Fix: rseq: arm branch to failure
[deliverable/linux.git] / tools / testing / selftests / rseq / rseq-arm.h
1 /*
2 * rseq-arm.h
3 *
4 * (C) Copyright 2016 - Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining a copy
7 * of this software and associated documentation files (the "Software"), to deal
8 * in the Software without restriction, including without limitation the rights
9 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
10 * copies of the Software, and to permit persons to whom the Software is
11 * furnished to do so, subject to the following conditions:
12 *
13 * The above copyright notice and this permission notice shall be included in
14 * all copies or substantial portions of the Software.
15 *
16 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
17 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
18 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
19 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
20 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
22 * SOFTWARE.
23 */
24
25 #define smp_mb() __asm__ __volatile__ ("dmb" : : : "memory")
26 #define smp_rmb() __asm__ __volatile__ ("dmb" : : : "memory")
27 #define smp_wmb() __asm__ __volatile__ ("dmb" : : : "memory")
28
29 #define smp_load_acquire(p) \
30 __extension__ ({ \
31 __typeof(*p) ____p1 = READ_ONCE(*p); \
32 smp_mb(); \
33 ____p1; \
34 })
35
36 #define smp_acquire__after_ctrl_dep() smp_rmb()
37
38 #define smp_store_release(p, v) \
39 do { \
40 smp_mb(); \
41 WRITE_ONCE(*p, v); \
42 } while (0)
43
44 #define has_fast_acquire_release() 0
45 #define has_single_copy_load_64() 1
46
47 /*
48 * The __rseq_table section can be used by debuggers to better handle
49 * single-stepping through the restartable critical sections.
50 *
51 * Load the immediate value 0 into register r1 right after the ldr
52 * instruction to improve instruction-level parallelism: load the
53 * constant while the processor is stalled waiting for the load to
54 * complete, which is required by the following comparison and branch.
55 */
56
57 #define RSEQ_FINISH_ASM(_target_final, _to_write_final, _start_value, \
58 _failure, _spec_store, _spec_input, \
59 _final_store, _final_input, _extra_clobber, \
60 _setup, _teardown, _scratch) \
61 do { \
62 _scratch \
63 __asm__ __volatile__ goto ( \
64 ".pushsection __rseq_table, \"aw\"\n\t" \
65 ".balign 32\n\t" \
66 ".word 1f, 0x0, 2f, 0x0, 5f, 0x0, 0x0, 0x0\n\t" \
67 ".popsection\n\t" \
68 "1:\n\t" \
69 _setup \
70 RSEQ_INJECT_ASM(1) \
71 "adr r0, 3f\n\t" \
72 "str r0, [%[rseq_cs]]\n\t" \
73 RSEQ_INJECT_ASM(2) \
74 "ldr r0, %[current_event_counter]\n\t" \
75 "mov r1, #0\n\t" \
76 "cmp %[start_event_counter], r0\n\t" \
77 "bne 5f\n\t" \
78 RSEQ_INJECT_ASM(3) \
79 _spec_store \
80 _final_store \
81 "2:\n\t" \
82 RSEQ_INJECT_ASM(5) \
83 "str r1, [%[rseq_cs]]\n\t" \
84 _teardown \
85 "b 4f\n\t" \
86 ".balign 32\n\t" \
87 "3:\n\t" \
88 ".word 1b, 0x0, 2b, 0x0, 5f, 0x0, 0x0, 0x0\n\t" \
89 "5:\n\t" \
90 "mov r1, #0\n\t" \
91 "str r1, [%[rseq_cs]]\n\t" \
92 _teardown \
93 "b %l[failure]\n\t" \
94 "4:\n\t" \
95 : /* gcc asm goto does not allow outputs */ \
96 : [start_event_counter]"r"((_start_value).event_counter), \
97 [current_event_counter]"m"((_start_value).rseqp->u.e.event_counter), \
98 [rseq_cs]"r"(&(_start_value).rseqp->rseq_cs) \
99 _spec_input \
100 _final_input \
101 RSEQ_INJECT_INPUT \
102 : "r0", "r1", "memory", "cc" \
103 _extra_clobber \
104 RSEQ_INJECT_CLOBBER \
105 : _failure \
106 ); \
107 } while (0)
108
109 #define RSEQ_FINISH_FINAL_STORE_ASM() \
110 "str %[to_write_final], [%[target_final]]\n\t"
111
112 #define RSEQ_FINISH_FINAL_STORE_RELEASE_ASM() \
113 "dmb\n\t" \
114 RSEQ_FINISH_FINAL_STORE_ASM()
115
116 #define RSEQ_FINISH_FINAL_STORE_INPUT(_target_final, _to_write_final) \
117 , [to_write_final]"r"(_to_write_final), \
118 [target_final]"r"(_target_final)
119
120 #define RSEQ_FINISH_SPECULATIVE_STORE_ASM() \
121 "str %[to_write_spec], [%[target_spec]]\n\t" \
122 RSEQ_INJECT_ASM(4)
123
124 #define RSEQ_FINISH_SPECULATIVE_STORE_INPUT(_target_spec, _to_write_spec) \
125 , [to_write_spec]"r"(_to_write_spec), \
126 [target_spec]"r"(_target_spec)
127
128 /* TODO: implement a faster memcpy. */
129 #define RSEQ_FINISH_MEMCPY_STORE_ASM() \
130 "cmp %[len_memcpy], #0\n\t" \
131 "beq 333f\n\t" \
132 "222:\n\t" \
133 "ldrb %%r0, [%[to_write_memcpy]]\n\t" \
134 "strb %%r0, [%[target_memcpy]]\n\t" \
135 "adds %[to_write_memcpy], #1\n\t" \
136 "adds %[target_memcpy], #1\n\t" \
137 "subs %[len_memcpy], #1\n\t" \
138 "bne 222b\n\t" \
139 "333:\n\t" \
140 RSEQ_INJECT_ASM(4)
141
142 #define RSEQ_FINISH_MEMCPY_STORE_INPUT(_target_memcpy, _to_write_memcpy, _len_memcpy) \
143 , [to_write_memcpy]"r"(_to_write_memcpy), \
144 [target_memcpy]"r"(_target_memcpy), \
145 [len_memcpy]"r"(_len_memcpy), \
146 [rseq_scratch0]"m"(rseq_scratch[0]), \
147 [rseq_scratch1]"m"(rseq_scratch[1]), \
148 [rseq_scratch2]"m"(rseq_scratch[2])
149
150 /* We can use r0. */
151 #define RSEQ_FINISH_MEMCPY_CLOBBER()
152
153 #define RSEQ_FINISH_MEMCPY_SCRATCH() \
154 uint32_t rseq_scratch[3];
155
156 /*
157 * We need to save and restore those input registers so they can be
158 * modified within the assembly.
159 */
160 #define RSEQ_FINISH_MEMCPY_SETUP() \
161 "str %[to_write_memcpy], %[rseq_scratch0]\n\t" \
162 "str %[target_memcpy], %[rseq_scratch1]\n\t" \
163 "str %[len_memcpy], %[rseq_scratch2]\n\t"
164
165 #define RSEQ_FINISH_MEMCPY_TEARDOWN() \
166 "ldr %[len_memcpy], %[rseq_scratch2]\n\t" \
167 "ldr %[target_memcpy], %[rseq_scratch1]\n\t" \
168 "ldr %[to_write_memcpy], %[rseq_scratch0]\n\t"
This page took 0.03297 seconds and 5 git commands to generate.