TPS6594-Q1: Linux Driver Initialization Procedure vs PMIC documentation

Part Number: TPS6594-Q1

Hello,

according to the PMIC documentation this is the required initialization procedure for the PMIC:

6.1 Initialization
Upon a successful power up, the BIST_PASS_INT and ENABLE_INT interrupts are set. Any other interrupts
indicate an issue but the automated recovery attempt was successful. The recommended procedure is to:
1. Interrogate the interrupts
2. Determine the course of action
3. Set the NSLEEP bits
4. Clear the interrupts

However in the driver, at least to my understanding, we have just this:

int tps6594_device_init(struct tps6594 *tps, bool enable_crc)
{
    ...
    
	/* Keep PMIC in ACTIVE state */
	ret = regmap_set_bits(tps->regmap, TPS6594_REG_FSM_NSLEEP_TRIGGERS,
			      TPS6594_BIT_NSLEEP1B | TPS6594_BIT_NSLEEP2B);
	if (ret)
		return dev_err_probe(dev, ret, "Failed to set PMIC state\n");

    ...

	return 0;
}

Can you confirm that the driver is fine?

7 months ago

0 fd 7 months ago

Intellectual 645 points

Just a small clarification, the concern is that no interrupt flag is cleared.

0 Nicholas McNamara 7 months ago in reply to fd

TI__Genius 14522 points

Hello FD,

I'm assuming this is the with the Toradex settings for the power button for the J7 processor correct?

Is there unexpected behavior associated with it?

Will reach out to the software engineer that understands the Linux driver side much better.

Thank you and BR,

Nicholas

0 fd 7 months ago in reply to Nicholas McNamara

Intellectual 645 points

Nicholas McNamara said:
I'm assuming this is the with the Toradex settings for the power button for the J7 processor correct?

Correct, this is the context.

Nicholas McNamara said:
Is there unexpected behavior associated with it?

We have no immediate way to test the complete PMIC / HW / SW combination.

While looking at the software implementation for such PMIC configuration we noticed this difference between the driver and your documentation, and we wanted to double-check with you.

0 Nicholas McNamara 6 months ago in reply to fd

TI__Genius 14522 points

Hello FD,

Yes the driver is fine, the sleep bits do not affect the FSM so the setting of these is not relevant for the Toradex PMIC code.

BR,

Nicholas

0 fd 6 months ago in reply to Nicholas McNamara

Intellectual 645 points

Nicholas McNamara

Nicholas McNamara said:
the sleep bits do not affect the FSM so the setting of these is not relevant for the Toradex PMIC code.

This was not my question, apologize if it was not clear.

According to the documentation, there is a specific initialization sequence described (see my first message) and the Linux driver does not implement it. Is this ok? Specifically setting the sleep bits is the only thing that is the same between the documentation and the code. The difference between the documentation and the code is about checking the interrupts status and clearing the interrupts

0 Nicholas McNamara 6 months ago in reply to fd

TI__Genius 14522 points

Hello FD,

now I understand the question does the driver handle the interrupts, the interrupts to get keep the device in the active state, but the other interrupts are stored.

The understanding is that the customer will implement their own code with the other interrupts, such as what to do when there's a thermal warning interrupt, short circuit, etc.

BR,

Nicholas

0 fd 6 months ago in reply to Nicholas McNamara

Intellectual 645 points

Nicholas McNamara said:
The understanding is that the customer will implement their own code with the other interrupts, such as what to do when there's a thermal warning interrupt, short circuit, etc.

Understood.

However this is not expected, I would have expected the driver provided by TI to implement the standard Linux PMIC error handling, see for example https://github.com/torvalds/linux/blob/master/include/linux/regulator/consumer.h#L100 as an entry point.

0 Nicholas McNamara 6 months ago in reply to fd

TI__Genius 14522 points

Hello FD,

Have you taken a look at: https://github.com/torvalds/linux/blob/master/drivers/regulator/tps6594-regulator.c

BR,

Nicholas

0 fd 6 months ago in reply to Nicholas McNamara

Intellectual 645 points

Nicholas McNamara said:
Have you taken a look at: https://github.com/torvalds/linux/blob/master/drivers/regulator/tps6594-regulator.c

whoops, I missed this. Please accept my apologies.

And what about clearing the interrupts? Again, the concern is just about a discrepancy between the documentation and the driver. I am fine with the answer "All good, the driver implements what the PMIC requires to function properly".

+1 Nicholas McNamara 6 months ago in reply to fd

TI__Genius 14522 points

Hello FD,

after speaking to the software team, the driver implements what PMIC requires to function properly.

BR,

Nicholas

Power management

Power management forum

TPS6594-Q1: Linux Driver Initialization Procedure vs PMIC documentation