1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
| /* SPDX-License-Identifier: GPL-2.0+
| *
| * $Id: checksum.S,v 1.10 2001/07/06 13:11:32 gniibe Exp $
| *
| * INET An implementation of the TCP/IP protocol suite for the LINUX
| * operating system. INET is implemented using the BSD Socket
| * interface as the means of communication with the user level.
| *
| * IP/TCP/UDP checksumming routines
| *
| * Authors: Jorge Cwik, <jorge@laser.satlink.net>
| * Arnt Gulbrandsen, <agulbra@nvg.unit.no>
| * Tom May, <ftom@netcom.com>
| * Pentium Pro/II routines:
| * Alexander Kjeldaas <astor@guardian.no>
| * Finn Arne Gangstad <finnag@guardian.no>
| * Lots of code moved from tcp.c and ip.c; see those files
| * for more names.
| *
| * Changes: Ingo Molnar, converted csum_partial_copy() to 2.1 exception
| * handling.
| * Andi Kleen, add zeroing on error
| * converted to pure assembler
| *
| * SuperH version: Copyright (C) 1999 Niibe Yutaka
| */
|
| #include <asm/errno.h>
| #include <linux/linkage.h>
|
| /*
| * computes a partial checksum, e.g. for TCP/UDP fragments
| */
|
| /*
| * asmlinkage __wsum csum_partial(const void *buf, int len, __wsum sum);
| */
|
| .text
| ENTRY(csum_partial)
| /*
| * Experiments with Ethernet and SLIP connections show that buff
| * is aligned on either a 2-byte or 4-byte boundary. We get at
| * least a twofold speedup on 486 and Pentium if it is 4-byte aligned.
| * Fortunately, it is easy to convert 2-byte alignment to 4-byte
| * alignment for the unrolled loop.
| */
| mov r4, r0
| tst #3, r0 ! Check alignment.
| bt/s 2f ! Jump if alignment is ok.
| mov r4, r7 ! Keep a copy to check for alignment
| !
| tst #1, r0 ! Check alignment.
| bt 21f ! Jump if alignment is boundary of 2bytes.
|
| ! buf is odd
| tst r5, r5
| add #-1, r5
| bt 9f
| mov.b @r4+, r0
| extu.b r0, r0
| addc r0, r6 ! t=0 from previous tst
| mov r6, r0
| shll8 r6
| shlr16 r0
| shlr8 r0
| or r0, r6
| mov r4, r0
| tst #2, r0
| bt 2f
| 21:
| ! buf is 2 byte aligned (len could be 0)
| add #-2, r5 ! Alignment uses up two bytes.
| cmp/pz r5 !
| bt/s 1f ! Jump if we had at least two bytes.
| clrt
| bra 6f
| add #2, r5 ! r5 was < 2. Deal with it.
| 1:
| mov.w @r4+, r0
| extu.w r0, r0
| addc r0, r6
| bf 2f
| add #1, r6
| 2:
| ! buf is 4 byte aligned (len could be 0)
| mov r5, r1
| mov #-5, r0
| shld r0, r1
| tst r1, r1
| bt/s 4f ! if it's =0, go to 4f
| clrt
| .align 2
| 3:
| mov.l @r4+, r0
| mov.l @r4+, r2
| mov.l @r4+, r3
| addc r0, r6
| mov.l @r4+, r0
| addc r2, r6
| mov.l @r4+, r2
| addc r3, r6
| mov.l @r4+, r3
| addc r0, r6
| mov.l @r4+, r0
| addc r2, r6
| mov.l @r4+, r2
| addc r3, r6
| addc r0, r6
| addc r2, r6
| movt r0
| dt r1
| bf/s 3b
| cmp/eq #1, r0
| ! here, we know r1==0
| addc r1, r6 ! add carry to r6
| 4:
| mov r5, r0
| and #0x1c, r0
| tst r0, r0
| bt 6f
| ! 4 bytes or more remaining
| mov r0, r1
| shlr2 r1
| mov #0, r2
| 5:
| addc r2, r6
| mov.l @r4+, r2
| movt r0
| dt r1
| bf/s 5b
| cmp/eq #1, r0
| addc r2, r6
| addc r1, r6 ! r1==0 here, so it means add carry-bit
| 6:
| ! 3 bytes or less remaining
| mov #3, r0
| and r0, r5
| tst r5, r5
| bt 9f ! if it's =0 go to 9f
| mov #2, r1
| cmp/hs r1, r5
| bf 7f
| mov.w @r4+, r0
| extu.w r0, r0
| cmp/eq r1, r5
| bt/s 8f
| clrt
| shll16 r0
| addc r0, r6
| 7:
| mov.b @r4+, r0
| extu.b r0, r0
| #ifndef __LITTLE_ENDIAN__
| shll8 r0
| #endif
| 8:
| addc r0, r6
| mov #0, r0
| addc r0, r6
| 9:
| ! Check if the buffer was misaligned, if so realign sum
| mov r7, r0
| tst #1, r0
| bt 10f
| mov r6, r0
| shll8 r6
| shlr16 r0
| shlr8 r0
| or r0, r6
| 10:
| rts
| mov r6, r0
|
| /*
| unsigned int csum_partial_copy_generic (const char *src, char *dst, int len)
| */
|
| /*
| * Copy from ds while checksumming, otherwise like csum_partial with initial
| * sum being ~0U
| */
|
| #define EXC(...) \
| 9999: __VA_ARGS__ ; \
| .section __ex_table, "a"; \
| .long 9999b, 6001f ; \
| .previous
|
| !
