Hi Manivannan, On Thu, Jul 26, 2018 at 09:16:02PM +0530, Manivannan Sadhasivam wrote: > Add Actions Semiconductor Owl family S900 I2C driver. > > Signed-off-by: Manivannan Sadhasivam > Acked-by: Peter Rosin Looks mostly good already. Thanks Peter for the initial review! > +static irqreturn_t owl_i2c_interrupt(int irq, void *_dev) > +{ > + struct owl_i2c_dev *i2c_dev = _dev; > + struct i2c_msg *msg = i2c_dev->msg; > + unsigned long flags; > + unsigned int stat, fifostat; > + > + spin_lock_irqsave(&i2c_dev->lock, flags); > + > + fifostat = readl(i2c_dev->base + OWL_I2C_REG_FIFOSTAT); > + if (fifostat & OWL_I2C_FIFOSTAT_RNB) { > + dev_dbg(&i2c_dev->adap.dev, "received NACK from device\n"); > + goto stop; > + } > + > + stat = readl(i2c_dev->base + OWL_I2C_REG_STAT); > + if (stat & OWL_I2C_STAT_BEB) { > + dev_dbg(&i2c_dev->adap.dev, "bus error\n"); > + goto stop; > + } I wonder if you can't pass back the different errors to the upper layers? Like -ENXIO for NACK and -EIO for bus error? We have a convention for that and it seems your HW can support it. The different error codes would then maybe also make the debug outputs obsolete. > + /* > + * Here, -ENXIO will be returned if interrupt occurred but no > + * read or write happened. Else if msg_ptr equals to message length, > + * message count will be returned. > + */ > + ret = i2c_dev->msg_ptr == msg->len ? num : -ENXIO; I'd think this kinda unusual construct could go then as well by just returning the error code derived from the interrupt handler above. > +static const struct i2c_adapter_quirks owl_i2c_quirks = { > + .flags = I2C_AQ_COMB | I2C_AQ_COMB_WRITE_FIRST, > + .max_read_len = 240, > + .max_write_len = 240, > + .max_comb_1st_msg_len = 6, > + .max_comb_2nd_msg_len = 240, > +}; Yay! Good use of the i2c_adapter_quirks struct :) Regards, Wolfram