-i2c/ZuQUfCK7d7DuR9eH
@smile.fi.intel.com/
- Andy Shevchenko : https://lore.kernel.org/linux-i2c/ZuQU2iIZwW4mAumo
@smile.fi.intel.com/
- Andy Shevchenko : https://lore.kernel.org/linux-i2c/ZuQWYabomv_xyYyj
@smile.fi.intel.com/

Changes since version 3:
- Remove "Bug fixes" in the cover letter title.
- Modify the term "function" to "func()" in the commit message and code comments.
- Correct the coding style.

Changes since version 2:
- Add more explanations in the commit message and code modification.
- Use lower character names for declarations.
- Remove Fixes tags in commits which are not to fix bugs.

Changes since version 1:
- Restore the npcm_i2caddr array length to fix the smatch warning.
- Remove unused variables.
- Handle the condition where scl_table_cnt reaches to the maximum value.
- Fix the checkpatch warning.

Charles Boyer (1):
i2c: npcm: Enable slave in eob interrupt

Tyrone Ting (5):
i2c: npcm: correct the read/write operation procedure
i2c: npcm: use a software flag to indicate a BER condition
i2c: npcm: Modify timeout evaluation mechanism
i2c: npcm: Modify the client address assignment
i2c: npcm: use i2c frequency table

 drivers/i2c/busses/i2c-npcm7xx.c | 425 +++++++++++++++++++++++--------
 1 file changed, 323 insertions(+), 102 deletions(-)

base-commit: f56f4ba2fc1dbefd3242946f2fad35338a60e3bc

Signature · 1 line
2.34.1
#1 lore raw

PATCH v4 1/6 i2c: npcm: correct the read/write operation procedure

From: Tyrone Ting <[email protected]>

From: Tyrone Ting <[email protected]>

Originally the driver uses the XMIT bit in SMBnST register to decide the upcoming i2c transaction. If XMIT bit is 1, then it will be an i2c write operation. If it's 0, then a read operation will be executed.

In slave mode the XMIT bit can simply be used directly to set the state. XMIT bit can be used as an indication to the current state of the state machine during slave operation. (meaning XMIT = 1 during writing and XMIT = 0 during reading).

In master operation XMIT is valid only if there are no bus errors. For example: in a multi master where the same module is switching from master to slave at runtime, and there are collisions, the XMIT bit cannot be trusted.

However the maser already "knows" what the bus state is, so this bit is not needed and the driver can just track what it is currently doing.

---
 drivers/i2c/busses/i2c-npcm7xx.c | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)
drivers/i2c/busses/i2c-npcm7xx.c +2 -5
index bbcb4d6668ce..2b76dbfba438 100644
@@ -1628,13 +1628,10 @@ static void npcm_i2c_irq_handle_sda(struct npcm_i2c *bus, u8 i2cst)
 			npcm_i2c_wr_byte(bus, bus->dest_addr | BIT(0));
 	/* SDA interrupt, after start\restart */
 	} else {
-		if (NPCM_I2CST_XMIT & i2cst) {
-			bus->operation = I2C_WRITE_OPER;
+		if (bus->operation == I2C_WRITE_OPER)
 			npcm_i2c_irq_master_handler_write(bus);
-		} else {
-			bus->operation = I2C_READ_OPER;
+		else if (bus->operation == I2C_READ_OPER)
 			npcm_i2c_irq_master_handler_read(bus);
-		}
 	}
 }
 
Signature · 1 line
2.34.1
#2 lore raw

Re: [PATCH v4 1/6] i2c: npcm: correct the read/write operation procedure

Quoted 3 lines · On Fri, Sep 20, 2024 at 06:18:15PM +0800, [email protected] wrote:

From: Tyrone Ting <[email protected]>

From: Tyrone Ting <[email protected]>

Something wrong with your email setup. This patch may not be applied in this form.

But before (re-)sending, wait for others to comment (up to 1 week).

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#3 lore raw

Re: [PATCH v4 1/6] i2c: npcm: correct the read/write operation procedure

Hi Andy:

Thank you for your feedback.

Andy Shevchenko <[email protected]> 於 2024年9月20日 週五 下午10:30寫道:

Quoted 10 lines · On Fri, Sep 20, 2024 at 06:18:15PM +0800, [email protected] wrote:
Quoted 3 lines

From: Tyrone Ting <[email protected]>

From: Tyrone Ting <[email protected]>

Something wrong with your email setup. This patch may not be applied in this form.

But before (re-)sending, wait for others to comment (up to 1 week).

I'll remove the "--from" parameter to git send-email command in next patch set.

Quoted 5 lines

--
With Best Regards,
Andy Shevchenko

Thank you.

Regards,
Tyrone

maintainer
#4 lore raw

Re: [PATCH v4 1/6] i2c: npcm: correct the read/write operation procedure

In reply to [email protected]

Hi Tyrone,

Quoted 3 lines · On Fri, Sep 20, 2024 at 06:18:15PM GMT, [email protected] wrote:

From: Tyrone Ting <[email protected]>

From: Tyrone Ting <[email protected]>

no worries, I can take care of this.

Quoted 14 lines

Originally the driver uses the XMIT bit in SMBnST register to decide the upcoming i2c transaction. If XMIT bit is 1, then it will be an i2c write operation. If it's 0, then a read operation will be executed.

In slave mode the XMIT bit can simply be used directly to set the state. XMIT bit can be used as an indication to the current state of the state machine during slave operation. (meaning XMIT = 1 during writing and XMIT = 0 during reading).

In master operation XMIT is valid only if there are no bus errors. For example: in a multi master where the same module is switching from master to slave at runtime, and there are collisions, the XMIT bit cannot be trusted.

Quoted 5 lines

However the maser already "knows" what the bus state is, so this bit is not needed and the driver can just track what it is currently doing.

This patch is independent from the rest of the series, can I start takin this in and unburden you from this?

Andi

#5 lore raw

Re: [PATCH v4 1/6] i2c: npcm: correct the read/write operation procedure

Hi Andi:

Thank you for your help.

Andi Shyti <[email protected]> 於 2024年9月26日 週四 上午4:19寫道:

Quoted 24 lines

Hi Tyrone,

Quoted 3 lines · On Fri, Sep 20, 2024 at 06:18:15PM GMT, [email protected] wrote:

From: Tyrone Ting <[email protected]>

From: Tyrone Ting <[email protected]>

no worries, I can take care of this.

Quoted 14 lines

Originally the driver uses the XMIT bit in SMBnST register to decide the upcoming i2c transaction. If XMIT bit is 1, then it will be an i2c write operation. If it's 0, then a read operation will be executed.

In slave mode the XMIT bit can simply be used directly to set the state. XMIT bit can be used as an indication to the current state of the state machine during slave operation. (meaning XMIT = 1 during writing and XMIT = 0 during reading).

