[deliverable/linux.git] / include / asm-sparc / checksum.h

/* $Id: checksum.h,v 1.33 2002/02/01 22:01:05 davem Exp $ */
#ifndef __SPARC_CHECKSUM_H
#define __SPARC_CHECKSUM_H

/*  checksum.h:  IP/UDP/TCP checksum routines on the Sparc.
 *
 *  Copyright(C) 1995 Linus Torvalds
 *  Copyright(C) 1995 Miguel de Icaza
 *  Copyright(C) 1996 David S. Miller
 *  Copyright(C) 1996 Eddie C. Dost
 *  Copyright(C) 1997 Jakub Jelinek
 *
 * derived from:
 *	Alpha checksum c-code
 *      ix86 inline assembly
 *      RFC1071 Computing the Internet Checksum
 */
 
#include <linux/in6.h>
#include <asm/uaccess.h>

/* computes the checksum of a memory block at buff, length len,
 * and adds in "sum" (32-bit)
 *
 * returns a 32-bit number suitable for feeding into itself
 * or csum_tcpudp_magic
 *
 * this function must be called with even lengths, except
 * for the last fragment, which may be odd
 *
 * it's best to have buff aligned on a 32-bit boundary
 */
extern __wsum csum_partial(const void *buff, int len, __wsum sum);

/* the same as csum_partial, but copies from fs:src while it
 * checksums
 *
 * here even more important to align src and dst on a 32-bit (or even
 * better 64-bit) boundary
 */

extern unsigned int __csum_partial_copy_sparc_generic (const unsigned char *, unsigned char *);

static inline __wsum
csum_partial_copy_nocheck(const void *src, void *dst, int len, __wsum sum)
{
	register unsigned int ret asm("o0") = (unsigned int)src;
	register char *d asm("o1") = dst;
	register int l asm("g1") = len;

	__asm__ __volatile__ (
		"call __csum_partial_copy_sparc_generic\n\t"
		" mov %6, %%g7\n"
	: "=&r" (ret), "=&r" (d), "=&r" (l)
	: "0" (ret), "1" (d), "2" (l), "r" (sum)
	: "o2", "o3", "o4", "o5", "o7",
	  "g2", "g3", "g4", "g5", "g7",
	  "memory", "cc");
	return (__force __wsum)ret;
}

static inline __wsum
csum_partial_copy_from_user(const void __user *src, void *dst, int len,
			    __wsum sum, int *err)
  {
	register unsigned long ret asm("o0") = (unsigned long)src;
	register char *d asm("o1") = dst;
	register int l asm("g1") = len;
	register __wsum s asm("g7") = sum;

	__asm__ __volatile__ (
	".section __ex_table,#alloc\n\t"
	".align 4\n\t"
	".word 1f,2\n\t"
	".previous\n"
	"1:\n\t"
	"call __csum_partial_copy_sparc_generic\n\t"
	" st %8, [%%sp + 64]\n"
	: "=&r" (ret), "=&r" (d), "=&r" (l), "=&r" (s)
	: "0" (ret), "1" (d), "2" (l), "3" (s), "r" (err)
	: "o2", "o3", "o4", "o5", "o7", "g2", "g3", "g4", "g5",
	  "cc", "memory");
	return (__force __wsum)ret;
}
  
static inline __wsum
csum_partial_copy_to_user(const void *src, void __user *dst, int len,
			  __wsum sum, int *err)
{
	if (!access_ok (VERIFY_WRITE, dst, len)) {
		*err = -EFAULT;
		return sum;
	} else {
		register unsigned long ret asm("o0") = (unsigned long)src;
		register char __user *d asm("o1") = dst;
		register int l asm("g1") = len;
		register __wsum s asm("g7") = sum;

		__asm__ __volatile__ (
		".section __ex_table,#alloc\n\t"
		".align 4\n\t"
		".word 1f,1\n\t"
		".previous\n"
		"1:\n\t"
		"call __csum_partial_copy_sparc_generic\n\t"
		" st %8, [%%sp + 64]\n"
		: "=&r" (ret), "=&r" (d), "=&r" (l), "=&r" (s)
		: "0" (ret), "1" (d), "2" (l), "3" (s), "r" (err)
		: "o2", "o3", "o4", "o5", "o7",
		  "g2", "g3", "g4", "g5",
		  "cc", "memory");
		return (__force __wsum)ret;
	}
}

