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_kdtrace.h"
53 #include "opt_ktrace.h"
54
55 #include <sys/param.h>
56 #include <sys/bus.h>
57 #include <sys/systm.h>
58 #include <sys/proc.h>
59 #include <sys/pioctl.h>
60 #include <sys/ptrace.h>
61 #include <sys/kdb.h>
62 #include <sys/kernel.h>
63 #include <sys/ktr.h>
64 #include <sys/lock.h>
65 #include <sys/mutex.h>
66 #include <sys/resourcevar.h>
67 #include <sys/signalvar.h>
68 #include <sys/syscall.h>
69 #include <sys/sysctl.h>
70 #include <sys/sysent.h>
71 #include <sys/uio.h>
72 #include <sys/vmmeter.h>
73 #ifdef KTRACE
74 #include <sys/ktrace.h>
75 #endif
76 #ifdef HWPMC_HOOKS
77 #include <sys/pmckern.h>
78 #endif
79 #include <security/audit/audit.h>
80
81 #include <vm/vm.h>
82 #include <vm/vm_param.h>
83 #include <vm/pmap.h>
84 #include <vm/vm_kern.h>
85 #include <vm/vm_map.h>
86 #include <vm/vm_page.h>
87 #include <vm/vm_extern.h>
88
89 #include <machine/cpu.h>
90 #include <machine/intr_machdep.h>
91 #include <machine/md_var.h>
92 #include <machine/pcb.h>
93 #ifdef SMP
94 #include <machine/smp.h>
95 #endif
96 #include <machine/tss.h>
97
98 #ifdef KDTRACE_HOOKS
99 #include <sys/dtrace_bsd.h>
100
101 /*
102 * This is a hook which is initialised by the dtrace module
103 * to handle traps which might occur during DTrace probe
104 * execution.
105 */
106 dtrace_trap_func_t dtrace_trap_func;
107
108 dtrace_doubletrap_func_t dtrace_doubletrap_func;
109
110 /*
111 * This is a hook which is initialised by the systrace module
112 * when it is loaded. This keeps the DTrace syscall provider
113 * implementation opaque.
114 */
115 systrace_probe_func_t systrace_probe_func;
116 #endif
117
118 extern void trap(struct trapframe *frame);
119 extern void syscall(struct trapframe *frame);
120 void dblfault_handler(struct trapframe *frame);
121
122 static int trap_pfault(struct trapframe *, int);
123 static void trap_fatal(struct trapframe *, vm_offset_t);
124
125 #define MAX_TRAP_MSG 30
126 static char *trap_msg[] = {
127 "", /* 0 unused */
128 "privileged instruction fault", /* 1 T_PRIVINFLT */
129 "", /* 2 unused */
130 "breakpoint instruction fault", /* 3 T_BPTFLT */
131 "", /* 4 unused */
132 "", /* 5 unused */
133 "arithmetic trap", /* 6 T_ARITHTRAP */
134 "", /* 7 unused */
135 "", /* 8 unused */
136 "general protection fault", /* 9 T_PROTFLT */
137 "trace trap", /* 10 T_TRCTRAP */
138 "", /* 11 unused */
139 "page fault", /* 12 T_PAGEFLT */
140 "", /* 13 unused */
141 "alignment fault", /* 14 T_ALIGNFLT */
142 "", /* 15 unused */
143 "", /* 16 unused */
144 "", /* 17 unused */
145 "integer divide fault", /* 18 T_DIVIDE */
146 "non-maskable interrupt trap", /* 19 T_NMI */
147 "overflow trap", /* 20 T_OFLOW */
148 "FPU bounds check fault", /* 21 T_BOUND */
149 "FPU device not available", /* 22 T_DNA */
150 "double fault", /* 23 T_DOUBLEFLT */
151 "FPU operand fetch fault", /* 24 T_FPOPFLT */
152 "invalid TSS fault", /* 25 T_TSSFLT */
153 "segment not present fault", /* 26 T_SEGNPFLT */
154 "stack fault", /* 27 T_STKFLT */
155 "machine check trap", /* 28 T_MCHK */
156 "SIMD floating-point exception", /* 29 T_XMMFLT */
157 "reserved (unknown) fault", /* 30 T_RESERVED */
158 };
159
160 #ifdef KDB
161 static int kdb_on_nmi = 1;
162 SYSCTL_INT(_machdep, OID_AUTO, kdb_on_nmi, CTLFLAG_RW,
163 &kdb_on_nmi, 0, "Go to KDB on NMI");
164 #endif
165 static int panic_on_nmi = 1;
166 SYSCTL_INT(_machdep, OID_AUTO, panic_on_nmi, CTLFLAG_RW,
167 &panic_on_nmi, 0, "Panic on NMI");
168 static int prot_fault_translation = 0;
169 SYSCTL_INT(_machdep, OID_AUTO, prot_fault_translation, CTLFLAG_RW,
170 &prot_fault_translation, 0, "Select signal to deliver on protection fault");
171
172 extern char *syscallnames[];
173
174 /*
175 * Exception, fault, and trap interface to the FreeBSD kernel.
