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
|
/* Copyright (c) 2009, 2010, 2011, 2012 ARM Ltd.
Permission is hereby granted, free of charge, to any person obtaining
a copy of this software and associated documentation files (the
``Software''), to deal in the Software without restriction, including
without limitation the rights to use, copy, modify, merge, publish,
distribute, sublicense, and/or sell copies of the Software, and to
permit persons to whom the Software is furnished to do so, subject to
the following conditions:
The above copyright notice and this permission notice shall be
included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED ``AS IS'', WITHOUT WARRANTY OF ANY KIND,
EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.
IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY
CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */
#define LIBFFI_ASM
#include <fficonfig.h>
#include <ffi.h>
#define cfi_adjust_cfa_offset(off) .cfi_adjust_cfa_offset off
#define cfi_rel_offset(reg, off) .cfi_rel_offset reg, off
#define cfi_restore(reg) .cfi_restore reg
#define cfi_def_cfa_register(reg) .cfi_def_cfa_register reg
.text
.globl ffi_call_SYSV
.type ffi_call_SYSV, #function
/* ffi_call_SYSV()
Create a stack frame, setup an argument context, call the callee
and extract the result.
The maximum required argument stack size is provided,
ffi_call_SYSV() allocates that stack space then calls the
prepare_fn to populate register context and stack. The
argument passing registers are loaded from the register
context and the callee called, on return the register passing
register are saved back to the context. Our caller will
extract the return value from the final state of the saved
register context.
Prototype:
extern unsigned
ffi_call_SYSV (void (*)(struct call_context *context, unsigned char *,
extended_cif *),
struct call_context *context,
extended_cif *,
unsigned required_stack_size,
void (*fn)(void));
Therefore on entry we have:
x0 prepare_fn
x1 &context
x2 &ecif
x3 bytes
x4 fn
This function uses the following stack frame layout:
==
saved x30(lr)
x29(fp)-> saved x29(fp)
saved x24
saved x23
saved x22
sp' -> saved x21
...
sp -> (constructed callee stack arguments)
==
Voila! */
#define ffi_call_SYSV_FS (8 * 4)
.cfi_startproc
ffi_call_SYSV:
stp x29, x30, [sp, #-16]!
cfi_adjust_cfa_offset (16)
cfi_rel_offset (x29, 0)
cfi_rel_offset (x30, 8)
mov x29, sp
cfi_def_cfa_register (x29)
sub sp, sp, #ffi_call_SYSV_FS
stp x21, x22, [sp, 0]
cfi_rel_offset (x21, 0 - ffi_call_SYSV_FS)
cfi_rel_offset (x22, 8 - ffi_call_SYSV_FS)
stp x23, x24, [sp, 16]
cfi_rel_offset (x23, 16 - ffi_call_SYSV_FS)
cfi_rel_offset (x24, 24 - ffi_call_SYSV_FS)
mov x21, x1
mov x22, x2
mov x24, x4
/* Allocate the stack space for the actual arguments, many
arguments will be passed in registers, but we assume
worst case and allocate sufficient stack for ALL of
the arguments. */
sub sp, sp, x3
/* unsigned (*prepare_fn) (struct call_context *context,
unsigned char *stack, extended_cif *ecif);
*/
mov x23, x0
mov x0, x1
mov x1, sp
/* x2 already in place */
blr x23
/* Preserve the flags returned. */
mov x23, x0
/* Figure out if we should touch the vector registers. */
tbz x23, #AARCH64_FFI_WITH_V_BIT, 1f
/* Load the vector argument passing registers. */
ldp q0, q1, [x21, #8*32 + 0]
ldp q2, q3, [x21, #8*32 + 32]
ldp q4, q5, [x21, #8*32 + 64]
ldp q6, q7, [x21, #8*32 + 96]
1:
/* Load the core argument passing registers. */
ldp x0, x1, [x21, #0]
ldp x2, x3, [x21, #16]
ldp x4, x5, [x21, #32]
ldp x6, x7, [x21, #48]
/* Don't forget x8 which may be holding the address of a return buffer.
