]> git.saurik.com Git - apple/xnu.git/blob - osfmk/i386/bsd_i386.c
xnu-3789.1.32.tar.gz
[apple/xnu.git] / osfmk / i386 / bsd_i386.c
1 /*
2 * Copyright (c) 2000-2016 Apple Inc. All rights reserved.
3 *
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
5 *
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
14 *
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
17 *
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
25 *
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
27 */
28 #ifdef MACH_BSD
29 #include <mach_rt.h>
30 #include <mach_debug.h>
31 #include <mach_ldebug.h>
32
33 #include <mach/kern_return.h>
34 #include <mach/mach_traps.h>
35 #include <mach/thread_status.h>
36 #include <mach/vm_param.h>
37
38 #include <kern/counters.h>
39 #include <kern/cpu_data.h>
40 #include <kern/mach_param.h>
41 #include <kern/task.h>
42 #include <kern/thread.h>
43 #include <kern/sched_prim.h>
44 #include <kern/misc_protos.h>
45 #include <kern/assert.h>
46 #include <kern/debug.h>
47 #include <kern/spl.h>
48 #include <kern/syscall_sw.h>
49 #include <ipc/ipc_port.h>
50 #include <vm/vm_kern.h>
51 #include <vm/pmap.h>
52
53 #include <i386/cpu_number.h>
54 #include <i386/eflags.h>
55 #include <i386/proc_reg.h>
56 #include <i386/tss.h>
57 #include <i386/user_ldt.h>
58 #include <i386/fpu.h>
59 #include <i386/machdep_call.h>
60 #include <i386/vmparam.h>
61 #include <i386/mp_desc.h>
62 #include <i386/misc_protos.h>
63 #include <i386/thread.h>
64 #include <i386/trap.h>
65 #include <i386/seg.h>
66 #include <mach/i386/syscall_sw.h>
67 #include <sys/syscall.h>
68 #include <sys/kdebug.h>
69 #include <sys/errno.h>
70 #include <../bsd/sys/sysent.h>
71
72 #ifdef MACH_BSD
73 extern void mach_kauth_cred_uthread_update(void);
74 extern void throttle_lowpri_io(int);
75 #endif
76
77 void * find_user_regs(thread_t);
78
79 unsigned int get_msr_exportmask(void);
80
81 unsigned int get_msr_nbits(void);
82
83 unsigned int get_msr_rbits(void);
84
85 /*
86 * thread_userstack:
87 *
88 * Return the user stack pointer from the machine
89 * dependent thread state info.
90 */
91 kern_return_t
92 thread_userstack(
93 __unused thread_t thread,
94 int flavor,
95 thread_state_t tstate,
96 __unused unsigned int count,
97 mach_vm_offset_t *user_stack,
98 int *customstack,
99 __unused boolean_t is64bit
100 )
101 {
102 if (customstack)
103 *customstack = 0;
104
105 switch (flavor) {
106 case x86_THREAD_STATE32:
107 {
108 x86_thread_state32_t *state25;
109
110 state25 = (x86_thread_state32_t *) tstate;
111
112 if (state25->esp) {
113 *user_stack = state25->esp;
114 if (customstack)
115 *customstack = 1;
116 } else {
117 *user_stack = VM_USRSTACK32;
118 if (customstack)
119 *customstack = 0;
120 }
121 break;
122 }
123
124 case x86_THREAD_STATE64:
125 {
126 x86_thread_state64_t *state25;
127
128 state25 = (x86_thread_state64_t *) tstate;
129
130 if (state25->rsp) {
131 *user_stack = state25->rsp;
132 if (customstack)
133 *customstack = 1;
134 } else {
135 *user_stack = VM_USRSTACK64;
136 if (customstack)
137 *customstack = 0;
138 }
139 break;
140 }
141
142 default:
143 return (KERN_INVALID_ARGUMENT);
144 }
145
146 return (KERN_SUCCESS);
147 }
148
149 /*
150 * thread_userstackdefault:
151 *
152 * Return the default stack location for the
153 * thread, if otherwise unknown.
154 */
155 kern_return_t
156 thread_userstackdefault(
157 mach_vm_offset_t *default_user_stack,
158 boolean_t is64bit)
159 {
160 if (is64bit) {
161 *default_user_stack = VM_USRSTACK64;
162 } else {
163 *default_user_stack = VM_USRSTACK32;
164 }
165 return (KERN_SUCCESS);
166 }
167
168 kern_return_t
169 thread_entrypoint(
170 __unused thread_t thread,
171 int flavor,
172 thread_state_t tstate,
173 __unused unsigned int count,
174 mach_vm_offset_t *entry_point
175 )
176 {
177 /*
178 * Set a default.
