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("$FreeBSD$");
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 #include <security/audit/audit.h>
79
80 #include <vm/vm.h>
81 #include <vm/vm_param.h>
82 #include <vm/pmap.h>
83 #include <vm/vm_kern.h>
84 #include <vm/vm_map.h>
85 #include <vm/vm_page.h>
86 #include <vm/vm_extern.h>
87
88 #include <machine/cpu.h>
89 #include <machine/intr_machdep.h>
90 #include <machine/md_var.h>
91 #include <machine/pcb.h>
92 #ifdef SMP
93 #include <machine/smp.h>
94 #endif
95 #include <machine/tss.h>
96
97 extern void trap(struct trapframe *frame);
98 extern void syscall(struct trapframe *frame);
99 void dblfault_handler(struct trapframe *frame);
100
101 static int trap_pfault(struct trapframe *, int);
102 static void trap_fatal(struct trapframe *, vm_offset_t);
103
104 #define MAX_TRAP_MSG 30
105 static char *trap_msg[] = {
106 "", /* 0 unused */
107 "privileged instruction fault", /* 1 T_PRIVINFLT */
108 "", /* 2 unused */
109 "breakpoint instruction fault", /* 3 T_BPTFLT */
110 "", /* 4 unused */
111 "", /* 5 unused */
112 "arithmetic trap", /* 6 T_ARITHTRAP */
113 "", /* 7 unused */
114 "", /* 8 unused */
115 "general protection fault", /* 9 T_PROTFLT */
116 "trace trap", /* 10 T_TRCTRAP */
117 "", /* 11 unused */
118 "page fault", /* 12 T_PAGEFLT */
119 "", /* 13 unused */
120 "alignment fault", /* 14 T_ALIGNFLT */
121 "", /* 15 unused */
122 "", /* 16 unused */
123 "", /* 17 unused */
124 "integer divide fault", /* 18 T_DIVIDE */
125 "non-maskable interrupt trap", /* 19 T_NMI */
126 "overflow trap", /* 20 T_OFLOW */
127 "FPU bounds check fault", /* 21 T_BOUND */
128 "FPU device not available", /* 22 T_DNA */
129 "double fault", /* 23 T_DOUBLEFLT */
130 "FPU operand fetch fault", /* 24 T_FPOPFLT */
131 "invalid TSS fault", /* 25 T_TSSFLT */
132 "segment not present fault", /* 26 T_SEGNPFLT */
133 "stack fault", /* 27 T_STKFLT */
134 "machine check trap", /* 28 T_MCHK */
135 "SIMD floating-point exception", /* 29 T_XMMFLT */
136 "reserved (unknown) fault", /* 30 T_RESERVED */
137 };
138
139 #ifdef KDB
140 static int kdb_on_nmi = 1;
141 SYSCTL_INT(_machdep, OID_AUTO, kdb_on_nmi, CTLFLAG_RW,
142 &kdb_on_nmi, 0, "Go to KDB on NMI");
143 #endif
144 static int panic_on_nmi = 1;
145 SYSCTL_INT(_machdep, OID_AUTO, panic_on_nmi, CTLFLAG_RW,
146 &panic_on_nmi, 0, "Panic on NMI");
147 static int prot_fault_translation = 0;
148 SYSCTL_INT(_machdep, OID_AUTO, prot_fault_translation, CTLFLAG_RW,
149 &prot_fault_translation, 0, "Select signal to deliver on protection fault");
150
151 extern char *syscallnames[];
152
153 /*
154 * Exception, fault, and trap interface to the FreeBSD kernel.
155 * This common code is called from assembly language IDT gate entry
156 * routines that prepare a suitable stack frame, and restore this
157 * frame after the exception has been processed.
158 */
159
160 void
161 trap(struct trapframe *frame)
162 {
163 struct thread *td = curthread;
164 struct proc *p = td->td_proc;
165 int i = 0, ucode = 0, code;
166 u_int type;
167 register_t addr = 0;
168 ksiginfo_t ksi;
169
170 PCPU_INC(cnt.v_trap);
171 type = frame->tf_trapno;
172
173 #ifdef SMP
174 #ifdef STOP_NMI
175 /* Handler for NMI IPIs used for stopping CPUs. */
176 if (type == T_NMI) {
177 if (ipi_nmi_handler() == 0)
178 goto out;
179 }
180 #endif /* STOP_NMI */
181 #endif /* SMP */
182
183 #ifdef KDB
184 if (kdb_active) {
185 kdb_reenter();
186 goto out;
187 }
188 #endif
189
190 #ifdef HWPMC_HOOKS
191 /*
192 * CPU PMCs interrupt using an NMI. If the PMC module is
193 * active, pass the 'rip' value to the PMC module's interrupt
194 * handler. A return value of '1' from the handler means that
195 * the NMI was handled by it and we can return immediately.
