/src/capstonenext/arch/BPF/BPFMapping.c
Line  | Count  | Source  | 
1  |  | /* Capstone Disassembly Engine */  | 
2  |  | /* BPF Backend by david942j <david942j@gmail.com>, 2019 */  | 
3  |  | /* SPDX-FileCopyrightText: 2024 Roee Toledano <roeetoledano10@gmail.com> */  | 
4  |  | /* SPDX-License-Identifier: BSD-3 */  | 
5  |  |  | 
6  |  | #include <string.h>  | 
7  |  |  | 
8  |  | #include "BPFConstants.h"  | 
9  |  | #include "BPFMapping.h"  | 
10  |  | #include "../../Mapping.h"  | 
11  |  | #include "../../utils.h"  | 
12  |  |  | 
13  |  | #ifndef CAPSTONE_DIET  | 
14  |  | static const name_map group_name_maps[] = { | 
15  |  |   { BPF_GRP_INVALID, NULL }, | 
16  |  |  | 
17  |  |   { BPF_GRP_LOAD, "load" },  { BPF_GRP_STORE, "store" }, | 
18  |  |   { BPF_GRP_ALU, "alu" },    { BPF_GRP_JUMP, "jump" }, | 
19  |  |   { BPF_GRP_CALL, "call" },  { BPF_GRP_RETURN, "return" }, | 
20  |  |   { BPF_GRP_MISC, "misc" }, | 
21  |  | };  | 
22  |  | #endif  | 
23  |  |  | 
24  |  | const char *BPF_group_name(csh handle, unsigned int id)  | 
25  | 34.2k  | { | 
26  | 34.2k  | #ifndef CAPSTONE_DIET  | 
27  | 34.2k  |   return id2name(group_name_maps, ARR_SIZE(group_name_maps), id);  | 
28  |  | #else  | 
29  |  |   return NULL;  | 
30  |  | #endif  | 
31  | 34.2k  | }  | 
32  |  |  | 
33  |  | #ifndef CAPSTONE_DIET  | 
34  |  | static const name_map insn_name_maps[BPF_INS_ENDING] = { | 
35  |  |   { BPF_INS_INVALID, NULL }, | 
36  |  |  | 
37  |  |   { BPF_INS_ADD, "add" },   { BPF_INS_SUB, "sub" }, | 
38  |  |   { BPF_INS_MUL, "mul" },   { BPF_INS_DIV, "div" }, | 
39  |  |   { BPF_INS_SDIV, "sdiv" }, { BPF_INS_OR, "or" }, | 
40  |  |   { BPF_INS_AND, "and" },   { BPF_INS_LSH, "lsh" }, | 
41  |  |   { BPF_INS_RSH, "rsh" },   { BPF_INS_NEG, "neg" }, | 
42  |  |   { BPF_INS_MOD, "mod" },   { BPF_INS_SMOD, "smod" }, | 
43  |  |   { BPF_INS_XOR, "xor" },   { BPF_INS_MOV, "mov" }, | 
44  |  |   { BPF_INS_MOVSB, "movsb" }, { BPF_INS_MOVSH, "movsh" }, | 
45  |  |   { BPF_INS_ARSH, "arsh" }, | 
46  |  |  | 
47  |  |   { BPF_INS_ADD64, "add64" }, { BPF_INS_SUB64, "sub64" }, | 
48  |  |   { BPF_INS_MUL64, "mul64" }, { BPF_INS_DIV64, "div64" }, | 
49  |  |   { BPF_INS_SDIV64, "sdiv64" }, { BPF_INS_OR64, "or64" }, | 
50  |  |   { BPF_INS_AND64, "and64" }, { BPF_INS_LSH64, "lsh64" }, | 
51  |  |   { BPF_INS_RSH64, "rsh64" }, { BPF_INS_NEG64, "neg64" }, | 
52  |  |   { BPF_INS_MOD64, "mod64" }, { BPF_INS_SMOD64, "smod64" }, | 
53  |  |   { BPF_INS_XOR64, "xor64" }, { BPF_INS_MOV64, "mov64" }, | 
54  |  |   { BPF_INS_MOVSB64, "movsb64" }, { BPF_INS_MOVSH64, "movsh64" }, | 
55  |  |   { BPF_INS_MOVSW64, "movsw64" }, { BPF_INS_ARSH64, "arsh64" }, | 
56  |  |  | 
57  |  |   { BPF_INS_LE16, "le16" }, { BPF_INS_LE32, "le32" }, | 
