FreeBSD/Linux Kernel Cross Reference
sys/amd64/amd64/trap.c
1 /*-
2 * Copyright (C) 1994, David Greenman
3 * Copyright (c) 1990, 1993
4 * The Regents of the University of California. All rights reserved.
5 *
6 * This code is derived from software contributed to Berkeley by
7 * the University of Utah, and William Jolitz.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
11 * are met:
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. All advertising materials mentioning features or use of this software
18 * must display the following acknowledgement:
19 * This product includes software developed by the University of
20 * California, Berkeley and its contributors.
21 * 4. Neither the name of the University nor the names of its contributors
22 * may be used to endorse or promote products derived from this software
23 * without specific prior written permission.
24 *
25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * SUCH DAMAGE.
36 *
37 * from: @(#)trap.c 7.4 (Berkeley) 5/13/91
38 */
39
40 #include <sys/cdefs.h>
41 __FBSDID("$FreeBSD: releng/5.3/sys/amd64/amd64/trap.c 134685 2004-09-03 06:40:26Z julian $");
42
43 /*
44 * AMD64 Trap and System call handling
45 */
46
47 #include "opt_clock.h"
48 #include "opt_cpu.h"
49 #include "opt_isa.h"
50 #include "opt_ktrace.h"
51
52 #include <sys/param.h>
53 #include <sys/bus.h>
54 #include <sys/systm.h>
55 #include <sys/proc.h>
56 #include <sys/pioctl.h>
57 #include <sys/ptrace.h>
58 #include <sys/kdb.h>
59 #include <sys/kernel.h>
60 #include <sys/ktr.h>
61 #include <sys/lock.h>
62 #include <sys/mutex.h>
63 #include <sys/resourcevar.h>
64 #include <sys/signalvar.h>
65 #include <sys/syscall.h>
66 #include <sys/sysctl.h>
67 #include <sys/sysent.h>
68 #include <sys/uio.h>
69 #include <sys/vmmeter.h>
70 #ifdef KTRACE
71 #include <sys/ktrace.h>
72 #endif
73
74 #include <vm/vm.h>
75 #include <vm/vm_param.h>
76 #include <vm/pmap.h>
77 #include <vm/vm_kern.h>
78 #include <vm/vm_map.h>
79 #include <vm/vm_page.h>
80 #include <vm/vm_extern.h>
81
82 #include <machine/cpu.h>
83 #include <machine/intr_machdep.h>
84 #include <machine/md_var.h>
85 #include <machine/pcb.h>
86 #ifdef SMP
87 #include <machine/smp.h>
88 #endif
89 #include <machine/tss.h>
90
91 extern void trap(struct trapframe frame);
92 extern void syscall(struct trapframe frame);
93
94 static int trap_pfault(struct trapframe *, int);
95 static void trap_fatal(struct trapframe *, vm_offset_t);
96 void dblfault_handler(void);
97
98 #define MAX_TRAP_MSG 28
99 static char *trap_msg[] = {
100 "", /* 0 unused */
101 "privileged instruction fault", /* 1 T_PRIVINFLT */
102 "", /* 2 unused */
103 "breakpoint instruction fault", /* 3 T_BPTFLT */
104 "", /* 4 unused */
105 "", /* 5 unused */
106 "arithmetic trap", /* 6 T_ARITHTRAP */
107 "", /* 7 unused */
108 "", /* 8 unused */
109 "general protection fault", /* 9 T_PROTFLT */
110 "trace trap", /* 10 T_TRCTRAP */
111 "", /* 11 unused */
112 "page fault", /* 12 T_PAGEFLT */
