Re: [PATCH v2 1/3] i3c: Drop support for I2C 10 bit addresing

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

On Wed, 27 Feb 2019 12:52:42 +0000
vitor <vitor.soares@synopsys.com> wrote:

> Hi,
> 
> On 27/02/19 12:10, Boris Brezillon wrote:
> > Hi Vitor,
> >
> > On Wed, 27 Feb 2019 12:05:30 +0000
> > vitor <vitor.soares@synopsys.com> wrote:
> >  
> >> On 26/02/19 14:28, Przemyslaw Gaj wrote:  
> >>> This patch dropps support for I2C devices with 10 bit addressing. When I2C
> >>> device with 10 bit address is defined in DT, I3C master registration fails.
> >>>
> >>> Signed-off-by: Przemyslaw Gaj <pgaj@cadence.com>
> >>>
> >>> ---
> >>>
> >>> Main changes between v1 and v2 are:
> >>> - Add error message when registering I2C device with 10 bit address.
> >>>
> >>> ---
> >>>  drivers/i3c/master.c | 9 +++++++++
> >>>  1 file changed, 9 insertions(+)
> >>>
> >>> diff --git a/drivers/i3c/master.c b/drivers/i3c/master.c
> >>> index 2dc628d..8c1e365 100644
> >>> --- a/drivers/i3c/master.c
> >>> +++ b/drivers/i3c/master.c
> >>> @@ -1962,6 +1962,15 @@ of_i3c_master_add_i2c_boardinfo(struct i3c_master_controller *master,
> >>>  	if (ret)
> >>>  		return ret;
> >>>  
> >>> +	/* The I3C Specification does not clearly say I2C devices with 10-bit

Nitpick: please do not use net-style comments in the I3C subsystem.

Use

/*
 * blablabla
 */

instead.

> >>> +	 * address are supported. These devices can't be passed properly through
> >>> +	 * DEFSLVS command.
> >>> +	 */    
> >> IMO we just need to say 10-bit address not used or not supported in i3c.  
> > I'd like to keep a clear comment on why it cannot supported right now
> > even though the spec is unclear about that. If the spec is updated to
> > state that I2C devs using extended addresses are forbidden, then we'll
> > update this comment accordingly.  
> 
> I'm not sure if the terms aren't the same,

Don't understand what you mean, sorry.

>but let's keep the comment.
> 
> >  
> >>> +	if (boardinfo->base.flags & I2C_CLIENT_TEN) {
> >>> +		dev_err(&master->dev, "I2C device with 10 bit address not supported.");
> >>> +		return -ENOTSUPP;
> >>> +	}
> >>> +
> >>>  	/* LVR is encoded in reg[2]. */
> >>>  	boardinfo->lvr = reg[2];
> >>>      
> >> Also need to change:
> >>
> >> #define I2C_MAX_ADDR            GENMASK(9, 0)
> >> to
> >> #define I2C_MAX_ADDR            GENMASK(6, 0)
> >> in master.h file  
> > Yep, you can reduce the address space.
> >  
> >> Not sure about:
> >> unsigned long addrslots[((I2C_MAX_ADDR + 1) * 2) / BITS_PER_LONG];
> >> @Boris can you check this part?  
> > This part is still valid, no need to update it as you already updated
> > I2C_MAX_ADDR.
> >
> > Thanks for the review.
> >
> > Boris  
> 
> Also we can make i2c_algorithm.functionality = I2C_FUNC_SMBUS_EMUL | I2C_FUNC_I2C, and in this way we won't need ->i2c_func() and avoid the I2C_FUNC_10BIT_ADDR flag on host controller side.

Ack. Let's rip ->i2c_func() out and return I2C_FUNC_SMBUS_EMUL |
I2C_FUNC_I2C for everyone. We'll add it back if some drivers want to
support SMBUS natively or if 10bit addressing appears to be needed at
some point.

_______________________________________________
linux-i3c mailing list
linux-i3c@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-i3c