On Tue, Jun 26, 2018 at 8:01 PM, Manivannan Sadhasivam
[off-list ref] wrote:

On Mon, Jun 25, 2018 at 12:38:50PM +0300, Andy Shevchenko wrote:

quoted

On Sat, Jun 23, 2018 at 7:11 PM, Manivannan Sadhasivam
[off-list ref] wrote:

quoted

quoted

+#define OWL_I2C_REG_CTL                (0x0000)
+#define OWL_I2C_REG_CLKDIV     (0x0004)
+#define OWL_I2C_REG_STAT       (0x0008)
+#define OWL_I2C_REG_ADDR       (0x000C)
+#define OWL_I2C_REG_TXDAT      (0x0010)
+#define OWL_I2C_REG_RXDAT      (0x0014)
+#define OWL_I2C_REG_CMD                (0x0018)
+#define OWL_I2C_REG_FIFOCTL    (0x001C)
+#define OWL_I2C_REG_FIFOSTAT   (0x0020)
+#define OWL_I2C_REG_DATCNT     (0x0024)
+#define OWL_I2C_REG_RCNT       (0x0028)

I don't understand why these have parents?

I'm not sure what you mean here! Can you please elaborate?

One has to read 'parens'. Sorry for confusion.

quoted

quoted

+#define OWL_I2C_TIMEOUT                (msecs_to_jiffies(4 * 1000))

Ditto.

Same as above

Ditto.

quoted

quoted

+       /* Wait until FIFO reset complete */
+       do {
+               val = readl(i2c_dev->base + OWL_I2C_REG_FIFOCTL);
+               if (!(val & (OWL_I2C_FIFOCTL_RFR | OWL_I2C_FIFOCTL_TFR)))
+                       break;
+       } while (1);

No way. Get rid of infinite loop.

Okay. Can I change it to max of 50 retries with 1ms delay?

Whatever suitable for HW. Rely on datasheet or other source of
information about this timeouts.

quoted

quoted

+static irqreturn_t owl_i2c_interrupt(int irq, void *_dev)

quoted

quoted

+               dev_warn(&i2c_dev->adap.dev, "received NACK from device");

And if it happens hundreds times in a row?

Should I change it to dev_dbg?

quoted

quoted

+               dev_warn(&i2c_dev->adap.dev, "bus error");

Same as above?

ratelimit, another level, tracepoints, getting rid of them completely
? your call.

-- 
With Best Regards,
Andy Shevchenko