113 "", /* 13 unused */
114 "alignment fault", /* 14 T_ALIGNFLT */
115 "", /* 15 unused */
116 "", /* 16 unused */
117 "", /* 17 unused */
118 "integer divide fault", /* 18 T_DIVIDE */
119 "non-maskable interrupt trap", /* 19 T_NMI */
120 "overflow trap", /* 20 T_OFLOW */
121 "FPU bounds check fault", /* 21 T_BOUND */
122 "FPU device not available", /* 22 T_DNA */
123 "double fault", /* 23 T_DOUBLEFLT */
124 "FPU operand fetch fault", /* 24 T_FPOPFLT */
125 "invalid TSS fault", /* 25 T_TSSFLT */
126 "segment not present fault", /* 26 T_SEGNPFLT */
127 "stack fault", /* 27 T_STKFLT */
128 "machine check trap", /* 28 T_MCHK */
129 };
130
131 #ifdef KDB
132 static int kdb_on_nmi = 1;
133 SYSCTL_INT(_machdep, OID_AUTO, kdb_on_nmi, CTLFLAG_RW,
134 &kdb_on_nmi, 0, "Go to KDB on NMI");
135 #endif
136 static int panic_on_nmi = 1;
137 SYSCTL_INT(_machdep, OID_AUTO, panic_on_nmi, CTLFLAG_RW,
138 &panic_on_nmi, 0, "Panic on NMI");
139
140 #ifdef WITNESS
141 extern char *syscallnames[];
142 #endif
143
144 #ifdef DEVICE_POLLING
145 extern u_int32_t poll_in_trap;
146 extern int ether_poll(int count);
147 #endif /* DEVICE_POLLING */
148
149 /*
150 * Exception, fault, and trap interface to the FreeBSD kernel.
151 * This common code is called from assembly language IDT gate entry
152 * routines that prepare a suitable stack frame, and restore this
153 * frame after the exception has been processed.
154 */
155
156 void
157 trap(frame)
158 struct trapframe frame;
159 {
160 struct thread *td = curthread;
161 struct proc *p = td->td_proc;
162 u_int sticks = 0;
163 int i = 0, ucode = 0, type, code;
164
165 atomic_add_int(&cnt.v_trap, 1);
166 type = frame.tf_trapno;
167
168 #ifdef KDB
169 if (kdb_active) {
170 kdb_reenter();
171 goto out;
172 }
173 #endif
174
175 if ((frame.tf_rflags & PSL_I) == 0) {
176 /*
177 * Buggy application or kernel code has disabled
178 * interrupts and then trapped. Enabling interrupts
179 * now is wrong, but it is better than running with
180 * interrupts disabled until they are accidentally
181 * enabled later.
182 */
183 if (ISPL(frame.tf_cs) == SEL_UPL)
184 printf(
185 "pid %ld (%s): trap %d with interrupts disabled\n",
186 (long)curproc->p_pid, curproc->p_comm, type);
187 else if (type != T_BPTFLT && type != T_TRCTRAP) {
188 /*
189 * XXX not quite right, since this may be for a
190 * multiple fault in user mode.
191 */
192 printf("kernel trap %d with interrupts disabled\n",
193 type);
194 /*
195 * We shouldn't enable interrupts while in a critical
196 * section.
197 */
198 if (td->td_critnest == 0)
199 enable_intr();
200 }
201 }
202
203 code = frame.tf_err;
204 if (type == T_PAGEFLT) {
205 /*
206 * If we get a page fault while in a critical section, then
207 * it is most likely a fatal kernel page fault. The kernel
208 * is already going to panic trying to get a sleep lock to
209 * do the VM lookup, so just consider it a fatal trap so the
210 * kernel can print out a useful trap message and even get
211 * to the debugger.