179 */
180 if (*entry_point == 0)
181 *entry_point = VM_MIN_ADDRESS;
182
183 switch (flavor) {
184 case x86_THREAD_STATE32:
185 {
186 x86_thread_state32_t *state25;
187
188 state25 = (i386_thread_state_t *) tstate;
189 *entry_point = state25->eip ? state25->eip: VM_MIN_ADDRESS;
190 break;
191 }
192
193 case x86_THREAD_STATE64:
194 {
195 x86_thread_state64_t *state25;
196
197 state25 = (x86_thread_state64_t *) tstate;
198 *entry_point = state25->rip ? state25->rip: VM_MIN_ADDRESS64;
199 break;
200 }
201 }
202 return (KERN_SUCCESS);
203 }
204
205 /*
206 * FIXME - thread_set_child
207 */
208
209 void thread_set_child(thread_t child, int pid);
210 void
211 thread_set_child(thread_t child, int pid)
212 {
213 pal_register_cache_state(child, DIRTY);
214
215 if (thread_is_64bit(child)) {
216 x86_saved_state64_t *iss64;
217
218 iss64 = USER_REGS64(child);
219
220 iss64->rax = pid;
221 iss64->rdx = 1;
222 iss64->isf.rflags &= ~EFL_CF;
223 } else {
224 x86_saved_state32_t *iss32;
225
226 iss32 = USER_REGS32(child);
227
228 iss32->eax = pid;
229 iss32->edx = 1;
230 iss32->efl &= ~EFL_CF;
231 }
232 }
233
234
235
236 /*
237 * System Call handling code
238 */
239
240 extern long fuword(vm_offset_t);
241
242 __attribute__((noreturn))
243 void
244 machdep_syscall(x86_saved_state_t *state)
245 {
246 int args[machdep_call_count];
247 int trapno;
248 int nargs;
249 const machdep_call_t *entry;
250 x86_saved_state32_t *regs;
251
252 assert(is_saved_state32(state));
253 regs = saved_state32(state);
254
255 trapno = regs->eax;
256 #if DEBUG_TRACE
257 kprintf("machdep_syscall(0x%08x) code=%d\n", regs, trapno);
258 #endif
259
260 DEBUG_KPRINT_SYSCALL_MDEP(
261 "machdep_syscall: trapno=%d\n", trapno);
262
263 if (trapno < 0 || trapno >= machdep_call_count) {
264 regs->eax = (unsigned int)kern_invalid(NULL);
265
266 thread_exception_return();
267 /* NOTREACHED */
268 }
269 entry = &machdep_call_table[trapno];
270 nargs = entry->nargs;
271
272 if (nargs != 0) {
273 if (copyin((user_addr_t) regs->uesp + sizeof (int),
274 (char *) args, (nargs * sizeof (int)))) {
275 regs->eax = KERN_INVALID_ADDRESS;
276
277 thread_exception_return();
278 /* NOTREACHED */
279 }
280 }
281 switch (nargs) {
282 case 0:
283 regs->eax = (*entry->routine.args_0)();
284 break;
285 case 1:
286 regs->eax = (*entry->routine.args_1)(args[0]);
287 break;
288 case 2:
289 regs->eax = (*entry->routine.args_2)(args[0],args[1]);
290 break;
291 case 3:
292 if (!entry->bsd_style)
293 regs->eax = (*entry->routine.args_3)(args[0],args[1],args[2]);
294 else {
295 int error;
296 uint32_t rval;
297
298 error = (*entry->routine.args_bsd_3)(&rval, args[0], args[1], args[2]);
299 if (error) {
300 regs->eax = error;
301 regs->efl |= EFL_CF; /* carry bit */
302 } else {
303 regs->eax = rval;
304 regs->efl &= ~EFL_CF;
305 }
306 }
307 break;
308 case 4:
309 regs->eax = (*entry->routine.args_4)(args[0], args[1], args[2], args[3]);
310 break;
311
312 default:
313 panic("machdep_syscall: too many args");
314 }
315
316 DEBUG_KPRINT_SYSCALL_MDEP("machdep_syscall: retval=%u\n", regs->eax);
317
318 throttle_lowpri_io(1);
319
320 thread_exception_return();
321 /* NOTREACHED */
322 }
323
324 __attribute__((noreturn))
325 void
326 machdep_syscall64(x86_saved_state_t *state)
327 {
328 int trapno;
329 const machdep_call_t *entry;
330 x86_saved_state64_t *regs;
331
332 assert(is_saved_state64(state));
333 regs = saved_state64(state);
334