58  |  |   { BPF_INS_LE64, "le64" }, { BPF_INS_BE16, "be16" }, | 
59  |  |   { BPF_INS_BE32, "be32" }, { BPF_INS_BE64, "be64" }, | 
60  |  |   { BPF_INS_BSWAP16, "bswap16" }, { BPF_INS_BSWAP32, "bswap32" }, | 
61  |  |   { BPF_INS_BSWAP64, "bswap64" }, | 
62  |  |  | 
63  |  |   { BPF_INS_LDW, "ldw" },   { BPF_INS_LDH, "ldh" }, | 
64  |  |   { BPF_INS_LDB, "ldb" },   { BPF_INS_LDDW, "lddw" }, | 
65  |  |   { BPF_INS_LDXW, "ldxw" }, { BPF_INS_LDXH, "ldxh" }, | 
66  |  |   { BPF_INS_LDXB, "ldxb" }, { BPF_INS_LDXDW, "ldxdw" }, | 
67  |  |   { BPF_INS_LDABSW, "ldabsw" }, { BPF_INS_LDABSH, "ldabsh" }, | 
68  |  |   { BPF_INS_LDABSB, "ldabsb" }, { BPF_INS_LDINDW, "ldindw" }, | 
69  |  |   { BPF_INS_LDINDH, "ldindh" }, { BPF_INS_LDINDB, "ldindb" }, | 
70  |  |  | 
71  |  |   { BPF_INS_STW, "stw" },   { BPF_INS_STH, "sth" }, | 
72  |  |   { BPF_INS_STB, "stb" },   { BPF_INS_STDW, "stdw" }, | 
73  |  |   { BPF_INS_STXW, "stxw" }, { BPF_INS_STXH, "stxh" }, | 
74  |  |   { BPF_INS_STXB, "stxb" }, { BPF_INS_STXDW, "stxdw" }, | 
75  |  |   { BPF_INS_XADDW, "xaddw" }, { BPF_INS_XADDDW, "xadddw" }, | 
76  |  |  | 
77  |  |   { BPF_INS_JA, "ja" },   { BPF_INS_JEQ, "jeq" }, | 
78  |  |   { BPF_INS_JGT, "jgt" },   { BPF_INS_JGE, "jge" }, | 
79  |  |   { BPF_INS_JSET, "jset" }, { BPF_INS_JNE, "jne" }, | 
80  |  |   { BPF_INS_JSGT, "jsgt" }, { BPF_INS_JSGE, "jsge" }, | 
81  |  |   { BPF_INS_CALL, "call" }, { BPF_INS_CALLX, "callx" }, | 
82  |  |   { BPF_INS_EXIT, "exit" }, { BPF_INS_JLT, "jlt" }, | 
83  |  |   { BPF_INS_JLE, "jle" },   { BPF_INS_JSLT, "jslt" }, | 
84  |  |   { BPF_INS_JSLE, "jsle" }, | 
85  |  |  | 
86  |  |   { BPF_INS_JAL, "jal" },   { BPF_INS_JEQ32, "jeq32" }, | 
87  |  |   { BPF_INS_JGT32, "jgt32" }, { BPF_INS_JGE32, "jge32" }, | 
88  |  |   { BPF_INS_JSET32, "jset32" }, { BPF_INS_JNE32, "jne32" }, | 
89  |  |   { BPF_INS_JSGT32, "jsgt32" }, { BPF_INS_JSGE32, "jsge32" }, | 
90  |  |   { BPF_INS_JLT32, "jlt32" }, { BPF_INS_JLE32, "jle32" }, | 
91  |  |   { BPF_INS_JSLT32, "jslt32" }, { BPF_INS_JSLE32, "jsle32" }, | 
92  |  |  | 
93  |  |   { BPF_INS_RET, "ret" }, | 
94  |  |  | 
95  |  |   { BPF_INS_AADD, "aadd" }, { BPF_INS_AOR, "aor" }, | 
96  |  |   { BPF_INS_AAND, "aand" }, { BPF_INS_AXOR, "axor" }, | 
97  |  |   { BPF_INS_AFADD, "afadd" }, { BPF_INS_AFOR, "afor" }, | 
98  |  |   { BPF_INS_AFAND, "afand" }, { BPF_INS_AFXOR, "afxor" }, | 
99  |  |  | 
100  |  |   { BPF_INS_AXCHG64, "axchg64" }, { BPF_INS_ACMPXCHG64, "acmpxchg64" }, | 
101  |  |   { BPF_INS_AADD64, "aadd64" }, { BPF_INS_AOR64, "aor64" }, | 
102  |  |   { BPF_INS_AAND64, "aand64" }, { BPF_INS_AXOR64, "axor64" }, | 
103  |  |   { BPF_INS_AFADD64, "afadd64" }, { BPF_INS_AFOR64, "afor64" }, | 
104  |  |   { BPF_INS_AFAND64, "afand64" }, { BPF_INS_AFXOR64, "afxor64" }, | 