In master operation XMIT is valid only if there are no bus errors. For example: in a multi master where the same module is switching from master to slave at runtime, and there are collisions, the XMIT bit cannot be trusted.

Quoted 9 lines
Quoted 5 lines

However the maser already "knows" what the bus state is, so this bit is not needed and the driver can just track what it is currently doing.

This patch is independent from the rest of the series, can I start takin this in and unburden you from this?

Sure, thank you again.

Quoted 1 line

Andi

Regards,
Tyrone

#6 lore raw

PATCH v4 2/6 i2c: npcm: use a software flag to indicate a BER condition

In reply to [email protected]

From: Tyrone Ting <[email protected]>

From: Tyrone Ting <[email protected]>

If not clearing the BB (bus busy) condition in the BER (bus error) interrupt, the driver causes a timeout and hence the i2c core doesn't do the i2c transfer retry but returns the driver's return value to the upper layer instead.

Clear the BB condition in the BER interrupt and a software flag is used. The driver does an i2c recovery without causing the timeout if the flag is set.

---
 drivers/i2c/busses/i2c-npcm7xx.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
drivers/i2c/busses/i2c-npcm7xx.c +5 -1
index 2b76dbfba438..2d034503d8bc 100644
@@ -334,6 +334,7 @@ struct npcm_i2c {
 	u64 nack_cnt;
 	u64 timeout_cnt;
 	u64 tx_complete_cnt;
+	bool ber_state;
 };
 
 static inline void npcm_i2c_select_bank(struct npcm_i2c *bus,
@@ -1521,6 +1522,7 @@ static void npcm_i2c_irq_handle_ber(struct npcm_i2c *bus)
 	if (npcm_i2c_is_master(bus)) {
 		npcm_i2c_master_abort(bus);
 	} else {
+		bus->ber_state = true;
 		npcm_i2c_clear_master_status(bus);
 
 		/* Clear BB (BUS BUSY) bit */
@@ -1699,6 +1701,7 @@ static int npcm_i2c_recovery_tgclk(struct i2c_adapter *_adap)
 		dev_dbg(bus->dev, "bus%d-0x%x recovery skipped, bus not stuck",
 			bus->num, bus->dest_addr);
 		npcm_i2c_reset(bus);
+		bus->ber_state = false;
 		return 0;
 	}
 
@@ -1763,6 +1766,7 @@ static int npcm_i2c_recovery_tgclk(struct i2c_adapter *_adap)
 		if (bus->rec_succ_cnt < ULLONG_MAX)
 			bus->rec_succ_cnt++;
 	}
+	bus->ber_state = false;
 	return status;
 }
 
@@ -2158,7 +2162,7 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
 
 	} while (time_is_after_jiffies(time_left) && bus_busy);
 
-	if (bus_busy) {
+	if (bus_busy || bus->ber_state) {
 		iowrite8(NPCM_I2CCST_BB, bus->reg + NPCM_I2CCST);
 		npcm_i2c_reset(bus);
 		i2c_recover_bus(adap);
Signature · 1 line
2.34.1
maintainer
#7 lore raw

Re: [PATCH v4 2/6] i2c: npcm: use a software flag to indicate a BER condition

Hi Tyrone,

patch looks good, but...

Quoted 24 lines · On Fri, Sep 20, 2024 at 06:18:16PM GMT, [email protected] wrote:

From: Tyrone Ting <[email protected]>

From: Tyrone Ting <[email protected]>

If not clearing the BB (bus busy) condition in the BER (bus error) interrupt, the driver causes a timeout and hence the i2c core doesn't do the i2c transfer retry but returns the driver's return value to the upper layer instead.

Clear the BB condition in the BER interrupt and a software flag is used. The driver does an i2c recovery without causing the timeout if the flag is set.

---
 drivers/i2c/busses/i2c-npcm7xx.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
drivers/i2c/busses/i2c-npcm7xx.c +0 -0
index 2b76dbfba438..2d034503d8bc 100644

drivers/i2c/busses/i2c-npcm7xx.c struct npcm_i2c { · line 337

Quoted 4 lines
drivers/i2c/busses/i2c-npcm7xx.c +1 -0
@@ -334,6 +334,7 @@ struct npcm_i2c {
 	u64 nack_cnt;
 	u64 timeout_cnt;
 	u64 tx_complete_cnt;
+	bool ber_state;

I need some comment here, what is this ber state? You described it in the commit log, but people won't be browsing much of the commit log rather than the code itself.

You could perhaps mention the meaning as:

bool ber_state; /* brief description */

and...

Quoted 28 lines
drivers/i2c/busses/i2c-npcm7xx.c +3 -0
 };
 
 static inline void npcm_i2c_select_bank(struct npcm_i2c *bus,
@@ -1521,6 +1522,7 @@ static void npcm_i2c_irq_handle_ber(struct npcm_i2c *bus)
 	if (npcm_i2c_is_master(bus)) {
 		npcm_i2c_master_abort(bus);
 	} else {
+		bus->ber_state = true;
 		npcm_i2c_clear_master_status(bus);
 
 		/* Clear BB (BUS BUSY) bit */
@@ -1699,6 +1701,7 @@ static int npcm_i2c_recovery_tgclk(struct i2c_adapter *_adap)
 		dev_dbg(bus->dev, "bus%d-0x%x recovery skipped, bus not stuck",
 			bus->num, bus->dest_addr);
 		npcm_i2c_reset(bus);
+		bus->ber_state = false;
 		return 0;
 	}
 
@@ -1763,6 +1766,7 @@ static int npcm_i2c_recovery_tgclk(struct i2c_adapter *_adap)
 		if (bus->rec_succ_cnt < ULLONG_MAX)
 			bus->rec_succ_cnt++;
 	}
+	bus->ber_state = false;
 	return status;
 }
 

drivers/i2c/busses/i2c-npcm7xx.c static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs, · line 2165

Quoted 5 lines
drivers/i2c/busses/i2c-npcm7xx.c +1 -1
@@ -2158,7 +2162,7 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
 
 	} while (time_is_after_jiffies(time_left) && bus_busy);
 
-	if (bus_busy) {
+	if (bus_busy || bus->ber_state) {

/*
* Check the BER (bus error) state, if it's true means that blah
* blah blah, while when it's false blah blah blah and we should
* or should not do blah blah blah
*/

if (bus_busy || bus->ber_state) {
	...
}

Thanks,
Andi

Quoted 6 lines
drivers/i2c/busses/i2c-npcm7xx.c +0 -0
 		iowrite8(NPCM_I2CCST_BB, bus->reg + NPCM_I2CCST);
 		npcm_i2c_reset(bus);
 		i2c_recover_bus(adap);
Signature · 1 line
2.34.1
#8 lore raw

Re: [PATCH v4 2/6] i2c: npcm: use a software flag to indicate a BER condition

Hi Andi:

Thank you for your feedback.

