[PATCH] sh: I/O routine cleanups and ioremap() overhaul

This introduces a few changes in the way that the I/O routines are defined on
SH, specifically so that things like the iomap API properly wrap through the
machvec for board-specific quirks.

In addition to this, the old p3_ioremap() work is converted to a more generic
__ioremap() that will map through the PMB if it's available, or fall back on
page tables for everything else.

An alpha-like IO_CONCAT is also added so we can start to clean up the
board-specific io.h mess, which will be handled in board update patches..

Signed-off-by: Paul Mundt <lethal@linux-sh.org>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
diff --git a/include/asm-sh/io.h b/include/asm-sh/io.h
index 6bc343fe..b0b2937 100644
--- a/include/asm-sh/io.h
+++ b/include/asm-sh/io.h
@@ -11,7 +11,7 @@
  * For read{b,w,l} and write{b,w,l} there are also __raw versions, which
  * do not have a memory barrier after them.
  *
- * In addition, we have 
+ * In addition, we have
  *   ctrl_in{b,w,l}/ctrl_out{b,w,l} for SuperH specific I/O.
  *   which are processor specific.
  */
@@ -23,19 +23,27 @@
  *  inb   by default expands to _inb, but the machine specific code may
  *        define it to __inb if it chooses.
  */
-
+#include <linux/config.h>
 #include <asm/cache.h>
 #include <asm/system.h>
 #include <asm/addrspace.h>
 #include <asm/machvec.h>
-#include <linux/config.h>
+#include <asm/pgtable.h>
+#include <asm-generic/iomap.h>
+
+#ifdef __KERNEL__
 
 /*
  * Depending on which platform we are running on, we need different
  * I/O functions.
  */
+#define __IO_PREFIX	generic
+#include <asm/io_generic.h>
 
-#ifdef __KERNEL__
+#define maybebadio(port) \
+  printk(KERN_ERR "bad PC-like io %s:%u for port 0x%lx at 0x%08x\n", \
+	 __FUNCTION__, __LINE__, (port), (u32)__builtin_return_address(0))
+
 /*
  * Since boards are able to define their own set of I/O routines through
  * their respective machine vector, we always wrap through the mv.
@@ -44,113 +52,120 @@
  * a given routine, it will be wrapped to generic code at run-time.
  */
 
-# define __inb(p)	sh_mv.mv_inb((p))
-# define __inw(p)	sh_mv.mv_inw((p))
-# define __inl(p)	sh_mv.mv_inl((p))
-# define __outb(x,p)	sh_mv.mv_outb((x),(p))
-# define __outw(x,p)	sh_mv.mv_outw((x),(p))
-# define __outl(x,p)	sh_mv.mv_outl((x),(p))
+#define __inb(p)	sh_mv.mv_inb((p))
+#define __inw(p)	sh_mv.mv_inw((p))
+#define __inl(p)	sh_mv.mv_inl((p))
+#define __outb(x,p)	sh_mv.mv_outb((x),(p))
+#define __outw(x,p)	sh_mv.mv_outw((x),(p))
+#define __outl(x,p)	sh_mv.mv_outl((x),(p))
 
-# define __inb_p(p)	sh_mv.mv_inb_p((p))
-# define __inw_p(p)	sh_mv.mv_inw_p((p))
-# define __inl_p(p)	sh_mv.mv_inl_p((p))
-# define __outb_p(x,p)	sh_mv.mv_outb_p((x),(p))
-# define __outw_p(x,p)	sh_mv.mv_outw_p((x),(p))
-# define __outl_p(x,p)	sh_mv.mv_outl_p((x),(p))
+#define __inb_p(p)	sh_mv.mv_inb_p((p))
+#define __inw_p(p)	sh_mv.mv_inw_p((p))
+#define __inl_p(p)	sh_mv.mv_inl_p((p))
+#define __outb_p(x,p)	sh_mv.mv_outb_p((x),(p))
+#define __outw_p(x,p)	sh_mv.mv_outw_p((x),(p))
+#define __outl_p(x,p)	sh_mv.mv_outl_p((x),(p))
 
-# define __insb(p,b,c)	sh_mv.mv_insb((p), (b), (c))
-# define __insw(p,b,c)	sh_mv.mv_insw((p), (b), (c))
-# define __insl(p,b,c)	sh_mv.mv_insl((p), (b), (c))
-# define __outsb(p,b,c)	sh_mv.mv_outsb((p), (b), (c))
-# define __outsw(p,b,c)	sh_mv.mv_outsw((p), (b), (c))
-# define __outsl(p,b,c)	sh_mv.mv_outsl((p), (b), (c))
+#define __insb(p,b,c)	sh_mv.mv_insb((p), (b), (c))
+#define __insw(p,b,c)	sh_mv.mv_insw((p), (b), (c))
+#define __insl(p,b,c)	sh_mv.mv_insl((p), (b), (c))
+#define __outsb(p,b,c)	sh_mv.mv_outsb((p), (b), (c))
+#define __outsw(p,b,c)	sh_mv.mv_outsw((p), (b), (c))
+#define __outsl(p,b,c)	sh_mv.mv_outsl((p), (b), (c))
 
