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
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
|
// Copyright 2009 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
//
// System calls and other sys.stuff for 386, Linux
//
#include "go_asm.h"
#include "go_tls.h"
#include "textflag.h"
// Most linux systems use glibc's dynamic linker, which puts the
// __kernel_vsyscall vdso helper at 0x10(GS) for easy access from position
// independent code and setldt in runtime does the same in the statically
// linked case. However, systems that use alternative libc such as Android's
// bionic and musl, do not save the helper anywhere, and so the only way to
// invoke a syscall from position independent code is boring old int $0x80
// (which is also what syscall wrappers in bionic/musl use).
//
// The benchmarks also showed that using int $0x80 is as fast as calling
// *%gs:0x10 except on AMD Opteron. See https://golang.org/cl/19833
// for the benchmark program and raw data.
//#define INVOKE_SYSCALL CALL 0x10(GS) // non-portable
#define INVOKE_SYSCALL INT $0x80
#define SYS_exit 1
#define SYS_read 3
#define SYS_write 4
#define SYS_open 5
#define SYS_close 6
#define SYS_getpid 20
#define SYS_access 33
#define SYS_kill 37
#define SYS_brk 45
#define SYS_munmap 91
#define SYS_socketcall 102
#define SYS_setittimer 104
#define SYS_clone 120
#define SYS_sched_yield 158
#define SYS_nanosleep 162
#define SYS_rt_sigreturn 173
#define SYS_rt_sigaction 174
#define SYS_rt_sigprocmask 175
#define SYS_sigaltstack 186
#define SYS_mmap2 192
#define SYS_mincore 218
#define SYS_madvise 219
#define SYS_gettid 224
#define SYS_futex 240
#define SYS_sched_getaffinity 242
#define SYS_set_thread_area 243
#define SYS_exit_group 252
#define SYS_timer_create 259
#define SYS_timer_settime 260
#define SYS_timer_delete 263
#define SYS_clock_gettime 265
#define SYS_tgkill 270
#define SYS_pipe2 331
TEXT runtime·exit(SB),NOSPLIT,$0
MOVL $SYS_exit_group, AX
MOVL code+0(FP), BX
INVOKE_SYSCALL
INT $3 // not reached
RET
TEXT exit1<>(SB),NOSPLIT,$0
MOVL $SYS_exit, AX
MOVL code+0(FP), BX
INVOKE_SYSCALL
INT $3 // not reached
RET
// func exitThread(wait *atomic.Uint32)
TEXT runtime·exitThread(SB),NOSPLIT,$0-4
MOVL wait+0(FP), AX
// We're done using the stack.
MOVL $0, (AX)
MOVL $1, AX // exit (just this thread)
MOVL $0, BX // exit code
INT $0x80 // no stack; must not use CALL
// We may not even have a stack any more.
INT $3
JMP 0(PC)
TEXT runtime·open(SB),NOSPLIT,$0
MOVL $SYS_open, AX
MOVL name+0(FP), BX
MOVL mode+4(FP), CX
MOVL perm+8(FP), DX
INVOKE_SYSCALL
CMPL AX, $0xfffff001
JLS 2(PC)
MOVL $-1, AX
MOVL AX, ret+12(FP)
RET
TEXT runtime·closefd(SB),NOSPLIT,$0
MOVL $SYS_close, AX
MOVL fd+0(FP), BX
INVOKE_SYSCALL
CMPL AX, $0xfffff001
JLS 2(PC)
MOVL $-1, AX
MOVL AX, ret+4(FP)
RET
TEXT runtime·write1(SB),NOSPLIT,$0
MOVL $SYS_write, AX