Andi Shyti <[email protected]> 於 2024年9月26日 週四 上午4:22寫道:

Quoted 39 lines

Hi Tyrone,

patch looks good, but...

Quoted 24 lines · On Fri, Sep 20, 2024 at 06:18:16PM GMT, [email protected] wrote:

From: Tyrone Ting <[email protected]>

From: Tyrone Ting <[email protected]>

If not clearing the BB (bus busy) condition in the BER (bus error) interrupt, the driver causes a timeout and hence the i2c core doesn't do the i2c transfer retry but returns the driver's return value to the upper layer instead.

Clear the BB condition in the BER interrupt and a software flag is used. The driver does an i2c recovery without causing the timeout if the flag is set.

---
 drivers/i2c/busses/i2c-npcm7xx.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)
drivers/i2c/busses/i2c-npcm7xx.c +0 -0
index 2b76dbfba438..2d034503d8bc 100644

drivers/i2c/busses/i2c-npcm7xx.c struct npcm_i2c { · line 337

Quoted 4 lines
drivers/i2c/busses/i2c-npcm7xx.c +1 -0
@@ -334,6 +334,7 @@ struct npcm_i2c {
      u64 nack_cnt;
      u64 timeout_cnt;
      u64 tx_complete_cnt;
+     bool ber_state;

I need some comment here, what is this ber state? You described it in the commit log, but people won't be browsing much of the commit log rather than the code itself.

Quoted 4 lines

You could perhaps mention the meaning as:

bool ber_state; /* brief description */

Understood. I'll add some descriptions.

Quoted 41 lines

and...

Quoted 28 lines
drivers/i2c/busses/i2c-npcm7xx.c +3 -0
 };
 
 static inline void npcm_i2c_select_bank(struct npcm_i2c *bus,
@@ -1521,6 +1522,7 @@ static void npcm_i2c_irq_handle_ber(struct npcm_i2c *bus)
      if (npcm_i2c_is_master(bus)) {
              npcm_i2c_master_abort(bus);
      } else {
+             bus->ber_state = true;
              npcm_i2c_clear_master_status(bus);
 
              /* Clear BB (BUS BUSY) bit */
@@ -1699,6 +1701,7 @@ static int npcm_i2c_recovery_tgclk(struct i2c_adapter *_adap)
              dev_dbg(bus->dev, "bus%d-0x%x recovery skipped, bus not stuck",
                      bus->num, bus->dest_addr);
              npcm_i2c_reset(bus);
+             bus->ber_state = false;
              return 0;
      }
 
@@ -1763,6 +1766,7 @@ static int npcm_i2c_recovery_tgclk(struct i2c_adapter *_adap)
              if (bus->rec_succ_cnt < ULLONG_MAX)
                      bus->rec_succ_cnt++;
      }
+     bus->ber_state = false;
      return status;
 }
 

drivers/i2c/busses/i2c-npcm7xx.c static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs, · line 2165

Quoted 5 lines
drivers/i2c/busses/i2c-npcm7xx.c +1 -1
@@ -2158,7 +2162,7 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
 
      } while (time_is_after_jiffies(time_left) && bus_busy);
 
-     if (bus_busy) {
+     if (bus_busy || bus->ber_state) {

/*
* Check the BER (bus error) state, if it's true means that blah
* blah blah, while when it's false blah blah blah and we should
* or should not do blah blah blah
*/

Quoted 4 lines
if (bus_busy || bus->ber_state) {
        ...
}

Understood. Your comments will be addressed.

Quoted 9 lines

Thanks,
Andi

Quoted 6 lines
drivers/i2c/busses/i2c-npcm7xx.c +0 -1
              iowrite8(NPCM_I2CCST_BB, bus->reg + NPCM_I2CCST);
              npcm_i2c_reset(bus);
              i2c_recover_bus(adap);
--

2.34.1

Thank you.

Regards,
Tyrone

#9 lore raw

PATCH v4 3/6 i2c: npcm: Modify timeout evaluation mechanism

In reply to [email protected]

From: Tyrone Ting <[email protected]>

From: Tyrone Ting <[email protected]>

The users want to connect a lot of masters on the same bus. This timeout is used to determine the time it takes to take bus ownership. The transactions are very long, so waiting 35ms is not enough.

Increase the timeout and treat it as the total timeout, including retries. The total timeout is 2 seconds now.

The i2c core layer will have chances to retry to call the i2c driver transfer function if the i2c driver reports that the bus is busy and returns EAGAIN.

---
 drivers/i2c/busses/i2c-npcm7xx.c | 19 ++++++++++---------
 1 file changed, 10 insertions(+), 9 deletions(-)
drivers/i2c/busses/i2c-npcm7xx.c +10 -9
index 2d034503d8bc..68f3d47323ab 100644
@@ -2132,19 +2132,12 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
 		}
 	}
 
-	/*
-	 * Adaptive TimeOut: estimated time in usec + 100% margin:
-	 * 2: double the timeout for clock stretching case
-	 * 9: bits per transaction (including the ack/nack)
-	 */
-	timeout_usec = (2 * 9 * USEC_PER_SEC / bus->bus_freq) * (2 + nread + nwrite);
-	timeout = max_t(unsigned long, bus->adap.timeout, usecs_to_jiffies(timeout_usec));
 	if (nwrite >= 32 * 1024 || nread >= 32 * 1024) {
 		dev_err(bus->dev, "i2c%d buffer too big\n", bus->num);
 		return -EINVAL;
 	}
 
-	time_left = jiffies + timeout + 1;
+	time_left = jiffies + bus->adap.timeout / bus->adap.retries + 1;
 	do {
 		/*
 		 * we must clear slave address immediately when the bus is not
@@ -2183,6 +2176,14 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
 	if (npcm_i2c_master_start_xmit(bus, slave_addr, nwrite, nread,
 				       write_data, read_data, read_PEC,
 				       read_block)) {
+		/*
+		 * Adaptive TimeOut: estimated time in usec + 100% margin:
+		 * 2: double the timeout for clock stretching case
+		 * 9: bits per transaction (including the ack/nack)
+		 */
+		timeout_usec = (2 * 9 * USEC_PER_SEC / bus->bus_freq) * (2 + nread + nwrite);
+		timeout = max_t(unsigned long, bus->adap.timeout / bus->adap.retries,
+				usecs_to_jiffies(timeout_usec));
 		time_left = wait_for_completion_timeout(&bus->cmd_complete,
 							timeout);
 
@@ -2308,7 +2309,7 @@ static int npcm_i2c_probe_bus(struct platform_device *pdev)
 	adap = &bus->adap;
 	adap->owner = THIS_MODULE;
 	adap->retries = 3;
-	adap->timeout = msecs_to_jiffies(35);
+	adap->timeout = 2 * HZ;
 	adap->algo = &npcm_i2c_algo;
 	adap->quirks = &npcm_i2c_quirks;
 	adap->algo_data = bus;
Signature · 1 line
2.34.1
maintainer
#10 lore raw

Re: [PATCH v4 3/6] i2c: npcm: Modify timeout evaluation mechanism

Hi Tyrone,

...