-# define __readb(a)	sh_mv.mv_readb((a))
-# define __readw(a)	sh_mv.mv_readw((a))
-# define __readl(a)	sh_mv.mv_readl((a))
-# define __writeb(v,a)	sh_mv.mv_writeb((v),(a))
-# define __writew(v,a)	sh_mv.mv_writew((v),(a))
-# define __writel(v,a)	sh_mv.mv_writel((v),(a))
+#define __readb(a)	sh_mv.mv_readb((a))
+#define __readw(a)	sh_mv.mv_readw((a))
+#define __readl(a)	sh_mv.mv_readl((a))
+#define __writeb(v,a)	sh_mv.mv_writeb((v),(a))
+#define __writew(v,a)	sh_mv.mv_writew((v),(a))
+#define __writel(v,a)	sh_mv.mv_writel((v),(a))
 
-# define __ioremap(a,s)	sh_mv.mv_ioremap((a), (s))
-# define __iounmap(a)	sh_mv.mv_iounmap((a))
+#define inb		__inb
+#define inw		__inw
+#define inl		__inl
+#define outb		__outb
+#define outw		__outw
+#define outl		__outl
 
-# define __isa_port2addr(a)	sh_mv.mv_isa_port2addr(a)
+#define inb_p		__inb_p
+#define inw_p		__inw_p
+#define inl_p		__inl_p
+#define outb_p		__outb_p
+#define outw_p		__outw_p
+#define outl_p		__outl_p
 
-# define inb		__inb
-# define inw		__inw
-# define inl		__inl
-# define outb		__outb
-# define outw		__outw
-# define outl		__outl
+#define insb		__insb
+#define insw		__insw
+#define insl		__insl
+#define outsb		__outsb
+#define outsw		__outsw
+#define outsl		__outsl
 
-# define inb_p		__inb_p
-# define inw_p		__inw_p
-# define inl_p		__inl_p
-# define outb_p		__outb_p
-# define outw_p		__outw_p
-# define outl_p		__outl_p
-
-# define insb		__insb
-# define insw		__insw
-# define insl		__insl
-# define outsb		__outsb
-# define outsw		__outsw
-# define outsl		__outsl
-
-# define __raw_readb	__readb
-# define __raw_readw	__readw
-# define __raw_readl	__readl
-# define __raw_writeb	__writeb
-# define __raw_writew	__writew
-# define __raw_writel	__writel
+#define __raw_readb(a)		__readb((void __iomem *)(a))
+#define __raw_readw(a)		__readw((void __iomem *)(a))
+#define __raw_readl(a)		__readl((void __iomem *)(a))
+#define __raw_writeb(v, a)	__writeb(v, (void __iomem *)(a))
+#define __raw_writew(v, a)	__writew(v, (void __iomem *)(a))
+#define __raw_writel(v, a)	__writel(v, (void __iomem *)(a))
 
 /*
  * The platform header files may define some of these macros to use
  * the inlined versions where appropriate.  These macros may also be
  * redefined by userlevel programs.
  */
-#ifdef __raw_readb
-# define readb(a)	({ unsigned long r_ = __raw_readb((unsigned long)a); mb(); r_; })
+#ifdef __readb
+# define readb(a)	({ unsigned long r_ = __raw_readb(a); mb(); r_; })
 #endif
 #ifdef __raw_readw
-# define readw(a)	({ unsigned long r_ = __raw_readw((unsigned long)a); mb(); r_; })
+# define readw(a)	({ unsigned long r_ = __raw_readw(a); mb(); r_; })
 #endif
 #ifdef __raw_readl
-# define readl(a)	({ unsigned long r_ = __raw_readl((unsigned long)a); mb(); r_; })
+# define readl(a)	({ unsigned long r_ = __raw_readl(a); mb(); r_; })
 #endif
 
 #ifdef __raw_writeb
-# define writeb(v,a)	({ __raw_writeb((v),(unsigned long)(a)); mb(); })
+# define writeb(v,a)	({ __raw_writeb((v),(a)); mb(); })
 #endif
 #ifdef __raw_writew
-# define writew(v,a)	({ __raw_writew((v),(unsigned long)(a)); mb(); })
+# define writew(v,a)	({ __raw_writew((v),(a)); mb(); })
 #endif
 #ifdef __raw_writel
-# define writel(v,a)	({ __raw_writel((v),(unsigned long)(a)); mb(); })
+# define writel(v,a)	({ __raw_writel((v),(a)); mb(); })
 #endif
 
