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/6.2/sys/amd64/amd64/trap.c 152905 2005-11-28 20:03:15Z jhb $");
42
43 /*
44 * AMD64 Trap and System call handling
45 */
46
47 #include "opt_clock.h"
48 #include "opt_cpu.h"
49 #include "opt_hwpmc_hooks.h"
50 #include "opt_isa.h"
51 #include "opt_kdb.h"
52 #include "opt_ktrace.h"
53
54 #include <sys/param.h>
55 #include <sys/bus.h>
56 #include <sys/systm.h>
57 #include <sys/proc.h>
58 #include <sys/pioctl.h>
59 #include <sys/ptrace.h>
60 #include <sys/kdb.h>
61 #include <sys/kernel.h>
62 #include <sys/ktr.h>
63 #include <sys/lock.h>
64 #include <sys/mutex.h>
65 #include <sys/resourcevar.h>
66 #include <sys/signalvar.h>
67 #include <sys/syscall.h>
68 #include <sys/sysctl.h>
69 #include <sys/sysent.h>
70 #include <sys/uio.h>
71 #include <sys/vmmeter.h>
72 #ifdef KTRACE
73 #include <sys/ktrace.h>
74 #endif
75 #ifdef HWPMC_HOOKS
76 #include <sys/pmckern.h>
77 #endif
78
79 #include <vm/vm.h>
80 #include <vm/vm_param.h>
81 #include <vm/pmap.h>
82 #include <vm/vm_kern.h>
83 #include <vm/vm_map.h>
84 #include <vm/vm_page.h>
85 #include <vm/vm_extern.h>
86
87 #include <machine/cpu.h>
88 #include <machine/intr_machdep.h>
89 #include <machine/md_var.h>
90 #include <machine/pcb.h>
91 #ifdef SMP
92 #include <machine/smp.h>
93 #endif
94 #include <machine/tss.h>
95
96 extern void trap(struct trapframe frame);
97 extern void syscall(struct trapframe frame);
98
99 static int trap_pfault(struct trapframe *, int);
100 static void trap_fatal(struct trapframe *, vm_offset_t);
101 void dblfault_handler(void);
102
103 #define MAX_TRAP_MSG 30
104 static char *trap_msg[] = {
105 "", /* 0 unused */
106 "privileged instruction fault", /* 1 T_PRIVINFLT */
107 "", /* 2 unused */
108 "breakpoint instruction fault", /* 3 T_BPTFLT */
109 "", /* 4 unused */
110 "", /* 5 unused */
111 "arithmetic trap", /* 6 T_ARITHTRAP */
112 "", /* 7 unused */
113 "", /* 8 unused */
114 "general protection fault", /* 9 T_PROTFLT */
115 "trace trap", /* 10 T_TRCTRAP */
116 "", /* 11 unused */
117 "page fault", /* 12 T_PAGEFLT */
118 "", /* 13 unused */
119 "alignment fault", /* 14 T_ALIGNFLT */
120 "", /* 15 unused */
121 "", /* 16 unused */
122 "", /* 17 unused */
123 "integer divide fault", /* 18 T_DIVIDE */
124 "non-maskable interrupt trap", /* 19 T_NMI */
125 "overflow trap", /* 20 T_OFLOW */
126 "FPU bounds check fault", /* 21 T_BOUND */
127 "FPU device not available", /* 22 T_DNA */
128 "double fault", /* 23 T_DOUBLEFLT */
129 "FPU operand fetch fault", /* 24 T_FPOPFLT */
130 "invalid TSS fault", /* 25 T_TSSFLT */
131 "segment not present fault", /* 26 T_SEGNPFLT */
132 "stack fault", /* 27 T_STKFLT */
133 "machine check trap", /* 28 T_MCHK */
134 "SIMD floating-point exception", /* 29 T_XMMFLT */
135 "reserved (unknown) fault", /* 30 T_RESERVED */
136 };
137
138 #ifdef KDB
139 static int kdb_on_nmi = 1;
140 SYSCTL_INT(_machdep, OID_AUTO, kdb_on_nmi, CTLFLAG_RW,
141 &kdb_on_nmi, 0, "Go to KDB on NMI");
142 #endif
143 static int panic_on_nmi = 1;
144 SYSCTL_INT(_machdep, OID_AUTO, panic_on_nmi, CTLFLAG_RW,
145 &panic_on_nmi, 0, "Panic on NMI");
146
147 #ifdef WITNESS
148 extern char *syscallnames[];
149 #endif
150
151 /*
152 * Exception, fault, and trap interface to the FreeBSD kernel.