212 */
213 if (td->td_critnest != 0)
214 trap_fatal(&frame, frame.tf_addr);
215 }
216
217 #ifdef DEVICE_POLLING
218 if (poll_in_trap)
219 ether_poll(poll_in_trap);
220 #endif /* DEVICE_POLLING */
221
222 if (ISPL(frame.tf_cs) == SEL_UPL) {
223 /* user trap */
224
225 sticks = td->td_sticks;
226 td->td_frame = &frame;
227 if (td->td_ucred != p->p_ucred)
228 cred_update_thread(td);
229
230 switch (type) {
231 case T_PRIVINFLT: /* privileged instruction fault */
232 ucode = type;
233 i = SIGILL;
234 break;
235
236 case T_BPTFLT: /* bpt instruction fault */
237 case T_TRCTRAP: /* trace trap */
238 enable_intr();
239 frame.tf_rflags &= ~PSL_T;
240 i = SIGTRAP;
241 break;
242
243 case T_ARITHTRAP: /* arithmetic trap */
244 ucode = fputrap();
245 if (ucode == -1)
246 goto userout;
247 i = SIGFPE;
248 break;
249
250 case T_PROTFLT: /* general protection fault */
251 case T_STKFLT: /* stack fault */
252 case T_SEGNPFLT: /* segment not present fault */
253 case T_TSSFLT: /* invalid TSS fault */
254 case T_DOUBLEFLT: /* double fault */
255 default:
256 ucode = code + BUS_SEGM_FAULT ;
257 i = SIGBUS;
258 break;
259
260 case T_PAGEFLT: /* page fault */
261 if (td->td_pflags & TDP_SA)
262 thread_user_enter(td);
263 i = trap_pfault(&frame, TRUE);
264 if (i == -1)
265 goto userout;
266 if (i == 0)
267 goto user;
268
269 ucode = T_PAGEFLT;
270 break;
271
272 case T_DIVIDE: /* integer divide fault */
273 ucode = FPE_INTDIV;
274 i = SIGFPE;
275 break;
276
277 #ifdef DEV_ISA
278 case T_NMI:
279 /* machine/parity/power fail/"kitchen sink" faults */
280 /* XXX Giant */
281 if (isa_nmi(code) == 0) {
282 #ifdef KDB
283 /*
284 * NMI can be hooked up to a pushbutton
285 * for debugging.
286 */
287 if (kdb_on_nmi) {
288 printf ("NMI ... going to debugger\n");
289 kdb_trap(type, 0, &frame);
290 }
291 #endif /* KDB */
292 goto userout;
293 } else if (panic_on_nmi)
294 panic("NMI indicates hardware failure");
295 break;
296 #endif /* DEV_ISA */
297
298 case T_OFLOW: /* integer overflow fault */
299 ucode = FPE_INTOVF;
300 i = SIGFPE;
301 break;
302
303 case T_BOUND: /* bounds check fault */
304 ucode = FPE_FLTSUB;
305 i = SIGFPE;
306 break;
307
308 case T_DNA:
309 /* transparent fault (due to context switch "late") */
310 if (fpudna())
311 goto userout;
312 i = SIGFPE;
313 ucode = FPE_FPU_NP_TRAP;
314 break;
315
316 case T_FPOPFLT: /* FPU operand fetch fault */
317 ucode = T_FPOPFLT;
318 i = SIGILL;
319 break;
320
321 case T_XMMFLT: /* SIMD floating-point exception */
322 ucode = 0; /* XXX */
323 i = SIGFPE;
324 break;
325 }
326 } else {
327 /* kernel trap */
328
329 KASSERT(cold || td->td_ucred != NULL,
330 ("kernel trap doesn't have ucred"));
331 switch (type) {
332 case T_PAGEFLT: /* page fault */
333 (void) trap_pfault(&frame, FALSE);
334 goto out;
335
336 case T_DNA:
337 /*
338 * The kernel is apparently using fpu for copying.
339 * XXX this should be fatal unless the kernel has
340 * registered such use.
341 */
342 if (fpudna()) {
343 printf("fpudna in kernel mode!\n");
344 goto out;
345 }
346 break;
347
348 case T_STKFLT: /* stack fault */
349 break;
350
351 case T_PROTFLT: /* general protection fault */
352 case T_SEGNPFLT: /* segment not present fault */
353 if (td->td_intr_nesting_level != 0)
354 break;
355
356 /*
357 * Invalid segment selectors and out of bounds
358 * %eip's and %esp's can be set up in user mode.
359 * This causes a fault in kernel mode when the
360 * kernel tries to return to user mode. We want
361 * to get this fault so that we can fix the
362 * problem here and not have to check all the
363 * selectors and pointers when the user changes
364 * them.