176 * This common code is called from assembly language IDT gate entry
177 * routines that prepare a suitable stack frame, and restore this
178 * frame after the exception has been processed.
179 */
180
181 void
182 trap(struct trapframe *frame)
183 {
184 struct thread *td = curthread;
185 struct proc *p = td->td_proc;
186 int i = 0, ucode = 0, code;
187 u_int type;
188 register_t addr = 0;
189 ksiginfo_t ksi;
190
191 PCPU_INC(cnt.v_trap);
192 type = frame->tf_trapno;
193
194 #ifdef SMP
195 #ifdef STOP_NMI
196 /* Handler for NMI IPIs used for stopping CPUs. */
197 if (type == T_NMI) {
198 if (ipi_nmi_handler() == 0)
199 goto out;
200 }
201 #endif /* STOP_NMI */
202 #endif /* SMP */
203
204 #ifdef KDB
205 if (kdb_active) {
206 kdb_reenter();
207 goto out;
208 }
209 #endif
210
211 #ifdef HWPMC_HOOKS
212 /*
213 * CPU PMCs interrupt using an NMI. If the PMC module is
214 * active, pass the 'rip' value to the PMC module's interrupt
215 * handler. A return value of '1' from the handler means that
216 * the NMI was handled by it and we can return immediately.
217 */
218 if (type == T_NMI && pmc_intr &&
219 (*pmc_intr)(PCPU_GET(cpuid), (uintptr_t) frame->tf_rip,
220 TRAPF_USERMODE(frame)))
221 goto out;
222 #endif
223
224 #ifdef KDTRACE_HOOKS
225 /*
226 * A trap can occur while DTrace executes a probe. Before
227 * executing the probe, DTrace blocks re-scheduling and sets
228 * a flag in it's per-cpu flags to indicate that it doesn't
229 * want to fault. On returning from the the probe, the no-fault
230 * flag is cleared and finally re-scheduling is enabled.
231 *
232 * If the DTrace kernel module has registered a trap handler,
233 * call it and if it returns non-zero, assume that it has
234 * handled the trap and modified the trap frame so that this
235 * function can return normally.
236 */
237 if (dtrace_trap_func != NULL)
238 if ((*dtrace_trap_func)(frame, type))
239 goto out;
240 #endif
241
242 if ((frame->tf_rflags & PSL_I) == 0) {
243 /*
244 * Buggy application or kernel code has disabled
245 * interrupts and then trapped. Enabling interrupts
246 * now is wrong, but it is better than running with
247 * interrupts disabled until they are accidentally
248 * enabled later.
249 */
250 if (ISPL(frame->tf_cs) == SEL_UPL)
251 printf(
252 "pid %ld (%s): trap %d with interrupts disabled\n",
253 (long)curproc->p_pid, curproc->p_comm, type);
254 else if (type != T_NMI && type != T_BPTFLT &&
255 type != T_TRCTRAP) {
256 /*
257 * XXX not quite right, since this may be for a
258 * multiple fault in user mode.
259 */
260 printf("kernel trap %d with interrupts disabled\n",
261 type);
262 /*
263 * We shouldn't enable interrupts while holding a
264 * spin lock or servicing an NMI.