MOVL fd+0(FP), BX
MOVL p+4(FP), CX
MOVL n+8(FP), DX
INVOKE_SYSCALL
MOVL AX, ret+12(FP)
RET
TEXT runtime·read(SB),NOSPLIT,$0
MOVL $SYS_read, AX
MOVL fd+0(FP), BX
MOVL p+4(FP), CX
MOVL n+8(FP), DX
INVOKE_SYSCALL
MOVL AX, ret+12(FP)
RET
// func pipe2(flags int32) (r, w int32, errno int32)
TEXT runtime·pipe2(SB),NOSPLIT,$0-16
MOVL $SYS_pipe2, AX
LEAL r+4(FP), BX
MOVL flags+0(FP), CX
INVOKE_SYSCALL
MOVL AX, errno+12(FP)
RET
TEXT runtime·usleep(SB),NOSPLIT,$8
MOVL $0, DX
MOVL usec+0(FP), AX
MOVL $1000000, CX
DIVL CX
MOVL AX, 0(SP)
MOVL $1000, AX // usec to nsec
MULL DX
MOVL AX, 4(SP)
// nanosleep(&ts, 0)
MOVL $SYS_nanosleep, AX
LEAL 0(SP), BX
MOVL $0, CX
INVOKE_SYSCALL
RET
TEXT runtime·gettid(SB),NOSPLIT,$0-4
MOVL $SYS_gettid, AX
INVOKE_SYSCALL
MOVL AX, ret+0(FP)
RET
TEXT runtime·raise(SB),NOSPLIT,$12
MOVL $SYS_getpid, AX
INVOKE_SYSCALL
MOVL AX, BX // arg 1 pid
MOVL $SYS_gettid, AX
INVOKE_SYSCALL
MOVL AX, CX // arg 2 tid
MOVL sig+0(FP), DX // arg 3 signal
MOVL $SYS_tgkill, AX
INVOKE_SYSCALL
RET
TEXT runtime·raiseproc(SB),NOSPLIT,$12
MOVL $SYS_getpid, AX
INVOKE_SYSCALL
MOVL AX, BX // arg 1 pid
MOVL sig+0(FP), CX // arg 2 signal
MOVL $SYS_kill, AX
INVOKE_SYSCALL
RET
TEXT ·getpid(SB),NOSPLIT,$0-4
MOVL $SYS_getpid, AX
INVOKE_SYSCALL
MOVL AX, ret+0(FP)
RET
TEXT ·tgkill(SB),NOSPLIT,$0
MOVL $SYS_tgkill, AX
MOVL tgid+0(FP), BX
MOVL tid+4(FP), CX
MOVL sig+8(FP), DX
INVOKE_SYSCALL
RET
TEXT runtime·setitimer(SB),NOSPLIT,$0-12
MOVL $SYS_setittimer, AX
MOVL mode+0(FP), BX
MOVL new+4(FP), CX
MOVL old+8(FP), DX
INVOKE_SYSCALL
RET
TEXT runtime·timer_create(SB),NOSPLIT,$0-16
MOVL $SYS_timer_create, AX
MOVL clockid+0(FP), BX
MOVL sevp+4(FP), CX
MOVL timerid+8(FP), DX
INVOKE_SYSCALL
MOVL AX, ret+12(FP)
RET
TEXT runtime·timer_settime(SB),NOSPLIT,$0-20
MOVL $SYS_timer_settime, AX
MOVL timerid+0(FP), BX
MOVL flags+4(FP), CX
MOVL new+8(FP), DX
MOVL old+12(FP), SI
INVOKE_SYSCALL
MOVL AX, ret+16(FP)
RET
TEXT runtime·timer_delete(SB),NOSPLIT,$0-8
MOVL $SYS_timer_delete, AX
MOVL timerid+0(FP), BX
INVOKE_SYSCALL
MOVL AX, ret+4(FP)
RET
TEXT runtime·mincore(SB),NOSPLIT,$0-16
MOVL $SYS_mincore, AX
MOVL addr+0(FP), BX
MOVL n+4(FP), CX
MOVL dst+8(FP), DX
INVOKE_SYSCALL
MOVL AX, ret+12(FP)
RET
// func walltime() (sec int64, nsec int32)
TEXT runtime·walltime(SB), NOSPLIT, $8-12
// We don't know how much stack space the VDSO code will need,
// so switch to g0.
MOVL SP, BP // Save old SP; BP unchanged by C code.
get_tls(CX)
MOVL g(CX), AX
MOVL g_m(AX), SI // SI unchanged by C code.
// Set vdsoPC and vdsoSP for SIGPROF traceback.
// Save the old values on stack and restore them on exit,
// so this function is reentrant.
MOVL m_vdsoPC(SI), CX
MOVL m_vdsoSP(SI), DX
MOVL CX, 0(SP)
MOVL DX, 4(SP)
LEAL sec+0(FP), DX
MOVL -4(DX), CX
MOVL CX, m_vdsoPC(SI)
MOVL DX, m_vdsoSP(SI)
CMPL AX, m_curg(SI) // Only switch if on curg.