 #define readb_relaxed(a) readb(a)
 #define readw_relaxed(a) readw(a)
 #define readl_relaxed(a) readl(a)
 
-#define mmiowb()
+/* Simple MMIO */
+#define ioread8(a)		readb(a)
+#define ioread16(a)		readw(a)
+#define ioread16be(a)		be16_to_cpu(__raw_readw((a)))
+#define ioread32(a)		readl(a)
+#define ioread32be(a)		be32_to_cpu(__raw_readl((a)))
 
-/*
- * If the platform has PC-like I/O, this function converts the offset into
- * an address.
- */
-static __inline__ unsigned long isa_port2addr(unsigned long offset)
-{
-	return __isa_port2addr(offset);
-}
+#define iowrite8(v,a)		writeb((v),(a))
+#define iowrite16(v,a)		writew((v),(a))
+#define iowrite16be(v,a)	__raw_writew(cpu_to_be16((v)),(a))
+#define iowrite32(v,a)		writel((v),(a))
+#define iowrite32be(v,a)	__raw_writel(cpu_to_be32((v)),(a))
+
+#define ioread8_rep(a,d,c)	insb((a),(d),(c))
+#define ioread16_rep(a,d,c)	insw((a),(d),(c))
+#define ioread32_rep(a,d,c)	insl((a),(d),(c))
+
+#define iowrite8_rep(a,s,c)	outsb((a),(s),(c))
+#define iowrite16_rep(a,s,c)	outsw((a),(s),(c))
+#define iowrite32_rep(a,s,c)	outsl((a),(s),(c))
+
+#define mmiowb()	wmb()	/* synco on SH-4A, otherwise a nop */
 
 /*
  * This function provides a method for the generic case where a board-specific
- * isa_port2addr simply needs to return the port + some arbitrary port base.
+ * ioport_map simply needs to return the port + some arbitrary port base.
  *
  * We use this at board setup time to implicitly set the port base, and
- * as a result, we can use the generic isa_port2addr.
+ * as a result, we can use the generic ioport_map.
  */
 static inline void __set_io_port_base(unsigned long pbase)
 {
@@ -159,51 +174,52 @@
 	generic_io_base = pbase;
 }
 
-#define isa_readb(a) readb(isa_port2addr(a))
-#define isa_readw(a) readw(isa_port2addr(a))
-#define isa_readl(a) readl(isa_port2addr(a))
-#define isa_writeb(b,a) writeb(b,isa_port2addr(a))
-#define isa_writew(w,a) writew(w,isa_port2addr(a))
-#define isa_writel(l,a) writel(l,isa_port2addr(a))
+#define isa_readb(a) readb(ioport_map(a, 1))
+#define isa_readw(a) readw(ioport_map(a, 2))
+#define isa_readl(a) readl(ioport_map(a, 4))
+#define isa_writeb(b,a) writeb(b,ioport_map(a, 1))
+#define isa_writew(w,a) writew(w,ioport_map(a, 2))
+#define isa_writel(l,a) writel(l,ioport_map(a, 4))
+
 #define isa_memset_io(a,b,c) \
-  memset((void *)(isa_port2addr((unsigned long)a)),(b),(c))
+  memset((void *)(ioport_map((unsigned long)(a), 1)),(b),(c))
 #define isa_memcpy_fromio(a,b,c) \
-  memcpy((a),(void *)(isa_port2addr((unsigned long)(b))),(c))
+  memcpy((a),(void *)(ioport_map((unsigned long)(b), 1)),(c))
 #define isa_memcpy_toio(a,b,c) \
-  memcpy((void *)(isa_port2addr((unsigned long)(a))),(b),(c))
+  memcpy((void *)(ioport_map((unsigned long)(a), 1)),(b),(c))
 
 /* We really want to try and get these to memcpy etc */
-extern void memcpy_fromio(void *, unsigned long, unsigned long);
-extern void memcpy_toio(unsigned long, const void *, unsigned long);
-extern void memset_io(unsigned long, int, unsigned long);
+extern void memcpy_fromio(void *, volatile void __iomem *, unsigned long);
+extern void memcpy_toio(volatile void __iomem *, const void *, unsigned long);
+extern void memset_io(volatile void __iomem *, int, unsigned long);
 
 /* SuperH on-chip I/O functions */
-static __inline__ unsigned char ctrl_inb(unsigned long addr)
+static inline unsigned char ctrl_inb(unsigned long addr)
 {
 	return *(volatile unsigned char*)addr;
 }
 