153 * This common code is called from assembly language IDT gate entry
154 * routines that prepare a suitable stack frame, and restore this
155 * frame after the exception has been processed.
156 */
157
158 void
159 trap(frame)
160 struct trapframe frame;
161 {
162 struct thread *td = curthread;
163 struct proc *p = td->td_proc;
164 u_int sticks = 0;
165 int i = 0, ucode = 0, type, code;
166
167 PCPU_LAZY_INC(cnt.v_trap);
168 type = frame.tf_trapno;
169
170 #ifdef KDB_STOP_NMI
171 /* Handler for NMI IPIs used for debugging */
172 if (type == T_NMI) {
173 if (ipi_nmi_handler() == 0)
174 goto out;
175 }
176 #endif /* KDB_STOP_NMI */
177
178 #ifdef KDB
179 if (kdb_active) {
180 kdb_reenter();
181 goto out;
182 }
183 #endif
184
185 #ifdef HWPMC_HOOKS
186 /*
187 * CPU PMCs interrupt using an NMI. If the PMC module is
188 * active, pass the 'rip' value to the PMC module's interrupt
189 * handler. A return value of '1' from the handler means that
190 * the NMI was handled by it and we can return immediately.
191 */
192 if (type == T_NMI && pmc_intr &&
193 (*pmc_intr)(PCPU_GET(cpuid), (uintptr_t) frame.tf_rip,
194 TRAPF_USERMODE(&frame)))
195 goto out;
196 #endif
197
198 if ((frame.tf_rflags & PSL_I) == 0) {
199 /*
200 * Buggy application or kernel code has disabled
201 * interrupts and then trapped. Enabling interrupts
202 * now is wrong, but it is better than running with
203 * interrupts disabled until they are accidentally
204 * enabled later.
205 */
206 if (ISPL(frame.tf_cs) == SEL_UPL)
207 printf(
208 "pid %ld (%s): trap %d with interrupts disabled\n",
209 (long)curproc->p_pid, curproc->p_comm, type);
210 else if (type != T_NMI && type != T_BPTFLT &&
211 type != T_TRCTRAP) {
212 /*
213 * XXX not quite right, since this may be for a
214 * multiple fault in user mode.
215 */
216 printf("kernel trap %d with interrupts disabled\n",
217 type);
218 /*
219 * We shouldn't enable interrupts while in a critical
220 * section.
221 */
222 if (td->td_critnest == 0)
223 enable_intr();
224 }
225 }
226
227 code = frame.tf_err;
228 if (type == T_PAGEFLT) {
229 /*
230 * If we get a page fault while in a critical section, then
231 * it is most likely a fatal kernel page fault. The kernel
232 * is already going to panic trying to get a sleep lock to
233 * do the VM lookup, so just consider it a fatal trap so the
234 * kernel can print out a useful trap message and even get
235 * to the debugger.