| ! r4: const char *SRC
| ! r5: char *DST
| ! r6: int LEN
| !
| ENTRY(csum_partial_copy_generic)
| mov #-1,r7
| mov #3,r0 ! Check src and dest are equally aligned
| mov r4,r1
| and r0,r1
| and r5,r0
| cmp/eq r1,r0
| bf 3f ! Different alignments, use slow version
| tst #1,r0 ! Check dest word aligned
| bf 3f ! If not, do it the slow way
|
| mov #2,r0
| tst r0,r5 ! Check dest alignment.
| bt 2f ! Jump if alignment is ok.
| add #-2,r6 ! Alignment uses up two bytes.
| cmp/pz r6 ! Jump if we had at least two bytes.
| bt/s 1f
| clrt
| add #2,r6 ! r6 was < 2. Deal with it.
| bra 4f
| mov r6,r2
|
| 3: ! Handle different src and dest alignments.
| ! This is not common, so simple byte by byte copy will do.
| mov r6,r2
| shlr r6
| tst r6,r6
| bt 4f
| clrt
| .align 2
| 5:
| EXC( mov.b @r4+,r1 )
| EXC( mov.b @r4+,r0 )
| extu.b r1,r1
| EXC( mov.b r1,@r5 )
| EXC( mov.b r0,@(1,r5) )
| extu.b r0,r0
| add #2,r5
|
| #ifdef __LITTLE_ENDIAN__
| shll8 r0
| #else
| shll8 r1
| #endif
| or r1,r0
|
| addc r0,r7
| movt r0
| dt r6
| bf/s 5b
| cmp/eq #1,r0
| mov #0,r0
| addc r0, r7
|
| mov r2, r0
| tst #1, r0
| bt 7f
| bra 5f
| clrt
|
| ! src and dest equally aligned, but to a two byte boundary.
| ! Handle first two bytes as a special case
| .align 2
| 1:
| EXC( mov.w @r4+,r0 )
| EXC( mov.w r0,@r5 )
| add #2,r5
| extu.w r0,r0
| addc r0,r7
| mov #0,r0
| addc r0,r7
| 2:
| mov r6,r2
| mov #-5,r0
| shld r0,r6
| tst r6,r6
| bt/s 2f
| clrt
| .align 2
| 1:
| EXC( mov.l @r4+,r0 )
| EXC( mov.l @r4+,r1 )
| addc r0,r7
| EXC( mov.l r0,@r5 )
| EXC( mov.l r1,@(4,r5) )
| addc r1,r7
|
| EXC( mov.l @r4+,r0 )
| EXC( mov.l @r4+,r1 )
| addc r0,r7
| EXC( mov.l r0,@(8,r5) )
| EXC( mov.l r1,@(12,r5) )
| addc r1,r7
|
| EXC( mov.l @r4+,r0 )
| EXC( mov.l @r4+,r1 )
| addc r0,r7
| EXC( mov.l r0,@(16,r5) )
| EXC( mov.l r1,@(20,r5) )
| addc r1,r7
|
| EXC( mov.l @r4+,r0 )
| EXC( mov.l @r4+,r1 )
| addc r0,r7
| EXC( mov.l r0,@(24,r5) )
| EXC( mov.l r1,@(28,r5) )
| addc r1,r7
| add #32,r5
| movt r0
| dt r6
| bf/s 1b
| cmp/eq #1,r0
| mov #0,r0
| addc r0,r7
|
| 2: mov r2,r6
| mov #0x1c,r0
| and r0,r6
| cmp/pl r6
| bf/s 4f
| clrt
| shlr2 r6
| 3:
| EXC( mov.l @r4+,r0 )
| addc r0,r7
| EXC( mov.l r0,@r5 )
| add #4,r5
| movt r0
| dt r6
| bf/s 3b
| cmp/eq #1,r0
| mov #0,r0
| addc r0,r7
| 4: mov r2,r6
| mov #3,r0
| and r0,r6
| cmp/pl r6
| bf 7f
| mov #2,r1
| cmp/hs r1,r6
| bf 5f
| EXC( mov.w @r4+,r0 )
| EXC( mov.w r0,@r5 )
| extu.w r0,r0
| add #2,r5
| cmp/eq r1,r6
| bt/s 6f
| clrt
| shll16 r0
| addc r0,r7
| 5:
| EXC( mov.b @r4+,r0 )
| EXC( mov.b r0,@r5 )
| extu.b r0,r0
| #ifndef __LITTLE_ENDIAN__
| shll8 r0
| #endif
| 6: addc r0,r7
| mov #0,r0
| addc r0,r7
| 7:
|
| # Exception handler:
| .section .fixup, "ax"
|
| 6001:
| rts
| mov #0,r0
| .previous
| rts
| mov r7,r0
|
|
