Commit | Line | Data |
---|---|---|
d95bbcb3 KG |
1 | #ifndef _ASM_POWERPC_CHECKSUM_H |
2 | #define _ASM_POWERPC_CHECKSUM_H | |
88ced031 | 3 | #ifdef __KERNEL__ |
1da177e4 LT |
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 | ||
7a332b0c AB |
12 | #ifdef CONFIG_GENERIC_CSUM |
13 | #include <asm-generic/checksum.h> | |
14 | #else | |
1da177e4 | 15 | /* |
d95bbcb3 KG |
16 | * Computes the checksum of a memory block at src, length len, |
17 | * and adds in "sum" (32-bit), while copying the block to dst. | |
18 | * If an access exception occurs on src or dst, it stores -EFAULT | |
19 | * to *src_err or *dst_err respectively (if that pointer is not | |
20 | * NULL), and, for an error on src, zeroes the rest of dst. | |
21 | * | |
22 | * Like csum_partial, this must be called with even lengths, | |
23 | * except for the last fragment. | |
1da177e4 | 24 | */ |
879178cf AV |
25 | extern __wsum csum_partial_copy_generic(const void *src, void *dst, |
26 | int len, __wsum sum, | |
1da177e4 | 27 | int *src_err, int *dst_err); |
fdd374b6 | 28 | |
fdd374b6 AB |
29 | #define _HAVE_ARCH_COPY_AND_CSUM_FROM_USER |
30 | extern __wsum csum_and_copy_from_user(const void __user *src, void *dst, | |
31 | int len, __wsum sum, int *err_ptr); | |
8c773914 AB |
32 | #define HAVE_CSUM_COPY_USER |
33 | extern __wsum csum_and_copy_to_user(const void *src, void __user *dst, | |
34 | int len, __wsum sum, int *err_ptr); | |
d95bbcb3 KG |
35 | |
36 | #define csum_partial_copy_nocheck(src, dst, len, sum) \ | |
37 | csum_partial_copy_generic((src), (dst), (len), (sum), NULL, NULL) | |
38 | ||
39 | ||
1da177e4 LT |
40 | /* |
41 | * turns a 32-bit partial checksum (e.g. from csum_partial) into a | |
42 | * 1's complement 16-bit checksum. | |
43 | */ | |
879178cf | 44 | static inline __sum16 csum_fold(__wsum sum) |
1da177e4 LT |
45 | { |
46 | unsigned int tmp; | |
47 | ||
48 | /* swap the two 16-bit halves of sum */ | |
49 | __asm__("rlwinm %0,%1,16,0,31" : "=r" (tmp) : "r" (sum)); | |
50 | /* if there is a carry from adding the two 16-bit halves, | |
51 | it will carry from the lower half into the upper half, | |
52 | giving us the correct sum in the upper half. */ | |
879178cf | 53 | return (__force __sum16)(~((__force u32)sum + tmp) >> 16); |
1da177e4 LT |
54 | } |
55 | ||
879178cf | 56 | static inline __wsum csum_tcpudp_nofold(__be32 saddr, __be32 daddr, |
1da177e4 LT |
57 | unsigned short len, |
58 | unsigned short proto, | |
879178cf | 59 | __wsum sum) |
1da177e4 | 60 | { |
879178cf AV |
61 | #ifdef __powerpc64__ |
62 | unsigned long s = (__force u32)sum; | |
1da177e4 | 63 | |
879178cf AV |
64 | s += (__force u32)saddr; |
65 | s += (__force u32)daddr; | |
66 | s += proto + len; | |
1da177e4 | 67 | s += (s >> 32); |
879178cf | 68 | return (__force __wsum) s; |
d95bbcb3 | 69 | #else |
d95bbcb3 KG |
70 | __asm__("\n\ |
71 | addc %0,%0,%1 \n\ | |
72 | adde %0,%0,%2 \n\ | |
73 | adde %0,%0,%3 \n\ | |
74 | addze %0,%0 \n\ | |
75 | " | |
76 | : "=r" (sum) | |
879178cf AV |
77 | : "r" (daddr), "r"(saddr), "r"(proto + len), "0"(sum)); |
78 | return sum; | |
1da177e4 | 79 | #endif |
879178cf | 80 | } |
7a332b0c | 81 | |
92c985f1 LC |
82 | /* |
83 | * computes the checksum of the TCP/UDP pseudo-header | |
84 | * returns a 16-bit checksum, already complemented | |
85 | */ | |
86 | static inline __sum16 csum_tcpudp_magic(__be32 saddr, __be32 daddr, | |
87 | unsigned short len, | |
88 | unsigned short proto, | |
89 | __wsum sum) | |
90 | { | |
91 | return csum_fold(csum_tcpudp_nofold(saddr, daddr, len, proto, sum)); | |
92 | } | |
93 | ||
501c8de7 LC |
94 | #define HAVE_ARCH_CSUM_ADD |
95 | static inline __wsum csum_add(__wsum csum, __wsum addend) | |
96 | { | |
97 | #ifdef __powerpc64__ | |
98 | u64 res = (__force u64)csum; | |
5a8847c8 CL |
99 | #endif |
100 | if (__builtin_constant_p(csum) && csum == 0) | |
101 | return addend; | |