JNE noswitch
MOVL m_g0(SI), DX
MOVL (g_sched+gobuf_sp)(DX), SP // Set SP to g0 stack
noswitch:
SUBL $16, SP // Space for results
ANDL $~15, SP // Align for C code
// Stack layout, depending on call path:
// x(SP) vDSO INVOKE_SYSCALL
// 12 ts.tv_nsec ts.tv_nsec
// 8 ts.tv_sec ts.tv_sec
// 4 &ts -
// 0 CLOCK_<id> -
MOVL runtime·vdsoClockgettimeSym(SB), AX
CMPL AX, $0
JEQ fallback
LEAL 8(SP), BX // &ts (struct timespec)
MOVL BX, 4(SP)
MOVL $0, 0(SP) // CLOCK_REALTIME
CALL AX
JMP finish
fallback:
MOVL $SYS_clock_gettime, AX
MOVL $0, BX // CLOCK_REALTIME
LEAL 8(SP), CX
INVOKE_SYSCALL
finish:
MOVL 8(SP), AX // sec
MOVL 12(SP), BX // nsec
MOVL BP, SP // Restore real SP
// Restore vdsoPC, vdsoSP
// We don't worry about being signaled between the two stores.
// If we are not in a signal handler, we'll restore vdsoSP to 0,
// and no one will care about vdsoPC. If we are in a signal handler,
// we cannot receive another signal.
MOVL 4(SP), CX
MOVL CX, m_vdsoSP(SI)
MOVL 0(SP), CX
MOVL CX, m_vdsoPC(SI)
// sec is in AX, nsec in BX
MOVL AX, sec_lo+0(FP)
MOVL $0, sec_hi+4(FP)
MOVL BX, nsec+8(FP)
RET
// int64 nanotime(void) so really
// void nanotime(int64 *nsec)
TEXT runtime·nanotime1(SB), NOSPLIT, $8-8
// Switch to g0 stack. See comment above in runtime·walltime.
MOVL SP, BP // Save old SP; BP unchanged by C code.
get_tls(CX)
MOVL g(CX), AX
MOVL g_m(AX), SI // SI unchanged by C code.
// Set vdsoPC and vdsoSP for SIGPROF traceback.
// Save the old values on stack and restore them on exit,
// so this function is reentrant.
MOVL m_vdsoPC(SI), CX
MOVL m_vdsoSP(SI), DX
MOVL CX, 0(SP)
MOVL DX, 4(SP)
LEAL ret+0(FP), DX
MOVL -4(DX), CX
MOVL CX, m_vdsoPC(SI)
MOVL DX, m_vdsoSP(SI)
CMPL AX, m_curg(SI) // Only switch if on curg.
JNE noswitch
MOVL m_g0(SI), DX
MOVL (g_sched+gobuf_sp)(DX), SP // Set SP to g0 stack
noswitch:
SUBL $16, SP // Space for results
ANDL $~15, SP // Align for C code
MOVL runtime·vdsoClockgettimeSym(SB), AX
CMPL AX, $0
JEQ fallback
LEAL 8(SP), BX // &ts (struct timespec)
MOVL BX, 4(SP)
MOVL $1, 0(SP) // CLOCK_MONOTONIC
CALL AX
JMP finish
fallback:
MOVL $SYS_clock_gettime, AX
MOVL $1, BX // CLOCK_MONOTONIC
LEAL 8(SP), CX
INVOKE_SYSCALL
finish:
MOVL 8(SP), AX // sec
MOVL 12(SP), BX // nsec
MOVL BP, SP // Restore real SP
// Restore vdsoPC, vdsoSP
// We don't worry about being signaled between the two stores.
// If we are not in a signal handler, we'll restore vdsoSP to 0,
// and no one will care about vdsoPC. If we are in a signal handler,
// we cannot receive another signal.