236 */
237 if (td->td_critnest != 0)
238 trap_fatal(&frame, frame.tf_addr);
239 }
240
241 if (ISPL(frame.tf_cs) == SEL_UPL) {
242 /* user trap */
243
244 sticks = td->td_sticks;
245 td->td_frame = &frame;
246 if (td->td_ucred != p->p_ucred)
247 cred_update_thread(td);
248
249 switch (type) {
250 case T_PRIVINFLT: /* privileged instruction fault */
251 ucode = type;
252 i = SIGILL;
253 break;
254
255 case T_BPTFLT: /* bpt instruction fault */
256 case T_TRCTRAP: /* trace trap */
257 enable_intr();
258 frame.tf_rflags &= ~PSL_T;
259 i = SIGTRAP;
260 break;
261
262 case T_ARITHTRAP: /* arithmetic trap */
263 ucode = fputrap();
264 if (ucode == -1)
265 goto userout;
266 i = SIGFPE;
267 break;
268
269 case T_PROTFLT: /* general protection fault */
270 case T_STKFLT: /* stack fault */
271 case T_SEGNPFLT: /* segment not present fault */
272 case T_TSSFLT: /* invalid TSS fault */
273 case T_DOUBLEFLT: /* double fault */
274 default:
275 ucode = code + BUS_SEGM_FAULT ;
276 i = SIGBUS;
277 break;
278
279 case T_PAGEFLT: /* page fault */
280 if (td->td_pflags & TDP_SA)
281 thread_user_enter(td);
282 i = trap_pfault(&frame, TRUE);
283 if (i == -1)
284 goto userout;
285 if (i == 0)
286 goto user;
287
288 ucode = T_PAGEFLT;
289 break;
290
291 case T_DIVIDE: /* integer divide fault */
292 ucode = FPE_INTDIV;
293 i = SIGFPE;
294 break;
295
296 #ifdef DEV_ISA
297 case T_NMI:
298 /* machine/parity/power fail/"kitchen sink" faults */
299 /* XXX Giant */
300 if (isa_nmi(code) == 0) {
301 #ifdef KDB
302 /*
303 * NMI can be hooked up to a pushbutton
304 * for debugging.
305 */
306 if (kdb_on_nmi) {
307 printf ("NMI ... going to debugger\n");
308 kdb_trap(type, 0, &frame);
309 }
310 #endif /* KDB */
311 goto userout;
312 } else if (panic_on_nmi)
313 panic("NMI indicates hardware failure");
314 break;
315 #endif /* DEV_ISA */
316
317 case T_OFLOW: /* integer overflow fault */
318 ucode = FPE_INTOVF;
319 i = SIGFPE;
320 break;
321
322 case T_BOUND: /* bounds check fault */
323 ucode = FPE_FLTSUB;
324 i = SIGFPE;
325 break;
326
327 case T_DNA:
328 /* transparent fault (due to context switch "late") */
329 if (fpudna())
330 goto userout;
331 i = SIGFPE;
332 ucode = FPE_FPU_NP_TRAP;
333 break;
334
335 case T_FPOPFLT: /* FPU operand fetch fault */
336 ucode = T_FPOPFLT;
337 i = SIGILL;
338 break;
339
340 case T_XMMFLT: /* SIMD floating-point exception */
341 ucode = 0; /* XXX */
342 i = SIGFPE;
343 break;
344 }
345 } else {
346 /* kernel trap */
347
348 KASSERT(cold || td->td_ucred != NULL,
349 ("kernel trap doesn't have ucred"));
350 switch (type) {
351 case T_PAGEFLT: /* page fault */
352 (void) trap_pfault(&frame, FALSE);
353 goto out;
354
355 case T_DNA:
356 /*
357 * The kernel is apparently using fpu for copying.
358 * XXX this should be fatal unless the kernel has
359 * registered such use.
360 */
361 if (fpudna()) {
362 printf("fpudna in kernel mode!\n");
363 goto out;
364 }
365 break;
366
367 case T_STKFLT: /* stack fault */
368 break;
369
370 case T_PROTFLT: /* general protection fault */
371 case T_SEGNPFLT: /* segment not present fault */
372 if (td->td_intr_nesting_level != 0)
373 break;
374
375 /*
376 * Invalid segment selectors and out of bounds
377 * %rip's and %rsp's can be set up in user mode.
378 * This causes a fault in kernel mode when the
379 * kernel tries to return to user mode. We want
380 * to get this fault so that we can fix the
381 * problem here and not have to check all the
382 * selectors and pointers when the user changes
383 * them.
