Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 1 | #ifdef __KERNEL__ |
| 2 | #ifndef _ASM_M32R_CHECKSUM_H |
| 3 | #define _ASM_M32R_CHECKSUM_H |
| 4 | |
| 5 | /* |
| 6 | * include/asm-m32r/checksum.h |
| 7 | * |
| 8 | * IP/TCP/UDP checksum routines |
| 9 | * |
| 10 | * This file is subject to the terms and conditions of the GNU General Public |
| 11 | * License. See the file "COPYING" in the main directory of this archive |
| 12 | * for more details. |
| 13 | * |
| 14 | * Some code taken from mips and parisc architecture. |
| 15 | * |
| 16 | * Copyright (C) 2001, 2002 Hiroyuki Kondo, Hirokazu Takata |
| 17 | * Copyright (C) 2004 Hirokazu Takata <takata at linux-m32r.org> |
| 18 | */ |
| 19 | |
| 20 | #include <linux/in6.h> |
| 21 | |
| 22 | /* |
| 23 | * computes the checksum of a memory block at buff, length len, |
| 24 | * and adds in "sum" (32-bit) |
| 25 | * |
| 26 | * returns a 32-bit number suitable for feeding into itself |
| 27 | * or csum_tcpudp_magic |
| 28 | * |
| 29 | * this function must be called with even lengths, except |
| 30 | * for the last fragment, which may be odd |
| 31 | * |
| 32 | * it's best to have buff aligned on a 32-bit boundary |
| 33 | */ |
| 34 | asmlinkage unsigned int csum_partial(const unsigned char *buff, |
| 35 | int len, unsigned int sum); |
| 36 | |
| 37 | /* |
| 38 | * The same as csum_partial, but copies from src while it checksums. |
| 39 | * |
| 40 | * Here even more important to align src and dst on a 32-bit (or even |
| 41 | * better 64-bit) boundary |
| 42 | */ |
| 43 | extern unsigned int csum_partial_copy_nocheck(const unsigned char *src, |
| 44 | unsigned char *dst, |
| 45 | int len, unsigned int sum); |
| 46 | |
| 47 | /* |
| 48 | * This is a new version of the above that records errors it finds in *errp, |
| 49 | * but continues and zeros thre rest of the buffer. |
| 50 | */ |
| 51 | extern unsigned int csum_partial_copy_from_user(const unsigned char __user *src, |
| 52 | unsigned char *dst, |
| 53 | int len, unsigned int sum, |
| 54 | int *err_ptr); |
| 55 | |
| 56 | /* |
| 57 | * Fold a partial checksum |
| 58 | */ |
| 59 | |
| 60 | static inline unsigned int csum_fold(unsigned int sum) |
| 61 | { |
| 62 | unsigned long tmpreg; |
| 63 | __asm__( |
| 64 | " sll3 %1, %0, #16 \n" |
| 65 | " cmp %0, %0 \n" |
| 66 | " addx %0, %1 \n" |
| 67 | " ldi %1, #0 \n" |
| 68 | " srli %0, #16 \n" |
| 69 | " addx %0, %1 \n" |
| 70 | " xor3 %0, %0, #0x0000ffff \n" |
| 71 | : "=r" (sum), "=&r" (tmpreg) |
| 72 | : "0" (sum) |
| 73 | : "cbit" |
| 74 | ); |
| 75 | return sum; |
| 76 | } |
| 77 | |
| 78 | /* |
| 79 | * This is a version of ip_compute_csum() optimized for IP headers, |
| 80 | * which always checksum on 4 octet boundaries. |
| 81 | */ |
| 82 | static inline unsigned short ip_fast_csum(unsigned char * iph, |
| 83 | unsigned int ihl) { |
| 84 | unsigned long sum, tmpreg0, tmpreg1; |
| 85 | |
| 86 | __asm__ __volatile__( |
| 87 | " ld %0, @%1+ \n" |
| 88 | " addi %2, #-4 \n" |
| 89 | "# bgez %2, 2f \n" |
| 90 | " cmp %0, %0 \n" |
| 91 | " ld %3, @%1+ \n" |
| 92 | " ld %4, @%1+ \n" |
| 93 | " addx %0, %3 \n" |
| 94 | " ld %3, @%1+ \n" |
| 95 | " addx %0, %4 \n" |
| 96 | " addx %0, %3 \n" |
| 97 | " .fillinsn\n" |
| 98 | "1: \n" |
| 99 | " ld %4, @%1+ \n" |
| 100 | " addi %2, #-1 \n" |
| 101 | " addx %0, %4 \n" |
| 102 | " bgtz %2, 1b \n" |
| 103 | "\n" |
| 104 | " ldi %3, #0 \n" |
| 105 | " addx %0, %3 \n" |
| 106 | " .fillinsn\n" |