105  |  |  | 
106  |  |   { BPF_INS_TAX, "tax" },   { BPF_INS_TXA, "txa" }, | 
107  |  | };  | 
108  |  | #endif  | 
109  |  |  | 
110  |  | bool BPF_getFeature(const cs_mode mode, const cs_mode feature)  | 
111  | 199k  | { | 
112  | 199k  |   return (mode & feature);  | 
113  | 199k  | }  | 
114  |  |  | 
115  |  | const char *BPF_insn_name(csh handle, unsigned int id)  | 
116  | 68.5k  | { | 
117  | 68.5k  | #ifndef CAPSTONE_DIET  | 
118  |  |   /* We have some special cases because 'ld' in cBPF is equivalent to 'ldw'  | 
119  |  |    * in eBPF, and we don't want to see 'ldw' appears in cBPF mode.  | 
120  |  |    */  | 
121  | 68.5k  |   if (!EBPF_MODE(((cs_struct *)handle)->mode)) { | 
122  | 23.9k  |     switch (id) { | 
123  | 4.65k  |     case BPF_INS_LD:  | 
124  | 4.65k  |       return "ld";  | 
125  | 2.17k  |     case BPF_INS_LDX:  | 
126  | 2.17k  |       return "ldx";  | 
127  | 908  |     case BPF_INS_ST:  | 
128  | 908  |       return "st";  | 
129  | 204  |     case BPF_INS_STX:  | 
130  | 204  |       return "stx";  | 
131  | 23.9k  |     }  | 
132  | 23.9k  |   }  | 
133  | 60.6k  |   return id2name(insn_name_maps, ARR_SIZE(insn_name_maps), id);  | 
134  |  | #else  | 
135  |  |   return NULL;  | 
136  |  | #endif  | 
137  | 68.5k  | }  | 
138  |  |  | 
139  |  | const char *BPF_reg_name(csh handle, unsigned int reg)  | 
140  | 45.5k  | { | 
141  | 45.5k  | #ifndef CAPSTONE_DIET  | 
142  | 45.5k  |   if (EBPF_MODE(((cs_struct *)handle)->mode)) { | 
143  | 27.7k  |     if (reg < BPF_REG_R0 || reg > BPF_REG_R10)  | 
144  | 0  |       return NULL;  | 
145  | 27.7k  |     static const char reg_names[11][4] = { "r0", "r1", "r2", "r3", | 
146  | 27.7k  |                    "r4", "r5", "r6", "r7",  | 
147  | 27.7k  |                    "r8", "r9", "r10" };  | 
148  | 27.7k  |     return reg_names[reg - BPF_REG_R0];  | 
149  | 27.7k  |   }  | 
150  |  |  | 
151  |  |   /* cBPF mode */  | 
152  | 17.8k  |   if (reg == BPF_REG_A)  | 
153  | 12.4k  |     return "a";  | 
154  | 5.41k  |   else if (reg == BPF_REG_X)  | 
155  | 5.41k  |     return "x";  | 
156  | 0  |   else  | 
157  | 0  |     return NULL;  | 
158  |  | #else  | 
159  |  |   return NULL;  | 
160  |  | #endif  | 
161  | 17.8k  | }  | 
162  |  |  | 
163  |  | void BPF_get_insn_id(cs_struct *h, cs_insn *insn, unsigned int id)  | 
164  | 19.8k  | { | 
165  |  |   // Not used by BPF. Information is set after disassembly.  | 
166  | 19.8k  | }  | 
167  |  |  | 
168  |  | static void sort_and_uniq(cs_regs arr, uint8_t n, uint8_t *new_n)  | 
169  | 0  | { | 
170  |  |   /* arr is always a tiny (usually n < 3) array,  | 
171  |  |    * a simple O(n^2) sort is efficient enough. */  | 
172  | 0  |   size_t iMin;  | 