-static __inline__ unsigned short ctrl_inw(unsigned long addr)
+static inline unsigned short ctrl_inw(unsigned long addr)
 {
 	return *(volatile unsigned short*)addr;
 }
 
-static __inline__ unsigned int ctrl_inl(unsigned long addr)
+static inline unsigned int ctrl_inl(unsigned long addr)
 {
 	return *(volatile unsigned long*)addr;
 }
 
-static __inline__ void ctrl_outb(unsigned char b, unsigned long addr)
+static inline void ctrl_outb(unsigned char b, unsigned long addr)
 {
 	*(volatile unsigned char*)addr = b;
 }
 
-static __inline__ void ctrl_outw(unsigned short b, unsigned long addr)
+static inline void ctrl_outw(unsigned short b, unsigned long addr)
 {
 	*(volatile unsigned short*)addr = b;
 }
 
-static __inline__ void ctrl_outl(unsigned int b, unsigned long addr)
+static inline void ctrl_outl(unsigned int b, unsigned long addr)
 {
         *(volatile unsigned long*)addr = b;
 }
@@ -214,12 +230,12 @@
  * Change virtual addresses to physical addresses and vv.
  * These are trivial on the 1:1 Linux/SuperH mapping
  */
-static __inline__ unsigned long virt_to_phys(volatile void * address)
+static inline unsigned long virt_to_phys(volatile void *address)
 {
 	return PHYSADDR(address);
 }
 
-static __inline__ void * phys_to_virt(unsigned long address)
+static inline void *phys_to_virt(unsigned long address)
 {
 	return (void *)P1SEGADDR(address);
 }
@@ -234,27 +250,60 @@
  * differently. On the x86 architecture, we just read/write the
  * memory location directly.
  *
- * On SH, we have the whole physical address space mapped at all times
- * (as MIPS does), so "ioremap()" and "iounmap()" do not need to do
- * anything.  (This isn't true for all machines but we still handle
- * these cases with wired TLB entries anyway ...)
+ * On SH, we traditionally have the whole physical address space mapped
+ * at all times (as MIPS does), so "ioremap()" and "iounmap()" do not
+ * need to do anything but place the address in the proper segment. This
+ * is true for P1 and P2 addresses, as well as some P3 ones. However,
+ * most of the P3 addresses and newer cores using extended addressing
+ * need to map through page tables, so the ioremap() implementation
+ * becomes a bit more complicated. See arch/sh/mm/ioremap.c for
+ * additional notes on this.
  *
  * We cheat a bit and always return uncachable areas until we've fixed
- * the drivers to handle caching properly.  
+ * the drivers to handle caching properly.
  */
-static __inline__ void * ioremap(unsigned long offset, unsigned long size)
+#ifdef CONFIG_MMU
+void __iomem *__ioremap(unsigned long offset, unsigned long size,
+			unsigned long flags);
+void __iounmap(void __iomem *addr);
+#else
+#define __ioremap(offset, size, flags)	((void __iomem *)(offset))
+#define __iounmap(addr)			do { } while (0)
+#endif /* CONFIG_MMU */
+
+static inline void __iomem *
+__ioremap_mode(unsigned long offset, unsigned long size, unsigned long flags)
 {
-	return __ioremap(offset, size);
+	unsigned long last_addr = offset + size - 1;
+
+	/*
+	 * For P1 and P2 space this is trivial, as everything is already
+	 * mapped. Uncached access for P1 addresses are done through P2.
+	 * In the P3 case or for addresses outside of the 29-bit space,
+	 * mapping must be done by the PMB or by using page tables.
+	 */
+	if (likely(PXSEG(offset) < P3SEG && PXSEG(last_addr) < P3SEG)) {
+		if (unlikely(flags & _PAGE_CACHABLE))
+			return (void __iomem *)P1SEGADDR(offset);
+
+		return (void __iomem *)P2SEGADDR(offset);
+	}
+
+	return __ioremap(offset, size, flags);
 }
 
-static __inline__ void iounmap(void *addr)
-{
-	return __iounmap(addr);
-}
+#define ioremap(offset, size)				\
+	__ioremap_mode((offset), (size), 0)
+#define ioremap_nocache(offset, size)			\
+	__ioremap_mode((offset), (size), 0)
+#define ioremap_cache(offset, size)			\
+	__ioremap_mode((offset), (size), _PAGE_CACHABLE)
+#define p3_ioremap(offset, size, flags)			\
+	__ioremap((offset), (size), (flags))
+#define iounmap(addr)					\
+	__iounmap((addr))
 
-#define ioremap_nocache(off,size) ioremap(off,size)
-
-static __inline__ int check_signature(unsigned long io_addr,
+static inline int check_signature(char __iomem *io_addr,
 			const unsigned char *signature, int length)
 {
 	int retval = 0;