265 */
266 if (type != T_NMI && td->td_md.md_spinlock_count == 0)
267 enable_intr();
268 }
269 }
270
271 code = frame->tf_err;
272 if (type == T_PAGEFLT) {
273 /*
274 * If we get a page fault while in a critical section, then
275 * it is most likely a fatal kernel page fault. The kernel
276 * is already going to panic trying to get a sleep lock to
277 * do the VM lookup, so just consider it a fatal trap so the
278 * kernel can print out a useful trap message and even get
279 * to the debugger.
280 *
281 * If we get a page fault while holding a non-sleepable
282 * lock, then it is most likely a fatal kernel page fault.
283 * If WITNESS is enabled, then it's going to whine about
284 * bogus LORs with various VM locks, so just skip to the
285 * fatal trap handling directly.
286 */
287 if (td->td_critnest != 0 ||
288 WITNESS_CHECK(WARN_SLEEPOK | WARN_GIANTOK, NULL,
289 "Kernel page fault") != 0)
290 trap_fatal(frame, frame->tf_addr);
291 }
292
293 if (ISPL(frame->tf_cs) == SEL_UPL) {
294 /* user trap */
295
296 td->td_pticks = 0;
297 td->td_frame = frame;
298 addr = frame->tf_rip;
299 if (td->td_ucred != p->p_ucred)
300 cred_update_thread(td);
301
302 switch (type) {
303 case T_PRIVINFLT: /* privileged instruction fault */
304 i = SIGILL;
305 ucode = ILL_PRVOPC;
306 break;
307
308 case T_BPTFLT: /* bpt instruction fault */
309 case T_TRCTRAP: /* trace trap */
310 enable_intr();
311 frame->tf_rflags &= ~PSL_T;
312 i = SIGTRAP;
313 ucode = (type == T_TRCTRAP ? TRAP_TRACE : TRAP_BRKPT);
314 break;
315
316 case T_ARITHTRAP: /* arithmetic trap */
317 ucode = fputrap();
318 if (ucode == -1)
319 goto userout;
320 i = SIGFPE;
321 break;
322
323 case T_PROTFLT: /* general protection fault */
324 i = SIGBUS;
325 ucode = BUS_OBJERR;
326 break;
327 case T_STKFLT: /* stack fault */
328 case T_SEGNPFLT: /* segment not present fault */
329 i = SIGBUS;
330 ucode = BUS_ADRERR;
331 break;
332 case T_TSSFLT: /* invalid TSS fault */
333 i = SIGBUS;
334 ucode = BUS_OBJERR;
335 break;
336 case T_DOUBLEFLT: /* double fault */
337 default:
338 i = SIGBUS;
339 ucode = BUS_OBJERR;
340 break;
341
342 case T_PAGEFLT: /* page fault */
343 addr = frame->tf_addr;
344 #ifdef KSE
345 if (td->td_pflags & TDP_SA)
346 thread_user_enter(td);
347 #endif
348 i = trap_pfault(frame, TRUE);
349 if (i == -1)
350 goto userout;
351 if (i == 0)
352 goto user;
353
354 if (i == SIGSEGV)
355 ucode = SEGV_MAPERR;
356 else {
357 if (prot_fault_translation == 0) {
358 /*
359 * Autodetect.
360 * This check also covers the images
361 * without the ABI-tag ELF note.
362 */
363 if (p->p_osrel >= 700004) {
364 i = SIGSEGV;
365 ucode = SEGV_ACCERR;
366 } else {
367 i = SIGBUS;
368 ucode = BUS_PAGE_FAULT;
369 }
370 } else if (prot_fault_translation == 1) {
371 /*
372 * Always compat mode.
373 */
374 i = SIGBUS;
375 ucode = BUS_PAGE_FAULT;
376 } else {
377 /*
378 * Always SIGSEGV mode.
379 */
380 i = SIGSEGV;
381 ucode = SEGV_ACCERR;
382 }
383 }
384 break;
385
386 case T_DIVIDE: /* integer divide fault */
387 ucode = FPE_INTDIV;
388 i = SIGFPE;
389 break;
390
391 #ifdef DEV_ISA
392 case T_NMI:
393 /* machine/parity/power fail/"kitchen sink" faults */
394 /* XXX Giant */
395 if (isa_nmi(code) == 0) {
396 #ifdef KDB
397 /*
398 * NMI can be hooked up to a pushbutton
399 * for debugging.