384 */
385 if (frame.tf_rip == (long)doreti_iret) {
386 frame.tf_rip = (long)doreti_iret_fault;
387 goto out;
388 }
389 if (PCPU_GET(curpcb)->pcb_onfault != NULL) {
390 frame.tf_rip =
391 (long)PCPU_GET(curpcb)->pcb_onfault;
392 goto out;
393 }
394 break;
395
396 case T_TSSFLT:
397 /*
398 * PSL_NT can be set in user mode and isn't cleared
399 * automatically when the kernel is entered. This
400 * causes a TSS fault when the kernel attempts to
401 * `iret' because the TSS link is uninitialized. We
402 * want to get this fault so that we can fix the
403 * problem here and not every time the kernel is
404 * entered.
405 */
406 if (frame.tf_rflags & PSL_NT) {
407 frame.tf_rflags &= ~PSL_NT;
408 goto out;
409 }
410 break;
411
412 case T_TRCTRAP: /* trace trap */
413 /*
414 * Ignore debug register trace traps due to
415 * accesses in the user's address space, which
416 * can happen under several conditions such as
417 * if a user sets a watchpoint on a buffer and
418 * then passes that buffer to a system call.
419 * We still want to get TRCTRAPS for addresses
420 * in kernel space because that is useful when
421 * debugging the kernel.
422 */
423 if (user_dbreg_trap()) {
424 /*
425 * Reset breakpoint bits because the
426 * processor doesn't
427 */
428 /* XXX check upper bits here */
429 load_dr6(rdr6() & 0xfffffff0);
430 goto out;
431 }
432 /*
433 * FALLTHROUGH (TRCTRAP kernel mode, kernel address)
434 */
435 case T_BPTFLT:
436 /*
437 * If KDB is enabled, let it handle the debugger trap.
438 * Otherwise, debugger traps "can't happen".
439 */
440 #ifdef KDB
441 /* XXX Giant */
442 if (kdb_trap(type, 0, &frame))
443 goto out;
444 #endif
445 break;
446
447 #ifdef DEV_ISA
448 case T_NMI:
449 /* XXX Giant */
450 /* machine/parity/power fail/"kitchen sink" faults */
451 if (isa_nmi(code) == 0) {
452 #ifdef KDB
453 /*
454 * NMI can be hooked up to a pushbutton
455 * for debugging.
456 */
457 if (kdb_on_nmi) {
458 printf ("NMI ... going to debugger\n");
459 kdb_trap(type, 0, &frame);
460 }
461 #endif /* KDB */
462 goto out;
463 } else if (panic_on_nmi == 0)
464 goto out;
465 /* FALLTHROUGH */
466 #endif /* DEV_ISA */
467 }
468
469 trap_fatal(&frame, 0);
470 goto out;
471 }
472
473 /* Translate fault for emulators (e.g. Linux) */
474 if (*p->p_sysent->sv_transtrap)
475 i = (*p->p_sysent->sv_transtrap)(i, type);
476
477 trapsignal(td, i, ucode);
478
479 #ifdef DEBUG
480 if (type <= MAX_TRAP_MSG) {
481 uprintf("fatal process exception: %s",
482 trap_msg[type]);
483 if ((type == T_PAGEFLT) || (type == T_PROTFLT))
484 uprintf(", fault VA = 0x%lx", frame.tf_addr);
485 uprintf("\n");
486 }
487 #endif
488
489 user:
490 userret(td, &frame, sticks);
491 mtx_assert(&Giant, MA_NOTOWNED);
492 userout:
493 out:
494 return;
495 }
496
497 static int
498 trap_pfault(frame, usermode)
499 struct trapframe *frame;
500 int usermode;
501 {
502 vm_offset_t va;
503 struct vmspace *vm = NULL;
504 vm_map_t map = 0;
505 int rv = 0;
506 vm_prot_t ftype;
507 struct thread *td = curthread;
508 struct proc *p = td->td_proc;
509 vm_offset_t eva = frame->tf_addr;
510
511 va = trunc_page(eva);
512 if (va >= KERNBASE) {
513 /*
514 * Don't allow user-mode faults in kernel address space.
515 */
516 if (usermode)
517 goto nogo;
518
519 map = kernel_map;
520 } else {
521 /*
522 * This is a fault on non-kernel virtual memory.