365 */
366 if (frame.tf_rip == (long)doreti_iret) {
367 frame.tf_rip = (long)doreti_iret_fault;
368 goto out;
369 }
370 if (PCPU_GET(curpcb)->pcb_onfault != NULL) {
371 frame.tf_rip =
372 (long)PCPU_GET(curpcb)->pcb_onfault;
373 goto out;
374 }
375 break;
376
377 case T_TSSFLT:
378 /*
379 * PSL_NT can be set in user mode and isn't cleared
380 * automatically when the kernel is entered. This
381 * causes a TSS fault when the kernel attempts to
382 * `iret' because the TSS link is uninitialized. We
383 * want to get this fault so that we can fix the
384 * problem here and not every time the kernel is
385 * entered.
386 */
387 if (frame.tf_rflags & PSL_NT) {
388 frame.tf_rflags &= ~PSL_NT;
389 goto out;
390 }
391 break;
392
393 case T_TRCTRAP: /* trace trap */
394 /*
395 * Ignore debug register trace traps due to
396 * accesses in the user's address space, which
397 * can happen under several conditions such as
398 * if a user sets a watchpoint on a buffer and
399 * then passes that buffer to a system call.
400 * We still want to get TRCTRAPS for addresses
401 * in kernel space because that is useful when
402 * debugging the kernel.
403 */
404 if (user_dbreg_trap()) {
405 /*
406 * Reset breakpoint bits because the
407 * processor doesn't
408 */
409 /* XXX check upper bits here */
410 load_dr6(rdr6() & 0xfffffff0);
411 goto out;
412 }
413 /*
414 * FALLTHROUGH (TRCTRAP kernel mode, kernel address)
415 */
416 case T_BPTFLT:
417 /*
418 * If KDB is enabled, let it handle the debugger trap.
419 * Otherwise, debugger traps "can't happen".
420 */
421 #ifdef KDB
422 /* XXX Giant */
423 if (kdb_trap(type, 0, &frame))
424 goto out;
425 #endif
426 break;
427
428 #ifdef DEV_ISA
429 case T_NMI:
430 /* XXX Giant */
431 /* machine/parity/power fail/"kitchen sink" faults */
432 if (isa_nmi(code) == 0) {
433 #ifdef KDB
434 /*
435 * NMI can be hooked up to a pushbutton
436 * for debugging.
437 */
438 if (kdb_on_nmi) {
439 printf ("NMI ... going to debugger\n");
440 kdb_trap(type, 0, &frame);
441 }
442 #endif /* KDB */
443 goto out;
444 } else if (panic_on_nmi == 0)
445 goto out;
446 /* FALLTHROUGH */
447 #endif /* DEV_ISA */
448 }
449
450 trap_fatal(&frame, 0);
451 goto out;
452 }
453
454 /* Translate fault for emulators (e.g. Linux) */
455 if (*p->p_sysent->sv_transtrap)
456 i = (*p->p_sysent->sv_transtrap)(i, type);
457
458 trapsignal(td, i, ucode);
459
460 #ifdef DEBUG
461 if (type <= MAX_TRAP_MSG) {
462 uprintf("fatal process exception: %s",
463 trap_msg[type]);
464 if ((type == T_PAGEFLT) || (type == T_PROTFLT))
465 uprintf(", fault VA = 0x%lx", frame.tf_addr);
466 uprintf("\n");
467 }
468 #endif
469
470 user:
471 userret(td, &frame, sticks);
472 mtx_assert(&Giant, MA_NOTOWNED);
473 userout:
474 out:
475 return;
476 }
477
478 static int
479 trap_pfault(frame, usermode)
480 struct trapframe *frame;
481 int usermode;
482 {
483 vm_offset_t va;
484 struct vmspace *vm = NULL;
485 vm_map_t map = 0;
486 int rv = 0;
487 vm_prot_t ftype;
488 struct thread *td = curthread;
489 struct proc *p = td->td_proc;
490 vm_offset_t eva = frame->tf_addr;
491
492 va = trunc_page(eva);
493 if (va >= KERNBASE) {
494 /*
495 * Don't allow user-mode faults in kernel address space.