335 trapno = (int)(regs->rax & SYSCALL_NUMBER_MASK);
336
337 DEBUG_KPRINT_SYSCALL_MDEP(
338 "machdep_syscall64: trapno=%d\n", trapno);
339
340 if (trapno < 0 || trapno >= machdep_call_count) {
341 regs->rax = (unsigned int)kern_invalid(NULL);
342
343 thread_exception_return();
344 /* NOTREACHED */
345 }
346 entry = &machdep_call_table64[trapno];
347
348 switch (entry->nargs) {
349 case 0:
350 regs->rax = (*entry->routine.args_0)();
351 break;
352 case 1:
353 regs->rax = (*entry->routine.args64_1)(regs->rdi);
354 break;
355 case 2:
356 regs->rax = (*entry->routine.args64_2)(regs->rdi, regs->rsi);
357 break;
358 default:
359 panic("machdep_syscall64: too many args");
360 }
361
362 DEBUG_KPRINT_SYSCALL_MDEP("machdep_syscall: retval=%llu\n", regs->rax);
363
364 throttle_lowpri_io(1);
365
366 thread_exception_return();
367 /* NOTREACHED */
368 }
369
370 #endif /* MACH_BSD */
371
372
373 typedef kern_return_t (*mach_call_t)(void *);
374
375 struct mach_call_args {
376 syscall_arg_t arg1;
377 syscall_arg_t arg2;
378 syscall_arg_t arg3;
379 syscall_arg_t arg4;
380 syscall_arg_t arg5;
381 syscall_arg_t arg6;
382 syscall_arg_t arg7;
383 syscall_arg_t arg8;
384 syscall_arg_t arg9;
385 };
386
387 static kern_return_t
388 mach_call_arg_munger32(uint32_t sp, struct mach_call_args *args, const mach_trap_t *trapp);
389
390
391 static kern_return_t
392 mach_call_arg_munger32(uint32_t sp, struct mach_call_args *args, const mach_trap_t *trapp)
393 {
394 if (copyin((user_addr_t)(sp + sizeof(int)), (char *)args, trapp->mach_trap_u32_words * sizeof (int)))
395 return KERN_INVALID_ARGUMENT;
396 #if CONFIG_REQUIRES_U32_MUNGING
397 trapp->mach_trap_arg_munge32(args);
398 #else
399 #error U32 mach traps on x86_64 kernel requires munging
400 #endif
401 return KERN_SUCCESS;
402 }
403
404
405 __private_extern__ void mach_call_munger(x86_saved_state_t *state);
406
407 extern const char *mach_syscall_name_table[];
408
409 __attribute__((noreturn))
410 void
411 mach_call_munger(x86_saved_state_t *state)
412 {
413 int argc;
414 int call_number;
415 mach_call_t mach_call;
416 kern_return_t retval;
417 struct mach_call_args args = { 0, 0, 0, 0, 0, 0, 0, 0, 0 };
418 x86_saved_state32_t *regs;
419
420 #if PROC_REF_DEBUG
421 struct uthread *ut = get_bsdthread_info(current_thread());
422
423 uthread_reset_proc_refcount(ut);
424 #endif
425
426 assert(is_saved_state32(state));
427 regs = saved_state32(state);
428
429 call_number = -(regs->eax);
430
431 DEBUG_KPRINT_SYSCALL_MACH(
432 "mach_call_munger: code=%d(%s)\n",
433 call_number, mach_syscall_name_table[call_number]);
434 #if DEBUG_TRACE
435 kprintf("mach_call_munger(0x%08x) code=%d\n", regs, call_number);
436 #endif
437
438 if (call_number < 0 || call_number >= mach_trap_count) {
439 i386_exception(EXC_SYSCALL, call_number, 1);
440 /* NOTREACHED */
441 }
442 mach_call = (mach_call_t)mach_trap_table[call_number].mach_trap_function;
443
444 if (mach_call == (mach_call_t)kern_invalid) {
445 DEBUG_KPRINT_SYSCALL_MACH(
446 "mach_call_munger: kern_invalid 0x%x\n", regs->eax);
447 i386_exception(EXC_SYSCALL, call_number, 1);
448 /* NOTREACHED */
449 }
450
451 argc = mach_trap_table[call_number].mach_trap_arg_count;
452 if (argc) {
453 retval = mach_call_arg_munger32(regs->uesp, &args, &mach_trap_table[call_number]);
454 if (retval != KERN_SUCCESS) {
455 regs->eax = retval;
456
457 DEBUG_KPRINT_SYSCALL_MACH(
458 "mach_call_munger: retval=0x%x\n", retval);
459
460 thread_exception_return();