173  | 0  |   size_t tmp;  | 
174  |  |  | 
175  |  |   /* a modified selection sort for sorting and making unique */  | 
176  | 0  |   for (size_t j = 0; j < n; j++) { | 
177  |  |     /* arr[iMin] will be min(arr[j .. n-1]) */  | 
178  | 0  |     iMin = j;  | 
179  | 0  |     for (size_t i = j + 1; i < n; i++) { | 
180  | 0  |       if (arr[i] < arr[iMin])  | 
181  | 0  |         iMin = i;  | 
182  | 0  |     }  | 
183  | 0  |     if (j != 0 && arr[iMin] == arr[j - 1]) { // duplicate ele found | 
184  | 0  |       arr[iMin] = arr[n - 1];  | 
185  | 0  |       --n;  | 
186  | 0  |     } else { | 
187  | 0  |       tmp = arr[iMin];  | 
188  | 0  |       arr[iMin] = arr[j];  | 
189  | 0  |       arr[j] = tmp;  | 
190  | 0  |     }  | 
191  | 0  |   }  | 
192  |  | 
  | 
193  | 0  |   *new_n = n;  | 
194  | 0  | }  | 
195  |  | void BPF_reg_access(const cs_insn *insn, cs_regs regs_read,  | 
196  |  |         uint8_t *regs_read_count, cs_regs regs_write,  | 
197  |  |         uint8_t *regs_write_count)  | 
198  | 0  | { | 
199  | 0  |   unsigned i;  | 
200  | 0  |   uint8_t read_count, write_count;  | 
201  | 0  |   const cs_bpf *bpf = &(insn->detail->bpf);  | 
202  |  | 
  | 
203  | 0  |   read_count = insn->detail->regs_read_count;  | 
204  | 0  |   write_count = insn->detail->regs_write_count;  | 
205  |  |  | 
206  |  |   // implicit registers  | 
207  | 0  |   memcpy(regs_read, insn->detail->regs_read,  | 
208  | 0  |          read_count * sizeof(insn->detail->regs_read[0]));  | 
209  | 0  |   memcpy(regs_write, insn->detail->regs_write,  | 
210  | 0  |          write_count * sizeof(insn->detail->regs_write[0]));  | 
211  |  | 
  | 
212  | 0  |   for (i = 0; i < bpf->op_count; i++) { | 
213  | 0  |     const cs_bpf_op *op = &(bpf->operands[i]);  | 
214  | 0  |     switch (op->type) { | 
215  | 0  |     default:  | 
216  | 0  |       break;  | 
217  | 0  |     case BPF_OP_REG:  | 
218  | 0  |       if (op->access & CS_AC_READ) { | 
219  | 0  |         regs_read[read_count] = op->reg;  | 
220  | 0  |         read_count++;  | 
221  | 0  |       }  | 
222  | 0  |       if (op->access & CS_AC_WRITE) { | 
223  | 0  |         regs_write[write_count] = op->reg;  | 
224  | 0  |         write_count++;  | 
225  | 0  |       }  | 
226  | 0  |       break;  | 
227  | 0  |     case BPF_OP_MEM:  | 
228  | 0  |       if (op->mem.base != BPF_REG_INVALID) { | 
229  | 0  |         regs_read[read_count] = op->mem.base;  | 
230  | 0  |         read_count++;  | 
231  | 0  |       }  | 
232  | 0  |       break;  | 
233  | 0  |     }  | 
234  | 0  |   }  | 
235  |  |  | 
236  | 0  |   sort_and_uniq(regs_read, read_count, regs_read_count);  | 
237  | 0  |   sort_and_uniq(regs_write, write_count, regs_write_count);  | 
238  | 0  | }  |