496 */
497 if (usermode)
498 goto nogo;
499
500 map = kernel_map;
501 } else {
502 /*
503 * This is a fault on non-kernel virtual memory.
504 * vm is initialized above to NULL. If curproc is NULL
505 * or curproc->p_vmspace is NULL the fault is fatal.
506 */
507 if (p != NULL)
508 vm = p->p_vmspace;
509
510 if (vm == NULL)
511 goto nogo;
512
513 map = &vm->vm_map;
514 }
515
516 if (frame->tf_err & PGEX_W)
517 ftype = VM_PROT_WRITE;
518 else
519 ftype = VM_PROT_READ;
520
521 if (map != kernel_map) {
522 /*
523 * Keep swapout from messing with us during this
524 * critical time.
525 */
526 PROC_LOCK(p);
527 ++p->p_lock;
528 PROC_UNLOCK(p);
529
530 /* Fault in the user page: */
531 rv = vm_fault(map, va, ftype,
532 (ftype & VM_PROT_WRITE) ? VM_FAULT_DIRTY
533 : VM_FAULT_NORMAL);
534
535 PROC_LOCK(p);
536 --p->p_lock;
537 PROC_UNLOCK(p);
538 } else {
539 /*
540 * Don't have to worry about process locking or stacks in the
541 * kernel.
542 */
543 rv = vm_fault(map, va, ftype, VM_FAULT_NORMAL);
544 }
545 if (rv == KERN_SUCCESS)
546 return (0);
547 nogo:
548 if (!usermode) {
549 if (td->td_intr_nesting_level == 0 &&
550 PCPU_GET(curpcb)->pcb_onfault != NULL) {
551 frame->tf_rip = (long)PCPU_GET(curpcb)->pcb_onfault;
552 return (0);
553 }
554 trap_fatal(frame, eva);
555 return (-1);
556 }
557
558 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV);
559 }
560
561 static void
562 trap_fatal(frame, eva)
563 struct trapframe *frame;
564 vm_offset_t eva;
565 {
566 int code, type, ss;
567 long esp;
568 struct soft_segment_descriptor softseg;
569
570 code = frame->tf_err;
571 type = frame->tf_trapno;
572 sdtossd(&gdt[IDXSEL(frame->tf_cs & 0xffff)], &softseg);
573
574 if (type <= MAX_TRAP_MSG)
575 printf("\n\nFatal trap %d: %s while in %s mode\n",
576 type, trap_msg[type],
577 ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel");
578 #ifdef SMP
579 /* two separate prints in case of a trap on an unmapped page */
580 printf("cpuid = %d; ", PCPU_GET(cpuid));
581 printf("apic id = %02x\n", PCPU_GET(apic_id));
582 #endif
583 if (type == T_PAGEFLT) {
584 printf("fault virtual address = 0x%lx\n", eva);
585 printf("fault code = %s %s, %s\n",
586 code & PGEX_U ? "user" : "supervisor",
587 code & PGEX_W ? "write" : "read",
588 code & PGEX_P ? "protection violation" : "page not present");
589 }
590 printf("instruction pointer = 0x%lx:0x%lx\n",
591 frame->tf_cs & 0xffff, frame->tf_rip);
592 if (ISPL(frame->tf_cs) == SEL_UPL) {
593 ss = frame->tf_ss & 0xffff;
594 esp = frame->tf_rsp;
595 } else {
596 ss = GSEL(GDATA_SEL, SEL_KPL);
597 esp = (long)&frame->tf_rsp;
598 }
599 printf("stack pointer = 0x%x:0x%lx\n", ss, esp);
600 printf("frame pointer = 0x%x:0x%lx\n", ss, frame->tf_rbp);
601 printf("code segment = base 0x%lx, limit 0x%lx, type 0x%x\n",
602 softseg.ssd_base, softseg.ssd_limit, softseg.ssd_type);
603 printf(" = DPL %d, pres %d, long %d, def32 %d, gran %d\n",
604 softseg.ssd_dpl, softseg.ssd_p, softseg.ssd_long, softseg.ssd_def32,
605 softseg.ssd_gran);
606 printf("processor eflags = ");
607 if (frame->tf_rflags & PSL_T)
608 printf("trace trap, ");
609 if (frame->tf_rflags & PSL_I)
610 printf("interrupt enabled, ");
611 if (frame->tf_rflags & PSL_NT)
612 printf("nested task, ");
613 if (frame->tf_rflags & PSL_RF)
614 printf("resume, ");
615 printf("IOPL = %ld\n", (frame->tf_rflags & PSL_IOPL) >> 12);
616 printf("current process = ");
617 if (curproc) {
618 printf("%lu (%s)\n",
619 (u_long)curproc->p_pid, curproc->p_comm ?