196 */
197 if (type == T_NMI && pmc_intr &&
198 (*pmc_intr)(PCPU_GET(cpuid), (uintptr_t) frame->tf_rip,
199 TRAPF_USERMODE(frame)))
200 goto out;
201 #endif
202
203 if ((frame->tf_rflags & PSL_I) == 0) {
204 /*
205 * Buggy application or kernel code has disabled
206 * interrupts and then trapped. Enabling interrupts
207 * now is wrong, but it is better than running with
208 * interrupts disabled until they are accidentally
209 * enabled later.
210 */
211 if (ISPL(frame->tf_cs) == SEL_UPL)
212 printf(
213 "pid %ld (%s): trap %d with interrupts disabled\n",
214 (long)curproc->p_pid, curproc->p_comm, type);
215 else if (type != T_NMI && type != T_BPTFLT &&
216 type != T_TRCTRAP) {
217 /*
218 * XXX not quite right, since this may be for a
219 * multiple fault in user mode.
220 */
221 printf("kernel trap %d with interrupts disabled\n",
222 type);
223 /*
224 * We shouldn't enable interrupts while holding a
225 * spin lock or servicing an NMI.
226 */
227 if (type != T_NMI && td->td_md.md_spinlock_count == 0)
228 enable_intr();
229 }
230 }
231
232 code = frame->tf_err;
233 if (type == T_PAGEFLT) {
234 /*
235 * If we get a page fault while in a critical section, then
236 * it is most likely a fatal kernel page fault. The kernel
237 * is already going to panic trying to get a sleep lock to
238 * do the VM lookup, so just consider it a fatal trap so the
239 * kernel can print out a useful trap message and even get
240 * to the debugger.
241 *
242 * If we get a page fault while holding a non-sleepable
243 * lock, then it is most likely a fatal kernel page fault.
244 * If WITNESS is enabled, then it's going to whine about
245 * bogus LORs with various VM locks, so just skip to the
246 * fatal trap handling directly.
247 */
248 if (td->td_critnest != 0 ||
249 WITNESS_CHECK(WARN_SLEEPOK | WARN_GIANTOK, NULL,
250 "Kernel page fault") != 0)
251 trap_fatal(frame, frame->tf_addr);
252 }
253
254 if (ISPL(frame->tf_cs) == SEL_UPL) {
255 /* user trap */
256
257 td->td_pticks = 0;
258 td->td_frame = frame;
259 addr = frame->tf_rip;
260 if (td->td_ucred != p->p_ucred)
261 cred_update_thread(td);
262
263 switch (type) {
264 case T_PRIVINFLT: /* privileged instruction fault */
265 i = SIGILL;
266 ucode = ILL_PRVOPC;
267 break;
268
269 case T_BPTFLT: /* bpt instruction fault */
270 case T_TRCTRAP: /* trace trap */
271 enable_intr();
272 frame->tf_rflags &= ~PSL_T;
273 i = SIGTRAP;
274 ucode = (type == T_TRCTRAP ? TRAP_TRACE : TRAP_BRKPT);
275 break;
276
277 case T_ARITHTRAP: /* arithmetic trap */
278 ucode = fputrap();
279 if (ucode == -1)
280 goto userout;
281 i = SIGFPE;
282 break;
283
284 case T_PROTFLT: /* general protection fault */
285 i = SIGBUS;
286 ucode = BUS_OBJERR;
287 break;
288 case T_STKFLT: /* stack fault */
289 case T_SEGNPFLT: /* segment not present fault */
290 i = SIGBUS;
291 ucode = BUS_ADRERR;
292 break;
293 case T_TSSFLT: /* invalid TSS fault */
294 i = SIGBUS;
295 ucode = BUS_OBJERR;
296 break;
297 case T_DOUBLEFLT: /* double fault */
298 default:
299 i = SIGBUS;
300 ucode = BUS_OBJERR;
301 break;
302
303 case T_PAGEFLT: /* page fault */
304 addr = frame->tf_addr;
305 #ifdef KSE
306 if (td->td_pflags & TDP_SA)
307 thread_user_enter(td);
308 #endif
309 i = trap_pfault(frame, TRUE);
310 if (i == -1)
311 goto userout;
312 if (i == 0)
313 goto user;
314
315 if (i == SIGSEGV)
316 ucode = SEGV_MAPERR;
317 else {
318 if (prot_fault_translation == 0) {
319 /*
320 * Autodetect.