Quoted 41 lines
drivers/i2c/busses/i2c-npcm7xx.c +9 -8
index 2d034503d8bc..68f3d47323ab 100644
@@ -2132,19 +2132,12 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
 		}
 	}
 
-	/*
-	 * Adaptive TimeOut: estimated time in usec + 100% margin:
-	 * 2: double the timeout for clock stretching case
-	 * 9: bits per transaction (including the ack/nack)
-	 */
-	timeout_usec = (2 * 9 * USEC_PER_SEC / bus->bus_freq) * (2 + nread + nwrite);
-	timeout = max_t(unsigned long, bus->adap.timeout, usecs_to_jiffies(timeout_usec));
 	if (nwrite >= 32 * 1024 || nread >= 32 * 1024) {
 		dev_err(bus->dev, "i2c%d buffer too big\n", bus->num);
 		return -EINVAL;
 	}
 
-	time_left = jiffies + timeout + 1;
+	time_left = jiffies + bus->adap.timeout / bus->adap.retries + 1;
 	do {
 		/*
 		 * we must clear slave address immediately when the bus is not
@@ -2183,6 +2176,14 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
 	if (npcm_i2c_master_start_xmit(bus, slave_addr, nwrite, nread,
 				       write_data, read_data, read_PEC,
 				       read_block)) {
+		/*
+		 * Adaptive TimeOut: estimated time in usec + 100% margin:
+		 * 2: double the timeout for clock stretching case
+		 * 9: bits per transaction (including the ack/nack)
+		 */
+		timeout_usec = (2 * 9 * USEC_PER_SEC / bus->bus_freq) * (2 + nread + nwrite);
+		timeout = max_t(unsigned long, bus->adap.timeout / bus->adap.retries,
+				usecs_to_jiffies(timeout_usec));
 		time_left = wait_for_completion_timeout(&bus->cmd_complete,
 							timeout);
 

drivers/i2c/busses/i2c-npcm7xx.c static int npcm_i2c_probe_bus(struct platform_device *pdev) · line 2312

Quoted 5 lines
drivers/i2c/busses/i2c-npcm7xx.c +1 -1
@@ -2308,7 +2309,7 @@ static int npcm_i2c_probe_bus(struct platform_device *pdev)
 	adap = &bus->adap;
 	adap->owner = THIS_MODULE;
 	adap->retries = 3;
-	adap->timeout = msecs_to_jiffies(35);
+	adap->timeout = 2 * HZ;

Same here, I need some good description of why timeout is set to 2 seconds. If the datasheet says 35ms, I do not exclude that someone in the future will send a patch saying "we don't need to wait 2 seconds, wait 35ms as per datasheet".

Thanks,
Andi

Quoted 6 lines
drivers/i2c/busses/i2c-npcm7xx.c +0 -0
 	adap->algo = &npcm_i2c_algo;
 	adap->quirks = &npcm_i2c_quirks;
 	adap->algo_data = bus;
Signature · 1 line
2.34.1
#11 lore raw

Re: [PATCH v4 3/6] i2c: npcm: Modify timeout evaluation mechanism

Hi Andi:

Thank you for your feedback.

Your comments will be addressed.

Andi Shyti <[email protected]> 於 2024年9月26日 週四 上午4:24寫道:

Quoted 57 lines

Hi Tyrone,

...

Quoted 41 lines
drivers/i2c/busses/i2c-npcm7xx.c +9 -8
index 2d034503d8bc..68f3d47323ab 100644
@@ -2132,19 +2132,12 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
              }
      }
 
-     /*
-      * Adaptive TimeOut: estimated time in usec + 100% margin:
-      * 2: double the timeout for clock stretching case
-      * 9: bits per transaction (including the ack/nack)
-      */
-     timeout_usec = (2 * 9 * USEC_PER_SEC / bus->bus_freq) * (2 + nread + nwrite);
-     timeout = max_t(unsigned long, bus->adap.timeout, usecs_to_jiffies(timeout_usec));
      if (nwrite >= 32 * 1024 || nread >= 32 * 1024) {
              dev_err(bus->dev, "i2c%d buffer too big\n", bus->num);
              return -EINVAL;
      }
 
-     time_left = jiffies + timeout + 1;
+     time_left = jiffies + bus->adap.timeout / bus->adap.retries + 1;
      do {
              /*
               * we must clear slave address immediately when the bus is not
@@ -2183,6 +2176,14 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
      if (npcm_i2c_master_start_xmit(bus, slave_addr, nwrite, nread,
                                     write_data, read_data, read_PEC,
                                     read_block)) {
+             /*
+              * Adaptive TimeOut: estimated time in usec + 100% margin:
+              * 2: double the timeout for clock stretching case
+              * 9: bits per transaction (including the ack/nack)
+              */
+             timeout_usec = (2 * 9 * USEC_PER_SEC / bus->bus_freq) * (2 + nread + nwrite);
+             timeout = max_t(unsigned long, bus->adap.timeout / bus->adap.retries,
+                             usecs_to_jiffies(timeout_usec));
              time_left = wait_for_completion_timeout(&bus->cmd_complete,
                                                      timeout);
 

drivers/i2c/busses/i2c-npcm7xx.c static int npcm_i2c_probe_bus(struct platform_device *pdev) · line 2312

Quoted 5 lines
drivers/i2c/busses/i2c-npcm7xx.c +1 -1
@@ -2308,7 +2309,7 @@ static int npcm_i2c_probe_bus(struct platform_device *pdev)
      adap = &bus->adap;
      adap->owner = THIS_MODULE;
      adap->retries = 3;
-     adap->timeout = msecs_to_jiffies(35);
+     adap->timeout = 2 * HZ;

Same here, I need some good description of why timeout is set to 2 seconds. If the datasheet says 35ms, I do not exclude that someone in the future will send a patch saying "we don't need to wait 2 seconds, wait 35ms as per datasheet".

Quoted 9 lines

Thanks,
Andi

Quoted 6 lines
drivers/i2c/busses/i2c-npcm7xx.c +0 -1
      adap->algo = &npcm_i2c_algo;
      adap->quirks = &npcm_i2c_quirks;
      adap->algo_data = bus;
--

2.34.1

Have a nice day.

