linux-sg2042/include/asm-m68k/raw_io.h

344 lines
7.7 KiB
C
Raw Normal View History

/*
* linux/include/asm-m68k/raw_io.h
*
* 10/20/00 RZ: - created from bits of io.h and ide.h to cleanup namespace
*
*/
#ifndef _RAW_IO_H
#define _RAW_IO_H
#ifdef __KERNEL__
#include <asm/types.h>
/* Values for nocacheflag and cmode */
#define IOMAP_FULL_CACHING 0
#define IOMAP_NOCACHE_SER 1
#define IOMAP_NOCACHE_NONSER 2
#define IOMAP_WRITETHROUGH 3
extern void iounmap(void __iomem *addr);
extern void __iomem *__ioremap(unsigned long physaddr, unsigned long size,
int cacheflag);
extern void __iounmap(void *addr, unsigned long size);
/* ++roman: The assignments to temp. vars avoid that gcc sometimes generates
* two accesses to memory, which may be undesirable for some devices.
*/
#define in_8(addr) \
({ u8 __v = (*(__force volatile u8 *) (addr)); __v; })
#define in_be16(addr) \
({ u16 __v = (*(__force volatile u16 *) (addr)); __v; })
#define in_be32(addr) \
({ u32 __v = (*(__force volatile u32 *) (addr)); __v; })
#define in_le16(addr) \
({ u16 __v = le16_to_cpu(*(__force volatile u16 *) (addr)); __v; })
#define in_le32(addr) \
({ u32 __v = le32_to_cpu(*(__force volatile u32 *) (addr)); __v; })
#define out_8(addr,b) (void)((*(__force volatile u8 *) (addr)) = (b))
#define out_be16(addr,w) (void)((*(__force volatile u16 *) (addr)) = (w))
#define out_be32(addr,l) (void)((*(__force volatile u32 *) (addr)) = (l))
#define out_le16(addr,w) (void)((*(__force volatile u16 *) (addr)) = cpu_to_le16(w))
#define out_le32(addr,l) (void)((*(__force volatile u32 *) (addr)) = cpu_to_le32(l))
#define raw_inb in_8
#define raw_inw in_be16
#define raw_inl in_be32
#define raw_outb(val,port) out_8((port),(val))
#define raw_outw(val,port) out_be16((port),(val))
#define raw_outl(val,port) out_be32((port),(val))
static inline void raw_insb(volatile u8 __iomem *port, u8 *buf, unsigned int len)
{
unsigned int i;
for (i = 0; i < len; i++)
*buf++ = in_8(port);
}
static inline void raw_outsb(volatile u8 __iomem *port, const u8 *buf,
unsigned int len)
{
unsigned int i;
for (i = 0; i < len; i++)
out_8(port, *buf++);
}
static inline void raw_insw(volatile u16 __iomem *port, u16 *buf, unsigned int nr)
{
unsigned int tmp;
if (nr & 15) {
tmp = (nr & 15) - 1;
asm volatile (
"1: movew %2@,%0@+; dbra %1,1b"
: "=a" (buf), "=d" (tmp)
: "a" (port), "0" (buf),
"1" (tmp));
}
if (nr >> 4) {
tmp = (nr >> 4) - 1;
