| 1 |
/* ----------------------------------------------------------------------- |
| 2 |
sysv.S - Copyright (c) 2013 Tensilica, Inc. |
| 3 |
|
| 4 |
XTENSA Foreign Function Interface |
| 5 |
|
| 6 |
Permission is hereby granted, free of charge, to any person obtaining |
| 7 |
a copy of this software and associated documentation files (the |
| 8 |
``Software''), to deal in the Software without restriction, including |
| 9 |
without limitation the rights to use, copy, modify, merge, publish, |
| 10 |
distribute, sublicense, and/or sell copies of the Software, and to |
| 11 |
permit persons to whom the Software is furnished to do so, subject to |
| 12 |
the following conditions: |
| 13 |
|
| 14 |
The above copyright notice and this permission notice shall be included |
| 15 |
in all copies or substantial portions of the Software. |
| 16 |
|
| 17 |
THE SOFTWARE IS PROVIDED ``AS IS'', WITHOUT WARRANTY OF ANY KIND, |
| 18 |
EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF |
| 19 |
MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND |
| 20 |
NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT |
| 21 |
HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, |
| 22 |
WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, |
| 23 |
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER |
| 24 |
DEALINGS IN THE SOFTWARE. |
| 25 |
----------------------------------------------------------------------- */ |
| 26 |
|
| 27 |
#define LIBFFI_ASM |
| 28 |
#include <fficonfig.h> |
| 29 |
#include <ffi.h> |
| 30 |
|
| 31 |
#define ENTRY(name) .text; .globl name; .type name,@function; .align 4; name: |
| 32 |
#define END(name) .size name , . - name |
| 33 |
|
| 34 |
/* Assert that the table below is in sync with ffi.h. */ |
| 35 |
|
| 36 |
#if FFI_TYPE_UINT8 != 5 \ |
| 37 |
|| FFI_TYPE_SINT8 != 6 \ |
| 38 |
|| FFI_TYPE_UINT16 != 7 \ |
| 39 |
|| FFI_TYPE_SINT16 != 8 \ |
| 40 |
|| FFI_TYPE_UINT32 != 9 \ |
| 41 |
|| FFI_TYPE_SINT32 != 10 \ |
| 42 |
|| FFI_TYPE_UINT64 != 11 |
| 43 |
#error "xtensa/sysv.S out of sync with ffi.h" |
| 44 |
#endif |
| 45 |
|
| 46 |
#define FFI_REGISTER_ARGS_OFFSET ((XTENSA_STACK_ALIGNMENT - \ |
| 47 |
FFI_REGISTER_NARGS * 4) & \ |
| 48 |
(XTENSA_STACK_ALIGNMENT - 1)) |
| 49 |
|
| 50 |
/* ffi_call_SYSV (rvalue, rbytes, flags, (*fnaddr)(), bytes, ecif) |
| 51 |
void *rvalue; a2 |
| 52 |
unsigned long rbytes; a3 |
| 53 |
unsigned flags; a4 |
| 54 |
void (*fnaddr)(); a5 |
| 55 |
unsigned long bytes; a6 |
| 56 |
extended_cif* ecif) a7 |
| 57 |
*/ |
| 58 |
|
| 59 |
ENTRY(ffi_call_SYSV) |
| 60 |
|
| 61 |
entry a1, 32 # 32 byte frame for using call8 below |
| 62 |
|
| 63 |
mov a10, a7 # a10(->arg0): ecif |
| 64 |
