Re: [PATCH 2/2] reset: Add Broadcom STB RESCAL reset controller

[Florian Fainelli <f.fainelli@gmail.com>]

On 12/11/2019 1:48 AM, Philipp Zabel wrote:
>> +#define BRCM_RESCAL_START	0
>> +#define	BRCM_RESCAL_START_BIT	BIT(0)
>> +#define BRCM_RESCAL_CTRL	4
>> +#define BRCM_RESCAL_STATUS	8
>> +#define BRCM_RESCAL_STATUS_BIT	BIT(0)
> 
> Is there any reason the start bit is indented but the status bit is not?

This is a convention we have tried to adopt to denote the definition
from a register word address/offset versus the definition for bits
within that register word.

> 
>> +
>> +struct brcm_rescal_reset {
>> +	void __iomem	*base;
>> +	struct device *dev;
>> +	struct reset_controller_dev rcdev;
>> +};
>> +
>> +static int brcm_rescal_reset_assert(struct reset_controller_dev *rcdev,
>> +				      unsigned long id)
>> +{
>> +	return 0;
>> +}
> 
> Please do not implement the assert operation if it doesn't cause a reset
> line to be asserted afterwards.
> The reset core will return 0 from reset_control_assert() for shared
> reset controls if .assert is not implemented.

OK, will drop it.

> 
>> +
>> +static int brcm_rescal_reset_deassert(struct reset_controller_dev *rcdev,
>> +				      unsigned long id)
>> +{
>> +	struct brcm_rescal_reset *data =
>> +		container_of(rcdev, struct brcm_rescal_reset, rcdev);
>> +	void __iomem *base = data->base;
>> +	const int NUM_RETRIES = 10;
>> +	u32 reg;
>> +	int i;
>> +
>> +	reg = readl(base + BRCM_RESCAL_START);
>> +	writel(reg | BRCM_RESCAL_START_BIT, base + BRCM_RESCAL_START);
>> +	reg = readl(base + BRCM_RESCAL_START);
> 
> Are there any other fields beside the START_BIT in this register?

This is the only bit actually.

> 
>> +	if (!(reg & BRCM_RESCAL_START_BIT)) {
>> +		dev_err(data->dev, "failed to start sata/pcie rescal\n");
>> +		return -EIO;
>> +	}
>> +
>> +	reg = readl(base + BRCM_RESCAL_STATUS);
>> +	for (i = NUM_RETRIES; i >= 0 &&  !(reg & BRCM_RESCAL_STATUS_BIT); i--) {
>> +		udelay(100);
>> +		reg = readl(base + BRCM_RESCAL_STATUS);
>> +	}
> 
> This timeout loop should be replaced by a single readl_poll_timeout().
> At 100 µs waits per iteration this could use the sleeping variant.

OK, will do.

> 
>> +	if (!(reg & BRCM_RESCAL_STATUS_BIT)) {
>> +		dev_err(data->dev, "timedout on sata/pcie rescal\n");
>> +		return -ETIMEDOUT;
>> +	}
>> +
>> +	reg = readl(base + BRCM_RESCAL_START);
>> +	writel(reg ^ BRCM_RESCAL_START_BIT, base + BRCM_RESCAL_START);
> 
> Please use &= ~BRCM_RESCAL_START_BIT instead.
> 

I think the idea was to avoid unconditionally clearing it, but based on
the documentation, I don't see this being harmful, Jim?

>> +	reg = readl(base + BRCM_RESCAL_START);
>> +	dev_dbg(data->dev, "sata/pcie rescal success\n");
>> +
>> +	return 0;
>> +}
> 
> This whole function looks a lot like it doesn't just deassert a reset
> line, but actually issues a complete reset procedure of some kind. Do
> you have some insight on what actually happens in the hardware when the
> start bit is triggered? I suspect this should be implemented with the
> .reset operation.

This hardware block is controlling the reset and calibration process of
the SATA/PCIe combo PHY analog front end, but is not technically part of
the PCIe or SATA PHY proper, it stands on its own, both functionally and
from a register space perspective. The motivation for modelling this as
a reset controller is that it does a reset (and a calibration) and this
is a shared reset line among 2/3 instances of another block. If you
think we should model this differently, please let us know.
-- 
Florian