461 /* NOTREACHED */
462 }
463 }
464
465 #ifdef MACH_BSD
466 mach_kauth_cred_uthread_update();
467 #endif
468
469 KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
470 MACHDBG_CODE(DBG_MACH_EXCP_SC, (call_number)) | DBG_FUNC_START,
471 args.arg1, args.arg2, args.arg3, args.arg4, 0);
472
473 retval = mach_call(&args);
474
475 DEBUG_KPRINT_SYSCALL_MACH("mach_call_munger: retval=0x%x\n", retval);
476
477 KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
478 MACHDBG_CODE(DBG_MACH_EXCP_SC,(call_number)) | DBG_FUNC_END,
479 retval, 0, 0, 0, 0);
480
481 regs->eax = retval;
482
483 throttle_lowpri_io(1);
484
485 #if PROC_REF_DEBUG
486 if (__improbable(uthread_get_proc_refcount(ut) != 0)) {
487 panic("system call returned with uu_proc_refcount != 0");
488 }
489 #endif
490
491 thread_exception_return();
492 /* NOTREACHED */
493 }
494
495
496 __private_extern__ void mach_call_munger64(x86_saved_state_t *regs);
497
498 __attribute__((noreturn))
499 void
500 mach_call_munger64(x86_saved_state_t *state)
501 {
502 int call_number;
503 int argc;
504 mach_call_t mach_call;
505 struct mach_call_args args = { 0, 0, 0, 0, 0, 0, 0, 0, 0 };
506 x86_saved_state64_t *regs;
507
508 #if PROC_REF_DEBUG
509 struct uthread *ut = get_bsdthread_info(current_thread());
510
511 uthread_reset_proc_refcount(ut);
512 #endif
513
514 assert(is_saved_state64(state));
515 regs = saved_state64(state);
516
517 call_number = (int)(regs->rax & SYSCALL_NUMBER_MASK);
518
519 DEBUG_KPRINT_SYSCALL_MACH(
520 "mach_call_munger64: code=%d(%s)\n",
521 call_number, mach_syscall_name_table[call_number]);
522
523 KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
524 MACHDBG_CODE(DBG_MACH_EXCP_SC,(call_number)) | DBG_FUNC_START,
525 regs->rdi, regs->rsi, regs->rdx, regs->r10, 0);
526
527 if (call_number < 0 || call_number >= mach_trap_count) {
528 i386_exception(EXC_SYSCALL, regs->rax, 1);
529 /* NOTREACHED */
530 }
531 mach_call = (mach_call_t)mach_trap_table[call_number].mach_trap_function;
532
533 if (mach_call == (mach_call_t)kern_invalid) {
534 i386_exception(EXC_SYSCALL, regs->rax, 1);
535 /* NOTREACHED */
536 }
537 argc = mach_trap_table[call_number].mach_trap_arg_count;
538 if (argc) {
539 int args_in_regs = MIN(6, argc);
540
541 memcpy(&args.arg1, &regs->rdi, args_in_regs * sizeof(syscall_arg_t));
542
543 if (argc > 6) {
544 int copyin_count;
545
546 assert(argc <= 9);
547 copyin_count = (argc - 6) * (int)sizeof(syscall_arg_t);
548
549 if (copyin((user_addr_t)(regs->isf.rsp + sizeof(user_addr_t)), (char *)&args.arg7, copyin_count)) {
550 regs->rax = KERN_INVALID_ARGUMENT;
551
552 thread_exception_return();
553 /* NOTREACHED */
554 }
555 }
556 }
557
558 #ifdef MACH_BSD
559 mach_kauth_cred_uthread_update();
560 #endif
561
562 regs->rax = (uint64_t)mach_call((void *)&args);
563
564 DEBUG_KPRINT_SYSCALL_MACH( "mach_call_munger64: retval=0x%llx\n", regs->rax);
565
566 KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
567 MACHDBG_CODE(DBG_MACH_EXCP_SC,(call_number)) | DBG_FUNC_END,
568 regs->rax, 0, 0, 0, 0);
569
570 throttle_lowpri_io(1);
571
572 #if PROC_REF_DEBUG
573 if (__improbable(uthread_get_proc_refcount(ut) != 0)) {
574 panic("system call returned with uu_proc_refcount != 0");
575 }
576 #endif
577
578 thread_exception_return();
579 /* NOTREACHED */
580 }
581
582
583 /*
584 * thread_setuserstack:
585 *
586 * Sets the user stack pointer into the machine
587 * dependent thread state info.