MOVL 4(SP), CX
MOVL CX, m_vdsoSP(SI)
MOVL 0(SP), CX
MOVL CX, m_vdsoPC(SI)
// sec is in AX, nsec in BX
// convert to DX:AX nsec
MOVL $1000000000, CX
MULL CX
ADDL BX, AX
ADCL $0, DX
MOVL AX, ret_lo+0(FP)
MOVL DX, ret_hi+4(FP)
RET
TEXT runtime·rtsigprocmask(SB),NOSPLIT,$0
MOVL $SYS_rt_sigprocmask, AX
MOVL how+0(FP), BX
MOVL new+4(FP), CX
MOVL old+8(FP), DX
MOVL size+12(FP), SI
INVOKE_SYSCALL
CMPL AX, $0xfffff001
JLS 2(PC)
INT $3
RET
TEXT runtime·rt_sigaction(SB),NOSPLIT,$0
MOVL $SYS_rt_sigaction, AX
MOVL sig+0(FP), BX
MOVL new+4(FP), CX
MOVL old+8(FP), DX
MOVL size+12(FP), SI
INVOKE_SYSCALL
MOVL AX, ret+16(FP)
RET
TEXT runtime·sigfwd(SB),NOSPLIT,$12-16
MOVL fn+0(FP), AX
MOVL sig+4(FP), BX
MOVL info+8(FP), CX
MOVL ctx+12(FP), DX
MOVL SP, SI
SUBL $32, SP
ANDL $-15, SP // align stack: handler might be a C function
MOVL BX, 0(SP)
MOVL CX, 4(SP)
MOVL DX, 8(SP)
MOVL SI, 12(SP) // save SI: handler might be a Go function
CALL AX
MOVL 12(SP), AX
MOVL AX, SP
RET
// Called using C ABI.
TEXT runtime·sigtramp(SB),NOSPLIT|TOPFRAME,$28
// Save callee-saved C registers, since the caller may be a C signal handler.
MOVL BX, bx-4(SP)
MOVL BP, bp-8(SP)
MOVL SI, si-12(SP)
MOVL DI, di-16(SP)
// We don't save mxcsr or the x87 control word because sigtrampgo doesn't
// modify them.
MOVL (28+4)(SP), BX
MOVL BX, 0(SP)
MOVL (28+8)(SP), BX
MOVL BX, 4(SP)
MOVL (28+12)(SP), BX
MOVL BX, 8(SP)
CALL runtime·sigtrampgo(SB)
MOVL di-16(SP), DI
MOVL si-12(SP), SI
MOVL bp-8(SP), BP
MOVL bx-4(SP), BX
RET
TEXT runtime·cgoSigtramp(SB),NOSPLIT,$0
JMP runtime·sigtramp(SB)
// For cgo unwinding to work, this function must look precisely like
// the one in glibc. The glibc source code is:
// https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/unix/sysv/linux/i386/libc_sigaction.c;h=0665b41bbcd0986f0b33bf19a7ecbcedf9961d0a#l59
// The code that cares about the precise instructions used is:
// https://gcc.gnu.org/git/?p=gcc.git;a=blob;f=libgcc/config/i386/linux-unwind.h;h=5486223d60272c73d5103b29ae592d2ee998e1cf#l136
//
// For gdb unwinding to work, this function must look precisely like the one in
// glibc and must be named "__restore_rt" or contain the string "sigaction" in
// the name. The gdb source code is:
// https://sourceware.org/git/?p=binutils-gdb.git;a=blob;f=gdb/i386-linux-tdep.c;h=a6adeca1b97416f7194341151a8ce30723a786a3#l168
TEXT runtime·sigreturn__sigaction(SB),NOSPLIT,$0
MOVL $SYS_rt_sigreturn, AX
// Sigreturn expects same SP as signal handler,
// so cannot CALL 0x10(GS) here.
INT $0x80
INT $3 // not reached
RET
TEXT runtime·mmap(SB),NOSPLIT,$0
MOVL $SYS_mmap2, AX
MOVL addr+0(FP), BX
MOVL n+4(FP), CX
MOVL prot+8(FP), DX
MOVL flags+12(FP), SI
MOVL fd+16(FP), DI
MOVL off+20(FP), BP
SHRL $12, BP
INVOKE_SYSCALL
CMPL AX, $0xfffff001
JLS ok
NOTL AX
INCL AX
MOVL $0, p+24(FP)
MOVL AX, err+28(FP)
RET
ok:
MOVL AX, p+24(FP)
MOVL $0, err+28(FP)
RET
TEXT runtime·munmap(SB),NOSPLIT,$0
MOVL $SYS_munmap, AX
MOVL addr+0(FP), BX
MOVL n+4(FP), CX
INVOKE_SYSCALL
CMPL AX, $0xfffff001
JLS 2(PC)
INT $3
RET
TEXT runtime·madvise(SB),NOSPLIT,$0
MOVL $SYS_madvise, AX
MOVL addr+0(FP), BX
MOVL n+4(FP), CX
MOVL flags+8(FP), DX
INVOKE_SYSCALL
MOVL AX, ret+12(FP)
RET
// int32 futex(int32 *uaddr, int32 op, int32 val,
// struct timespec *timeout, int32 *uaddr2, int32 val2);
TEXT runtime·futex(SB),NOSPLIT,$0
MOVL $SYS_futex, AX
MOVL addr+0(FP), BX
MOVL op+4(FP), CX
MOVL val+8(FP), DX
MOVL ts+12(FP), SI
MOVL addr2+16(FP), DI
MOVL val3+20(FP), BP
INVOKE_SYSCALL
MOVL AX, ret+24(FP)
RET
// int32 clone(int32 flags, void *stack, M *mp, G *gp, void (*fn)(void));
TEXT runtime·clone(SB),NOSPLIT,$0
MOVL $SYS_clone, AX
MOVL flags+0(FP), BX
MOVL stk+4(FP), CX
MOVL $0, DX // parent tid ptr
MOVL $0, DI // child tid ptr
// Copy mp, gp, fn off parent stack for use by child.