sub a11, a1, a6 # a11(->arg1): stack pointer |
| 65 |
mov a7, a1 # fp |
| 66 |
movsp a1, a11 # set new sp = old_sp - bytes |
| 67 |
|
| 68 |
# align ffi_prep_args stack argument so that arguments |
| 69 |
# passed on stack if any start on 16-byte aligned boundary |
| 70 |
|
| 71 |
addi a11, a11, FFI_REGISTER_ARGS_OFFSET |
| 72 |
|
| 73 |
movi a8, ffi_prep_args |
| 74 |
callx8 a8 # ffi_prep_args(ecif, stack) |
| 75 |
|
| 76 |
# prepare to move stack pointer back |
| 77 |
# to point to arguments passed on stack |
| 78 |
|
| 79 |
addi a6, a1, FFI_REGISTER_ARGS_SPACE |
| 80 |
|
| 81 |
# we can pass up to 6 arguments in registers |
| 82 |
# for simplicity, just load 6 arguments |
| 83 |
|
| 84 |
l32i a10, a1, FFI_REGISTER_ARGS_OFFSET + 0 |
| 85 |
l32i a11, a1, FFI_REGISTER_ARGS_OFFSET + 4 |
| 86 |
l32i a12, a1, FFI_REGISTER_ARGS_OFFSET + 8 |
| 87 |
l32i a13, a1, FFI_REGISTER_ARGS_OFFSET + 12 |
| 88 |
l32i a14, a1, FFI_REGISTER_ARGS_OFFSET + 16 |
| 89 |
l32i a15, a1, FFI_REGISTER_ARGS_OFFSET + 20 |
| 90 |
|
| 91 |
# move stack pointer |
| 92 |
|
| 93 |
movsp a1, a6 |
| 94 |
|
| 95 |
callx8 a5 # (*fn)(args...) |
| 96 |
|
| 97 |
# Handle return value(s) |
| 98 |
|
| 99 |
beqz a2, .Lexit |
| 100 |
|
| 101 |
movi a5, FFI_TYPE_STRUCT |
| 102 |
bne a4, a5, .Lstore |
| 103 |
movi a5, 16 |
| 104 |
blt a5, a3, .Lexit |
| 105 |
|
| 106 |
s32i a10, a2, 0 |
| 107 |
blti a3, 5, .Lexit |
| 108 |
addi a3, a3, -1 |
| 109 |
s32i a11, a2, 4 |
| 110 |
blti a3, 8, .Lexit |
| 111 |
s32i a12, a2, 8 |
| 112 |
blti a3, 12, .Lexit |
| 113 |
s32i a13, a2, 12 |
| 114 |
|
| 115 |
.Lexit: retw |
| 116 |
|
| 117 |
.Lstore: |
| 118 |
addi a4, a4, -FFI_TYPE_UINT8 |
| 119 |
bgei a4, 7, .Lexit # should never happen |
| 120 |
movi a6, store_calls |
| 121 |
add a4, a4, a4 |
| 122 |
addx4 a6, a4, a6 # store_table + idx * 8 |
| 123 |
jx a6 |
| 124 |
|
| 125 |
.align 8 |
| 126 |
store_calls: |
| 127 |
# UINT8 |
| 128 |
s8i a10, a2, 0 |
| 129 |
retw |
| 130 |
|
| 131 |
# SINT8 |
| 132 |
.align 8 |
| 133 |
s8i a10, a2, 0 |
| 134 |
retw |
| 135 |
|
| 136 |
# UINT16 |
| 137 |
.align 8 |
| 138 |
s16i a10, a2, 0 |
| 139 |
retw |
| 140 |
|
| 141 |
# SINT16 |
| 142 |
.align 8 |
| 143 |
s16i a10, a2, 0 |
| 144 |
retw |
| 145 |
|
| 146 |
# UINT32 |
| 147 |
.align 8 |
| 148 |
s32i a10, a2, 0 |
| 149 |
retw |
| 150 |
|
| 151 |
# SINT32 |
| 152 |
.align 8 |
| 153 |
s32i a10, a2, 0 |
| 154 |
retw |
| 155 |
|
| 156 |
# UINT64 |
| 157 |
.align 8 |
| 158 |
s32i a10, a2, 0 |
| 159 |
s32i a11, a2, 4 |
| 160 |
retw |
| 161 |
|
| 162 |
END(ffi_call_SYSV) |
| 163 |
|
| 164 |
|
| 165 |
/* |
| 166 |
* void ffi_cacheflush (unsigned long start, unsigned long end) |
| 167 |
*/ |
| 168 |
|
| 169 |