*/
ldr x8, [x21, #8*8]
blr x24
/* Save the core argument passing registers. */
stp x0, x1, [x21, #0]
stp x2, x3, [x21, #16]
stp x4, x5, [x21, #32]
stp x6, x7, [x21, #48]
/* Note nothing useful ever comes back in x8! */
/* Figure out if we should touch the vector registers. */
tbz x23, #AARCH64_FFI_WITH_V_BIT, 1f
/* Save the vector argument passing registers. */
stp q0, q1, [x21, #8*32 + 0]
stp q2, q3, [x21, #8*32 + 32]
stp q4, q5, [x21, #8*32 + 64]
stp q6, q7, [x21, #8*32 + 96]
1:
/* All done, unwind our stack frame. */
ldp x21, x22, [x29, # - ffi_call_SYSV_FS]
cfi_restore (x21)
cfi_restore (x22)
ldp x23, x24, [x29, # - ffi_call_SYSV_FS + 16]
cfi_restore (x23)
cfi_restore (x24)
mov sp, x29
cfi_def_cfa_register (sp)
ldp x29, x30, [sp], #16
cfi_adjust_cfa_offset (-16)
cfi_restore (x29)
cfi_restore (x30)
ret
.cfi_endproc
.size ffi_call_SYSV, .-ffi_call_SYSV
#define ffi_closure_SYSV_FS (8 * 2 + AARCH64_CALL_CONTEXT_SIZE)
/* ffi_closure_SYSV
Closure invocation glue. This is the low level code invoked directly by
the closure trampoline to setup and call a closure.
On entry x17 points to a struct trampoline_data, x16 has been clobbered
all other registers are preserved.
We allocate a call context and save the argument passing registers,
then invoked the generic C ffi_closure_SYSV_inner() function to do all
the real work, on return we load the result passing registers back from
the call context.
On entry
extern void
ffi_closure_SYSV (struct trampoline_data *);
struct trampoline_data
{
UINT64 *ffi_closure;
UINT64 flags;
};
This function uses the following stack frame layout:
==
saved x30(lr)
x29(fp)-> saved x29(fp)
saved x22
saved x21
...
sp -> call_context
==
Voila! */
.text
.globl ffi_closure_SYSV
.cfi_startproc
ffi_closure_SYSV:
stp x29, x30, [sp, #-16]!
cfi_adjust_cfa_offset (16)
cfi_rel_offset (x29, 0)
cfi_rel_offset (x30, 8)
mov x29, sp
sub sp, sp, #ffi_closure_SYSV_FS
cfi_adjust_cfa_offset (ffi_closure_SYSV_FS)
stp x21, x22, [x29, #-16]
cfi_rel_offset (x21, 0)
cfi_rel_offset (x22, 8)
/* Load x21 with &call_context. */
mov x21, sp
/* Preserve our struct trampoline_data * */
mov x22, x17
/* Save the rest of the argument passing registers. */
stp x0, x1, [x21, #0]
stp x2, x3, [x21, #16]
stp x4, x5, [x21, #32]
stp x6, x7, [x21, #48]
/* Don't forget we may have been given a result scratch pad address.
*/
str x8, [x21, #64]
/* Figure out if we should touch the vector registers. */
ldr x0, [x22, #8]
tbz x0, #AARCH64_FFI_WITH_V_BIT, 1f
/* Save the argument passing vector registers. */
stp q0, q1, [x21, #8*32 + 0]
stp q2, q3, [x21, #8*32 + 32]
stp q4, q5, [x21, #8*32 + 64]
stp q6, q7, [x21, #8*32 + 96]
1:
/* Load &ffi_closure.. */
ldr x0, [x22, #0]
mov x1, x21
/* Compute the location of the stack at the point that the
trampoline was called. */
add x2, x29, #16
bl ffi_closure_SYSV_inner
/* Figure out if we should touch the vector registers. */
ldr x0, [x22, #8]
tbz x0, #AARCH64_FFI_WITH_V_BIT, 1f
/* Load the result passing vector registers. */
ldp q0, q1, [x21, #8*32 + 0]
ldp q2, q3, [x21, #8*32 + 32]
ldp q4, q5, [x21, #8*32 + 64]
ldp q6, q7, [x21, #8*32 + 96]
1:
/* Load the result passing core registers. */
ldp x0, x1, [x21, #0]
ldp x2, x3, [x21, #16]
ldp x4, x5, [x21, #32]
ldp x6, x7, [x21, #48]
/* Note nothing usefull is returned in x8. */
/* We are done, unwind our frame. */
ldp x21, x22, [x29, #-16]
cfi_restore (x21)
cfi_restore (x22)
mov sp, x29
cfi_adjust_cfa_offset (-ffi_closure_SYSV_FS)
ldp x29, x30, [sp], #16
cfi_adjust_cfa_offset (-16)
cfi_restore (x29)
cfi_restore (x30)
ret
.cfi_endproc
.size ffi_closure_SYSV, .-ffi_closure_SYSV
|