400 */
401 if (kdb_on_nmi) {
402 printf ("NMI ... going to debugger\n");
403 kdb_trap(type, 0, frame);
404 }
405 #endif /* KDB */
406 goto userout;
407 } else if (panic_on_nmi)
408 panic("NMI indicates hardware failure");
409 break;
410 #endif /* DEV_ISA */
411
412 case T_OFLOW: /* integer overflow fault */
413 ucode = FPE_INTOVF;
414 i = SIGFPE;
415 break;
416
417 case T_BOUND: /* bounds check fault */
418 ucode = FPE_FLTSUB;
419 i = SIGFPE;
420 break;
421
422 case T_DNA:
423 /* transparent fault (due to context switch "late") */
424 fpudna();
425 goto userout;
426
427 case T_FPOPFLT: /* FPU operand fetch fault */
428 ucode = ILL_COPROC;
429 i = SIGILL;
430 break;
431
432 case T_XMMFLT: /* SIMD floating-point exception */
433 ucode = 0; /* XXX */
434 i = SIGFPE;
435 break;
436 }
437 } else {
438 /* kernel trap */
439
440 KASSERT(cold || td->td_ucred != NULL,
441 ("kernel trap doesn't have ucred"));
442 switch (type) {
443 case T_PAGEFLT: /* page fault */
444 (void) trap_pfault(frame, FALSE);
445 goto out;
446
447 case T_DNA:
448 /*
449 * The kernel is apparently using fpu for copying.
450 * XXX this should be fatal unless the kernel has
451 * registered such use.
452 */
453 fpudna();
454 printf("fpudna in kernel mode!\n");
455 goto out;
456
457 case T_STKFLT: /* stack fault */
458 break;
459
460 case T_PROTFLT: /* general protection fault */
461 case T_SEGNPFLT: /* segment not present fault */
462 if (td->td_intr_nesting_level != 0)
463 break;
464
465 /*
466 * Invalid segment selectors and out of bounds
467 * %rip's and %rsp's can be set up in user mode.
468 * This causes a fault in kernel mode when the
469 * kernel tries to return to user mode. We want
470 * to get this fault so that we can fix the
471 * problem here and not have to check all the
472 * selectors and pointers when the user changes
473 * them.
474 */
475 if (frame->tf_rip == (long)doreti_iret) {
476 frame->tf_rip = (long)doreti_iret_fault;
477 goto out;
478 }
479 if (PCPU_GET(curpcb)->pcb_onfault != NULL) {
480 frame->tf_rip =
481 (long)PCPU_GET(curpcb)->pcb_onfault;
482 goto out;
483 }
484 break;
485
486 case T_TSSFLT:
487 /*
488 * PSL_NT can be set in user mode and isn't cleared
489 * automatically when the kernel is entered. This
490 * causes a TSS fault when the kernel attempts to
491 * `iret' because the TSS link is uninitialized. We
492 * want to get this fault so that we can fix the
493 * problem here and not every time the kernel is
494 * entered.
495 */
496 if (frame->tf_rflags & PSL_NT) {
497 frame->tf_rflags &= ~PSL_NT;
498 goto out;
499 }
500 break;
501
502 case T_TRCTRAP: /* trace trap */
503 /*
504 * Ignore debug register trace traps due to
505 * accesses in the user's address space, which
506 * can happen under several conditions such as
507 * if a user sets a watchpoint on a buffer and
508 * then passes that buffer to a system call.
509 * We still want to get TRCTRAPS for addresses
510 * in kernel space because that is useful when
511 * debugging the kernel.
512 */
513 if (user_dbreg_trap()) {
514 /*
515 * Reset breakpoint bits because the
516 * processor doesn't
517 */
518 /* XXX check upper bits here */
519 load_dr6(rdr6() & 0xfffffff0);
520 goto out;
521 }
522 /*
523 * FALLTHROUGH (TRCTRAP kernel mode, kernel address)
524 */
525 case T_BPTFLT:
526 /*
527 * If KDB is enabled, let it handle the debugger trap.
