Re: [PATCH v6 01/10] i3c: Add core I3C infrastructure

[Boris Brezillon <boris.brezillon@bootlin.com>]

Hi Vitor,

On Fri, 24 Aug 2018 18:52:52 +0100
vitor <Vitor.Soares@synopsys.com> wrote:

> Hi Boris,
> 
> 
> On 24-08-2018 13:39, Boris Brezillon wrote:
> > Hi Vitor,
> >
> > On Wed, 22 Aug 2018 17:43:34 +0100
> > vitor <Vitor.Soares@synopsys.com> wrote:
> >  
> >> Hi Boris,
> >>
> >>
> >> On 19-07-2018 16:29, Boris Brezillon wrote:  
> >>> +int i3c_bus_register(struct i3c_bus *i3cbus)
> >>> +{
> >>> +	struct i2c_dev_desc *desc;
> >>> +
> >>> +	i3c_bus_for_each_i2cdev(i3cbus, desc) {
> >>> +		switch (desc->boardinfo->lvr & I3C_LVR_I2C_INDEX_MASK) {
> >>> +		case I3C_LVR_I2C_INDEX(0):
> >>> +			if (i3cbus->mode < I3C_BUS_MODE_MIXED_FAST)
> >>> +				i3cbus->mode = I3C_BUS_MODE_MIXED_FAST;
> >>> +			break;
> >>> +
> >>> +		case I3C_LVR_I2C_INDEX(1):
> >>> +		case I3C_LVR_I2C_INDEX(2):
> >>> +			if (i3cbus->mode < I3C_BUS_MODE_MIXED_SLOW)
> >>> +				i3cbus->mode = I3C_BUS_MODE_MIXED_SLOW;
> >>> +			break;
> >>> +
> >>> +		default:
> >>> +			return -EINVAL;
> >>> +		}
> >>> +	}
> >>> +
> >>> +	if (!i3cbus->scl_rate.i3c)
> >>> +		i3cbus->scl_rate.i3c = I3C_BUS_TYP_I3C_SCL_RATE;
> >>> +
> >>> +	if (!i3cbus->scl_rate.i2c) {
> >>> +		if (i3cbus->mode == I3C_BUS_MODE_MIXED_SLOW)
> >>> +			i3cbus->scl_rate.i2c = I3C_BUS_I2C_FM_SCL_RATE;
> >>> +		else
> >>> +			i3cbus->scl_rate.i2c = I3C_BUS_I2C_FM_PLUS_SCL_RATE;
> >>> +	}
> >>> +
> >>> +	/*
> >>> +	 * I3C/I2C frequency may have been overridden, check that user-provided
> >>> +	 * values are not exceeding max possible frequency.
> >>> +	 */
> >>> +	if (i3cbus->scl_rate.i3c > I3C_BUS_MAX_I3C_SCL_RATE ||
> >>> +	    i3cbus->scl_rate.i2c > I3C_BUS_I2C_FM_PLUS_SCL_RATE) {
> >>> +		return -EINVAL;
> >>> +	}
> >>> +
> >>> +	dev_set_name(&i3cbus->dev, "i3c-%d", i3cbus->id);
> >>> +
> >>> +	return device_add(&i3cbus->dev);
> >>> +}  
> >> During the tests of the bus with i2c devices I found the i2c_dev_desc
> >> objects aren't allocated before this function. This cause i3cbus->mode =
> >> I3C_BUS_MODE_PURE.  
> > I just checked and DT parsing (+ I2C descs creation) is done before
> > i3c_bus_register() is called, so we should be good. How did you declare
> > your I2C devices (right now, only DT declaration is supported).  
> During the DT parsing, you create the i2c_dev_boardinfo. the 
> i2c_dev_desc is created in i3c_master_bus_init() which is after the 
> i3c_mater_create_bus().

Oops, you're right.

> One possible way to fix this is to pass master 
> also to i3c_bus_register and iterate over i2c_dev_board_info list.

Yes, that's the proper fix. I'll do that in v7.

> 
> >> I want to do something for the slave and secondary master, do you
> >> already have infrastructure that you can share?  
> > What do you mean?
> >
> > Regards,
> >
> > Boris  
> 
> I want start to add the secondary master functionality but it is also 
> necessary to add the infrastructure to the subsystem.
> So, to avoid duplicated work can you share your plans for the secondary 
> master?

Well, before even considering supporting secondary master registration,
we need to handle mastership handover. As for the DAA operation, it's
likely to be host specific, so we'll have to add a new hook to the
i3c_master_controller_ops struct.

Once you've done that, we'll have trigger a mastership handover
everytime an I3C driver tries to send a frame on the bus, and the
master this frame should do through is not in control of the bus. That
should be pretty easy for the nominal case, but error cases are likely
to be hard to deal with.
Note that I have a ->cur_master field in the i3c_bus object which
stores allows us to track whose the currently active master. If
master->this != master->bus->cur_master that means you need to start a
mastership handover procedure.

That's all I thought about for now, and we'll probably face other
problems when implementing it. Let me know if you have other questions,
and don't hesitate to share your code early during the development
phase.

Also note that the bus representation is likely to change based on
Arnd's feedback, so you might have to rework your implementation a bit
at some point.

Regards,

Boris