Regards,
Tyrone

#12 lore raw

PATCH v4 4/6 i2c: npcm: Modify the client address assignment

In reply to [email protected]

From: Tyrone Ting <[email protected]>

From: Tyrone Ting <[email protected]>

Store the client address earlier since it might get called in the i2c_recover_bus logic flow at the early stage of the func() npcm_i2c_master_xfer.

---
 drivers/i2c/busses/i2c-npcm7xx.c | 13 ++++++++++++-
 1 file changed, 12 insertions(+), 1 deletion(-)
drivers/i2c/busses/i2c-npcm7xx.c +12 -1
index 68f3d47323ab..248835220c86 100644
@@ -2155,6 +2155,18 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
 
 	} while (time_is_after_jiffies(time_left) && bus_busy);
 
+	/*
+	 * Previously, the address was stored w/o left-shift by one bit and
+	 * with that shift in the following call to npcm_i2c_master_start_xmit.
+	 *
+	 * Since there are cases that the i2c_recover_bus gets called at the
+	 * early stage of the func() npcm_i2c_master_xfer, the address is
+	 * stored with the shift and used in the i2c_recover_bus call.
+	 *
+	 * The address is stored from bit 1 to bit 7 in the register for
+	 * sending the i2c address later so it's left-shifted by 1 bit.
+	 */
+	bus->dest_addr = slave_addr << 1;
 	if (bus_busy || bus->ber_state) {
 		iowrite8(NPCM_I2CCST_BB, bus->reg + NPCM_I2CCST);
 		npcm_i2c_reset(bus);
@@ -2163,7 +2175,6 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
 	}
 
 	npcm_i2c_init_params(bus);
-	bus->dest_addr = slave_addr;
 	bus->msgs = msgs;
 	bus->msgs_num = num;
 	bus->cmd_err = 0;
Signature · 1 line
2.34.1
#13 lore raw

Re: [PATCH v4 4/6] i2c: npcm: Modify the client address assignment

Quoted 3 lines · On Fri, Sep 20, 2024 at 06:18:18PM +0800, [email protected] wrote:

From: Tyrone Ting <[email protected]>

From: Tyrone Ting <[email protected]>

It seems all your mails have an issue.

Quoted 3 lines

Store the client address earlier since it might get called in the i2c_recover_bus logic flow at the early stage of the func() npcm_i2c_master_xfer.

You got my comment really wrong.

func() in my example was to refer to _a_ function mentioned in the text. And IIRC I even posted the example, like: i2c_recover_bus().

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#14 lore raw

Re: [PATCH v4 4/6] i2c: npcm: Modify the client address assignment

Hi Andy:

Thank you for your feedback.

Andy Shevchenko <[email protected]> 於 2024年9月20日 週五 下午10:33寫道:

Quoted 8 lines · On Fri, Sep 20, 2024 at 06:18:18PM +0800, [email protected] wrote:
Quoted 3 lines

From: Tyrone Ting <[email protected]>

From: Tyrone Ting <[email protected]>

It seems all your mails have an issue.

Acked in the previous email.

Quoted 9 lines
Quoted 3 lines

Store the client address earlier since it might get called in the i2c_recover_bus logic flow at the early stage of the func() npcm_i2c_master_xfer.

You got my comment really wrong.

func() in my example was to refer to _a_ function mentioned in the text. And IIRC I even posted the example, like: i2c_recover_bus().

So the commit message in next patch set would be like:
Store the client address earlier since it might get called in the i2c_recover_bus() logic flow at the early stage of npcm_i2c_master_xfer().

The code comment would be like:
Previously, the address was stored w/o left-shift by one bit and with that shift in the following call to npcm_i2c_master_start_xmit().

Since there are cases that the i2c_recover_bus() gets called at the early stage of npcm_i2c_master_xfer(), the address is stored with the shift and used in the i2c_recover_bus call().

Quoted 5 lines

--
With Best Regards,
Andy Shevchenko

Have a nice day.

Regards,
Tyrone

#15 lore raw

Re: [PATCH v4 4/6] i2c: npcm: Modify the client address assignment

Quoted 2 lines · Tyrone Ting

Andy Shevchenko <[email protected]> 於 2024年9月20日 週五 下午10:33寫道:

Quoted 1 line · On Fri, Sep 20, 2024 at 06:18:18PM +0800, [email protected] wrote:

...

Quoted 10 lines
Quoted 8 lines
Quoted 3 lines

Store the client address earlier since it might get called in the i2c_recover_bus logic flow at the early stage of the func() npcm_i2c_master_xfer.

You got my comment really wrong.

func() in my example was to refer to _a_ function mentioned in the text. And IIRC I even posted the example, like: i2c_recover_bus().

So the commit message in next patch set would be like:

Yes.

Quoted 5 lines

Store the client address earlier since it might get called in the i2c_recover_bus() logic flow at the early stage of npcm_i2c_master_xfer().

The code comment would be like:

Yes!

Quoted 6 lines

Previously, the address was stored w/o left-shift by one bit and with that shift in the following call to npcm_i2c_master_start_xmit().

Since there are cases that the i2c_recover_bus() gets called at the early stage of npcm_i2c_master_xfer(), the address is stored with the shift and used in the i2c_recover_bus call().

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#16 lore raw

PATCH v4 5/6 i2c: npcm: use i2c frequency table

In reply to [email protected]

From: Tyrone Ting <[email protected]>

From: Tyrone Ting <[email protected]>

Modify i2c frequency from table parameters
for NPCM i2c modules.

Supported frequencies are:

1. 100KHz
2. 400KHz
3. 1MHz

The original equations were tested on a variety of chips and base clocks. Since we added devices that use higher frequencies of the module we saw that there is a mismatch between the equation and the actual results on the bus itself, measured on scope.

Meanwhile, the equations were not accurate to begin with. They are an approximation of the ideal value. The ideal value is calculated per frequency of the core module.

So instead of using the equations we did an optimization per module frequency, verified on a device.

Most of the work was focused on the rise time of the SCL and SDA, which depends on external load of the bus and PU.

Different PCB designs, or specifically to this case: the number and type of targets on the bus, impact the required values for the timing registers.

Users can recalculate the numbers for each bus and get an even better optimization, but our users chose not to.

We manually picked values per frequency that match the entire valid range of targets (from 1 to max number). Then we check against the AMR described in SMB spec and make sure that none of the values is exceeding.

This process was led by the chip architect and included a lot of testing.