SUBL $16, CX
MOVL mp+8(FP), SI
MOVL SI, 0(CX)
MOVL gp+12(FP), SI
MOVL SI, 4(CX)
MOVL fn+16(FP), SI
MOVL SI, 8(CX)
MOVL $1234, 12(CX)
// cannot use CALL 0x10(GS) here, because the stack changes during the
// system call (after CALL 0x10(GS), the child is still using the
// parent's stack when executing its RET instruction).
INT $0x80
// In parent, return.
CMPL AX, $0
JEQ 3(PC)
MOVL AX, ret+20(FP)
RET
// Paranoia: check that SP is as we expect.
NOP SP // tell vet SP changed - stop checking offsets
MOVL 12(SP), BP
CMPL BP, $1234
JEQ 2(PC)
INT $3
// Initialize AX to Linux tid
MOVL $SYS_gettid, AX
INVOKE_SYSCALL
MOVL 0(SP), BX // m
MOVL 4(SP), DX // g
MOVL 8(SP), SI // fn
CMPL BX, $0
JEQ nog
CMPL DX, $0
JEQ nog
MOVL AX, m_procid(BX) // save tid as m->procid
// set up ldt 7+id to point at m->tls.
LEAL m_tls(BX), BP
MOVL m_id(BX), DI
ADDL $7, DI // m0 is LDT#7. count up.
// setldt(tls#, &tls, sizeof tls)
PUSHAL // save registers
PUSHL $32 // sizeof tls
PUSHL BP // &tls
PUSHL DI // tls #
CALL runtime·setldt(SB)
POPL AX
POPL AX
POPL AX
POPAL
// Now segment is established. Initialize m, g.
get_tls(AX)
MOVL DX, g(AX)
MOVL BX, g_m(DX)
CALL runtime·stackcheck(SB) // smashes AX, CX
MOVL 0(DX), DX // paranoia; check they are not nil
MOVL 0(BX), BX
// more paranoia; check that stack splitting code works
PUSHAL
CALL runtime·emptyfunc(SB)
POPAL
nog:
CALL SI // fn()
CALL exit1<>(SB)
MOVL $0x1234, 0x1005
TEXT runtime·sigaltstack(SB),NOSPLIT,$-8
MOVL $SYS_sigaltstack, AX
MOVL new+0(FP), BX
MOVL old+4(FP), CX
INVOKE_SYSCALL
CMPL AX, $0xfffff001
JLS 2(PC)
INT $3
RET
// <asm-i386/ldt.h>
// struct user_desc {
// unsigned int entry_number;
// unsigned long base_addr;
// unsigned int limit;
// unsigned int seg_32bit:1;
// unsigned int contents:2;
// unsigned int read_exec_only:1;
// unsigned int limit_in_pages:1;
// unsigned int seg_not_present:1;
// unsigned int useable:1;
// };
#define SEG_32BIT 0x01
// contents are the 2 bits 0x02 and 0x04.
#define CONTENTS_DATA 0x00
#define CONTENTS_STACK 0x02
#define CONTENTS_CODE 0x04
#define READ_EXEC_ONLY 0x08
#define LIMIT_IN_PAGES 0x10
#define SEG_NOT_PRESENT 0x20
#define USEABLE 0x40
// `-1` means the kernel will pick a TLS entry on the first setldt call,
// which happens during runtime init, and that we'll store back the saved
// entry and reuse that on subsequent calls when creating new threads.
