Re: [PATCH] mmc: tmio: add barriers to IO operations

[Paul Mundt <lethal@linux-sh.org>]

On Thu, Jan 24, 2013 at 12:20:19PM +0100, Guennadi Liakhovetski wrote:
> On Thu, 24 Jan 2013, Paul Mundt wrote:
> 
> > On Wed, Jan 23, 2013 at 05:46:18PM +0100, Guennadi Liakhovetski wrote:
> > > Without barriers SDIO operations fail with runtime PM enabled.
> > > 
> > > Signed-off-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
> > > ---
> > >  drivers/mmc/host/tmio_mmc.h |   14 ++++++++------
> > >  1 files changed, 8 insertions(+), 6 deletions(-)
> > > 
> > > diff --git a/drivers/mmc/host/tmio_mmc.h b/drivers/mmc/host/tmio_mmc.h
> > > index d857f5c..ad1a1c6 100644
> > > --- a/drivers/mmc/host/tmio_mmc.h
> > > +++ b/drivers/mmc/host/tmio_mmc.h
> > > @@ -159,19 +159,20 @@ int tmio_mmc_host_runtime_resume(struct device *dev);
> > >  
> > >  static inline u16 sd_ctrl_read16(struct tmio_mmc_host *host, int addr)
> > >  {
> > > -	return readw(host->ctl + (addr << host->bus_shift));
> > > +	return ioread16(host->ctl + (addr << host->bus_shift));
> > >  }
> > >  
> > >  static inline void sd_ctrl_read16_rep(struct tmio_mmc_host *host, int addr,
> > >  		u16 *buf, int count)
> > >  {
> > > +	wmb();
> > >  	readsw(host->ctl + (addr << host->bus_shift), buf, count);
> > >  }
> > >  
> > Some reason to not use ioread16_rep() and friends?
> 
> Aren't they "raw," i.e. without barriers:
> 
> #define ioread16_rep(p,d,c)	__raw_readsw(p,d,c)

I wasn't talking about the barrier semantics, I meant for consistency, as
you are converting to the ioread/write API for everything else.