---
 drivers/i2c/busses/i2c-npcm7xx.c | 374 ++++++++++++++++++++++++-------
 1 file changed, 288 insertions(+), 86 deletions(-)
drivers/i2c/busses/i2c-npcm7xx.c +288 -86
index 248835220c86..9d533873e477 100644
@@ -263,6 +263,265 @@ static const int npcm_i2caddr[I2C_NUM_OWN_ADDR] = {
 #define I2C_FREQ_MIN_HZ			10000
 #define I2C_FREQ_MAX_HZ			I2C_MAX_FAST_MODE_PLUS_FREQ
 
+struct smb_timing_t {
+	u32 core_clk;
+	u8 hldt;
+	u8 dbcnt;
+	u16 sclfrq;
+	u8 scllt;
+	u8 sclht;
+	bool fast_mode;
+};
+
+static struct smb_timing_t smb_timing_100khz[] = {
+	{
+		.core_clk = 100000000, .hldt = 0x2A, .dbcnt = 0x4,
+		.sclfrq = 0xFB, .scllt = 0x0, .sclht = 0x0,
+		.fast_mode = false,
+	},
+	{
+		.core_clk = 62500000, .hldt = 0x2A, .dbcnt = 0x1,
+		.sclfrq = 0x9D, .scllt = 0x0, .sclht = 0x0,
+		.fast_mode = false,
+	},
+	{
+		.core_clk = 50000000, .hldt = 0x2A, .dbcnt = 0x1,
+		.sclfrq = 0x7E, .scllt = 0x0, .sclht = 0x0,
+		.fast_mode = false,
+	},
+	{
+		.core_clk = 48000000, .hldt = 0x2A, .dbcnt = 0x1,
+		.sclfrq = 0x79, .scllt = 0x0, .sclht = 0x0,
+		.fast_mode = false,
+	},
+	{
+		.core_clk = 40000000, .hldt = 0x2A, .dbcnt = 0x1,
+		.sclfrq = 0x65, .scllt = 0x0, .sclht = 0x0,
+		.fast_mode = false,
+	},
+	{
+		.core_clk = 30000000, .hldt = 0x2A, .dbcnt = 0x1,
+		.sclfrq = 0x4C, .scllt = 0x0, .sclht = 0x0,
+		.fast_mode = false,
+	},
+	{
+		.core_clk = 29000000, .hldt = 0x2A, .dbcnt = 0x1,
+		.sclfrq = 0x49, .scllt = 0x0, .sclht = 0x0,
+		.fast_mode = false,
+	},
+	{
+		.core_clk = 26000000, .hldt = 0x2A, .dbcnt = 0x1,
+		.sclfrq = 0x42, .scllt = 0x0, .sclht = 0x0,
+		.fast_mode = false,
+	},
+	{
+		.core_clk = 25000000, .hldt = 0x2A, .dbcnt = 0x1,
+		.sclfrq = 0x3F, .scllt = 0x0, .sclht = 0x0,
+		.fast_mode = false,
+	},
+	{
+		.core_clk = 24000000, .hldt = 0x2A, .dbcnt = 0x1,
+		.sclfrq = 0x3D, .scllt = 0x0, .sclht = 0x0,
+		.fast_mode = false,
+	},
+	{
+		.core_clk = 20000000, .hldt = 0x2A, .dbcnt = 0x1,
+		.sclfrq = 0x33, .scllt = 0x0, .sclht = 0x0,
+		.fast_mode = false,
+	},
+	{
+		.core_clk = 16180000, .hldt = 0x2A, .dbcnt = 0x1,
+		.sclfrq = 0x29, .scllt = 0x0, .sclht = 0x0,
+		.fast_mode = false,
+	},
+	{
+		.core_clk = 15000000, .hldt = 0x23, .dbcnt = 0x1,
+		.sclfrq = 0x26, .scllt = 0x0, .sclht = 0x0,
+		.fast_mode = false,
+	},
+	{
+		.core_clk = 13000000, .hldt = 0x1D, .dbcnt = 0x1,
+		.sclfrq = 0x21, .scllt = 0x0, .sclht = 0x0,
+		.fast_mode = false,
+	},
+	{
+		.core_clk = 12000000, .hldt = 0x1B, .dbcnt = 0x1,
+		.sclfrq = 0x1F, .scllt = 0x0, .sclht = 0x0,
+		.fast_mode = false,
+	},
+	{
+		.core_clk = 10000000, .hldt = 0x18, .dbcnt = 0x1,
+		.sclfrq = 0x1A, .scllt = 0x0, .sclht = 0x0,
+		.fast_mode = false,
+	},
+	{
+		.core_clk = 9000000, .hldt = 0x16, .dbcnt = 0x1,
+		.sclfrq = 0x17, .scllt = 0x0, .sclht = 0x0,
+		.fast_mode = false,
+	},
+	{
+		.core_clk = 8090000, .hldt = 0x14, .dbcnt = 0x1,
+		.sclfrq = 0x15, .scllt = 0x0, .sclht = 0x0,
+		.fast_mode = false,
+	},
+	{
+		.core_clk = 7500000, .hldt = 0x7, .dbcnt = 0x1,
+		.sclfrq = 0x13, .scllt = 0x0, .sclht = 0x0,
+		.fast_mode = false,
+	},
+	{
+		.core_clk = 6500000, .hldt = 0xE, .dbcnt = 0x1,
+		.sclfrq = 0x11, .scllt = 0x0, .sclht = 0x0,
+		.fast_mode = false,
+	},
+	{
+		.core_clk = 4000000, .hldt = 0x9, .dbcnt = 0x1,
+		.sclfrq = 0xB, .scllt = 0x0, .sclht = 0x0,
+		.fast_mode = false,
+	},
+};
+
+static struct smb_timing_t smb_timing_400khz[] = {
+	{
+		.core_clk = 100000000, .hldt = 0x2A, .dbcnt = 0x3,
+		.sclfrq = 0x0, .scllt = 0x47, .sclht = 0x35,
+		.fast_mode = true,
+	},
+	{
+		.core_clk = 62500000, .hldt = 0x2A, .dbcnt = 0x2,
+		.sclfrq = 0x0, .scllt = 0x2C, .sclht = 0x22,
+		.fast_mode = true,
+	},
+	{
+		.core_clk = 50000000, .hldt = 0x21, .dbcnt = 0x1,