#define HAVE_CSUM_COPY_USER
#define csum_and_copy_to_user csum_partial_copy_to_user

/* ihl is always 5 or greater, almost always is 5, and iph is word aligned
 * the majority of the time.
 */
static inline __sum16 ip_fast_csum(const void *iph, unsigned int ihl)
{
	__sum16 sum;

	/* Note: We must read %2 before we touch %0 for the first time,
	 *       because GCC can legitimately use the same register for
	 *       both operands.
	 */
	__asm__ __volatile__("sub\t%2, 4, %%g4\n\t"
			     "ld\t[%1 + 0x00], %0\n\t"
			     "ld\t[%1 + 0x04], %%g2\n\t"
			     "ld\t[%1 + 0x08], %%g3\n\t"
			     "addcc\t%%g2, %0, %0\n\t"
			     "addxcc\t%%g3, %0, %0\n\t"
			     "ld\t[%1 + 0x0c], %%g2\n\t"
			     "ld\t[%1 + 0x10], %%g3\n\t"
			     "addxcc\t%%g2, %0, %0\n\t"
			     "addx\t%0, %%g0, %0\n"
			     "1:\taddcc\t%%g3, %0, %0\n\t"
			     "add\t%1, 4, %1\n\t"
			     "addxcc\t%0, %%g0, %0\n\t"
			     "subcc\t%%g4, 1, %%g4\n\t"
			     "be,a\t2f\n\t"
			     "sll\t%0, 16, %%g2\n\t"
			     "b\t1b\n\t"
			     "ld\t[%1 + 0x10], %%g3\n"
			     "2:\taddcc\t%0, %%g2, %%g2\n\t"
			     "srl\t%%g2, 16, %0\n\t"
			     "addx\t%0, %%g0, %0\n\t"
			     "xnor\t%%g0, %0, %0"
			     : "=r" (sum), "=&r" (iph)
			     : "r" (ihl), "1" (iph)
			     : "g2", "g3", "g4", "cc", "memory");
	return sum;
}

/* Fold a partial checksum without adding pseudo headers. */
static inline __sum16 csum_fold(__wsum sum)
{
	unsigned int tmp;

	__asm__ __volatile__("addcc\t%0, %1, %1\n\t"
			     "srl\t%1, 16, %1\n\t"
			     "addx\t%1, %%g0, %1\n\t"
			     "xnor\t%%g0, %1, %0"
			     : "=&r" (sum), "=r" (tmp)
			     : "0" (sum), "1" ((__force u32)sum<<16)
			     : "cc");
	return (__force __sum16)sum;
}

static inline __wsum csum_tcpudp_nofold(__be32 saddr, __be32 daddr,
					       unsigned short len,
					       unsigned short proto,
					       __wsum sum)
{
	__asm__ __volatile__("addcc\t%1, %0, %0\n\t"
			     "addxcc\t%2, %0, %0\n\t"
			     "addxcc\t%3, %0, %0\n\t"
			     "addx\t%0, %%g0, %0\n\t"
			     : "=r" (sum), "=r" (saddr)
			     : "r" (daddr), "r" (proto + len), "0" (sum),
			       "1" (saddr)
			     : "cc");
	return sum;
}

/*
 * computes the checksum of the TCP/UDP pseudo-header
 * returns a 16-bit checksum, already complemented
 */
static inline __sum16 csum_tcpudp_magic(__be32 saddr, __be32 daddr,
						   unsigned short len,
						   unsigned short proto,
						   __wsum sum)
{
	return csum_fold(csum_tcpudp_nofold(saddr,daddr,len,proto,sum));
}

