]>
Commit | Line | Data |
---|---|---|
b2441318 | 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
6bc9a396 CL |
2 | #ifndef _ASM_SCORE_CHECKSUM_H |
3 | #define _ASM_SCORE_CHECKSUM_H | |
4 | ||
5 | #include <linux/in6.h> | |
7c0f6ba6 | 6 | #include <linux/uaccess.h> |
6bc9a396 CL |
7 | |
8 | /* | |
9 | * computes the checksum of a memory block at buff, length len, | |
10 | * and adds in "sum" (32-bit) | |
11 | * | |
12 | * returns a 32-bit number suitable for feeding into itself | |
13 | * or csum_tcpudp_magic | |
14 | * | |
15 | * this function must be called with even lengths, except | |
16 | * for the last fragment, which may be odd | |
17 | * | |
18 | * it's best to have buff aligned on a 32-bit boundary | |
19 | */ | |
20 | unsigned int csum_partial(const void *buff, int len, __wsum sum); | |
21 | unsigned int csum_partial_copy_from_user(const char *src, char *dst, int len, | |
22 | unsigned int sum, int *csum_err); | |
23 | unsigned int csum_partial_copy(const char *src, char *dst, | |
24 | int len, unsigned int sum); | |
25 | ||
26 | /* | |
27 | * this is a new version of the above that records errors it finds in *errp, | |
28 | * but continues and zeros the rest of the buffer. | |
29 | */ | |
30 | ||
31 | /* | |
32 | * Copy and checksum to user | |
33 | */ | |
34 | #define HAVE_CSUM_COPY_USER | |
35 | static inline | |
36 | __wsum csum_and_copy_to_user(const void *src, void __user *dst, int len, | |
37 | __wsum sum, int *err_ptr) | |
38 | { | |
39 | sum = csum_partial(src, len, sum); | |
40 | if (copy_to_user(dst, src, len)) { | |
41 | *err_ptr = -EFAULT; | |
42 | return (__force __wsum) -1; /* invalid checksum */ | |
43 | } | |
44 | return sum; | |
45 | } | |
46 | ||
47 | ||
48 | #define csum_partial_copy_nocheck csum_partial_copy | |
49 | /* | |
50 | * Fold a partial checksum without adding pseudo headers | |
51 | */ | |
52 | ||
53 | static inline __sum16 csum_fold(__wsum sum) | |
54 | { | |
55 | /* the while loop is unnecessary really, it's always enough with two | |
56 | iterations */ | |
57 | __asm__ __volatile__( | |
58 | ".set volatile\n\t" | |
59 | ".set\tr1\n\t" | |
60 | "slli\tr1,%0, 16\n\t" | |
61 | "add\t%0,%0, r1\n\t" | |
62 | "cmp.c\tr1, %0\n\t" | |
63 | "srli\t%0, %0, 16\n\t" | |
64 | "bleu\t1f\n\t" | |
65 | "addi\t%0, 0x1\n\t" | |
66 | "1:ldi\tr30, 0xffff\n\t" | |
67 | "xor\t%0, %0, r30\n\t" | |
68 | "slli\t%0, %0, 16\n\t" | |
69 | "srli\t%0, %0, 16\n\t" | |
70 | ".set\tnor1\n\t" | |
71 | ".set optimize\n\t" | |
72 | : "=r" (sum) | |
73 | : "0" (sum)); | |
74 | return sum; | |
75 | } | |
76 | ||
77 | /* | |
78 | * This is a version of ip_compute_csum() optimized for IP headers, | |