321 * This check also covers the images
322 * without the ABI-tag ELF note.
323 */
324 if (p->p_osrel >= 700004) {
325 i = SIGSEGV;
326 ucode = SEGV_ACCERR;
327 } else {
328 i = SIGBUS;
329 ucode = BUS_PAGE_FAULT;
330 }
331 } else if (prot_fault_translation == 1) {
332 /*
333 * Always compat mode.
334 */
335 i = SIGBUS;
336 ucode = BUS_PAGE_FAULT;
337 } else {
338 /*
339 * Always SIGSEGV mode.
340 */
341 i = SIGSEGV;
342 ucode = SEGV_ACCERR;
343 }
344 }
345 break;
346
347 case T_DIVIDE: /* integer divide fault */
348 ucode = FPE_INTDIV;
349 i = SIGFPE;
350 break;
351
352 #ifdef DEV_ISA
353 case T_NMI:
354 /* machine/parity/power fail/"kitchen sink" faults */
355 /* XXX Giant */
356 if (isa_nmi(code) == 0) {
357 #ifdef KDB
358 /*
359 * NMI can be hooked up to a pushbutton
360 * for debugging.
361 */
362 if (kdb_on_nmi) {
363 printf ("NMI ... going to debugger\n");
364 kdb_trap(type, 0, frame);
365 }
366 #endif /* KDB */
367 goto userout;
368 } else if (panic_on_nmi)
369 panic("NMI indicates hardware failure");
370 break;
371 #endif /* DEV_ISA */
372
373 case T_OFLOW: /* integer overflow fault */
374 ucode = FPE_INTOVF;
375 i = SIGFPE;
376 break;
377
378 case T_BOUND: /* bounds check fault */
379 ucode = FPE_FLTSUB;
380 i = SIGFPE;
381 break;
382
383 case T_DNA:
384 /* transparent fault (due to context switch "late") */
385 if (fpudna())
386 goto userout;
387 printf("pid %d killed due to lack of floating point\n",
388 p->p_pid);
389 i = SIGKILL;
390 ucode = 0;
391 break;
392
393 case T_FPOPFLT: /* FPU operand fetch fault */
394 ucode = ILL_COPROC;
395 i = SIGILL;
396 break;
397
398 case T_XMMFLT: /* SIMD floating-point exception */
399 ucode = 0; /* XXX */
400 i = SIGFPE;
401 break;
402 }
403 } else {
404 /* kernel trap */
405
406 KASSERT(cold || td->td_ucred != NULL,
407 ("kernel trap doesn't have ucred"));
408 switch (type) {
409 case T_PAGEFLT: /* page fault */
410 (void) trap_pfault(frame, FALSE);
411 goto out;
412
413 case T_DNA:
414 /*
415 * The kernel is apparently using fpu for copying.
416 * XXX this should be fatal unless the kernel has
417 * registered such use.
418 */
419 if (fpudna()) {
420 printf("fpudna in kernel mode!\n");
421 goto out;
422 }
423 break;
424
425 case T_STKFLT: /* stack fault */
426 break;
427
428 case T_PROTFLT: /* general protection fault */
429 case T_SEGNPFLT: /* segment not present fault */
430 if (td->td_intr_nesting_level != 0)
431 break;
432
433 /*
434 * Invalid segment selectors and out of bounds
435 * %rip's and %rsp's can be set up in user mode.
436 * This causes a fault in kernel mode when the
437 * kernel tries to return to user mode. We want
438 * to get this fault so that we can fix the
439 * problem here and not have to check all the
440 * selectors and pointers when the user changes
441 * them.
