arch/powerpc/include/asm/checksum.h

   1 #ifndef _ASM_POWERPC_CHECKSUM_H
   2 #define _ASM_POWERPC_CHECKSUM_H
   3 #ifdef __KERNEL__
   4
   5 /*
   6  * This program is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU General Public License
   8  * as published by the Free Software Foundation; either version
   9  * 2 of the License, or (at your option) any later version.
  10  */
  11
  12 #include <linux/bitops.h>
  13 #include <linux/in6.h>
  14 /*
  15  * Computes the checksum of a memory block at src, length len,
  16  * and adds in "sum" (32-bit), while copying the block to dst.
  17  * If an access exception occurs on src or dst, it stores -EFAULT
  18  * to *src_err or *dst_err respectively (if that pointer is not
  19  * NULL), and, for an error on src, zeroes the rest of dst.
  20  *
  21  * Like csum_partial, this must be called with even lengths,
  22  * except for the last fragment.
  23  */
  24 extern __wsum csum_partial_copy_generic(const void *src, void *dst,
  25                                               int len, __wsum sum,
  26                                               int *src_err, int *dst_err);
  27
  28 #define _HAVE_ARCH_COPY_AND_CSUM_FROM_USER
  29 extern __wsum csum_and_copy_from_user(const void __user *src, void *dst,
  30                                       int len, __wsum sum, int *err_ptr);
  31 #define HAVE_CSUM_COPY_USER
  32 extern __wsum csum_and_copy_to_user(const void *src, void __user *dst,
  33                                     int len, __wsum sum, int *err_ptr);
  34
  35 #define csum_partial_copy_nocheck(src, dst, len, sum)   \
  36         csum_partial_copy_generic((src), (dst), (len), (sum), NULL, NULL)
  37
  38
  39 /*
  40  * turns a 32-bit partial checksum (e.g. from csum_partial) into a
  41  * 1's complement 16-bit checksum.
  42  */
  43 static inline __sum16 csum_fold(__wsum sum)
  44 {
  45         unsigned int tmp;
  46
  47         /* swap the two 16-bit halves of sum */
  48         __asm__("rlwinm %0,%1,16,0,31" : "=r" (tmp) : "r" (sum));
  49         /* if there is a carry from adding the two 16-bit halves,
  50            it will carry from the lower half into the upper half,
  51            giving us the correct sum in the upper half. */
  52         return (__force __sum16)(~((__force u32)sum + tmp) >> 16);
  53 }
  54
  55 static inline u32 from64to32(u64 x)
  56 {
  57         return (x + ror64(x, 32)) >> 32;
  58 }
  59
  60 static inline __wsum csum_tcpudp_nofold(__be32 saddr, __be32 daddr, __u32 len,
  61                                         __u8 proto, __wsum sum)
  62 {
  63 #ifdef __powerpc64__
  64         u64 s = (__force u32)sum;
  65
  66         s += (__force u32)saddr;
  67         s += (__force u32)daddr;
  68 #ifdef __BIG_ENDIAN__
  69         s += proto + len;
  70 #else
  71         s += (proto + len) << 8;
  72 #endif
  73         return (__force __wsum) from64to32(s);
  74 #else
  75     __asm__("\n\
  76         addc %0,%0,%1 \n\
  77         adde %0,%0,%2 \n\
  78         adde %0,%0,%3 \n\
  79         addze %0,%0 \n\
  80         "
  81         : "=r" (sum)
  82         : "r" (daddr), "r"(saddr), "r"(proto + len), "0"(sum));
  83         return sum;
  84 #endif
  85 }
  86
  87 /*
  88  * computes the checksum of the TCP/UDP pseudo-header
  89  * returns a 16-bit checksum, already complemented
  90  */
  91 static inline __sum16 csum_tcpudp_magic(__be32 saddr, __be32 daddr, __u32 len,
  92                                         __u8 proto, __wsum sum)
  93 {
  94         return csum_fold(csum_tcpudp_nofold(saddr, daddr, len, proto, sum));
  95 }
  96
  97 #define HAVE_ARCH_CSUM_ADD
  98 static inline __wsum csum_add(__wsum csum, __wsum addend)
  99 {
 100 #ifdef __powerpc64__
 101         u64 res = (__force u64)csum;
 102 #endif
 103         if (__builtin_constant_p(csum) && csum == 0)
 104                 return addend;
 105         if (__builtin_constant_p(addend) && addend == 0)
 106                 return csum;
 107
 108 #ifdef __powerpc64__
 109         res += (__force u64)addend;
 110         return (__force __wsum)((u32)res + (res >> 32));
 111 #else
 112         asm("addc %0,%0,%1;"
 113             "addze %0,%0;"
 114             : "+r" (csum) : "r" (addend) : "xer");
 115         return csum;
 116 #endif
 117 }
 118
 119 /*
 120  * This is a version of ip_compute_csum() optimized for IP headers,
 121  * which always checksum on 4 octet boundaries.  ihl is the number
 122  * of 32-bit words and is always >= 5.