#define _HAVE_ARCH_IPV6_CSUM

static inline __sum16 csum_ipv6_magic(const struct in6_addr *saddr,
				      const struct in6_addr *daddr,
				      __u32 len, unsigned short proto,
				      __wsum sum)
{
	__asm__ __volatile__ (
		"addcc	%3, %4, %%g4\n\t"
		"addxcc	%5, %%g4, %%g4\n\t"
		"ld	[%2 + 0x0c], %%g2\n\t"
		"ld	[%2 + 0x08], %%g3\n\t"
		"addxcc	%%g2, %%g4, %%g4\n\t"
		"ld	[%2 + 0x04], %%g2\n\t"
		"addxcc	%%g3, %%g4, %%g4\n\t"
		"ld	[%2 + 0x00], %%g3\n\t"
		"addxcc	%%g2, %%g4, %%g4\n\t"
		"ld	[%1 + 0x0c], %%g2\n\t"
		"addxcc	%%g3, %%g4, %%g4\n\t"
		"ld	[%1 + 0x08], %%g3\n\t"
		"addxcc	%%g2, %%g4, %%g4\n\t"
		"ld	[%1 + 0x04], %%g2\n\t"
		"addxcc	%%g3, %%g4, %%g4\n\t"
		"ld	[%1 + 0x00], %%g3\n\t"
		"addxcc	%%g2, %%g4, %%g4\n\t"
		"addxcc	%%g3, %%g4, %0\n\t"
		"addx	0, %0, %0\n"
		: "=&r" (sum)
		: "r" (saddr), "r" (daddr), 
		  "r"(htonl(len)), "r"(htonl(proto)), "r"(sum)
		: "g2", "g3", "g4", "cc");

	return csum_fold(sum);
}

/* this routine is used for miscellaneous IP-like checksums, mainly in icmp.c */
static inline __sum16 ip_compute_csum(const void *buff, int len)
{
	return csum_fold(csum_partial(buff, len, 0));
}