442 */
443 if (frame->tf_rip == (long)doreti_iret) {
444 frame->tf_rip = (long)doreti_iret_fault;
445 goto out;
446 }
447 if (PCPU_GET(curpcb)->pcb_onfault != NULL) {
448 frame->tf_rip =
449 (long)PCPU_GET(curpcb)->pcb_onfault;
450 goto out;
451 }
452 break;
453
454 case T_TSSFLT:
455 /*
456 * PSL_NT can be set in user mode and isn't cleared
457 * automatically when the kernel is entered. This
458 * causes a TSS fault when the kernel attempts to
459 * `iret' because the TSS link is uninitialized. We
460 * want to get this fault so that we can fix the
461 * problem here and not every time the kernel is
462 * entered.
463 */
464 if (frame->tf_rflags & PSL_NT) {
465 frame->tf_rflags &= ~PSL_NT;
466 goto out;
467 }
468 break;
469
470 case T_TRCTRAP: /* trace trap */
471 /*
472 * Ignore debug register trace traps due to
473 * accesses in the user's address space, which
474 * can happen under several conditions such as
475 * if a user sets a watchpoint on a buffer and
476 * then passes that buffer to a system call.
477 * We still want to get TRCTRAPS for addresses
478 * in kernel space because that is useful when
479 * debugging the kernel.
480 */
481 if (user_dbreg_trap()) {
482 /*
483 * Reset breakpoint bits because the
484 * processor doesn't
485 */
486 /* XXX check upper bits here */
487 load_dr6(rdr6() & 0xfffffff0);
488 goto out;
489 }
490 /*
491 * FALLTHROUGH (TRCTRAP kernel mode, kernel address)
492 */
493 case T_BPTFLT:
494 /*
495 * If KDB is enabled, let it handle the debugger trap.
496 * Otherwise, debugger traps "can't happen".
497 */
498 #ifdef KDB
499 if (kdb_trap(type, 0, frame))
500 goto out;
501 #endif
502 break;
503
504 #ifdef DEV_ISA
505 case T_NMI:
506 /* XXX Giant */
507 /* machine/parity/power fail/"kitchen sink" faults */
508 if (isa_nmi(code) == 0) {
509 #ifdef KDB
510 /*
511 * NMI can be hooked up to a pushbutton
512 * for debugging.
513 */
514 if (kdb_on_nmi) {
515 printf ("NMI ... going to debugger\n");
516 kdb_trap(type, 0, frame);
517 }
518 #endif /* KDB */
519 goto out;
520 } else if (panic_on_nmi == 0)
521 goto out;
522 /* FALLTHROUGH */
523 #endif /* DEV_ISA */
524 }
525
526 trap_fatal(frame, 0);
527 goto out;
528 }
529
530 /* Translate fault for emulators (e.g. Linux) */
531 if (*p->p_sysent->sv_transtrap)
532 i = (*p->p_sysent->sv_transtrap)(i, type);
533
534 ksiginfo_init_trap(&ksi);
535 ksi.ksi_signo = i;
536 ksi.ksi_code = ucode;
537 ksi.ksi_trapno = type;
538 ksi.ksi_addr = (void *)addr;
539 trapsignal(td, &ksi);
540
541 #ifdef DEBUG
542 if (type <= MAX_TRAP_MSG) {
543 uprintf("fatal process exception: %s",
544 trap_msg[type]);
545 if ((type == T_PAGEFLT) || (type == T_PROTFLT))
546 uprintf(", fault VA = 0x%lx", frame->tf_addr);
547 uprintf("\n");
548 }
549 #endif
550
551 user:
552 userret(td, frame);
553 mtx_assert(&Giant, MA_NOTOWNED);
554 userout:
555 out:
556 return;
557 }
558
559 static int
560 trap_pfault(frame, usermode)
561 struct trapframe *frame;
562 int usermode;
563 {
564 vm_offset_t va;
565 struct vmspace *vm = NULL;
566 vm_map_t map;
567 int rv = 0;
568 vm_prot_t ftype;
569 struct thread *td = curthread;
570 struct proc *p = td->td_proc;
571 vm_offset_t eva = frame->tf_addr;
572
573 va = trunc_page(eva);
574 if (va >= KERNBASE) {
575 /*
576 * Don't allow user-mode faults in kernel address space.