 123  */
 124 static inline __wsum ip_fast_csum_nofold(const void *iph, unsigned int ihl)
 125 {
 126         const u32 *ptr = (const u32 *)iph + 1;
 127 #ifdef __powerpc64__
 128         unsigned int i;
 129         u64 s = *(const u32 *)iph;
 130
 131         for (i = 0; i < ihl - 1; i++, ptr++)
 132                 s += *ptr;
 133         return (__force __wsum)from64to32(s);
 134 #else
 135         __wsum sum, tmp;
 136
 137         asm("mtctr %3;"
 138             "addc %0,%4,%5;"
 139             "1: lwzu %1, 4(%2);"
 140             "adde %0,%0,%1;"
 141             "bdnz 1b;"
 142             "addze %0,%0;"
 143             : "=r" (sum), "=r" (tmp), "+b" (ptr)
 144             : "r" (ihl - 2), "r" (*(const u32 *)iph), "r" (*ptr)
 145             : "ctr", "xer", "memory");
 146
 147         return sum;
 148 #endif
 149 }
 150
 151 static inline __sum16 ip_fast_csum(const void *iph, unsigned int ihl)
 152 {
 153         return csum_fold(ip_fast_csum_nofold(iph, ihl));
 154 }
 155
 156 /*
 157  * computes the checksum of a memory block at buff, length len,
 158  * and adds in "sum" (32-bit)
 159  *
 160  * returns a 32-bit number suitable for feeding into itself
 161  * or csum_tcpudp_magic
 162  *
 163  * this function must be called with even lengths, except
 164  * for the last fragment, which may be odd
 165  *
 166  * it's best to have buff aligned on a 32-bit boundary
 167  */
 168 __wsum __csum_partial(const void *buff, int len, __wsum sum);
 169
 170 static inline __wsum csum_partial(const void *buff, int len, __wsum sum)
 171 {
 172         if (__builtin_constant_p(len) && len <= 16 && (len & 1) == 0) {
 173                 if (len == 2)
 174                         sum = csum_add(sum, (__force __wsum)*(const u16 *)buff);
 175                 if (len >= 4)
 176                         sum = csum_add(sum, (__force __wsum)*(const u32 *)buff);
 177                 if (len == 6)
 178                         sum = csum_add(sum, (__force __wsum)
 179                                             *(const u16 *)(buff + 4));
 180                 if (len >= 8)
 181                         sum = csum_add(sum, (__force __wsum)
 182                                             *(const u32 *)(buff + 4));
 183                 if (len == 10)
 184                         sum = csum_add(sum, (__force __wsum)
 185                                             *(const u16 *)(buff + 8));
 186                 if (len >= 12)
 187                         sum = csum_add(sum, (__force __wsum)
 188                                             *(const u32 *)(buff + 8));
 189                 if (len == 14)
 190                         sum = csum_add(sum, (__force __wsum)
 191                                             *(const u16 *)(buff + 12));
 192                 if (len >= 16)
 193                         sum = csum_add(sum, (__force __wsum)
 194                                             *(const u32 *)(buff + 12));
 195         } else if (__builtin_constant_p(len) && (len & 3) == 0) {
 196                 sum = csum_add(sum, ip_fast_csum_nofold(buff, len >> 2));
 197         } else {
 198                 sum = __csum_partial(buff, len, sum);
 199         }
 200         return sum;
 201 }
 202
 203 /*
 204  * this routine is used for miscellaneous IP-like checksums, mainly
 205  * in icmp.c
 206  */
 207 static inline __sum16 ip_compute_csum(const void *buff, int len)
 208 {
 209         return csum_fold(csum_partial(buff, len, 0));
 210 }
 211
 212 #define _HAVE_ARCH_IPV6_CSUM
 213 __sum16 csum_ipv6_magic(const struct in6_addr *saddr,
 214                         const struct in6_addr *daddr,
 215                         __u32 len, __u8 proto, __wsum sum);
 216
 217 #endif /* __KERNEL__ */
 218 #endif