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
|
// SPDX-License-Identifier: GPL-2.0
/*
* Copyright (c) 2021, 2024, Oracle and/or its affiliates.
*/
#include <bpf_asm_helpers.h>
.text
/*
* For two buffers, return:
* 1 if the two buffers match in every bit
* 0 if the two buffers mismatch in any bit
* To help the BPF verifier, minimize branching by:
* - using a len whose exact value is known to the BPF verifier
* - operating on 64 bits at a time (len is a multiple of 8)
* - replacing conditional branching with arithmetic operations
* such as ^ | & etc.
*
* uint64_t dt_index_match(char *tmp1, char *tmp2, uint64_t len)
* {
* r0 = 0; // accumulate bits indicating mismatches
* r6 = 0; // loop counter
* L1:
* if (r6 >= len) goto L2;
*
* r4 = *((uint64_t*)&tmp1[r6]);
* r5 = *((uint64_t*)&tmp2[r6]);
* r0 |= (r4 ^ r5);
*
* r6 += 8;
* goto L1;
*
* L2:
* // value of r0
* // perfect match any mismatches
* // == 0 != 0
* r0 |= (r0 >> 32); // == 0 != 0
* r0 &= -1; // == 0 > 0
* r0 -= 1; // < 0 >= 0
* r0 >>= 63; // == 1 == 0
*
* return r0;
* }
*/
.align 4
.global dt_index_match
.type dt_index_match, @function
dt_index_match:
mov %r0, 0
mov %r6, 0
.L1:
jge %r6, %r3, .L2
mov %r4, %r1
add %r4, %r6
ldxdw %r4, [%r4+0]
mov %r5, %r2
add %r5, %r6
ldxdw %r5, [%r5+0]
xor %r4, %r5
or %r0, %r4
add %r6, 8
ja .L1
.L2:
mov %r4, %r0
rsh %r4, 32
or %r0, %r4
and %r0, -1
sub %r0, 1
rsh %r0, 63
exit
.size dt_index_match, .-dt_index_match
/*
* int dt_index(const char *s, const char *t, int start, char *tmp1, char *tmp2)
* {
* uint64_t r0, tlen;
* uint64_t buflen;
*
* // determine actual start index
* if (start < 0) start = 0;
*
* // round buflen for dt_index_match()
* buflen = STRSZ rounded up to multiple of 8;
*
* // keep a copy of t in tmp2
* tlen = bpf_probe_read_str(tmp2, buflen, t);
*
* // determine maximum possible index value
* maxi = bpf_probe_read_str(tmp1, buflen, s);
* maxi -= tlen;
*
* // drop terminating NULL
* tlen--;
*
* // Fill end of tmp1 with contents from tmp2
* // to suppress spurious mismatches.
* bpf_probe_read(tmp1 + tlen, buflen - tlen, tmp2 + tlen);
*
* Lloop:
* // check loop
* if (start > maxi) return -1;
*
* // fill start of tmp1 with s, starting at the proposed index
* bpf_probe_read(tmp1, tlen, s + start);
*
* // keep looping if not a match
* r0 = dt_index_match(tmp1, tmp2, buflen);
* start++;
* if (r0 == 0) goto Lloop;
*
* start--;
* return start;
* }
*
* Some variables are kept in registers or spilled to the stack:
* r6 = start [%fp+-8] = s
* r7 = tmp1 [%fp+-16] = buflen
* r8 = tmp2 [%fp+-24] = maxi
* r9 = tlen
* but t is not needed once we have copied its contents to tmp2.
*/
.align 4
.global dt_index
.type dt_index, @function
dt_index:
jsge %r3, 0, 1
mov %r3, 0 /* if (start < 0) start = 0 */
lddw %r6, STRSZ
add %r6, 7
and %r6, -8
stxdw [%fp+-16], %r6 /* buflen = STRSZ rounded up to multiple of 8 */
stxdw [%fp+-8], %r1 /* stash copies of some variables */
mov %r6, %r3
mov %r7, %r4
mov %r8, %r5
mov %r3, %r2
ldxdw %r2, [%fp+-16]
mov %r1, %r8
call BPF_FUNC_probe_read_str /* tlen = bpf_probe_read_str(tmp2, buflen, t) */
jsle %r0, 0, .Lerror
mov %r9, %r0
mov %r1, %r7
ldxdw %r2, [%fp+-16]
ldxdw %r3, [%fp+-8]
call BPF_FUNC_probe_read_str /* maxi = bpf_probe_read_str(tmp1, buflen, s) */
sub %r0, %r9 /* maxi -= tlen */
jslt %r0, 0, .Lerror
stxdw [%fp+-24], %r0
sub %r9, 1 /* tlen-- */
mov %r1, %r7
add %r1, %r9
ldxdw %r2, [%fp+-16]
sub %r2, %r9
mov %r3, %r8
add %r3, %r9
call BPF_FUNC_probe_read /* bpf_probe_read(tmp1 + tlen, buflen - tlen, tmp2 + tlen) */
.Lloop:
/* help the BPF verifier */
ldxdw %r0, [%fp+-16]
jge %r6, %r0, .Lerror /* if (start >= buflen) goto Lerror */
ldxdw %r0, [%fp+-24]
jgt %r6, %r0, .Lerror /* if (start > maxi) goto Lerror */
mov %r1, %r7
mov %r2, %r9
ldxdw %r3, [%fp+-8]
add %r3, %r6
call BPF_FUNC_probe_read /* bpf_probe_read(tmp1, tlen, s + start) */
mov %r1, %r7
mov %r2, %r8
ldxdw %r3, [%fp+-16]
call dt_index_match /* r0 = dt_index_match(tmp1, tmp2, buflen) */
add %r6, 1 /* start++ */
jeq %r0, 0, .Lloop /* if (r0 == 0) goto Lloop */
/* done */
sub %r6, 1 /* start-- */
mov %r0, %r6 /* return start */
exit
.Lerror:
mov %r0, -1
exit
.size dt_index, .-dt_index
|