79 | * which always checksum on 4 octet boundaries. | |
80 | * | |
81 | * By Jorge Cwik <jorge@laser.satlink.net>, adapted for linux by | |
82 | * Arnt Gulbrandsen. | |
83 | */ | |
84 | static inline __sum16 ip_fast_csum(const void *iph, unsigned int ihl) | |
85 | { | |
86 | unsigned int sum; | |
87 | unsigned long dummy; | |
88 | ||
89 | __asm__ __volatile__( | |
90 | ".set volatile\n\t" | |
91 | ".set\tnor1\n\t" | |
92 | "lw\t%0, [%1]\n\t" | |
93 | "subri\t%2, %2, 4\n\t" | |
94 | "slli\t%2, %2, 2\n\t" | |
95 | "lw\t%3, [%1, 4]\n\t" | |
96 | "add\t%2, %2, %1\n\t" | |
97 | "add\t%0, %0, %3\n\t" | |
98 | "cmp.c\t%3, %0\n\t" | |
99 | "lw\t%3, [%1, 8]\n\t" | |
100 | "bleu\t1f\n\t" | |
101 | "addi\t%0, 0x1\n\t" | |
102 | "1:\n\t" | |
103 | "add\t%0, %0, %3\n\t" | |
104 | "cmp.c\t%3, %0\n\t" | |
105 | "lw\t%3, [%1, 12]\n\t" | |
106 | "bleu\t1f\n\t" | |
107 | "addi\t%0, 0x1\n\t" | |
108 | "1:add\t%0, %0, %3\n\t" | |
109 | "cmp.c\t%3, %0\n\t" | |
110 | "bleu\t1f\n\t" | |
111 | "addi\t%0, 0x1\n" | |
112 | ||
113 | "1:\tlw\t%3, [%1, 16]\n\t" | |
114 | "addi\t%1, 4\n\t" | |
115 | "add\t%0, %0, %3\n\t" | |
116 | "cmp.c\t%3, %0\n\t" | |
117 | "bleu\t2f\n\t" | |
118 | "addi\t%0, 0x1\n" | |
119 | "2:cmp.c\t%2, %1\n\t" | |
120 | "bne\t1b\n\t" | |
121 | ||
122 | ".set\tr1\n\t" | |
123 | ".set optimize\n\t" | |
124 | : "=&r" (sum), "=&r" (iph), "=&r" (ihl), "=&r" (dummy) | |
125 | : "1" (iph), "2" (ihl)); | |
126 | ||
127 | return csum_fold(sum); | |
128 | } | |
129 | ||
130 | static inline __wsum | |
01cfbad7 AD |
131 | csum_tcpudp_nofold(__be32 saddr, __be32 daddr, __u32 len, |
132 | __u8 proto, __wsum sum) | |
6bc9a396 | 133 | { |
01cfbad7 | 134 | unsigned long tmp = (len + proto) << 8; |
6bc9a396 CL |
135 | __asm__ __volatile__( |
136 | ".set volatile\n\t" | |
137 | "add\t%0, %0, %2\n\t" | |
138 | "cmp.c\t%2, %0\n\t" | |
139 | "bleu\t1f\n\t" | |
140 | "addi\t%0, 0x1\n\t" | |
141 | "1:\n\t" | |
142 | "add\t%0, %0, %3\n\t" | |
143 | "cmp.c\t%3, %0\n\t" | |
144 | "bleu\t1f\n\t" | |
145 | "addi\t%0, 0x1\n\t" | |
146 | "1:\n\t" | |
147 | "add\t%0, %0, %4\n\t" | |
148 | "cmp.c\t%4, %0\n\t" | |
149 | "bleu\t1f\n\t" | |
150 | "addi\t%0, 0x1\n\t" | |
151 | "1:\n\t" | |
152 | ".set optimize\n\t" | |
153 | : "=r" (sum) | |
154 | : "0" (daddr), "r"(saddr), | |
155 | "r" (tmp), | |
156 | "r" (sum)); | |
157 | return sum; | |
158 | } | |
159 | ||
160 | /* | |
161 | * computes the checksum of the TCP/UDP pseudo-header | |
162 | * returns a 16-bit checksum, already complemented | |
163 | */ | |
164 | static inline __sum16 | |
01cfbad7 AD |
165 | csum_tcpudp_magic(__be32 saddr, __be32 daddr, __u32 len, |