577 */
578 if (usermode)
579 goto nogo;
580
581 map = kernel_map;
582 } else {
583 /*
584 * This is a fault on non-kernel virtual memory.
585 * vm is initialized above to NULL. If curproc is NULL
586 * or curproc->p_vmspace is NULL the fault is fatal.
587 */
588 if (p != NULL)
589 vm = p->p_vmspace;
590
591 if (vm == NULL)
592 goto nogo;
593
594 map = &vm->vm_map;
595 }
596
597 /*
598 * PGEX_I is defined only if the execute disable bit capability is
599 * supported and enabled.
600 */
601 if (frame->tf_err & PGEX_W)
602 ftype = VM_PROT_WRITE;
603 else if ((frame->tf_err & PGEX_I) && pg_nx != 0)
604 ftype = VM_PROT_EXECUTE;
605 else
606 ftype = VM_PROT_READ;
607
608 if (map != kernel_map) {
609 /*
610 * Keep swapout from messing with us during this
611 * critical time.
612 */
613 PROC_LOCK(p);
614 ++p->p_lock;
615 PROC_UNLOCK(p);
616
617 /* Fault in the user page: */
618 rv = vm_fault(map, va, ftype,
619 (ftype & VM_PROT_WRITE) ? VM_FAULT_DIRTY
620 : VM_FAULT_NORMAL);
621
622 PROC_LOCK(p);
623 --p->p_lock;
624 PROC_UNLOCK(p);
625 } else {
626 /*
627 * Don't have to worry about process locking or stacks in the
628 * kernel.
629 */
630 rv = vm_fault(map, va, ftype, VM_FAULT_NORMAL);
631 }
632 if (rv == KERN_SUCCESS)
633 return (0);
634 nogo:
635 if (!usermode) {
636 if (td->td_intr_nesting_level == 0 &&
637 PCPU_GET(curpcb)->pcb_onfault != NULL) {
638 frame->tf_rip = (long)PCPU_GET(curpcb)->pcb_onfault;
639 return (0);
640 }
641 trap_fatal(frame, eva);
642 return (-1);
643 }
644
645 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV);
646 }
647
648 static void
649 trap_fatal(frame, eva)
650 struct trapframe *frame;
651 vm_offset_t eva;
652 {
653 int code, ss;
654 u_int type;
655 long esp;
656 struct soft_segment_descriptor softseg;
657 char *msg;
658
659 code = frame->tf_err;
660 type = frame->tf_trapno;
661 sdtossd(&gdt[IDXSEL(frame->tf_cs & 0xffff)], &softseg);
662
663 if (type <= MAX_TRAP_MSG)
664 msg = trap_msg[type];
665 else
666 msg = "UNKNOWN";
667 printf("\n\nFatal trap %d: %s while in %s mode\n", type, msg,
668 ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel");
669 #ifdef SMP
670 /* two separate prints in case of a trap on an unmapped page */
671 printf("cpuid = %d; ", PCPU_GET(cpuid));
672 printf("apic id = %02x\n", PCPU_GET(apic_id));
673 #endif
674 if (type == T_PAGEFLT) {
675 printf("fault virtual address = 0x%lx\n", eva);
676 printf("fault code = %s %s %s, %s\n",
677 code & PGEX_U ? "user" : "supervisor",
678 code & PGEX_W ? "write" : "read",
679 code & PGEX_I ? "instruction" : "data",
680 code & PGEX_P ? "protection violation" : "page not present");
681 }
682 printf("instruction pointer = 0x%lx:0x%lx\n",
683 frame->tf_cs & 0xffff, frame->tf_rip);
684 if (ISPL(frame->tf_cs) == SEL_UPL) {
685 ss = frame->tf_ss & 0xffff;
686 esp = frame->tf_rsp;
687 } else {
688 ss = GSEL(GDATA_SEL, SEL_KPL);
689 esp = (long)&frame->tf_rsp;
690 }
691 printf("stack pointer = 0x%x:0x%lx\n", ss, esp);
692 printf("frame pointer = 0x%x:0x%lx\n", ss, frame->tf_rbp);
693 printf("code segment = base 0x%lx, limit 0x%lx, type 0x%x\n",
694 softseg.ssd_base, softseg.ssd_limit, softseg.ssd_type);
695 printf(" = DPL %d, pres %d, long %d, def32 %d, gran %d\n",
696 softseg.ssd_dpl, softseg.ssd_p, softseg.ssd_long, softseg.ssd_def32,
697 softseg.ssd_gran);
698 printf("processor eflags = ");
699 if (frame->tf_rflags & PSL_T)
700 printf("trace trap, ");
701 if (frame->tf_rflags & PSL_I)
702 printf("interrupt enabled, ");
703 if (frame->tf_rflags & PSL_NT)
704 printf("nested task, ");
705 if (frame->tf_rflags & PSL_RF)
706 printf("resume, ");
707 printf("IOPL = %ld\n", (frame->tf_rflags & PSL_IOPL) >> 12);
708 printf("current process = ");
709 if (curproc) {
710 printf("%lu (%s)\n",
711 (u_long)curproc->p_pid, curproc->p_comm ?