523 * vm is initialized above to NULL. If curproc is NULL
524 * or curproc->p_vmspace is NULL the fault is fatal.
525 */
526 if (p != NULL)
527 vm = p->p_vmspace;
528
529 if (vm == NULL)
530 goto nogo;
531
532 map = &vm->vm_map;
533 }
534
535 if (frame->tf_err & PGEX_W)
536 ftype = VM_PROT_WRITE;
537 else
538 ftype = VM_PROT_READ;
539
540 if (map != kernel_map) {
541 /*
542 * Keep swapout from messing with us during this
543 * critical time.
544 */
545 PROC_LOCK(p);
546 ++p->p_lock;
547 PROC_UNLOCK(p);
548
549 /* Fault in the user page: */
550 rv = vm_fault(map, va, ftype,
551 (ftype & VM_PROT_WRITE) ? VM_FAULT_DIRTY
552 : VM_FAULT_NORMAL);
553
554 PROC_LOCK(p);
555 --p->p_lock;
556 PROC_UNLOCK(p);
557 } else {
558 /*
559 * Don't have to worry about process locking or stacks in the
560 * kernel.
561 */
562 rv = vm_fault(map, va, ftype, VM_FAULT_NORMAL);
563 }
564 if (rv == KERN_SUCCESS)
565 return (0);
566 nogo:
567 if (!usermode) {
568 if (td->td_intr_nesting_level == 0 &&
569 PCPU_GET(curpcb)->pcb_onfault != NULL) {
570 frame->tf_rip = (long)PCPU_GET(curpcb)->pcb_onfault;
571 return (0);
572 }
573 trap_fatal(frame, eva);
574 return (-1);
575 }
576
577 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV);
578 }
579
580 static void
581 trap_fatal(frame, eva)
582 struct trapframe *frame;
583 vm_offset_t eva;
584 {
585 int code, type, ss;
586 long esp;
587 struct soft_segment_descriptor softseg;
588 char *msg;
589
590 code = frame->tf_err;
591 type = frame->tf_trapno;
592 sdtossd(&gdt[IDXSEL(frame->tf_cs & 0xffff)], &softseg);
593
594 if (type <= MAX_TRAP_MSG)
595 msg = trap_msg[type];
596 else
597 msg = "UNKNOWN";
598 printf("\n\nFatal trap %d: %s while in %s mode\n", type, msg,
599 ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel");
600 #ifdef SMP
601 /* two separate prints in case of a trap on an unmapped page */
602 printf("cpuid = %d; ", PCPU_GET(cpuid));
603 printf("apic id = %02x\n", PCPU_GET(apic_id));
604 #endif
605 if (type == T_PAGEFLT) {
606 printf("fault virtual address = 0x%lx\n", eva);
607 printf("fault code = %s %s, %s\n",
608 code & PGEX_U ? "user" : "supervisor",
609 code & PGEX_W ? "write" : "read",
610 code & PGEX_P ? "protection violation" : "page not present");
611 }
612 printf("instruction pointer = 0x%lx:0x%lx\n",
613 frame->tf_cs & 0xffff, frame->tf_rip);
614 if (ISPL(frame->tf_cs) == SEL_UPL) {
615 ss = frame->tf_ss & 0xffff;
616 esp = frame->tf_rsp;
617 } else {
618 ss = GSEL(GDATA_SEL, SEL_KPL);
619 esp = (long)&frame->tf_rsp;
620 }
621 printf("stack pointer = 0x%x:0x%lx\n", ss, esp);
622 printf("frame pointer = 0x%x:0x%lx\n", ss, frame->tf_rbp);
623 printf("code segment = base 0x%lx, limit 0x%lx, type 0x%x\n",
624 softseg.ssd_base, softseg.ssd_limit, softseg.ssd_type);
625 printf(" = DPL %d, pres %d, long %d, def32 %d, gran %d\n",
626 softseg.ssd_dpl, softseg.ssd_p, softseg.ssd_long, softseg.ssd_def32,
627 softseg.ssd_gran);
628 printf("processor eflags = ");
629 if (frame->tf_rflags & PSL_T)
630 printf("trace trap, ");
631 if (frame->tf_rflags & PSL_I)
632 printf("interrupt enabled, ");
633 if (frame->tf_rflags & PSL_NT)
634 printf("nested task, ");
635 if (frame->tf_rflags & PSL_RF)
636 printf("resume, ");
637 printf("IOPL = %ld\n", (frame->tf_rflags & PSL_IOPL) >> 12);
638 printf("current process = ");
639 if (curproc) {
640 printf("%lu (%s)\n",
641 (u_long)curproc->p_pid, curproc->p_comm ?