528 * Otherwise, debugger traps "can't happen".
529 */
530 #ifdef KDB
531 if (kdb_trap(type, 0, frame))
532 goto out;
533 #endif
534 break;
535
536 #ifdef DEV_ISA
537 case T_NMI:
538 /* XXX Giant */
539 /* machine/parity/power fail/"kitchen sink" faults */
540 if (isa_nmi(code) == 0) {
541 #ifdef KDB
542 /*
543 * NMI can be hooked up to a pushbutton
544 * for debugging.
545 */
546 if (kdb_on_nmi) {
547 printf ("NMI ... going to debugger\n");
548 kdb_trap(type, 0, frame);
549 }
550 #endif /* KDB */
551 goto out;
552 } else if (panic_on_nmi == 0)
553 goto out;
554 /* FALLTHROUGH */
555 #endif /* DEV_ISA */
556 }
557
558 trap_fatal(frame, 0);
559 goto out;
560 }
561
562 /* Translate fault for emulators (e.g. Linux) */
563 if (*p->p_sysent->sv_transtrap)
564 i = (*p->p_sysent->sv_transtrap)(i, type);
565
566 ksiginfo_init_trap(&ksi);
567 ksi.ksi_signo = i;
568 ksi.ksi_code = ucode;
569 ksi.ksi_trapno = type;
570 ksi.ksi_addr = (void *)addr;
571 trapsignal(td, &ksi);
572
573 #ifdef DEBUG
574 if (type <= MAX_TRAP_MSG) {
575 uprintf("fatal process exception: %s",
576 trap_msg[type]);
577 if ((type == T_PAGEFLT) || (type == T_PROTFLT))
578 uprintf(", fault VA = 0x%lx", frame->tf_addr);
579 uprintf("\n");
580 }
581 #endif
582
583 user:
584 userret(td, frame);
585 mtx_assert(&Giant, MA_NOTOWNED);
586 userout:
587 out:
588 return;
589 }
590
591 static int
592 trap_pfault(frame, usermode)
593 struct trapframe *frame;
594 int usermode;
595 {
596 vm_offset_t va;
597 struct vmspace *vm = NULL;
598 vm_map_t map;
599 int rv = 0;
600 vm_prot_t ftype;
601 struct thread *td = curthread;
602 struct proc *p = td->td_proc;
603 vm_offset_t eva = frame->tf_addr;
604
605 va = trunc_page(eva);
606 if (va >= VM_MIN_KERNEL_ADDRESS) {
607 /*
608 * Don't allow user-mode faults in kernel address space.
609 */
610 if (usermode)
611 goto nogo;
612
613 map = kernel_map;
614 } else {
615 /*
616 * This is a fault on non-kernel virtual memory.
617 * vm is initialized above to NULL. If curproc is NULL
618 * or curproc->p_vmspace is NULL the fault is fatal.
619 */
620 if (p != NULL)
621 vm = p->p_vmspace;
622
623 if (vm == NULL)
624 goto nogo;
625
626 map = &vm->vm_map;
627 }
628
629 /*
630 * PGEX_I is defined only if the execute disable bit capability is
631 * supported and enabled.
632 */
633 if (frame->tf_err & PGEX_W)
634 ftype = VM_PROT_WRITE;
635 else if ((frame->tf_err & PGEX_I) && pg_nx != 0)
636 ftype = VM_PROT_EXECUTE;
637 else
638 ftype = VM_PROT_READ;
639
640 if (map != kernel_map) {
641 /*
642 * Keep swapout from messing with us during this
643 * critical time.
644 */
645 PROC_LOCK(p);
646 ++p->p_lock;
647 PROC_UNLOCK(p);
648
649 /* Fault in the user page: */
650 rv = vm_fault(map, va, ftype,
651 (ftype & VM_PROT_WRITE) ? VM_FAULT_DIRTY
652 : VM_FAULT_NORMAL);
653
654 PROC_LOCK(p);
655 --p->p_lock;
656 PROC_UNLOCK(p);
657 } else {
658 /*
659 * Don't have to worry about process locking or stacks in the
660 * kernel.