712 curproc->p_comm : "");
713 } else {
714 printf("Idle\n");
715 }
716
717 #ifdef KDB
718 if (debugger_on_panic || kdb_active)
719 if (kdb_trap(type, 0, frame))
720 return;
721 #endif
722 printf("trap number = %d\n", type);
723 if (type <= MAX_TRAP_MSG)
724 panic("%s", trap_msg[type]);
725 else
726 panic("unknown/reserved trap");
727 }
728
729 /*
730 * Double fault handler. Called when a fault occurs while writing
731 * a frame for a trap/exception onto the stack. This usually occurs
732 * when the stack overflows (such is the case with infinite recursion,
733 * for example).
734 */
735 void
736 dblfault_handler(struct trapframe *frame)
737 {
738 printf("\nFatal double fault\n");
739 printf("rip = 0x%lx\n", frame->tf_rip);
740 printf("rsp = 0x%lx\n", frame->tf_rsp);
741 printf("rbp = 0x%lx\n", frame->tf_rbp);
742 #ifdef SMP
743 /* two separate prints in case of a trap on an unmapped page */
744 printf("cpuid = %d; ", PCPU_GET(cpuid));
745 printf("apic id = %02x\n", PCPU_GET(apic_id));
746 #endif
747 panic("double fault");
748 }
749
750 /*
751 * syscall - system call request C handler
752 *
753 * A system call is essentially treated as a trap.
754 */
755 void
756 syscall(struct trapframe *frame)
757 {
758 caddr_t params;
759 struct sysent *callp;
760 struct thread *td = curthread;
761 struct proc *p = td->td_proc;
762 register_t orig_tf_rflags;
763 int error;
764 int narg;
765 register_t args[8];
766 register_t *argp;
767 u_int code;
768 int reg, regcnt;
769 ksiginfo_t ksi;
770
771 PCPU_INC(cnt.v_syscall);
772
773 #ifdef DIAGNOSTIC
774 if (ISPL(frame->tf_cs) != SEL_UPL) {
775 panic("syscall");
776 /* NOT REACHED */
777 }
778 #endif
779
780 reg = 0;
781 regcnt = 6;
782 td->td_pticks = 0;
783 td->td_frame = frame;
784 if (td->td_ucred != p->p_ucred)
785 cred_update_thread(td);
786 #ifdef KSE
787 if (p->p_flag & P_SA)
788 thread_user_enter(td);
789 #endif
790 params = (caddr_t)frame->tf_rsp + sizeof(register_t);
791 code = frame->tf_rax;
792 orig_tf_rflags = frame->tf_rflags;
793
794 if (p->p_sysent->sv_prepsyscall) {
795 /*
796 * The prep code is MP aware.