+		.sclfrq = 0x0, .scllt = 0x24, .sclht = 0x1B,
+		.fast_mode = true,
+	},
+	{
+		.core_clk = 48000000, .hldt = 0x1E, .dbcnt = 0x1,
+		.sclfrq = 0x0, .scllt = 0x24, .sclht = 0x19,
+		.fast_mode = true,
+	},
+	{
+		.core_clk = 40000000, .hldt = 0x1B, .dbcnt = 0x1,
+		.sclfrq = 0x0, .scllt = 0x1E, .sclht = 0x14,
+		.fast_mode = true,
+	},
+	{
+		.core_clk = 33000000, .hldt = 0x15, .dbcnt = 0x1,
+		.sclfrq = 0x0, .scllt = 0x19, .sclht = 0x11,
+		.fast_mode = true,
+	},
+	{
+		.core_clk = 30000000, .hldt = 0x15, .dbcnt = 0x1,
+		.sclfrq = 0x0, .scllt = 0x19, .sclht = 0xD,
+		.fast_mode = true,
+	},
+	{
+		.core_clk = 29000000, .hldt = 0x11, .dbcnt = 0x1,
+		.sclfrq = 0x0, .scllt = 0x15, .sclht = 0x10,
+		.fast_mode = true,
+	},
+	{
+		.core_clk = 26000000, .hldt = 0x10, .dbcnt = 0x1,
+		.sclfrq = 0x0, .scllt = 0x13, .sclht = 0xE,
+		.fast_mode = true,
+	},
+	{
+		.core_clk = 25000000, .hldt = 0xF, .dbcnt = 0x1,
+		.sclfrq = 0x0, .scllt = 0x13, .sclht = 0xD,
+		.fast_mode = true,
+	},
+	{
+		.core_clk = 24000000, .hldt = 0xD, .dbcnt = 0x1,
+		.sclfrq = 0x0, .scllt = 0x12, .sclht = 0xD,
+		.fast_mode = true,
+	},
+	{
+		.core_clk = 20000000, .hldt = 0xB, .dbcnt = 0x1,
+		.sclfrq = 0x0, .scllt = 0xF, .sclht = 0xA,
+		.fast_mode = true,
+	},
+	{
+		.core_clk = 16180000, .hldt = 0xA, .dbcnt = 0x1,
+		.sclfrq = 0x0, .scllt = 0xC, .sclht = 0x9,
+		.fast_mode = true,
+	},
+	{
+		.core_clk = 15000000, .hldt = 0x9, .dbcnt = 0x1,
+		.sclfrq = 0x0, .scllt = 0xB, .sclht = 0x8,
+		.fast_mode = true,
+	},
+	{
+		.core_clk = 13000000, .hldt = 0x7, .dbcnt = 0x1,
+		.sclfrq = 0x0, .scllt = 0xA, .sclht = 0x7,
+		.fast_mode = true,
+	},
+	{
+		.core_clk = 12000000, .hldt = 0x7, .dbcnt = 0x1,
+		.sclfrq = 0x0, .scllt = 0xA, .sclht = 0x6,
+		.fast_mode = true,
+	},
+	{
+		.core_clk = 10000000, .hldt = 0x6, .dbcnt = 0x1,
+		.sclfrq = 0x0, .scllt = 0x8, .sclht = 0x5,
+		.fast_mode = true,
+	},
+};
+
+static struct smb_timing_t smb_timing_1000khz[] = {
+	{
+		.core_clk = 100000000, .hldt = 0x15, .dbcnt = 0x4,
+		.sclfrq = 0x0, .scllt = 0x1C, .sclht = 0x15,
+		.fast_mode = true,
+	},
+	{
+		.core_clk = 62500000, .hldt = 0xF, .dbcnt = 0x3,
+		.sclfrq = 0x0, .scllt = 0x11, .sclht = 0xE,
+		.fast_mode = true,
+	},
+	{
+		.core_clk = 50000000, .hldt = 0xA, .dbcnt = 0x2,
+		.sclfrq = 0x0, .scllt = 0xE, .sclht = 0xB,
+		.fast_mode = true,
+	},
+	{
+		.core_clk = 48000000, .hldt = 0x9, .dbcnt = 0x2,
+		.sclfrq = 0x0, .scllt = 0xD, .sclht = 0xB,
+		.fast_mode = true,
+	},
+	{
+		.core_clk = 41000000, .hldt = 0x9, .dbcnt = 0x2,
+		.sclfrq = 0x0, .scllt = 0xC, .sclht = 0x9,
+		.fast_mode = true,
+	},
+	{
+		.core_clk = 40000000, .hldt = 0x8, .dbcnt = 0x2,
+		.sclfrq = 0x0, .scllt = 0xB, .sclht = 0x9,
+		.fast_mode = true,
+	},
+	{
+		.core_clk = 33000000, .hldt = 0x7, .dbcnt = 0x1,
+		.sclfrq = 0x0, .scllt = 0xA, .sclht = 0x7,
+		.fast_mode = true,
+	},
+	{
+		.core_clk = 25000000, .hldt = 0x4, .dbcnt = 0x1,
+		.sclfrq = 0x0, .scllt = 0x7, .sclht = 0x6,
+		.fast_mode = true,
+	},
+	{
+		.core_clk = 24000000, .hldt = 0x7, .dbcnt = 0x1,
+		.sclfrq = 0x0, .scllt = 0x8, .sclht = 0x5,
+		.fast_mode = true,
+	},
+	{
+		.core_clk = 20000000, .hldt = 0x4, .dbcnt = 0x1,
+		.sclfrq = 0x0, .scllt = 0x6, .sclht = 0x4,
+		.fast_mode = true,
+	},
+};
+
 struct npcm_i2c_data {
 	u8 fifo_size;
 	u32 segctl_init_val;
@@ -1805,102 +2064,45 @@ static void npcm_i2c_recovery_init(struct i2c_adapter *_adap)
  */
 static int npcm_i2c_init_clk(struct npcm_i2c *bus, u32 bus_freq_hz)
 {
-	u32  k1 = 0;
-	u32  k2 = 0;
-	u8   dbnct = 0;
-	u32  sclfrq = 0;
-	u8   hldt = 7;
+	struct  smb_timing_t *smb_timing;
+	u8   scl_table_cnt = 0, table_size = 0;
 	u8   fast_mode = 0;
-	u32  src_clk_khz;
-	u32  bus_freq_khz;
 