661 */
662 rv = vm_fault(map, va, ftype, VM_FAULT_NORMAL);
663 }
664 if (rv == KERN_SUCCESS)
665 return (0);
666 nogo:
667 if (!usermode) {
668 if (td->td_intr_nesting_level == 0 &&
669 PCPU_GET(curpcb)->pcb_onfault != NULL) {
670 frame->tf_rip = (long)PCPU_GET(curpcb)->pcb_onfault;
671 return (0);
672 }
673 trap_fatal(frame, eva);
674 return (-1);
675 }
676
677 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV);
678 }
679
680 static void
681 trap_fatal(frame, eva)
682 struct trapframe *frame;
683 vm_offset_t eva;
684 {
685 int code, ss;
686 u_int type;
687 long esp;
688 struct soft_segment_descriptor softseg;
689 char *msg;
690
691 code = frame->tf_err;
692 type = frame->tf_trapno;
693 sdtossd(&gdt[NGDT * PCPU_GET(cpuid) + IDXSEL(frame->tf_cs & 0xffff)],
694 &softseg);
695
696 if (type <= MAX_TRAP_MSG)
697 msg = trap_msg[type];
698 else
699 msg = "UNKNOWN";
700 printf("\n\nFatal trap %d: %s while in %s mode\n", type, msg,
701 ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel");
702 #ifdef SMP
703 /* two separate prints in case of a trap on an unmapped page */
704 printf("cpuid = %d; ", PCPU_GET(cpuid));
705 printf("apic id = %02x\n", PCPU_GET(apic_id));
706 #endif
707 if (type == T_PAGEFLT) {
708 printf("fault virtual address = 0x%lx\n", eva);
709 printf("fault code = %s %s %s, %s\n",
710 code & PGEX_U ? "user" : "supervisor",
711 code & PGEX_W ? "write" : "read",
712 code & PGEX_I ? "instruction" : "data",
713 code & PGEX_P ? "protection violation" : "page not present");
714 }
715 printf("instruction pointer = 0x%lx:0x%lx\n",
716 frame->tf_cs & 0xffff, frame->tf_rip);
717 if (ISPL(frame->tf_cs) == SEL_UPL) {
718 ss = frame->tf_ss & 0xffff;
719 esp = frame->tf_rsp;
720 } else {
721 ss = GSEL(GDATA_SEL, SEL_KPL);
722 esp = (long)&frame->tf_rsp;
723 }
724 printf("stack pointer = 0x%x:0x%lx\n", ss, esp);
725 printf("frame pointer = 0x%x:0x%lx\n", ss, frame->tf_rbp);
726 printf("code segment = base 0x%lx, limit 0x%lx, type 0x%x\n",
727 softseg.ssd_base, softseg.ssd_limit, softseg.ssd_type);
728 printf(" = DPL %d, pres %d, long %d, def32 %d, gran %d\n",
729 softseg.ssd_dpl, softseg.ssd_p, softseg.ssd_long, softseg.ssd_def32,
730 softseg.ssd_gran);
731 printf("processor eflags = ");
732 if (frame->tf_rflags & PSL_T)
733 printf("trace trap, ");
734 if (frame->tf_rflags & PSL_I)
735 printf("interrupt enabled, ");
736 if (frame->tf_rflags & PSL_NT)
737 printf("nested task, ");
738 if (frame->tf_rflags & PSL_RF)
739 printf("resume, ");
740 printf("IOPL = %ld\n", (frame->tf_rflags & PSL_IOPL) >> 12);
741 printf("current process = ");
742 if (curproc) {
743 printf("%lu (%s)\n",
744 (u_long)curproc->p_pid, curproc->p_comm ?
745 curproc->p_comm : "");
746 } else {
747 printf("Idle\n");
748 }
749
750 #ifdef KDB
751 if (debugger_on_panic || kdb_active)
752 if (kdb_trap(type, 0, frame))
753 return;
754 #endif
755 printf("trap number = %d\n", type);
756 if (type <= MAX_TRAP_MSG)
757 panic("%s", trap_msg[type]);
758 else
759 panic("unknown/reserved trap");
760 }
761
762 /*
763 * Double fault handler. Called when a fault occurs while writing
764 * a frame for a trap/exception onto the stack. This usually occurs
765 * when the stack overflows (such is the case with infinite recursion,
766 * for example).
