mirror of
https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git
synced 2024-08-21 16:29:59 +00:00
![Linus Torvalds](/assets/img/avatar_default.png)
Initial git repository build. I'm not bothering with the full history, even though we have it. We can create a separate "historical" git archive of that later if we want to, and in the meantime it's about 3.2GB when imported into git - space that would just make the early git days unnecessarily complicated, when we don't have a lot of good infrastructure for it. Let it rip!
488 lines
9.7 KiB
C
488 lines
9.7 KiB
C
/*
|
|
* arch/parisc/lib/io.c
|
|
*
|
|
* Copyright (c) Matthew Wilcox 2001 for Hewlett-Packard
|
|
* Copyright (c) Randolph Chung 2001 <tausq@debian.org>
|
|
*
|
|
* IO accessing functions which shouldn't be inlined because they're too big
|
|
*/
|
|
|
|
#include <linux/kernel.h>
|
|
#include <linux/module.h>
|
|
#include <asm/io.h>
|
|
|
|
/* Copies a block of memory to a device in an efficient manner.
|
|
* Assumes the device can cope with 32-bit transfers. If it can't,
|
|
* don't use this function.
|
|
*/
|
|
void memcpy_toio(volatile void __iomem *dst, const void *src, int count)
|
|
{
|
|
if (((unsigned long)dst & 3) != ((unsigned long)src & 3))
|
|
goto bytecopy;
|
|
while ((unsigned long)dst & 3) {
|
|
writeb(*(char *)src, dst++);
|
|
src++;
|
|
count--;
|
|
}
|
|
while (count > 3) {
|
|
__raw_writel(*(u32 *)src, dst);
|
|
src += 4;
|
|
dst += 4;
|
|
count -= 4;
|
|
}
|
|
bytecopy:
|
|
while (count--) {
|
|
writeb(*(char *)src, dst++);
|
|
src++;
|
|
}
|
|
}
|
|
|
|
/*
|
|
** Copies a block of memory from a device in an efficient manner.
|
|
** Assumes the device can cope with 32-bit transfers. If it can't,
|
|
** don't use this function.
|
|
**
|
|
** CR16 counts on C3000 reading 256 bytes from Symbios 896 RAM:
|
|
** 27341/64 = 427 cyc per int
|
|
** 61311/128 = 478 cyc per short
|
|
** 122637/256 = 479 cyc per byte
|
|
** Ergo bus latencies dominant (not transfer size).
|
|
** Minimize total number of transfers at cost of CPU cycles.
|
|
** TODO: only look at src alignment and adjust the stores to dest.
|
|
*/
|
|
void memcpy_fromio(void *dst, const volatile void __iomem *src, int count)
|
|
{
|
|
/* first compare alignment of src/dst */
|
|
if ( (((unsigned long)dst ^ (unsigned long)src) & 1) || (count < 2) )
|
|
goto bytecopy;
|
|
|
|
if ( (((unsigned long)dst ^ (unsigned long)src) & 2) || (count < 4) )
|
|
goto shortcopy;
|
|
|
|
/* Then check for misaligned start address */
|
|
if ((unsigned long)src & 1) {
|
|
*(u8 *)dst = readb(src);
|
|
src++;
|
|
dst++;
|
|
count--;
|
|
if (count < 2) goto bytecopy;
|
|
}
|
|
|
|
if ((unsigned long)src & 2) {
|
|
*(u16 *)dst = __raw_readw(src);
|
|
src += 2;
|
|
dst += 2;
|
|
count -= 2;
|
|
}
|
|
|
|
while (count > 3) {
|
|
*(u32 *)dst = __raw_readl(src);
|
|
dst += 4;
|
|
src += 4;
|
|
count -= 4;
|
|
}
|
|
|
|
shortcopy:
|
|
while (count > 1) {
|
|
*(u16 *)dst = __raw_readw(src);
|
|
src += 2;
|
|
dst += 2;
|
|
count -= 2;
|
|
}
|
|
|
|
bytecopy:
|
|
while (count--) {
|
|
*(char *)dst = readb(src);
|
|
src++;
|
|
dst++;
|
|
}
|
|
}
|
|
|
|
/* Sets a block of memory on a device to a given value.
|
|
* Assumes the device can cope with 32-bit transfers. If it can't,
|
|
* don't use this function.
