Re: [PATCH 00/19] sh-sci: Platform data simplification

[Laurent Pinchart <laurent.pinchart@ideasonboard.com>]

Hi Greg,

On Wednesday 11 Jan 2017 09:04:37 Greg KH wrote:
> On Wed, Jan 04, 2017 at 01:06:20AM +0200, Laurent Pinchart wrote:
> > Hello,
> > 
> > Most of the patches in this series have been sitting in my development
> > tree for three years now. While rebasing all my development branches I
> > decided it was time to send them out in case they're considered useful.
> > I've then thrown a few additional patches in to move the driver further in
> > the right direction (or at least the direction I consider right).
> > 
> > The idea behind this series is to move towards removal of the sh-sci
> > platform data. This won't be possible until SH platforms move to DT (and
> > I won't hold my breath), but we can get rid of most of the platform data
> > fields already and cleanup the sh-sci driver at the same time.
> > 
> > The series already removes the capabilities, port_reg, regshift,
> > dma_slave_tx and dma_slave_rx fields. We should be able to go further:
> > 
> > - The flags field is only used by sh5 to set UPF_IOREMAP. It should be
> >   possible to find another way to determine in the driver whether to
> >   ioremap the registers.
> > 
> > - The sampling_rate field is only used by sh7723 and sh7724 to override
> >   the value hardcoded in the driver, even though the datasheets of those
> >   SoCs don't mention any restriction. This might not be needed and should
> >   be investigated.
> > 
> > - The scscr REIE bit should be possible to compute in the driver, as it's
> >   set for all SCIF ports except SCIx_SH3_SCIF_REGTYPE,
> >   SCIx_SH7705_SCIF_REGTYPE,
> >   and the sh7343 SCIx_SH4_SCIF_REGTYPE. The latter will be trickier to
> >   handle.
> >
> > - Now that we have support for baud rate generators and externel clocks,
> >   the scscr CKE1 bit should be computed in the driver in all cases.
> > 
> > - The scscr TOIE bit doesn't seem to follow a clear rule, but given that
> >   it's the last scscr bit to handle I'm sure we'll find something :-)
> > 
> > - It might make sense to merge the type and regtype fields in a single
> >   type.
> > 
> > Finally, it might be time to drop support for the SH platforms that would
> > get in the way of sh-sci cleanup.
> 
> I've merged patches 1, 2, 12, 13, 14, 15, 16, 17, and 18 in my tty tree.
> 
> If I should take any others there as well, please let me know.

The other patches are needed, otherwise you'll break compilation of arch/sh/. 
I'm waiting for an ack from Geert on two patches and I'll then repost the 
whole series.

-- 
Regards,

Laurent Pinchart