-	src_clk_khz = bus->apb_clk / 1000;
-	bus_freq_khz = bus_freq_hz / 1000;
 	bus->bus_freq = bus_freq_hz;
 
-	/* 100KHz and below: */
-	if (bus_freq_hz <= I2C_MAX_STANDARD_MODE_FREQ) {
-		sclfrq = src_clk_khz / (bus_freq_khz * 4);
-
-		if (sclfrq < SCLFRQ_MIN || sclfrq > SCLFRQ_MAX)
-			return -EDOM;
-
-		if (src_clk_khz >= 40000)
-			hldt = 17;
-		else if (src_clk_khz >= 12500)
-			hldt = 15;
-		else
-			hldt = 7;
-	}
-
-	/* 400KHz: */
-	else if (bus_freq_hz <= I2C_MAX_FAST_MODE_FREQ) {
-		sclfrq = 0;
+	switch (bus_freq_hz) {
+	case I2C_MAX_STANDARD_MODE_FREQ:
+		smb_timing = smb_timing_100khz;
+		table_size = ARRAY_SIZE(smb_timing_100khz);
+		break;
+	case I2C_MAX_FAST_MODE_FREQ:
+		smb_timing = smb_timing_400khz;
+		table_size = ARRAY_SIZE(smb_timing_400khz);
 		fast_mode = I2CCTL3_400K_MODE;
-
-		if (src_clk_khz < 7500)
-			/* 400KHZ cannot be supported for core clock < 7.5MHz */
-			return -EDOM;
-
-		else if (src_clk_khz >= 50000) {
-			k1 = 80;
-			k2 = 48;
-			hldt = 12;
-			dbnct = 7;
-		}
-
-		/* Master or Slave with frequency > 25MHz */
-		else if (src_clk_khz > 25000) {
-			hldt = clk_coef(src_clk_khz, 300) + 7;
-			k1 = clk_coef(src_clk_khz, 1600);
-			k2 = clk_coef(src_clk_khz, 900);
-		}
-	}
-
-	/* 1MHz: */
-	else if (bus_freq_hz <= I2C_MAX_FAST_MODE_PLUS_FREQ) {
-		sclfrq = 0;
+		break;
+	case I2C_MAX_FAST_MODE_PLUS_FREQ:
+		smb_timing = smb_timing_1000khz;
+		table_size = ARRAY_SIZE(smb_timing_1000khz);
 		fast_mode = I2CCTL3_400K_MODE;
-
-		/* 1MHZ cannot be supported for core clock < 24 MHz */
-		if (src_clk_khz < 24000)
-			return -EDOM;
-
-		k1 = clk_coef(src_clk_khz, 620);
-		k2 = clk_coef(src_clk_khz, 380);
-
-		/* Core clk > 40 MHz */
-		if (src_clk_khz > 40000) {
-			/*
-			 * Set HLDT:
-			 * SDA hold time:  (HLDT-7) * T(CLK) >= 120
-			 * HLDT = 120/T(CLK) + 7 = 120 * FREQ(CLK) + 7
-			 */
-			hldt = clk_coef(src_clk_khz, 120) + 7;
-		} else {
-			hldt = 7;
-			dbnct = 2;
-		}
+		break;
+	default:
+		return -EINVAL;
 	}
 
-	/* Frequency larger than 1 MHz is not supported */
-	else
-		return -EINVAL;
+	for (scl_table_cnt = 0; scl_table_cnt < table_size; scl_table_cnt++)
+		if (bus->apb_clk >= smb_timing[scl_table_cnt].core_clk)
+			break;
 
-	if (bus_freq_hz >= I2C_MAX_FAST_MODE_FREQ) {
-		k1 = round_up(k1, 2);
-		k2 = round_up(k2 + 1, 2);
-		if (k1 < SCLFRQ_MIN || k1 > SCLFRQ_MAX ||
-		    k2 < SCLFRQ_MIN || k2 > SCLFRQ_MAX)
-			return -EDOM;
-	}
+	if (scl_table_cnt == table_size)
+		return -EINVAL;
 
 	/* write sclfrq value. bits [6:0] are in I2CCTL2 reg */
-	iowrite8(FIELD_PREP(I2CCTL2_SCLFRQ6_0, sclfrq & 0x7F),
+	iowrite8(FIELD_PREP(I2CCTL2_SCLFRQ6_0, smb_timing[scl_table_cnt].sclfrq & 0x7F),
 		 bus->reg + NPCM_I2CCTL2);
 
 	/* bits [8:7] are in I2CCTL3 reg */
-	iowrite8(fast_mode | FIELD_PREP(I2CCTL3_SCLFRQ8_7, (sclfrq >> 7) & 0x3),
+	iowrite8(FIELD_PREP(I2CCTL3_SCLFRQ8_7, (smb_timing[scl_table_cnt].sclfrq >> 7) & 0x3) |
+		 fast_mode,
 		 bus->reg + NPCM_I2CCTL3);
 
 	/* Select Bank 0 to access NPCM_I2CCTL4/NPCM_I2CCTL5 */
@@ -1912,13 +2114,13 @@ static int npcm_i2c_init_clk(struct npcm_i2c *bus, u32 bus_freq_hz)
 		 * k1 = 2 * SCLLT7-0 -> Low Time  = k1 / 2
 		 * k2 = 2 * SCLLT7-0 -> High Time = k2 / 2
 		 */
-		iowrite8(k1 / 2, bus->reg + NPCM_I2CSCLLT);
-		iowrite8(k2 / 2, bus->reg + NPCM_I2CSCLHT);
+		iowrite8(smb_timing[scl_table_cnt].scllt, bus->reg + NPCM_I2CSCLLT);
+		iowrite8(smb_timing[scl_table_cnt].sclht, bus->reg + NPCM_I2CSCLHT);
 
-		iowrite8(dbnct, bus->reg + NPCM_I2CCTL5);
+		iowrite8(smb_timing[scl_table_cnt].dbcnt, bus->reg + NPCM_I2CCTL5);
 	}
 
-	iowrite8(hldt, bus->reg + NPCM_I2CCTL4);
+	iowrite8(smb_timing[scl_table_cnt].hldt, bus->reg + NPCM_I2CCTL4);
 
 	/* Return to Bank 1, and stay there by default: */
 	npcm_i2c_select_bank(bus, I2C_BANK_1);
Signature · 1 line
2.34.1
#17 lore raw

PATCH v4 6/6 i2c: npcm: Enable slave in eob interrupt

In reply to [email protected]

From: Charles Boyer <[email protected]>

Nuvoton slave enable was in user space API call master_xfer, so it is subject to delays from the OS scheduler. If the BMC is not enabled for slave mode in time for master to send response, then it will NAK the address match. Then the PLDM request timeout occurs.

If the slave enable is moved to the EOB interrupt service routine, then the BMC can be ready in slave mode by the time it needs to receive a response.

---
 drivers/i2c/busses/i2c-npcm7xx.c | 6 ++++++
 1 file changed, 6 insertions(+)
drivers/i2c/busses/i2c-npcm7xx.c +6 -0
index 9d533873e477..aa3310afbf6d 100644
@@ -1925,6 +1925,12 @@ static int npcm_i2c_int_master_handler(struct npcm_i2c *bus)
 	    (FIELD_GET(NPCM_I2CCST3_EO_BUSY,
 		       ioread8(bus->reg + NPCM_I2CCST3)))) {
 		npcm_i2c_irq_handle_eob(bus);
+#if IS_ENABLED(CONFIG_I2C_SLAVE)
+		/* reenable slave if it was enabled */
+		if (bus->slave)
+			iowrite8((bus->slave->addr & 0x7F) | NPCM_I2CADDR_SAEN,
+				 bus->reg + NPCM_I2CADDR1);
+#endif
 		return 0;
 	}
 
Signature · 1 line
2.34.1