|
|
*/
|
|
void memset_io(volatile void __iomem *addr, unsigned char val, int count)
|
|
{
|
|
u32 val32 = (val << 24) | (val << 16) | (val << 8) | val;
|
|
while ((unsigned long)addr & 3) {
|
|
writeb(val, addr++);
|
|
count--;
|
|
}
|
|
while (count > 3) {
|
|
__raw_writel(val32, addr);
|
|
addr += 4;
|
|
count -= 4;
|
|
}
|
|
while (count--) {
|
|
writeb(val, addr++);
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Read COUNT 8-bit bytes from port PORT into memory starting at
|
|
* SRC.
|
|
*/
|
|
void insb (unsigned long port, void *dst, unsigned long count)
|
|
{
|
|
unsigned char *p;
|
|
|
|
p = (unsigned char *)dst;
|
|
|
|
while (((unsigned long)p) & 0x3) {
|
|
if (!count)
|
|
return;
|
|
count--;
|
|
*p = inb(port);
|
|
p++;
|
|
}
|
|
|
|
while (count >= 4) {
|
|
unsigned int w;
|
|
count -= 4;
|
|
w = inb(port) << 24;
|
|
w |= inb(port) << 16;
|
|
w |= inb(port) << 8;
|
|
w |= inb(port);
|
|
*(unsigned int *) p = w;
|
|
p += 4;
|
|
}
|
|
|
|
while (count) {
|
|
--count;
|
|
*p = inb(port);
|
|
p++;
|
|
}
|
|
}
|
|
|
|
|
|
/*
|
|
* Read COUNT 16-bit words from port PORT into memory starting at
|
|
* SRC. SRC must be at least short aligned. This is used by the
|
|
* IDE driver to read disk sectors. Performance is important, but
|
|
* the interfaces seems to be slow: just using the inlined version
|
|
* of the inw() breaks things.
|
|
*/
|
|
void insw (unsigned long port, void *dst, unsigned long count)
|
|
{
|
|
unsigned int l = 0, l2;
|
|
unsigned char *p;
|
|
|
|
p = (unsigned char *)dst;
|
|
|
|
if (!count)
|
|
return;
|
|
|
|
switch (((unsigned long)p) & 0x3)
|
|
{
|
|
case 0x00: /* Buffer 32-bit aligned */
|
|
while (count>=2) {
|
|
|
|
count -= 2;
|
|
l = cpu_to_le16(inw(port)) << 16;
|
|
l |= cpu_to_le16(inw(port));
|
|
*(unsigned int *)p = l;
|
|
p += 4;
|
|
}
|
|
if (count) {
|
|
*(unsigned short *)p = cpu_to_le16(inw(port));
|
|
}
|
|
break;
|
|
|
|
case 0x02: /* Buffer 16-bit aligned */
|
|
*(unsigned short *)p = cpu_to_le16(inw(port));
|
|
p += 2;
|
|
count--;
|
|
while (count>=2) {
|
|
|
|
count -= 2;
|
|
l = cpu_to_le16(inw(port)) << 16;
|
|
l |= cpu_to_le16(inw(port));
|
|
*(unsigned int *)p = l;
|
|
p += 4;
|
|
}
|
|
if (count) {
|
|
*(unsigned short *)p = cpu_to_le16(inw(port));
|
|
}
|
|
break;
|
|
|
|
case 0x01: /* Buffer 8-bit aligned */
|
|
case 0x03:
|
|
/* I don't bother with 32bit transfers
|
|
* in this case, 16bit will have to do -- DE */
|
|
--count;
|
|
|
|
l = cpu_to_le16(inw(port));
|
|
*p = l >> 8;
|
|
p++;
|
|
while (count--)
|
|
{
|
|
l2 = cpu_to_le16(inw(port));
|
|
*(unsigned short *)p = (l & 0xff) << 8 | (l2 >> 8);
|
|
p += 2;
|
|
l = l2;
|
|
}
|
|
*p = l & 0xff;
|
|
break;
|
|
}
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
* Read COUNT 32-bit words from port PORT into memory starting at
|
|
* SRC. Now works with any alignment in SRC. Performance is important,
|
|
* but the interfaces seems to be slow: just using the inlined version
|
|
* of the inl() breaks things.
