[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20160925235542.GD7312@infradead.org>
Date: Sun, 25 Sep 2016 16:55:42 -0700
From: Christoph Hellwig <hch@...radead.org>
To: Ondrej Zary <linux@...nbow-software.org>
Cc: Christoph Hellwig <hch@...radead.org>,
Finn Thain <fthain@...egraphics.com.au>,
linux-scsi@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 4/3] g_NCR5380: Merge g_NCR5380 and g_NCR5380_mmio
On Sun, Sep 25, 2016 at 09:39:20PM +0200, Ondrej Zary wrote:
> Merge the PIO and MMIO code (with the help of ioport_map) in g_NCR5380 and
> delete g_NCR5380_mmio.
Awesome! this looks great to me, but we probably should thrown
in a MODULE_ALIAS for g_NCR5380_mmio so that existing module
configurations keep working.
> #define NCR5380_read(reg) \
> + ioread8(((struct NCR5380_hostdata *)shost_priv(instance))->iomem + \
> + ((struct NCR5380_hostdata *)shost_priv(instance))->offset + \
> + (reg))
> #define NCR5380_write(reg, value) \
> - outb(value, instance->io_port + (reg))
> + iowrite8(value, ((struct NCR5380_hostdata *)shost_priv(instance))->iomem + \
> + ((struct NCR5380_hostdata *)shost_priv(instance))->offset + \
> + (reg))
Can we make these #defines point to inline helper functions that avoid
the casting magic, e.g. something like
static inline void g_NCR5380_write(struct Scsi_Host *instance, u32 reg,
u8 value)
{
struct NCR5380_hostdata *host = shost_priv(instance);
iowrite8(value, host->iomem, host->offset + reg);
}
#define NCR5380_write(reg, value)
g_NCR5380_write(instance, reg, value)
Powered by blists - more mailing lists