asm volatile (
"1: "
"movew %2@,%0@+; "
"movew %2@,%0@+; "
"movew %2@,%0@+; "
"movew %2@,%0@+; "
"movew %2@,%0@+; "
"movew %2@,%0@+; "
"movew %2@,%0@+; "
"movew %2@,%0@+; "
"movew %2@,%0@+; "
"movew %2@,%0@+; "
"movew %2@,%0@+; "
"movew %2@,%0@+; "
"movew %2@,%0@+; "
"movew %2@,%0@+; "
"movew %2@,%0@+; "
"movew %2@,%0@+; "
"dbra %1,1b"
: "=a" (buf), "=d" (tmp)
: "a" (port), "0" (buf),
"1" (tmp));
}
}
static inline void raw_outsw(volatile u16 __iomem *port, const u16 *buf,
unsigned int nr)
{
unsigned int tmp;
if (nr & 15) {
tmp = (nr & 15) - 1;
asm volatile (
"1: movew %0@+,%2@; dbra %1,1b"
: "=a" (buf), "=d" (tmp)
: "a" (port), "0" (buf),
"1" (tmp));
}
if (nr >> 4) {
tmp = (nr >> 4) - 1;
asm volatile (
"1: "
"movew %0@+,%2@; "
"movew %0@+,%2@; "
"movew %0@+,%2@; "
"movew %0@+,%2@; "
"movew %0@+,%2@; "
"movew %0@+,%2@; "
"movew %0@+,%2@; "
"movew %0@+,%2@; "
"movew %0@+,%2@; "
"movew %0@+,%2@; "
"movew %0@+,%2@; "
"movew %0@+,%2@; "
"movew %0@+,%2@; "
"movew %0@+,%2@; "
"movew %0@+,%2@; "
"movew %0@+,%2@; "
"dbra %1,1b"
: "=a" (buf), "=d" (tmp)
: "a" (port), "0" (buf),
"1" (tmp));
}
}
static inline void raw_insl(volatile u32 __iomem *port, u32 *buf, unsigned int nr)
{
unsigned int tmp;
if (nr & 15) {
tmp = (nr & 15) - 1;
asm volatile (
"1: movel %2@,%0@+; dbra %1,1b"
: "=a" (buf), "=d" (tmp)
: "a" (port), "0" (buf),
"1" (tmp));
}
if (nr >> 4) {
tmp = (nr >> 4) - 1;
asm volatile (
"1: "
"movel %2@,%0@+; "
"movel %2@,%0@+; "
"movel %2@,%0@+; "
"movel %2@,%0@+; "
"movel %2@,%0@+; "
"movel %2@,%0@+; "
"movel %2@,%0@+; "
"movel %2@,%0@+; "
"movel %2@,%0@+; "
"movel %2@,%0@+; "
"movel %2@,%0@+; "
"movel %2@,%0@+; "
"movel %2@,%0@+; "
"movel %2@,%0@+; "
"movel %2@,%0@+; "
"movel %2@,%0@+; "
"dbra %1,1b"
: "=a" (buf), "=d" (tmp)
: "a" (port), "0" (buf),
"1" (tmp));
}
}
static inline void raw_outsl(volatile u32 __iomem *port, const u32 *buf,
unsigned int nr)
{
unsigned int tmp;
if (nr & 15) {
tmp = (nr & 15) - 1;
asm volatile (
"1: movel %0@+,%2@; dbra %1,1b"
: "=a" (buf), "=d" (tmp)
: "a" (port), "0" (buf),
"1" (tmp));
}
if (nr >> 4) {
tmp = (nr >> 4) - 1;
asm volatile (
"1: "
"movel %0@+,%2@; "
"movel %0@+,%2@; "
"movel %0@+,%2@; "
"movel %0@+,%2@; "
"movel %0@+,%2@; "
"movel %0@+,%2@; "
"movel %0@+,%2@; "
"movel %0@+,%2@; "
"movel %0@+,%2@; "
"movel %0@+,%2@; "
"movel %0@+,%2@; "
"movel %0@+,%2@; "
"movel %0@+,%2@; "
"movel %0@+,%2@; "