642 curproc->p_comm : "");
643 } else {
644 printf("Idle\n");
645 }
646
647 #ifdef KDB
648 if (debugger_on_panic || kdb_active) {
649 register_t rflags;
650 rflags = intr_disable();
651 if (kdb_trap(type, 0, frame)) {
652 intr_restore(rflags);
653 return;
654 }
655 intr_restore(rflags);
656 }
657 #endif
658 printf("trap number = %d\n", type);
659 if (type <= MAX_TRAP_MSG)
660 panic("%s", trap_msg[type]);
661 else
662 panic("unknown/reserved trap");
663 }
664
665 /*
666 * Double fault handler. Called when a fault occurs while writing
667 * a frame for a trap/exception onto the stack. This usually occurs
668 * when the stack overflows (such is the case with infinite recursion,
669 * for example).
670 */
671 void
672 dblfault_handler()
673 {
674 printf("\nFatal double fault\n");
675 #ifdef SMP
676 /* two separate prints in case of a trap on an unmapped page */
677 printf("cpuid = %d; ", PCPU_GET(cpuid));
678 printf("apic id = %02x\n", PCPU_GET(apic_id));
679 #endif
680 panic("double fault");
681 }
682
683 /*
684 * syscall - system call request C handler
685 *
686 * A system call is essentially treated as a trap.
687 */
688 void
689 syscall(frame)
690 struct trapframe frame;
691 {
692 caddr_t params;
693 struct sysent *callp;
694 struct thread *td = curthread;
695 struct proc *p = td->td_proc;
696 register_t orig_tf_rflags;
697 u_int sticks;
698 int error;
699 int narg;
700 register_t args[8];
701 register_t *argp;
702 u_int code;
703 int reg, regcnt;
704
705 /*
706 * note: PCPU_LAZY_INC() can only be used if we can afford
707 * occassional inaccuracy in the count.
708 */
709 PCPU_LAZY_INC(cnt.v_syscall);
710
711 #ifdef DIAGNOSTIC
712 if (ISPL(frame.tf_cs) != SEL_UPL) {
713 mtx_lock(&Giant); /* try to stabilize the system XXX */
714 panic("syscall");
715 /* NOT REACHED */
716 mtx_unlock(&Giant);
717 }
718 #endif
719
720 reg = 0;
721 regcnt = 6;
722 sticks = td->td_sticks;
723 td->td_frame = &frame;
724 if (td->td_ucred != p->p_ucred)
725 cred_update_thread(td);
726 if (p->p_flag & P_SA)
727 thread_user_enter(td);
728 params = (caddr_t)frame.tf_rsp + sizeof(register_t);
729 code = frame.tf_rax;
730 orig_tf_rflags = frame.tf_rflags;
731
732 if (p->p_sysent->sv_prepsyscall) {
733 /*
734 * The prep code is MP aware.