|
|
*/
|
|
void insl (unsigned long port, void *dst, unsigned long count)
|
|
{
|
|
unsigned int l = 0, l2;
|
|
unsigned char *p;
|
|
|
|
p = (unsigned char *)dst;
|
|
|
|
if (!count)
|
|
return;
|
|
|
|
switch (((unsigned long) dst) & 0x3)
|
|
{
|
|
case 0x00: /* Buffer 32-bit aligned */
|
|
while (count--)
|
|
{
|
|
*(unsigned int *)p = cpu_to_le32(inl(port));
|
|
p += 4;
|
|
}
|
|
break;
|
|
|
|
case 0x02: /* Buffer 16-bit aligned */
|
|
--count;
|
|
|
|
l = cpu_to_le32(inl(port));
|
|
*(unsigned short *)p = l >> 16;
|
|
p += 2;
|
|
|
|
while (count--)
|
|
{
|
|
l2 = cpu_to_le32(inl(port));
|
|
*(unsigned int *)p = (l & 0xffff) << 16 | (l2 >> 16);
|
|
p += 4;
|
|
l = l2;
|
|
}
|
|
*(unsigned short *)p = l & 0xffff;
|
|
break;
|
|
case 0x01: /* Buffer 8-bit aligned */
|
|
--count;
|
|
|
|
l = cpu_to_le32(inl(port));
|
|
*(unsigned char *)p = l >> 24;
|
|
p++;
|
|
*(unsigned short *)p = (l >> 8) & 0xffff;
|
|
p += 2;
|
|
while (count--)
|
|
{
|
|
l2 = cpu_to_le32(inl(port));
|
|
*(unsigned int *)p = (l & 0xff) << 24 | (l2 >> 8);
|
|
p += 4;
|
|
l = l2;
|
|
}
|
|
*p = l & 0xff;
|
|
break;
|
|
case 0x03: /* Buffer 8-bit aligned */
|
|
--count;
|
|
|
|
l = cpu_to_le32(inl(port));
|
|
*p = l >> 24;
|
|
p++;
|
|
while (count--)
|
|
{
|
|
l2 = cpu_to_le32(inl(port));
|
|
*(unsigned int *)p = (l & 0xffffff) << 8 | l2 >> 24;
|
|
p += 4;
|
|
l = l2;
|
|
}
|
|
*(unsigned short *)p = (l >> 8) & 0xffff;
|
|
p += 2;
|
|
*p = l & 0xff;
|
|
break;
|
|
}
|
|
}
|
|
|
|
|
|
/*
|
|
* Like insb but in the opposite direction.
|
|
* Don't worry as much about doing aligned memory transfers:
|
|
* doing byte reads the "slow" way isn't nearly as slow as
|
|
* doing byte writes the slow way (no r-m-w cycle).
|
|
*/
|
|
void outsb(unsigned long port, const void * src, unsigned long count)
|
|
{
|
|
const unsigned char *p;
|
|
|
|
p = (const unsigned char *)src;
|
|
while (count) {
|
|
count--;
|
|
outb(*p, port);
|
|
p++;
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Like insw but in the opposite direction. This is used by the IDE
|
|
* driver to write disk sectors. Performance is important, but the
|
|
* interfaces seems to be slow: just using the inlined version of the
|
|
* outw() breaks things.