102 | if (__builtin_constant_p(addend) && addend == 0) | |
103 | return csum; | |
501c8de7 | 104 | |
5a8847c8 | 105 | #ifdef __powerpc64__ |
501c8de7 LC |
106 | res += (__force u64)addend; |
107 | return (__force __wsum)((u32)res + (res >> 32)); | |
108 | #else | |
109 | asm("addc %0,%0,%1;" | |
110 | "addze %0,%0;" | |
11dfbf58 | 111 | : "+r" (csum) : "r" (addend) : "xer"); |
501c8de7 LC |
112 | return csum; |
113 | #endif | |
114 | } | |
115 | ||
37e08cad CL |
116 | /* |
117 | * This is a version of ip_compute_csum() optimized for IP headers, | |
118 | * which always checksum on 4 octet boundaries. ihl is the number | |
119 | * of 32-bit words and is always >= 5. | |
120 | */ | |
121 | static inline __wsum ip_fast_csum_nofold(const void *iph, unsigned int ihl) | |
122 | { | |
123 | const u32 *ptr = (const u32 *)iph + 1; | |
124 | #ifdef __powerpc64__ | |
125 | unsigned int i; | |
126 | u64 s = *(const u32 *)iph; | |
127 | ||
128 | for (i = 0; i < ihl - 1; i++, ptr++) | |
129 | s += *ptr; | |
130 | s += (s >> 32); | |
131 | return (__force __wsum)s; | |
132 | #else | |
133 | __wsum sum, tmp; | |
134 | ||
135 | asm("mtctr %3;" | |
136 | "addc %0,%4,%5;" | |
137 | "1: lwzu %1, 4(%2);" | |
138 | "adde %0,%0,%1;" | |
139 | "bdnz 1b;" | |
140 | "addze %0,%0;" | |
141 | : "=r" (sum), "=r" (tmp), "+b" (ptr) | |
142 | : "r" (ihl - 2), "r" (*(const u32 *)iph), "r" (*ptr) | |
143 | : "ctr", "xer", "memory"); | |
144 | ||
145 | return sum; | |
146 | #endif | |
147 | } | |
148 | ||
149 | static inline __sum16 ip_fast_csum(const void *iph, unsigned int ihl) | |
150 | { | |
151 | return csum_fold(ip_fast_csum_nofold(iph, ihl)); | |
152 | } | |
153 | ||
7e393220 CL |
154 | /* |
155 | * computes the checksum of a memory block at buff, length len, | |
156 | * and adds in "sum" (32-bit) | |
157 | * | |
158 | * returns a 32-bit number suitable for feeding into itself | |
159 | * or csum_tcpudp_magic | |
160 | * | |
161 | * this function must be called with even lengths, except | |
162 | * for the last fragment, which may be odd | |
163 | * | |
164 | * it's best to have buff aligned on a 32-bit boundary | |
165 | */ | |
166 | __wsum __csum_partial(const void *buff, int len, __wsum sum); | |
167 | ||
168 | static inline __wsum csum_partial(const void *buff, int len, __wsum sum) | |
169 | { | |
170 | if (__builtin_constant_p(len) && len <= 16 && (len & 1) == 0) { | |
171 | if (len == 2) | |
172 | sum = csum_add(sum, (__force __wsum)*(const u16 *)buff); | |
173 | if (len >= 4) | |
174 | sum = csum_add(sum, (__force __wsum)*(const u32 *)buff); | |
175 | if (len == 6) | |
176 | sum = csum_add(sum, (__force __wsum) | |
177 | *(const u16 *)(buff + 4)); | |
178 | if (len >= 8) | |
179 | sum = csum_add(sum, (__force __wsum) | |
180 | *(const u32 *)(buff + 4)); | |
181 | if (len == 10) | |
182 | sum = csum_add(sum, (__force __wsum) | |
183 | *(const u16 *)(buff + 8)); | |
184 | if (len >= 12) | |
185 | sum = csum_add(sum, (__force __wsum) | |
186 | *(const u32 *)(buff + 8)); | |
187 | if (len == 14) | |
188 | sum = csum_add(sum, (__force __wsum) | |
189 | *(const u16 *)(buff + 12)); | |
190 | if (len >= 16) | |
191 | sum = csum_add(sum, (__force __wsum) | |
192 | *(const u32 *)(buff + 12)); | |
193 | } else if (__builtin_constant_p(len) && (len & 3) == 0) { | |
194 | sum = csum_add(sum, ip_fast_csum_nofold(buff, len >> 2)); | |
195 | } else { | |
196 | sum = __csum_partial(buff, len, sum); | |
197 | } | |
198 | return sum; | |
199 | } | |
200 | ||
201 | /* | |
202 | * this routine is used for miscellaneous IP-like checksums, mainly | |
203 | * in icmp.c | |
204 | */ | |
205 | static inline __sum16 ip_compute_csum(const void *buff, int len) | |
206 | { | |
207 | return csum_fold(csum_partial(buff, len, 0)); | |
208 | } | |
209 | ||
7a332b0c | 210 | #endif |
88ced031 | 211 | #endif /* __KERNEL__ */ |
d95bbcb3 | 212 | #endif |