735 */
736 (*p->p_sysent->sv_prepsyscall)(&frame, (int *)args, &code, ¶ms);
737 } else {
738 if (code == SYS_syscall || code == SYS___syscall) {
739 code = frame.tf_rdi;
740 reg++;
741 regcnt--;
742 }
743 }
744
745 if (p->p_sysent->sv_mask)
746 code &= p->p_sysent->sv_mask;
747
748 if (code >= p->p_sysent->sv_size)
749 callp = &p->p_sysent->sv_table[0];
750 else
751 callp = &p->p_sysent->sv_table[code];
752
753 narg = callp->sy_narg & SYF_ARGMASK;
754
755 /*
756 * copyin and the ktrsyscall()/ktrsysret() code is MP-aware
757 */
758 KASSERT(narg <= sizeof(args) / sizeof(args[0]),
759 ("Too many syscall arguments!"));
760 error = 0;
761 argp = &frame.tf_rdi;
762 argp += reg;
763 bcopy(argp, args, sizeof(args[0]) * regcnt);
764 if (narg > regcnt) {
765 KASSERT(params != NULL, ("copyin args with no params!"));
766 error = copyin(params, &args[regcnt],
767 (narg - regcnt) * sizeof(args[0]));
768 }
769 argp = &args[0];
770
771 #ifdef KTRACE
772 if (KTRPOINT(td, KTR_SYSCALL))
773 ktrsyscall(code, narg, argp);
774 #endif
775
776 CTR4(KTR_SYSC, "syscall enter thread %p pid %d proc %s code %d", td,
777 td->td_proc->p_pid, td->td_proc->p_comm, code);
778
779 if (error == 0) {
780 td->td_retval[0] = 0;
781 td->td_retval[1] = frame.tf_rdx;
782
783 STOPEVENT(p, S_SCE, narg);
784
785 PTRACESTOP_SC(p, td, S_PT_SCE);
786
787 if ((callp->sy_narg & SYF_MPSAFE) == 0) {
788 mtx_lock(&Giant);
789 error = (*callp->sy_call)(td, argp);
790 mtx_unlock(&Giant);
791 } else
792 error = (*callp->sy_call)(td, argp);
793 }
794
795 switch (error) {
796 case 0:
797 frame.tf_rax = td->td_retval[0];
798 frame.tf_rdx = td->td_retval[1];
799 frame.tf_rflags &= ~PSL_C;
800 break;
801
802 case ERESTART:
803 /*
804 * Reconstruct pc, we know that 'syscall' is 2 bytes.
805 * We have to do a full context restore so that %r10
806 * (which was holding the value of %rcx) is restored for
807 * the next iteration.
808 */
809 frame.tf_rip -= frame.tf_err;
810 frame.tf_r10 = frame.tf_rcx;
811 td->td_pcb->pcb_flags |= PCB_FULLCTX;
812 break;
813
814 case EJUSTRETURN:
815 break;
816
817 default:
818 if (p->p_sysent->sv_errsize) {
819 if (error >= p->p_sysent->sv_errsize)
820 error = -1; /* XXX */
821 else
822 error = p->p_sysent->sv_errtbl[error];
823 }
824 frame.tf_rax = error;
825 frame.tf_rflags |= PSL_C;
826 break;
827 }
828
829 /*
830 * Traced syscall.
831 */
832 if (orig_tf_rflags & PSL_T) {
833 frame.tf_rflags &= ~PSL_T;
834 trapsignal(td, SIGTRAP, 0);
835 }
836
837 /*
838 * Handle reschedule and other end-of-syscall issues
839 */
840 userret(td, &frame, sticks);
841
842 CTR4(KTR_SYSC, "syscall exit thread %p pid %d proc %s code %d", td,
843 td->td_proc->p_pid, td->td_proc->p_comm, code);
844
845 #ifdef KTRACE
846 if (KTRPOINT(td, KTR_SYSRET))
847 ktrsysret(code, error, td->td_retval[0]);
848 #endif
849
850 /*
851 * This works because errno is findable through the
852 * register set. If we ever support an emulation where this
853 * is not the case, this code will need to be revisited.
854 */
855 STOPEVENT(p, S_SCX, code);
856
857 PTRACESTOP_SC(p, td, S_PT_SCX);
858
859 WITNESS_WARN(WARN_PANIC, NULL, "System call %s returning",
860 (code >= 0 && code < SYS_MAXSYSCALL) ? syscallnames[code] : "???");
861 mtx_assert(&sched_lock, MA_NOTOWNED);
862 mtx_assert(&Giant, MA_NOTOWNED);
863 }
Cache object: 449eba4511a8bfcca18fca3fcb749a1f
|