"movel %0@+,%2@; "
"movel %0@+,%2@; "
"dbra %1,1b"
: "=a" (buf), "=d" (tmp)
: "a" (port), "0" (buf),
"1" (tmp));
}
}
static inline void raw_insw_swapw(volatile u16 __iomem *port, u16 *buf,
unsigned int nr)
{
if ((nr) % 8)
__asm__ __volatile__
("\tmovel %0,%/a0\n\t"
"movel %1,%/a1\n\t"
"movel %2,%/d6\n\t"
"subql #1,%/d6\n"
"1:\tmovew %/a0@,%/d0\n\t"
"rolw #8,%/d0\n\t"
"movew %/d0,%/a1@+\n\t"
"dbra %/d6,1b"
:
: "g" (port), "g" (buf), "g" (nr)
: "d0", "a0", "a1", "d6");
else
__asm__ __volatile__
("movel %0,%/a0\n\t"
"movel %1,%/a1\n\t"
"movel %2,%/d6\n\t"
"lsrl #3,%/d6\n\t"
"subql #1,%/d6\n"
"1:\tmovew %/a0@,%/d0\n\t"
"rolw #8,%/d0\n\t"
"movew %/d0,%/a1@+\n\t"
"movew %/a0@,%/d0\n\t"
"rolw #8,%/d0\n\t"
"movew %/d0,%/a1@+\n\t"
"movew %/a0@,%/d0\n\t"
"rolw #8,%/d0\n\t"
"movew %/d0,%/a1@+\n\t"
"movew %/a0@,%/d0\n\t"
"rolw #8,%/d0\n\t"
"movew %/d0,%/a1@+\n\t"
"movew %/a0@,%/d0\n\t"
"rolw #8,%/d0\n\t"
"movew %/d0,%/a1@+\n\t"
"movew %/a0@,%/d0\n\t"
"rolw #8,%/d0\n\t"
"movew %/d0,%/a1@+\n\t"
"movew %/a0@,%/d0\n\t"
"rolw #8,%/d0\n\t"
"movew %/d0,%/a1@+\n\t"
"movew %/a0@,%/d0\n\t"
"rolw #8,%/d0\n\t"
"movew %/d0,%/a1@+\n\t"
"dbra %/d6,1b"
:
: "g" (port), "g" (buf), "g" (nr)
: "d0", "a0", "a1", "d6");
}
static inline void raw_outsw_swapw(volatile u16 __iomem *port, const u16 *buf,
unsigned int nr)
{
if ((nr) % 8)
__asm__ __volatile__
("movel %0,%/a0\n\t"
"movel %1,%/a1\n\t"
"movel %2,%/d6\n\t"
"subql #1,%/d6\n"
"1:\tmovew %/a1@+,%/d0\n\t"
"rolw #8,%/d0\n\t"
"movew %/d0,%/a0@\n\t"
"dbra %/d6,1b"
:
: "g" (port), "g" (buf), "g" (nr)
: "d0", "a0", "a1", "d6");
else
__asm__ __volatile__
("movel %0,%/a0\n\t"
"movel %1,%/a1\n\t"
"movel %2,%/d6\n\t"
"lsrl #3,%/d6\n\t"
"subql #1,%/d6\n"
"1:\tmovew %/a1@+,%/d0\n\t"
"rolw #8,%/d0\n\t"
"movew %/d0,%/a0@\n\t"
"movew %/a1@+,%/d0\n\t"
"rolw #8,%/d0\n\t"
"movew %/d0,%/a0@\n\t"
"movew %/a1@+,%/d0\n\t"
"rolw #8,%/d0\n\t"
"movew %/d0,%/a0@\n\t"
"movew %/a1@+,%/d0\n\t"
"rolw #8,%/d0\n\t"
"movew %/d0,%/a0@\n\t"
"movew %/a1@+,%/d0\n\t"
"rolw #8,%/d0\n\t"
"movew %/d0,%/a0@\n\t"
"movew %/a1@+,%/d0\n\t"
"rolw #8,%/d0\n\t"
"movew %/d0,%/a0@\n\t"
"movew %/a1@+,%/d0\n\t"
"rolw #8,%/d0\n\t"
"movew %/d0,%/a0@\n\t"
"movew %/a1@+,%/d0\n\t"
"rolw #8,%/d0\n\t"
"movew %/d0,%/a0@\n\t"
"dbra %/d6,1b"
:
: "g" (port), "g" (buf), "g" (nr)
: "d0", "a0", "a1", "d6");
}
#define __raw_writel raw_outl
#endif /* __KERNEL__ */
#endif /* _RAW_IO_H */