|
|
*/
|
|
void outsw (unsigned long port, const void *src, unsigned long count)
|
|
{
|
|
unsigned int l = 0, l2;
|
|
const unsigned char *p;
|
|
|
|
p = (const unsigned char *)src;
|
|
|
|
if (!count)
|
|
return;
|
|
|
|
switch (((unsigned long)p) & 0x3)
|
|
{
|
|
case 0x00: /* Buffer 32-bit aligned */
|
|
while (count>=2) {
|
|
count -= 2;
|
|
l = *(unsigned int *)p;
|
|
p += 4;
|
|
outw(le16_to_cpu(l >> 16), port);
|
|
outw(le16_to_cpu(l & 0xffff), port);
|
|
}
|
|
if (count) {
|
|
outw(le16_to_cpu(*(unsigned short*)p), port);
|
|
}
|
|
break;
|
|
|
|
case 0x02: /* Buffer 16-bit aligned */
|
|
|
|
outw(le16_to_cpu(*(unsigned short*)p), port);
|
|
p += 2;
|
|
count--;
|
|
|
|
while (count>=2) {
|
|
count -= 2;
|
|
l = *(unsigned int *)p;
|
|
p += 4;
|
|
outw(le16_to_cpu(l >> 16), port);
|
|
outw(le16_to_cpu(l & 0xffff), port);
|
|
}
|
|
if (count) {
|
|
outw(le16_to_cpu(*(unsigned short *)p), port);
|
|
}
|
|
break;
|
|
|
|
case 0x01: /* Buffer 8-bit aligned */
|
|
/* I don't bother with 32bit transfers
|
|
* in this case, 16bit will have to do -- DE */
|
|
|
|
l = *p << 8;
|
|
p++;
|
|
count--;
|
|
while (count)
|
|
{
|
|
count--;
|
|
l2 = *(unsigned short *)p;
|
|
p += 2;
|
|
outw(le16_to_cpu(l | l2 >> 8), port);
|
|
l = l2 << 8;
|
|
}
|
|
l2 = *(unsigned char *)p;
|
|
outw (le16_to_cpu(l | l2>>8), port);
|
|
break;
|
|
|
|
}
|
|
}
|
|
|
|
|
|
/*
|
|
* Like insl but in the opposite direction. This is used by the IDE
|
|
* driver to write disk sectors. Works with any alignment in SRC.
|
|
* Performance is important, but the interfaces seems to be slow:
|
|
* just using the inlined version of the outl() breaks things.
|
|
*/
|
|
void outsl (unsigned long port, const void *src, unsigned long count)
|
|
{
|
|
unsigned int l = 0, l2;
|
|
const unsigned char *p;
|
|
|
|
p = (const unsigned char *)src;
|
|
|
|
if (!count)
|
|
return;
|
|
|
|
switch (((unsigned long)p) & 0x3)
|
|
{
|
|
case 0x00: /* Buffer 32-bit aligned */
|
|
while (count--)
|
|
{
|
|
outl(le32_to_cpu(*(unsigned int *)p), port);
|
|
p += 4;
|
|
}
|
|
break;
|
|
|
|
case 0x02: /* Buffer 16-bit aligned */
|
|
--count;
|
|
|
|
l = *(unsigned short *)p;
|
|
p += 2;
|
|
|
|
while (count--)
|
|
{
|
|
l2 = *(unsigned int *)p;
|
|
p += 4;
|
|
outl (le32_to_cpu(l << 16 | l2 >> 16), port);
|
|
l = l2;
|
|
}
|
|
l2 = *(unsigned short *)p;
|
|
outl (le32_to_cpu(l << 16 | l2), port);
|
|
break;
|
|
case 0x01: /* Buffer 8-bit aligned */
|
|
--count;
|
|
|
|
l = *p << 24;
|
|
p++;
|
|
l |= *(unsigned short *)p << 8;
|
|
p += 2;
|
|
|
|
while (count--)
|
|
{
|
|
l2 = *(unsigned int *)p;
|
|
p += 4;
|
|
outl (le32_to_cpu(l | l2 >> 24), port);
|
|
l = l2 << 8;
|
|
}
|
|
l2 = *p;
|
|
outl (le32_to_cpu(l | l2), port);
|
|
break;
|
|
case 0x03: /* Buffer 8-bit aligned */
|
|
--count;
|
|
|
|
l = *p << 24;
|
|
p++;
|
|
|
|
while (count--)
|
|
{
|
|
l2 = *(unsigned int *)p;
|
|
p += 4;
|
|
outl (le32_to_cpu(l | l2 >> 8), port);
|
|
l = l2 << 24;
|
|
}
|
|
l2 = *(unsigned short *)p << 16;
|
|
p += 2;
|
|
l2 |= *p;
|
|
outl (le32_to_cpu(l | l2), port);
|
|
break;
|
|
}
|
|
}
|
|
|
|
EXPORT_SYMBOL(insb);
|
|
EXPORT_SYMBOL(insw);
|
|
EXPORT_SYMBOL(insl);
|
|
EXPORT_SYMBOL(outsb);
|
|
EXPORT_SYMBOL(outsw);
|
|
EXPORT_SYMBOL(outsl);
|