767 */
768 void
769 dblfault_handler(struct trapframe *frame)
770 {
771 #ifdef KDTRACE_HOOKS
772 if (dtrace_doubletrap_func != NULL)
773 (*dtrace_doubletrap_func)();
774 #endif
775 printf("\nFatal double fault\n");
776 printf("rip = 0x%lx\n", frame->tf_rip);
777 printf("rsp = 0x%lx\n", frame->tf_rsp);
778 printf("rbp = 0x%lx\n", frame->tf_rbp);
779 #ifdef SMP
780 /* two separate prints in case of a trap on an unmapped page */
781 printf("cpuid = %d; ", PCPU_GET(cpuid));
782 printf("apic id = %02x\n", PCPU_GET(apic_id));
783 #endif
784 panic("double fault");
785 }
786
787 /*
788 * syscall - system call request C handler
789 *
790 * A system call is essentially treated as a trap.
791 */
792 void
793 syscall(struct trapframe *frame)
794 {
795 caddr_t params;
796 struct sysent *callp;
797 struct thread *td = curthread;
798 struct proc *p = td->td_proc;
799 register_t orig_tf_rflags;
800 int error;
801 int narg;
802 register_t args[8];
803 register_t *argp;
804 u_int code;
805 int reg, regcnt;
806 ksiginfo_t ksi;
807
808 PCPU_INC(cnt.v_syscall);
809
810 #ifdef DIAGNOSTIC
811 if (ISPL(frame->tf_cs) != SEL_UPL) {
812 panic("syscall");
813 /* NOT REACHED */
814 }
815 #endif
816
817 reg = 0;
818 regcnt = 6;
819 td->td_pticks = 0;
820 td->td_frame = frame;
821 if (td->td_ucred != p->p_ucred)
822 cred_update_thread(td);
823 #ifdef KSE
824 if (p->p_flag & P_SA)
825 thread_user_enter(td);
826 #endif
827 params = (caddr_t)frame->tf_rsp + sizeof(register_t);
828 code = frame->tf_rax;
829 orig_tf_rflags = frame->tf_rflags;
830
831 if (p->p_sysent->sv_prepsyscall) {
832 /*
833 * The prep code is MP aware.
834 */
835 (*p->p_sysent->sv_prepsyscall)(frame, (int *)args, &code, ¶ms);
836 } else {
837 if (code == SYS_syscall || code == SYS___syscall) {
838 code = frame->tf_rdi;
839 reg++;
840 regcnt--;
841 }
842 }
843
844 if (p->p_sysent->sv_mask)
845 code &= p->p_sysent->sv_mask;
846
847 if (code >= p->p_sysent->sv_size)
848 callp = &p->p_sysent->sv_table[0];
849 else
850 callp = &p->p_sysent->sv_table[code];
851
852 narg = callp->sy_narg;
853
854 /*
855 * copyin and the ktrsyscall()/ktrsysret() code is MP-aware
856 */
857 KASSERT(narg <= sizeof(args) / sizeof(args[0]),
858 ("Too many syscall arguments!"));
859 error = 0;
860 argp = &frame->tf_rdi;
861 argp += reg;
862 bcopy(argp, args, sizeof(args[0]) * regcnt);
863 if (narg > regcnt) {
864 KASSERT(params != NULL, ("copyin args with no params!"));
865 error = copyin(params, &args[regcnt],
866 (narg - regcnt) * sizeof(args[0]));
867 }
868 argp = &args[0];
869
870 #ifdef KTRACE
871 if (KTRPOINT(td, KTR_SYSCALL))
872 ktrsyscall(code, narg, argp);
873 #endif
874
875 CTR4(KTR_SYSC, "syscall enter thread %p pid %d proc %s code %d", td,
876 td->td_proc->p_pid, td->td_proc->p_comm, code);
877
878 td->td_syscalls++;
879
880 if (error == 0) {
881 td->td_retval[0] = 0;
882 td->td_retval[1] = frame->tf_rdx;
883
884 STOPEVENT(p, S_SCE, narg);
885
886 PTRACESTOP_SC(p, td, S_PT_SCE);
887
888 #ifdef KDTRACE_HOOKS
889 /*
890 * If the systrace module has registered it's probe
891 * callback and if there is a probe active for the
892 * syscall 'entry', process the probe.
