Use rseq for cpu_id in libringbuffer
[lttng-ust.git] / libringbuffer / rseq-arm.h
CommitLineData
b76e5200
MD
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
b76e5200
MD
25#define has_fast_acquire_release() 0
26#define has_single_copy_load_64() 1
27
28/*
29 * The __rseq_table section can be used by debuggers to better handle
30 * single-stepping through the restartable critical sections.
31 *
32 * Load the immediate value 0 into register r1 right after the ldr
33 * instruction to improve instruction-level parallelism: load the
34 * constant while the processor is stalled waiting for the load to
35 * complete, which is required by the following comparison and branch.
36 */
37
38#define RSEQ_FINISH_ASM(_target_final, _to_write_final, _start_value, \
39 _failure, _spec_store, _spec_input, \
40 _final_store, _final_input, _extra_clobber, \
41 _setup, _teardown, _scratch) \
42do { \
43 _scratch \
44 __asm__ __volatile__ goto ( \
45 ".pushsection __rseq_table, \"aw\"\n\t" \
46 ".balign 32\n\t" \
47 ".word 1f, 0x0, 2f, 0x0, 5f, 0x0, 0x0, 0x0\n\t" \
48 ".popsection\n\t" \
49 "1:\n\t" \
50 _setup \
51 RSEQ_INJECT_ASM(1) \
52 "adr r0, 3f\n\t" \
53 "str r0, [%[rseq_cs]]\n\t" \
54 RSEQ_INJECT_ASM(2) \
55 "ldr r0, %[current_event_counter]\n\t" \
56 "mov r1, #0\n\t" \
57 "cmp %[start_event_counter], r0\n\t" \
58 "bne 5f\n\t" \
59 RSEQ_INJECT_ASM(3) \
60 _spec_store \
61 _final_store \
62 "2:\n\t" \
63 RSEQ_INJECT_ASM(5) \
64 "str r1, [%[rseq_cs]]\n\t" \
65 _teardown \
66 "b 4f\n\t" \
67 ".balign 32\n\t" \
68 "3:\n\t" \
69 ".word 1b, 0x0, 2b, 0x0, 5f, 0x0, 0x0, 0x0\n\t" \
70 "5:\n\t" \
71 "mov r1, #0\n\t" \
72 "str r1, [%[rseq_cs]]\n\t" \
73 _teardown \
74 "b %l[failure]\n\t" \
75 "4:\n\t" \
26cc635c 76 : /* gcc asm goto does not allow outputs */ \
b76e5200
MD
77 : [start_event_counter]"r"((_start_value).event_counter), \
78 [current_event_counter]"m"((_start_value).rseqp->u.e.event_counter), \
79 [rseq_cs]"r"(&(_start_value).rseqp->rseq_cs) \
80 _spec_input \
81 _final_input \
82 RSEQ_INJECT_INPUT \
83 : "r0", "r1", "memory", "cc" \
84 _extra_clobber \
85 RSEQ_INJECT_CLOBBER \
86 : _failure \
87 ); \
88} while (0)
89
90#define RSEQ_FINISH_FINAL_STORE_ASM() \
91 "str %[to_write_final], [%[target_final]]\n\t"
92
93#define RSEQ_FINISH_FINAL_STORE_RELEASE_ASM() \
94 "dmb\n\t" \
95 RSEQ_FINISH_FINAL_STORE_ASM()
96
97#define RSEQ_FINISH_FINAL_STORE_INPUT(_target_final, _to_write_final) \
98 , [to_write_final]"r"(_to_write_final), \
99 [target_final]"r"(_target_final)
100
101#define RSEQ_FINISH_SPECULATIVE_STORE_ASM() \
102 "str %[to_write_spec], [%[target_spec]]\n\t" \
103 RSEQ_INJECT_ASM(4)
104
105#define RSEQ_FINISH_SPECULATIVE_STORE_INPUT(_target_spec, _to_write_spec) \
106 , [to_write_spec]"r"(_to_write_spec), \
107 [target_spec]"r"(_target_spec)
108
109/* TODO: implement a faster memcpy. */
110#define RSEQ_FINISH_MEMCPY_STORE_ASM() \
111 "cmp %[len_memcpy], #0\n\t" \
112 "beq 333f\n\t" \
113 "222:\n\t" \
114 "ldrb %%r0, [%[to_write_memcpy]]\n\t" \
115 "strb %%r0, [%[target_memcpy]]\n\t" \
116 "adds %[to_write_memcpy], #1\n\t" \
117 "adds %[target_memcpy], #1\n\t" \
118 "subs %[len_memcpy], #1\n\t" \
119 "bne 222b\n\t" \
120 "333:\n\t" \
121 RSEQ_INJECT_ASM(4)
122
123#define RSEQ_FINISH_MEMCPY_STORE_INPUT(_target_memcpy, _to_write_memcpy, _len_memcpy) \
124 , [to_write_memcpy]"r"(_to_write_memcpy), \
125 [target_memcpy]"r"(_target_memcpy), \
126 [len_memcpy]"r"(_len_memcpy), \
127 [rseq_scratch0]"m"(rseq_scratch[0]), \
128 [rseq_scratch1]"m"(rseq_scratch[1]), \
129 [rseq_scratch2]"m"(rseq_scratch[2])
130
131/* We can use r0. */
132#define RSEQ_FINISH_MEMCPY_CLOBBER()
133
134#define RSEQ_FINISH_MEMCPY_SCRATCH() \
135 uint32_t rseq_scratch[3];
136
137/*
138 * We need to save and restore those input registers so they can be
139 * modified within the assembly.
140 */
141#define RSEQ_FINISH_MEMCPY_SETUP() \
142 "str %[to_write_memcpy], %[rseq_scratch0]\n\t" \
143 "str %[target_memcpy], %[rseq_scratch1]\n\t" \
144 "str %[len_memcpy], %[rseq_scratch2]\n\t"
145
146#define RSEQ_FINISH_MEMCPY_TEARDOWN() \
147 "ldr %[len_memcpy], %[rseq_scratch2]\n\t" \
148 "ldr %[target_memcpy], %[rseq_scratch1]\n\t" \
149 "ldr %[to_write_memcpy], %[rseq_scratch0]\n\t"
This page took 0.029377 seconds and 5 git commands to generate.