797 */
798 (*p->p_sysent->sv_prepsyscall)(frame, (int *)args, &code, ¶ms);
799 } else {
800 if (code == SYS_syscall || code == SYS___syscall) {
801 code = frame->tf_rdi;
802 reg++;
803 regcnt--;
804 }
805 }
806
807 if (p->p_sysent->sv_mask)
808 code &= p->p_sysent->sv_mask;
809
810 if (code >= p->p_sysent->sv_size)
811 callp = &p->p_sysent->sv_table[0];
812 else
813 callp = &p->p_sysent->sv_table[code];
814
815 narg = callp->sy_narg;
816
817 /*
818 * copyin and the ktrsyscall()/ktrsysret() code is MP-aware
819 */
820 KASSERT(narg <= sizeof(args) / sizeof(args[0]),
821 ("Too many syscall arguments!"));
822 error = 0;
823 argp = &frame->tf_rdi;
824 argp += reg;
825 bcopy(argp, args, sizeof(args[0]) * regcnt);
826 if (narg > regcnt) {
827 KASSERT(params != NULL, ("copyin args with no params!"));
828 error = copyin(params, &args[regcnt],
829 (narg - regcnt) * sizeof(args[0]));
830 }
831 argp = &args[0];
832
833 #ifdef KTRACE
834 if (KTRPOINT(td, KTR_SYSCALL))
835 ktrsyscall(code, narg, argp);
836 #endif
837
838 CTR4(KTR_SYSC, "syscall enter thread %p pid %d proc %s code %d", td,
839 td->td_proc->p_pid, td->td_proc->p_comm, code);
840
841 td->td_syscalls++;
842
843 if (error == 0) {
844 td->td_retval[0] = 0;
845 td->td_retval[1] = frame->tf_rdx;
846
847 STOPEVENT(p, S_SCE, narg);
848
849 PTRACESTOP_SC(p, td, S_PT_SCE);
850
851 AUDIT_SYSCALL_ENTER(code, td);
852 error = (*callp->sy_call)(td, argp);
853 AUDIT_SYSCALL_EXIT(error, td);
854 }
855
856 switch (error) {
857 case 0:
858 frame->tf_rax = td->td_retval[0];
859 frame->tf_rdx = td->td_retval[1];
860 frame->tf_rflags &= ~PSL_C;
861 break;
862
863 case ERESTART:
864 /*
865 * Reconstruct pc, we know that 'syscall' is 2 bytes.
866 * We have to do a full context restore so that %r10
867 * (which was holding the value of %rcx) is restored for
868 * the next iteration.
869 */
870 frame->tf_rip -= frame->tf_err;
871 frame->tf_r10 = frame->tf_rcx;
872 td->td_pcb->pcb_flags |= PCB_FULLCTX;
873 break;
874
875 case EJUSTRETURN:
876 break;
877
878 default:
879 if (p->p_sysent->sv_errsize) {
880 if (error >= p->p_sysent->sv_errsize)
881 error = -1; /* XXX */
882 else
883 error = p->p_sysent->sv_errtbl[error];
884 }
885 frame->tf_rax = error;
886 frame->tf_rflags |= PSL_C;
887 break;
888 }
889
890 /*
891 * Traced syscall.
892 */
893 if (orig_tf_rflags & PSL_T) {
894 frame->tf_rflags &= ~PSL_T;
895 ksiginfo_init_trap(&ksi);
896 ksi.ksi_signo = SIGTRAP;
897 ksi.ksi_code = TRAP_TRACE;
898 ksi.ksi_addr = (void *)frame->tf_rip;
899 trapsignal(td, &ksi);
900 }
901
902 /*
903 * Check for misbehavior.
904 */
905 WITNESS_WARN(WARN_PANIC, NULL, "System call %s returning",
906 (code >= 0 && code < SYS_MAXSYSCALL) ? syscallnames[code] : "???");
907 KASSERT(td->td_critnest == 0,
908 ("System call %s returning in a critical section",
909 (code >= 0 && code < SYS_MAXSYSCALL) ? syscallnames[code] : "???"));
910 KASSERT(td->td_locks == 0,
911 ("System call %s returning with %d locks held",
912 (code >= 0 && code < SYS_MAXSYSCALL) ? syscallnames[code] : "???",
913 td->td_locks));
914
915 /*
916 * Handle reschedule and other end-of-syscall issues
917 */
918 userret(td, frame);
919
920 CTR4(KTR_SYSC, "syscall exit thread %p pid %d proc %s code %d", td,
921 td->td_proc->p_pid, td->td_proc->p_comm, code);
922
923 #ifdef KTRACE
924 if (KTRPOINT(td, KTR_SYSRET))
925 ktrsysret(code, error, td->td_retval[0]);
926 #endif
927
928 /*
929 * This works because errno is findable through the
930 * register set. If we ever support an emulation where this
931 * is not the case, this code will need to be revisited.
932 */
933 STOPEVENT(p, S_SCX, code);
934
935 PTRACESTOP_SC(p, td, S_PT_SCX);
936 }
Cache object: aa78912c590cba279b7337e2e9d0a16a
|