#endif /* !(__SPARC_CHECKSUM_H) */
Commit	Line	Data
1da177e4 LT	1	/* $Id: checksum.h,v 1.33 2002/02/01 22:01:05 davem Exp $ */
	2	#ifndef __SPARC_CHECKSUM_H
	3	#define __SPARC_CHECKSUM_H
	4
	5	/* checksum.h: IP/UDP/TCP checksum routines on the Sparc.
	6	*
	7	* Copyright(C) 1995 Linus Torvalds
	8	* Copyright(C) 1995 Miguel de Icaza
	9	* Copyright(C) 1996 David S. Miller
	10	* Copyright(C) 1996 Eddie C. Dost
	11	* Copyright(C) 1997 Jakub Jelinek
	12	*
	13	* derived from:
	14	* Alpha checksum c-code
	15	* ix86 inline assembly
	16	* RFC1071 Computing the Internet Checksum
	17	*/
	18
	19	#include <linux/in6.h>
	20	#include <asm/uaccess.h>
	21
	22	/* computes the checksum of a memory block at buff, length len,
	23	* and adds in "sum" (32-bit)
	24	*
	25	* returns a 32-bit number suitable for feeding into itself
	26	* or csum_tcpudp_magic
	27	*
	28	* this function must be called with even lengths, except
	29	* for the last fragment, which may be odd
	30	*
	31	* it's best to have buff aligned on a 32-bit boundary
	32	*/
16ec7e16	33	extern __wsum csum_partial(const void *buff, int len, __wsum sum);
1da177e4 LT	34
	35	/* the same as csum_partial, but copies from fs:src while it
	36	* checksums
	37	*
	38	* here even more important to align src and dst on a 32-bit (or even
	39	* better 64-bit) boundary
	40	*/
	41
	42	extern unsigned int __csum_partial_copy_sparc_generic (const unsigned char , unsigned char );
	43
16ec7e16 AV	44	static inline __wsum
16ec7e16 AV	45	csum_partial_copy_nocheck(const void src, void dst, int len, __wsum sum)
1da177e4 LT	46	{
	47	register unsigned int ret asm("o0") = (unsigned int)src;
	48	register char *d asm("o1") = dst;
	49	register int l asm("g1") = len;
	50
	51	__asm__ __volatile__ (
	52	"call __csum_partial_copy_sparc_generic\n\t"
	53	" mov %6, %%g7\n"
	54	: "=&r" (ret), "=&r" (d), "=&r" (l)
	55	: "0" (ret), "1" (d), "2" (l), "r" (sum)
	56	: "o2", "o3", "o4", "o5", "o7",
	57	"g2", "g3", "g4", "g5", "g7",
	58	"memory", "cc");
16ec7e16	59	return (__force __wsum)ret;
1da177e4 LT	60	}
1da177e4 LT	61
16ec7e16 AV	62	static inline __wsum
	63	csum_partial_copy_from_user(const void __user src, void dst, int len,
	64	__wsum sum, int *err)
1da177e4	65	{
16ec7e16 AV	66	register unsigned long ret asm("o0") = (unsigned long)src;
	67	register char *d asm("o1") = dst;
	68	register int l asm("g1") = len;
	69	register __wsum s asm("g7") = sum;
1da177e4	70
16ec7e16 AV	71	__asm__ __volatile__ (
	72	".section __ex_table,#alloc\n\t"
	73	".align 4\n\t"
	74	".word 1f,2\n\t"
	75	".previous\n"
	76	"1:\n\t"
	77	"call __csum_partial_copy_sparc_generic\n\t"
	78	" st %8, [%%sp + 64]\n"
	79	: "=&r" (ret), "=&r" (d), "=&r" (l), "=&r" (s)
	80	: "0" (ret), "1" (d), "2" (l), "3" (s), "r" (err)
	81	: "o2", "o3", "o4", "o5", "o7", "g2", "g3", "g4", "g5",
	82	"cc", "memory");
	83	return (__force __wsum)ret;
	84	}
1da177e4	85
16ec7e16 AV	86	static inline __wsum
	87	csum_partial_copy_to_user(const void src, void __user dst, int len,
	88	__wsum sum, int *err)
1da177e4 LT	89	{
	90	if (!access_ok (VERIFY_WRITE, dst, len)) {
	91	*err = -EFAULT;
	92	return sum;
	93	} else {
	94	register unsigned long ret asm("o0") = (unsigned long)src;
	95	register char __user *d asm("o1") = dst;
	96	register int l asm("g1") = len;
16ec7e16	97	register __wsum s asm("g7") = sum;
1da177e4 LT	98
	99	__asm__ __volatile__ (
	100	".section __ex_table,#alloc\n\t"
	101	".align 4\n\t"
	102	".word 1f,1\n\t"
	103	".previous\n"
	104	"1:\n\t"
	105	"call __csum_partial_copy_sparc_generic\n\t"
	106	" st %8, [%%sp + 64]\n"
	107	: "=&r" (ret), "=&r" (d), "=&r" (l), "=&r" (s)
	108	: "0" (ret), "1" (d), "2" (l), "3" (s), "r" (err)
	109	: "o2", "o3", "o4", "o5", "o7",
	110	"g2", "g3", "g4", "g5",
	111	"cc", "memory");
16ec7e16	112	return (__force __wsum)ret;
1da177e4 LT	113	}
	114	}
	115
	116	#define HAVE_CSUM_COPY_USER
	117	#define csum_and_copy_to_user csum_partial_copy_to_user
	118
	119	/* ihl is always 5 or greater, almost always is 5, and iph is word aligned