166 | __u8 proto, __wsum sum) | |
6bc9a396 CL |
167 | { |
168 | return csum_fold(csum_tcpudp_nofold(saddr, daddr, len, proto, sum)); | |
169 | } | |
170 | ||
171 | /* | |
172 | * this routine is used for miscellaneous IP-like checksums, mainly | |
173 | * in icmp.c | |
174 | */ | |
175 | ||
176 | static inline unsigned short ip_compute_csum(const void *buff, int len) | |
177 | { | |
178 | return csum_fold(csum_partial(buff, len, 0)); | |
179 | } | |
180 | ||
181 | #define _HAVE_ARCH_IPV6_CSUM | |
182 | static inline __sum16 csum_ipv6_magic(const struct in6_addr *saddr, | |
1e940829 AD |
183 | const struct in6_addr *daddr, |
184 | __u32 len, __u8 proto, __wsum sum) | |
6bc9a396 CL |
185 | { |
186 | __asm__ __volatile__( | |
1ed62ca6 LW |
187 | ".set\tvolatile\t\t\t# csum_ipv6_magic\n\t" |
188 | "add\t%0, %0, %5\t\t\t# proto (long in network byte order)\n\t" | |
189 | "cmp.c\t%5, %0\n\t" | |
190 | "bleu 1f\n\t" | |
191 | "addi\t%0, 0x1\n\t" | |
192 | "1:add\t%0, %0, %6\t\t\t# csum\n\t" | |
193 | "cmp.c\t%6, %0\n\t" | |
194 | "lw\t%1, [%2, 0]\t\t\t# four words source address\n\t" | |
195 | "bleu 1f\n\t" | |
196 | "addi\t%0, 0x1\n\t" | |
197 | "1:add\t%0, %0, %1\n\t" | |
198 | "cmp.c\t%1, %0\n\t" | |
199 | "1:lw\t%1, [%2, 4]\n\t" | |
200 | "bleu 1f\n\t" | |
201 | "addi\t%0, 0x1\n\t" | |
202 | "1:add\t%0, %0, %1\n\t" | |
203 | "cmp.c\t%1, %0\n\t" | |
204 | "lw\t%1, [%2,8]\n\t" | |
205 | "bleu 1f\n\t" | |
206 | "addi\t%0, 0x1\n\t" | |
207 | "1:add\t%0, %0, %1\n\t" | |
208 | "cmp.c\t%1, %0\n\t" | |
209 | "lw\t%1, [%2, 12]\n\t" | |
210 | "bleu 1f\n\t" | |
211 | "addi\t%0, 0x1\n\t" | |
212 | "1:add\t%0, %0,%1\n\t" | |
213 | "cmp.c\t%1, %0\n\t" | |
214 | "lw\t%1, [%3, 0]\n\t" | |
215 | "bleu 1f\n\t" | |
216 | "addi\t%0, 0x1\n\t" | |
217 | "1:add\t%0, %0, %1\n\t" | |
218 | "cmp.c\t%1, %0\n\t" | |
219 | "lw\t%1, [%3, 4]\n\t" | |
220 | "bleu 1f\n\t" | |
221 | "addi\t%0, 0x1\n\t" | |
222 | "1:add\t%0, %0, %1\n\t" | |
223 | "cmp.c\t%1, %0\n\t" | |
224 | "lw\t%1, [%3, 8]\n\t" | |
225 | "bleu 1f\n\t" | |
226 | "addi\t%0, 0x1\n\t" | |
227 | "1:add\t%0, %0, %1\n\t" | |
228 | "cmp.c\t%1, %0\n\t" | |
229 | "lw\t%1, [%3, 12]\n\t" | |
230 | "bleu 1f\n\t" | |
231 | "addi\t%0, 0x1\n\t" | |
232 | "1:add\t%0, %0, %1\n\t" | |
233 | "cmp.c\t%1, %0\n\t" | |
234 | "bleu 1f\n\t" | |
235 | "addi\t%0, 0x1\n\t" | |
236 | "1:\n\t" | |
237 | ".set\toptimize" | |
6bc9a396 CL |
238 | : "=r" (sum), "=r" (proto) |
239 | : "r" (saddr), "r" (daddr), | |
240 | "0" (htonl(len)), "1" (htonl(proto)), "r" (sum)); | |
241 | ||
242 | return csum_fold(sum); | |
243 | } | |
244 | #endif /* _ASM_SCORE_CHECKSUM_H */ |