620 curproc->p_comm : "");
621 } else {
622 printf("Idle\n");
623 }
624
625 #ifdef KDB
626 if (kdb_trap(type, 0, frame))
627 return;
628 #endif
629 printf("trap number = %d\n", type);
630 if (type <= MAX_TRAP_MSG)
631 panic("%s", trap_msg[type]);
632 else
633 panic("unknown/reserved trap");
634 }
635
636 /*
637 * Double fault handler. Called when a fault occurs while writing
638 * a frame for a trap/exception onto the stack. This usually occurs
639 * when the stack overflows (such is the case with infinite recursion,
640 * for example).
641 */
642 void
643 dblfault_handler()
644 {
645 printf("\nFatal double fault\n");
646 #ifdef SMP
647 /* two separate prints in case of a trap on an unmapped page */
648 printf("cpuid = %d; ", PCPU_GET(cpuid));
649 printf("apic id = %02x\n", PCPU_GET(apic_id));
650 #endif
651 panic("double fault");
652 }
653
654 /*
655 * syscall - system call request C handler
656 *
657 * A system call is essentially treated as a trap.
658 */
659 void
660 syscall(frame)
661 struct trapframe frame;
662 {
663 caddr_t params;
664 struct sysent *callp;
665 struct thread *td = curthread;
666 struct proc *p = td->td_proc;
667 register_t orig_tf_rflags;
668 u_int sticks;
669 int error;
670 int narg;
671 register_t args[8];
672 register_t *argp;
673 u_int code;
674 int reg, regcnt;
675
676 /*
677 * note: PCPU_LAZY_INC() can only be used if we can afford
678 * occassional inaccuracy in the count.
679 */
680 PCPU_LAZY_INC(cnt.v_syscall);
681
682 #ifdef DIAGNOSTIC
683 if (ISPL(frame.tf_cs) != SEL_UPL) {
684 mtx_lock(&Giant); /* try to stabilize the system XXX */
685 panic("syscall");
686 /* NOT REACHED */
687 mtx_unlock(&Giant);
688 }
689 #endif
690
691 reg = 0;
692 regcnt = 6;
693 sticks = td->td_sticks;
694 td->td_frame = &frame;
695 if (td->td_ucred != p->p_ucred)
696 cred_update_thread(td);
697 if (p->p_flag & P_SA)
698 thread_user_enter(td);
699 params = (caddr_t)frame.tf_rsp + sizeof(register_t);
700 code = frame.tf_rax;
701 orig_tf_rflags = frame.tf_rflags;
702
703 if (p->p_sysent->sv_prepsyscall) {
704 /*
705 * The prep code is MP aware.