	120	* the majority of the time.
	121	*/
16ec7e16	122	static inline __sum16 ip_fast_csum(const void *iph, unsigned int ihl)
1da177e4	123	{
16ec7e16	124	__sum16 sum;
1da177e4 LT	125
	126	/* Note: We must read %2 before we touch %0 for the first time,
	127	* because GCC can legitimately use the same register for
	128	* both operands.
	129	*/
	130	__asm__ __volatile__("sub\t%2, 4, %%g4\n\t"
	131	"ld\t[%1 + 0x00], %0\n\t"
	132	"ld\t[%1 + 0x04], %%g2\n\t"
	133	"ld\t[%1 + 0x08], %%g3\n\t"
	134	"addcc\t%%g2, %0, %0\n\t"
	135	"addxcc\t%%g3, %0, %0\n\t"
	136	"ld\t[%1 + 0x0c], %%g2\n\t"
	137	"ld\t[%1 + 0x10], %%g3\n\t"
	138	"addxcc\t%%g2, %0, %0\n\t"
	139	"addx\t%0, %%g0, %0\n"
	140	"1:\taddcc\t%%g3, %0, %0\n\t"
	141	"add\t%1, 4, %1\n\t"
	142	"addxcc\t%0, %%g0, %0\n\t"
	143	"subcc\t%%g4, 1, %%g4\n\t"
	144	"be,a\t2f\n\t"
	145	"sll\t%0, 16, %%g2\n\t"
	146	"b\t1b\n\t"
	147	"ld\t[%1 + 0x10], %%g3\n"
	148	"2:\taddcc\t%0, %%g2, %%g2\n\t"
	149	"srl\t%%g2, 16, %0\n\t"
	150	"addx\t%0, %%g0, %0\n\t"
	151	"xnor\t%%g0, %0, %0"
	152	: "=r" (sum), "=&r" (iph)
	153	: "r" (ihl), "1" (iph)
51bcf092	154	: "g2", "g3", "g4", "cc", "memory");
1da177e4 LT	155	return sum;
	156	}
	157
	158	/* Fold a partial checksum without adding pseudo headers. */
16ec7e16	159	static inline __sum16 csum_fold(__wsum sum)
1da177e4 LT	160	{
	161	unsigned int tmp;
	162
	163	__asm__ __volatile__("addcc\t%0, %1, %1\n\t"
	164	"srl\t%1, 16, %1\n\t"
	165	"addx\t%1, %%g0, %1\n\t"
	166	"xnor\t%%g0, %1, %0"
	167	: "=&r" (sum), "=r" (tmp)
16ec7e16	168	: "0" (sum), "1" ((__force u32)sum<<16)
1da177e4	169	: "cc");
16ec7e16	170	return (__force __sum16)sum;
1da177e4 LT	171	}
1da177e4 LT	172
16ec7e16 AV	173	static inline __wsum csum_tcpudp_nofold(__be32 saddr, __be32 daddr,
16ec7e16 AV	174	unsigned short len,
1da177e4	175	unsigned short proto,
16ec7e16	176	__wsum sum)
1da177e4 LT	177	{
	178	__asm__ __volatile__("addcc\t%1, %0, %0\n\t"
	179	"addxcc\t%2, %0, %0\n\t"
	180	"addxcc\t%3, %0, %0\n\t"
	181	"addx\t%0, %%g0, %0\n\t"
	182	: "=r" (sum), "=r" (saddr)
16ec7e16	183	: "r" (daddr), "r" (proto + len), "0" (sum),
1da177e4 LT	184	"1" (saddr)
	185	: "cc");
	186	return sum;
	187	}
	188
	189	/*
	190	* computes the checksum of the TCP/UDP pseudo-header
	191	* returns a 16-bit checksum, already complemented
	192	*/
16ec7e16	193	static inline __sum16 csum_tcpudp_magic(__be32 saddr, __be32 daddr,
1da177e4 LT	194	unsigned short len,
1da177e4 LT	195	unsigned short proto,
16ec7e16	196	__wsum sum)
1da177e4 LT	197	{
	198	return csum_fold(csum_tcpudp_nofold(saddr,daddr,len,proto,sum));
	199	}
	200
	201	#define _HAVE_ARCH_IPV6_CSUM
	202
16ec7e16 AV	203	static inline __sum16 csum_ipv6_magic(const struct in6_addr *saddr,
	204	const struct in6_addr *daddr,
	205	__u32 len, unsigned short proto,
	206	__wsum sum)
1da177e4 LT	207	{
	208	__asm__ __volatile__ (
	209	"addcc %3, %4, %%g4\n\t"
	210	"addxcc %5, %%g4, %%g4\n\t"
	211	"ld [%2 + 0x0c], %%g2\n\t"
	212	"ld [%2 + 0x08], %%g3\n\t"
	213	"addxcc %%g2, %%g4, %%g4\n\t"
	214	"ld [%2 + 0x04], %%g2\n\t"
	215	"addxcc %%g3, %%g4, %%g4\n\t"
	216	"ld [%2 + 0x00], %%g3\n\t"
	217	"addxcc %%g2, %%g4, %%g4\n\t"
	218	"ld [%1 + 0x0c], %%g2\n\t"
	219	"addxcc %%g3, %%g4, %%g4\n\t"
	220	"ld [%1 + 0x08], %%g3\n\t"
	221	"addxcc %%g2, %%g4, %%g4\n\t"
	222	"ld [%1 + 0x04], %%g2\n\t"
	223	"addxcc %%g3, %%g4, %%g4\n\t"
	224	"ld [%1 + 0x00], %%g3\n\t"
	225	"addxcc %%g2, %%g4, %%g4\n\t"
	226	"addxcc %%g3, %%g4, %0\n\t"
	227	"addx 0, %0, %0\n"
	228	: "=&r" (sum)
	229	: "r" (saddr), "r" (daddr),
	230	"r"(htonl(len)), "r"(htonl(proto)), "r"(sum)
	231	: "g2", "g3", "g4", "cc");
	232
	233	return csum_fold(sum);
	234	}
	235
	236	/* this routine is used for miscellaneous IP-like checksums, mainly in icmp.c */
16ec7e16	237	static inline __sum16 ip_compute_csum(const void *buff, int len)
1da177e4 LT	238	{
	239	return csum_fold(csum_partial(buff, len, 0));
	240	}
	241
	242	#endif /* !(__SPARC_CHECKSUM_H) */