On Wednesday 12 November 2014 10:03:58 Jiri Slaby wrote:
> On 11/12/2014, 09:46 AM, Kevin Cernekee wrote:
> > With a few tweaks, the PXA serial driver can handle other 16550A clones.
> > Add a fifo-size DT property to override the FIFO depth (BCM7xxx uses 32),
> > and {native,big}-endian properties similar to regmap to support SoCs that
> > have BE or "automagic endian" registers.
> >
> > Signed-off-by: Kevin Cernekee <cernekee@xxxxxxxxx>
> > ---
> > drivers/tty/serial/pxa.c | 16 ++++++++++++++--
> > 1 file changed, 14 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/tty/serial/pxa.c b/drivers/tty/serial/pxa.c
> > index 21b7d8b..78ed7ee 100644
> > --- a/drivers/tty/serial/pxa.c
> > +++ b/drivers/tty/serial/pxa.c
> > @@ -60,13 +60,19 @@ struct uart_pxa_port {
> > static inline unsigned int serial_in(struct uart_pxa_port *up, int offset)
> > {
> > offset <<= 2;
> > - return readl(up->port.membase + offset);
> > + if (!up->port.big_endian)
> > + return readl(up->port.membase + offset);
> > + else
> > + return ioread32be(up->port.membase + offset);
>
> This needn't fly IMO, unless you map the space using iomap (not ioremap).
For all I know, the ioread family is required to work with tokens returned
from ioremap on all architectures. The difference to readl is that it
also works on tokens returned from ioport_map or one of the wrappers
around it.
Arnd