DATA runtime·tls_entry_number+0(SB)/4, $-1
GLOBL runtime·tls_entry_number(SB), NOPTR, $4
// setldt(int entry, int address, int limit)
// We use set_thread_area, which mucks with the GDT, instead of modify_ldt,
// which would modify the LDT, but is disabled on some kernels.
// The name, setldt, is a misnomer, although we leave this name as it is for
// the compatibility with other platforms.
TEXT runtime·setldt(SB),NOSPLIT,$32
MOVL base+4(FP), DX
#ifdef GOOS_android
// Android stores the TLS offset in runtime·tls_g.
SUBL runtime·tls_g(SB), DX
MOVL DX, 0(DX)
#else
/*
* When linking against the system libraries,
* we use its pthread_create and let it set up %gs
* for us. When we do that, the private storage
* we get is not at 0(GS), but -4(GS).
* To insulate the rest of the tool chain from this
* ugliness, 8l rewrites 0(TLS) into -4(GS) for us.
* To accommodate that rewrite, we translate
* the address here and bump the limit to 0xffffffff (no limit)
* so that -4(GS) maps to 0(address).
* Also, the final 0(GS) (current 4(DX)) has to point
* to itself, to mimic ELF.
*/
ADDL $0x4, DX // address
MOVL DX, 0(DX)
#endif
// get entry number
MOVL runtime·tls_entry_number(SB), CX
// set up user_desc
LEAL 16(SP), AX // struct user_desc
MOVL CX, 0(AX) // unsigned int entry_number
MOVL DX, 4(AX) // unsigned long base_addr
MOVL $0xfffff, 8(AX) // unsigned int limit
MOVL $(SEG_32BIT|LIMIT_IN_PAGES|USEABLE|CONTENTS_DATA), 12(AX) // flag bits
// call set_thread_area
MOVL AX, BX // user_desc
MOVL $SYS_set_thread_area, AX
// We can't call this via 0x10(GS) because this is called from setldt0 to set that up.
INT $0x80
// breakpoint on error
CMPL AX, $0xfffff001
JLS 2(PC)
INT $3
// read allocated entry number back out of user_desc
LEAL 16(SP), AX // get our user_desc back
MOVL 0(AX), AX
// store entry number if the kernel allocated it
CMPL CX, $-1
JNE 2(PC)
MOVL AX, runtime·tls_entry_number(SB)
// compute segment selector - (entry*8+3)
SHLL $3, AX
ADDL $3, AX
MOVW AX, GS
RET
TEXT runtime·osyield(SB),NOSPLIT,$0
MOVL $SYS_sched_yield, AX
INVOKE_SYSCALL
RET
TEXT runtime·sched_getaffinity(SB),NOSPLIT,$0
MOVL $SYS_sched_getaffinity, AX
MOVL pid+0(FP), BX
MOVL len+4(FP), CX
MOVL buf+8(FP), DX
INVOKE_SYSCALL
MOVL AX, ret+12(FP)
RET
// int access(const char *name, int mode)
TEXT runtime·access(SB),NOSPLIT,$0
MOVL $SYS_access, AX
MOVL name+0(FP), BX
MOVL mode+4(FP), CX
INVOKE_SYSCALL
MOVL AX, ret+8(FP)
RET
// int connect(int fd, const struct sockaddr *addr, socklen_t addrlen)
TEXT runtime·connect(SB),NOSPLIT,$0-16
// connect is implemented as socketcall(NR_socket, 3, *(rest of args))
// stack already should have fd, addr, addrlen.
MOVL $SYS_socketcall, AX
MOVL $3, BX // connect
LEAL fd+0(FP), CX
INVOKE_SYSCALL
MOVL AX, ret+12(FP)
RET
// int socket(int domain, int type, int protocol)
TEXT runtime·socket(SB),NOSPLIT,$0-16
// socket is implemented as socketcall(NR_socket, 1, *(rest of args))
// stack already should have domain, type, protocol.
MOVL $SYS_socketcall, AX
MOVL $1, BX // socket
LEAL domain+0(FP), CX
INVOKE_SYSCALL
MOVL AX, ret+12(FP)
RET
// func sbrk0() uintptr
TEXT runtime·sbrk0(SB),NOSPLIT,$0-4
// Implemented as brk(NULL).
MOVL $SYS_brk, AX
MOVL $0, BX // NULL
INVOKE_SYSCALL
MOVL AX, ret+0(FP)
RET
|