#define EXTRA_ARGS_SIZE 24 |
| 170 |
|
| 171 |
ENTRY(ffi_cacheflush) |
| 172 |
|
| 173 |
entry a1, 32 |
| 174 |
|
| 175 |
1: |
| 176 |
#if XCHAL_DCACHE_SIZE |
| 177 |
dhwbi a2, 0 |
| 178 |
#endif |
| 179 |
#if XCHAL_ICACHE_SIZE |
| 180 |
ihi a2, 0 |
| 181 |
#endif |
| 182 |
addi a2, a2, 4 |
| 183 |
blt a2, a3, 1b |
| 184 |
|
| 185 |
retw |
| 186 |
|
| 187 |
END(ffi_cacheflush) |
| 188 |
|
| 189 |
/* ffi_trampoline is copied to the stack */ |
| 190 |
|
| 191 |
ENTRY(ffi_trampoline) |
| 192 |
|
| 193 |
/* 32 bytes for spill + spill overflow area of a frame that uses |
| 194 |
call8, |
| 195 |
FFI_REGISTER_NARGS * 4 bytes for arguments passed in registers, |
| 196 |
aligned up to 4 to maintain 16 byte stack alignment, |
| 197 |
4 * 4 bytes for the result. |
| 198 |
This size must be in sync with ffi_closure_SYSV_inner logic. |
| 199 |
*/ |
| 200 |
entry a1, 32 + FFI_REGISTER_ARGS_SPACE + (4 * 4) # [ 0] |
| 201 |
j 2f # [ 3] |
| 202 |
.align 4 # [ 6] |
| 203 |
1: .long 0 # [ 8] |
| 204 |
2: l32r a15, 1b # [12] |
| 205 |
_mov a14, a0 # [15] |
| 206 |
callx0 a15 # [18] |
| 207 |
# [21] |
| 208 |
END(ffi_trampoline) |
| 209 |
|
| 210 |
/* |
| 211 |
* ffi_closure() |
| 212 |
* |
| 213 |
* a0: closure + 21 |
| 214 |
* a14: return address (a0) |
| 215 |
*/ |
| 216 |
|
| 217 |
ENTRY(ffi_closure_SYSV) |
| 218 |
|
| 219 |
/* intentionally omitting entry here */ |
| 220 |
|
| 221 |
# restore return address (a0) and move pointer to closure to a10 |
| 222 |
addi a10, a0, -21 |
| 223 |
mov a0, a14 |
| 224 |
|
| 225 |
# allow up to 4 arguments as return values |
| 226 |
addi a11, a1, 4 * 4 |
| 227 |
|
| 228 |
# save up to 6 arguments to stack (allocated by entry below) |
| 229 |
s32i a2, a11, 0 |
| 230 |
s32i a3, a11, 4 |
| 231 |
s32i a4, a11, 8 |
| 232 |
s32i a5, a11, 12 |
| 233 |
s32i a6, a11, 16 |
| 234 |
s32i a7, a11, 20 |
| 235 |
|
| 236 |
movi a8, ffi_closure_SYSV_inner |
| 237 |
mov a12, a1 |
| 238 |
callx8 a8 # .._inner(*closure, **avalue, *rvalue) |
| 239 |
|
| 240 |
# load up to four return arguments |
| 241 |
l32i a2, a1, 0 |
| 242 |
l32i a3, a1, 4 |
| 243 |
l32i a4, a1, 8 |
| 244 |
l32i a5, a1, 12 |
| 245 |
|
| 246 |
# (sign-)extend return value |
| 247 |
movi a11, FFI_TYPE_UINT8 |
| 248 |
bne a10, a11, 1f |
| 249 |
extui a2, a2, 0, 8 |
| 250 |
retw |
| 251 |
|
| 252 |
1: movi a11, FFI_TYPE_SINT8 |
| 253 |
bne a10, a11, 1f |
| 254 |
sext a2, a2, 7 |
| 255 |
retw |
| 256 |
|
| 257 |
1: movi a11, FFI_TYPE_UINT16 |
| 258 |
bne a10, a11, 1f |
| 259 |
extui a2, a2, 0, 16 |
| 260 |
retw |
| 261 |
|
| 262 |
1: movi a11, FFI_TYPE_SINT16 |
| 263 |
bne a10, a11, 1f |
| 264 |
sext a2, a2, 15 |
| 265 |
|
| 266 |
1: retw |
| 267 |
|
| 268 |
END(ffi_closure_SYSV) |