893 */
894 if (systrace_probe_func != NULL && callp->sy_entry != 0)
895 (*systrace_probe_func)(callp->sy_entry, code, callp,
896 args);
897 #endif
898
899 AUDIT_SYSCALL_ENTER(code, td);
900 error = (*callp->sy_call)(td, argp);
901 AUDIT_SYSCALL_EXIT(error, td);
902
903 /* Save the latest error return value. */
904 td->td_errno = error;
905
906 #ifdef KDTRACE_HOOKS
907 /*
908 * If the systrace module has registered it's probe
909 * callback and if there is a probe active for the
910 * syscall 'return', process the probe.
911 */
912 if (systrace_probe_func != NULL && callp->sy_return != 0)
913 (*systrace_probe_func)(callp->sy_return, code, callp,
914 args);
915 #endif
916 }
917
918 switch (error) {
919 case 0:
920 frame->tf_rax = td->td_retval[0];
921 frame->tf_rdx = td->td_retval[1];
922 frame->tf_rflags &= ~PSL_C;
923 break;
924
925 case ERESTART:
926 /*
927 * Reconstruct pc, we know that 'syscall' is 2 bytes.
928 * We have to do a full context restore so that %r10
929 * (which was holding the value of %rcx) is restored for
930 * the next iteration.
931 */
932 frame->tf_rip -= frame->tf_err;
933 frame->tf_r10 = frame->tf_rcx;
934 td->td_pcb->pcb_flags |= PCB_FULLCTX;
935 break;
936
937 case EJUSTRETURN:
938 break;
939
940 default:
941 if (p->p_sysent->sv_errsize) {
942 if (error >= p->p_sysent->sv_errsize)
943 error = -1; /* XXX */
944 else
945 error = p->p_sysent->sv_errtbl[error];
946 }
947 frame->tf_rax = error;
948 frame->tf_rflags |= PSL_C;
949 break;
950 }
951
952 /*
953 * Traced syscall.
954 */
955 if (orig_tf_rflags & PSL_T) {
956 frame->tf_rflags &= ~PSL_T;
957 ksiginfo_init_trap(&ksi);
958 ksi.ksi_signo = SIGTRAP;
959 ksi.ksi_code = TRAP_TRACE;
960 ksi.ksi_addr = (void *)frame->tf_rip;
961 trapsignal(td, &ksi);
962 }
963
964 /*
965 * Check for misbehavior.
966 */
967 WITNESS_WARN(WARN_PANIC, NULL, "System call %s returning",
968 (code >= 0 && code < SYS_MAXSYSCALL) ? syscallnames[code] : "???");
969 KASSERT(td->td_critnest == 0,
970 ("System call %s returning in a critical section",
971 (code >= 0 && code < SYS_MAXSYSCALL) ? syscallnames[code] : "???"));
972 KASSERT(td->td_locks == 0,
973 ("System call %s returning with %d locks held",
974 (code >= 0 && code < SYS_MAXSYSCALL) ? syscallnames[code] : "???",
975 td->td_locks));
976
977 /*
978 * Handle reschedule and other end-of-syscall issues
979 */
980 userret(td, frame);
981
982 CTR4(KTR_SYSC, "syscall exit thread %p pid %d proc %s code %d", td,
983 td->td_proc->p_pid, td->td_proc->p_comm, code);
984
985 #ifdef KTRACE
986 if (KTRPOINT(td, KTR_SYSRET))
987 ktrsysret(code, error, td->td_retval[0]);
988 #endif
989
990 /*
991 * This works because errno is findable through the
992 * register set. If we ever support an emulation where this
993 * is not the case, this code will need to be revisited.
994 */
995 STOPEVENT(p, S_SCX, code);
996
997 PTRACESTOP_SC(p, td, S_PT_SCX);
998 }
Cache object: aa98e49d212122ac3270cad774a3e9cb
|