706 */
707 (*p->p_sysent->sv_prepsyscall)(&frame, (int *)args, &code, ¶ms);
708 } else {
709 if (code == SYS_syscall || code == SYS___syscall) {
710 code = frame.tf_rdi;
711 reg++;
712 regcnt--;
713 }
714 }
715
716 if (p->p_sysent->sv_mask)
717 code &= p->p_sysent->sv_mask;
718
719 if (code >= p->p_sysent->sv_size)
720 callp = &p->p_sysent->sv_table[0];
721 else
722 callp = &p->p_sysent->sv_table[code];
723
724 narg = callp->sy_narg & SYF_ARGMASK;
725
726 /*
727 * copyin and the ktrsyscall()/ktrsysret() code is MP-aware
728 */
729 KASSERT(narg <= sizeof(args) / sizeof(args[0]),
730 ("Too many syscall arguments!"));
731 error = 0;
732 argp = &frame.tf_rdi;
733 argp += reg;
734 bcopy(argp, args, sizeof(args[0]) * regcnt);
735 if (narg > regcnt) {
736 KASSERT(params != NULL, ("copyin args with no params!"));
737 error = copyin(params, &args[regcnt],
738 (narg - regcnt) * sizeof(args[0]));
739 }
740 argp = &args[0];
741
742 #ifdef KTRACE
743 if (KTRPOINT(td, KTR_SYSCALL))
744 ktrsyscall(code, narg, argp);
745 #endif
746
747 CTR4(KTR_SYSC, "syscall enter thread %p pid %d proc %s code %d", td,
748 td->td_proc->p_pid, td->td_proc->p_comm, code);
749
750 if (error == 0) {
751 td->td_retval[0] = 0;
752 td->td_retval[1] = frame.tf_rdx;
753
754 STOPEVENT(p, S_SCE, narg);
755
756 PTRACESTOP_SC(p, td, S_PT_SCE);
757
758 if ((callp->sy_narg & SYF_MPSAFE) == 0) {
759 mtx_lock(&Giant);
760 error = (*callp->sy_call)(td, argp);
761 mtx_unlock(&Giant);
762 } else
763 error = (*callp->sy_call)(td, argp);
764 }
765
766 switch (error) {
767 case 0:
768 frame.tf_rax = td->td_retval[0];
769 frame.tf_rdx = td->td_retval[1];
770 frame.tf_rflags &= ~PSL_C;
771 break;
772
773 case ERESTART:
774 /*
775 * Reconstruct pc, we know that 'syscall' is 2 bytes.
776 * We have to do a full context restore so that %r10
777 * (which was holding the value of %rcx) is restored for
778 * the next iteration.
779 */
780 frame.tf_rip -= frame.tf_err;
781 frame.tf_r10 = frame.tf_rcx;
782 td->td_pcb->pcb_flags |= PCB_FULLCTX;
783 break;
784
785 case EJUSTRETURN:
786 break;
787
788 default:
789 if (p->p_sysent->sv_errsize) {
790 if (error >= p->p_sysent->sv_errsize)
791 error = -1; /* XXX */
792 else
793 error = p->p_sysent->sv_errtbl[error];
794 }
795 frame.tf_rax = error;
796 frame.tf_rflags |= PSL_C;
797 break;
798 }
799
800 /*
801 * Traced syscall.
802 */
803 if (orig_tf_rflags & PSL_T) {
804 frame.tf_rflags &= ~PSL_T;
805 trapsignal(td, SIGTRAP, 0);
806 }
807
808 /*
809 * Handle reschedule and other end-of-syscall issues
810 */
811 userret(td, &frame, sticks);
812
813 CTR4(KTR_SYSC, "syscall exit thread %p pid %d proc %s code %d", td,
814 td->td_proc->p_pid, td->td_proc->p_comm, code);
815
816 #ifdef KTRACE
817 if (KTRPOINT(td, KTR_SYSRET))
818 ktrsysret(code, error, td->td_retval[0]);
819 #endif
820
821 /*
822 * This works because errno is findable through the
823 * register set. If we ever support an emulation where this
824 * is not the case, this code will need to be revisited.
825 */
826 STOPEVENT(p, S_SCX, code);
827
828 PTRACESTOP_SC(p, td, S_PT_SCX);
829
830 WITNESS_WARN(WARN_PANIC, NULL, "System call %s returning",
831 (code >= 0 && code < SYS_MAXSYSCALL) ? syscallnames[code] : "???");
832 mtx_assert(&sched_lock, MA_NOTOWNED);
833 mtx_assert(&Giant, MA_NOTOWNED);
834 }
Cache object: 898b016cc77ccec6766349d0731d4b03
|