588 */
589 void
590 thread_setuserstack(
591 thread_t thread,
592 mach_vm_address_t user_stack)
593 {
594 pal_register_cache_state(thread, DIRTY);
595 if (thread_is_64bit(thread)) {
596 x86_saved_state64_t *iss64;
597
598 iss64 = USER_REGS64(thread);
599
600 iss64->isf.rsp = (uint64_t)user_stack;
601 } else {
602 x86_saved_state32_t *iss32;
603
604 iss32 = USER_REGS32(thread);
605
606 iss32->uesp = CAST_DOWN_EXPLICIT(unsigned int, user_stack);
607 }
608 }
609
610 /*
611 * thread_adjuserstack:
612 *
613 * Returns the adjusted user stack pointer from the machine
614 * dependent thread state info. Used for small (<2G) deltas.
615 */
616 uint64_t
617 thread_adjuserstack(
618 thread_t thread,
619 int adjust)
620 {
621 pal_register_cache_state(thread, DIRTY);
622 if (thread_is_64bit(thread)) {
623 x86_saved_state64_t *iss64;
624
625 iss64 = USER_REGS64(thread);
626
627 iss64->isf.rsp += adjust;
628
629 return iss64->isf.rsp;
630 } else {
631 x86_saved_state32_t *iss32;
632
633 iss32 = USER_REGS32(thread);
634
635 iss32->uesp += adjust;
636
637 return CAST_USER_ADDR_T(iss32->uesp);
638 }
639 }
640
641 /*
642 * thread_setentrypoint:
643 *
644 * Sets the user PC into the machine
645 * dependent thread state info.
646 */
647 void
648 thread_setentrypoint(thread_t thread, mach_vm_address_t entry)
649 {
650 pal_register_cache_state(thread, DIRTY);
651 if (thread_is_64bit(thread)) {
652 x86_saved_state64_t *iss64;
653
654 iss64 = USER_REGS64(thread);
655
656 iss64->isf.rip = (uint64_t)entry;
657 } else {
658 x86_saved_state32_t *iss32;
659
660 iss32 = USER_REGS32(thread);
661
662 iss32->eip = CAST_DOWN_EXPLICIT(unsigned int, entry);
663 }
664 }
665
666
667 kern_return_t
668 thread_setsinglestep(thread_t thread, int on)
669 {
670 pal_register_cache_state(thread, DIRTY);
671 if (thread_is_64bit(thread)) {
672 x86_saved_state64_t *iss64;
673
674 iss64 = USER_REGS64(thread);
675
676 if (on)
677 iss64->isf.rflags |= EFL_TF;
678 else
679 iss64->isf.rflags &= ~EFL_TF;
680 } else {
681 x86_saved_state32_t *iss32;
682
683 iss32 = USER_REGS32(thread);
684
685 if (on) {
686 iss32->efl |= EFL_TF;
687 /* Ensure IRET */
688 if (iss32->cs == SYSENTER_CS)
689 iss32->cs = SYSENTER_TF_CS;
690 }
691 else
692 iss32->efl &= ~EFL_TF;
693 }
694
695 return (KERN_SUCCESS);
696 }
697
698 void *
699 get_user_regs(thread_t th)
700 {
701 pal_register_cache_state(th, DIRTY);
702 return(USER_STATE(th));
703 }
704
705 void *
706 find_user_regs(thread_t thread)
707 {
708 return get_user_regs(thread);
709 }
710
711 #if CONFIG_DTRACE
712 /*
713 * DTrace would like to have a peek at the kernel interrupt state, if available.
714 * Based on osfmk/chud/i386/chud_thread_i386.c:chudxnu_thread_get_state(), which see.
715 */
716 x86_saved_state_t *find_kern_regs(thread_t);
717
718 x86_saved_state_t *
719 find_kern_regs(thread_t thread)
720 {
721 if (thread == current_thread() &&
722 NULL != current_cpu_datap()->cpu_int_state &&
723 !(USER_STATE(thread) == current_cpu_datap()->cpu_int_state &&
724 current_cpu_datap()->cpu_interrupt_level == 1)) {
725
726 return current_cpu_datap()->cpu_int_state;
727 } else {
728 return NULL;
729 }
730 }
731
732 vm_offset_t dtrace_get_cpu_int_stack_top(void);
733
734 vm_offset_t
735 dtrace_get_cpu_int_stack_top(void)
736 {
737 return current_cpu_datap()->cpu_int_stack_top;
738 }
739 #endif