| 107 | "2: \n" |
Thomas Graf | 2c65649 | 2005-08-20 17:24:25 -0700 | [diff] [blame] | 108 | /* Since the input registers which are loaded with iph and ihl |
Linus Torvalds | 1da177e | 2005-04-16 15:20:36 -0700 | [diff] [blame] | 109 | are modified, we must also specify them as outputs, or gcc |
| 110 | will assume they contain their original values. */ |
| 111 | : "=&r" (sum), "=r" (iph), "=r" (ihl), "=&r" (tmpreg0), "=&r" (tmpreg1) |
| 112 | : "1" (iph), "2" (ihl) |
| 113 | : "cbit", "memory"); |
| 114 | |
| 115 | return csum_fold(sum); |
| 116 | } |
| 117 | |
| 118 | static inline unsigned long csum_tcpudp_nofold(unsigned long saddr, |
| 119 | unsigned long daddr, |
| 120 | unsigned short len, |
| 121 | unsigned short proto, |
| 122 | unsigned int sum) |
| 123 | { |
| 124 | #if defined(__LITTLE_ENDIAN) |
| 125 | unsigned long len_proto = (ntohs(len)<<16)+proto*256; |
| 126 | #else |
| 127 | unsigned long len_proto = (proto<<16)+len; |
| 128 | #endif |
| 129 | unsigned long tmpreg; |
| 130 | |
| 131 | __asm__( |
| 132 | " cmp %0, %0 \n" |
| 133 | " addx %0, %2 \n" |
| 134 | " addx %0, %3 \n" |
| 135 | " addx %0, %4 \n" |
| 136 | " ldi %1, #0 \n" |
| 137 | " addx %0, %1 \n" |
| 138 | : "=r" (sum), "=&r" (tmpreg) |
| 139 | : "r" (daddr), "r" (saddr), "r" (len_proto), "0" (sum) |
| 140 | : "cbit" |
| 141 | ); |
| 142 | |
| 143 | return sum; |
| 144 | } |
| 145 | |
| 146 | /* |
| 147 | * computes the checksum of the TCP/UDP pseudo-header |
| 148 | * returns a 16-bit checksum, already complemented |
| 149 | */ |
| 150 | static inline unsigned short int csum_tcpudp_magic(unsigned long saddr, |
| 151 | unsigned long daddr, |
| 152 | unsigned short len, |
| 153 | unsigned short proto, |
| 154 | unsigned int sum) |
| 155 | { |
| 156 | return csum_fold(csum_tcpudp_nofold(saddr,daddr,len,proto,sum)); |
| 157 | } |
| 158 | |
| 159 | /* |
| 160 | * this routine is used for miscellaneous IP-like checksums, mainly |
| 161 | * in icmp.c |
| 162 | */ |
| 163 | |
| 164 | static inline unsigned short ip_compute_csum(unsigned char * buff, int len) { |
| 165 | return csum_fold (csum_partial(buff, len, 0)); |
| 166 | } |
| 167 | |
| 168 | #define _HAVE_ARCH_IPV6_CSUM |
| 169 | static inline unsigned short int csum_ipv6_magic(struct in6_addr *saddr, |
| 170 | struct in6_addr *daddr, |
| 171 | __u16 len, |
| 172 | unsigned short proto, |
| 173 | unsigned int sum) |
| 174 | { |
| 175 | unsigned long tmpreg0, tmpreg1, tmpreg2, tmpreg3; |
| 176 | __asm__( |
| 177 | " ld %1, @(%5) \n" |
| 178 | " ld %2, @(4,%5) \n" |
| 179 | " ld %3, @(8,%5) \n" |
| 180 | " ld %4, @(12,%5) \n" |
| 181 | " add %0, %1 \n" |
| 182 | " addx %0, %2 \n" |
| 183 | " addx %0, %3 \n" |
| 184 | " addx %0, %4 \n" |
| 185 | " ld %1, @(%6) \n" |
| 186 | " ld %2, @(4,%6) \n" |
| 187 | " ld %3, @(8,%6) \n" |
| 188 | " ld %4, @(12,%6) \n" |
| 189 | " addx %0, %1 \n" |
| 190 | " addx %0, %2 \n" |
| 191 | " addx %0, %3 \n" |
| 192 | " addx %0, %4 \n" |
| 193 | " addx %0, %7 \n" |
| 194 | " addx %0, %8 \n" |
| 195 | " ldi %1, #0 \n" |
| 196 | " addx %0, %1 \n" |
| 197 | : "=&r" (sum), "=&r" (tmpreg0), "=&r" (tmpreg1), |
| 198 | "=&r" (tmpreg2), "=&r" (tmpreg3) |
| 199 | : "r" (saddr), "r" (daddr), |
| 200 | "r" (htonl((__u32) (len))), "r" (htonl(proto)), "0" (sum) |
| 201 | : "cbit" |
| 202 | ); |
| 203 | |
| 204 | return csum_fold(sum); |
| 205 | } |
| 206 | |
| 207 | #endif /* _ASM_M32R_CHECKSUM_H */ |
| 208 | #endif /* __KERNEL__ */ |