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/8.3/sys/amd64/amd64/trap.c 236953 2012-06-12 12:10:10Z bz $");
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
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/mca.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
117 /*
118 * These hooks are necessary for the pid, usdt and fasttrap providers.
119 */
120 dtrace_fasttrap_probe_ptr_t dtrace_fasttrap_probe_ptr;
121 dtrace_pid_probe_ptr_t dtrace_pid_probe_ptr;
122 dtrace_return_probe_ptr_t dtrace_return_probe_ptr;
123 #endif
124
125 extern void trap(struct trapframe *frame);
126 extern void syscall(struct trapframe *frame);
127 void dblfault_handler(struct trapframe *frame);
128
129 static int trap_pfault(struct trapframe *, int);
130 static void trap_fatal(struct trapframe *, vm_offset_t);
131
132 #define MAX_TRAP_MSG 33
133 static char *trap_msg[] = {
134 "", /* 0 unused */
135 "privileged instruction fault", /* 1 T_PRIVINFLT */
136 "", /* 2 unused */
137 "breakpoint instruction fault", /* 3 T_BPTFLT */
138 "", /* 4 unused */
139 "", /* 5 unused */
140 "arithmetic trap", /* 6 T_ARITHTRAP */
141 "", /* 7 unused */
142 "", /* 8 unused */
143 "general protection fault", /* 9 T_PROTFLT */
144 "trace trap", /* 10 T_TRCTRAP */
145 "", /* 11 unused */
146 "page fault", /* 12 T_PAGEFLT */
147 "", /* 13 unused */
148 "alignment fault", /* 14 T_ALIGNFLT */
149 "", /* 15 unused */
150 "", /* 16 unused */
151 "", /* 17 unused */
152 "integer divide fault", /* 18 T_DIVIDE */
153 "non-maskable interrupt trap", /* 19 T_NMI */
154 "overflow trap", /* 20 T_OFLOW */
155 "FPU bounds check fault", /* 21 T_BOUND */
156 "FPU device not available", /* 22 T_DNA */
157 "double fault", /* 23 T_DOUBLEFLT */
158 "FPU operand fetch fault", /* 24 T_FPOPFLT */
159 "invalid TSS fault", /* 25 T_TSSFLT */
160 "segment not present fault", /* 26 T_SEGNPFLT */
161 "stack fault", /* 27 T_STKFLT */
162 "machine check trap", /* 28 T_MCHK */
163 "SIMD floating-point exception", /* 29 T_XMMFLT */
164 "reserved (unknown) fault", /* 30 T_RESERVED */
165 "", /* 31 unused (reserved) */
166 "DTrace pid return trap", /* 32 T_DTRACE_RET */
167 "DTrace fasttrap probe trap", /* 33 T_DTRACE_PROBE */
168 };
169
170 #ifdef KDB
171 static int kdb_on_nmi = 1;
172 SYSCTL_INT(_machdep, OID_AUTO, kdb_on_nmi, CTLFLAG_RW,
173 &kdb_on_nmi, 0, "Go to KDB on NMI");
174 TUNABLE_INT("machdep.kdb_on_nmi", &kdb_on_nmi);
175 #endif
176 static int panic_on_nmi = 1;
177 SYSCTL_INT(_machdep, OID_AUTO, panic_on_nmi, CTLFLAG_RW,
178 &panic_on_nmi, 0, "Panic on NMI");
179 TUNABLE_INT("machdep.panic_on_nmi", &panic_on_nmi);
180 static int prot_fault_translation = 0;
181 SYSCTL_INT(_machdep, OID_AUTO, prot_fault_translation, CTLFLAG_RW,
182 &prot_fault_translation, 0, "Select signal to deliver on protection fault");
183
184 /*
185 * Exception, fault, and trap interface to the FreeBSD kernel.
186 * This common code is called from assembly language IDT gate entry
187 * routines that prepare a suitable stack frame, and restore this
188 * frame after the exception has been processed.
189 */
190
191 void
192 trap(struct trapframe *frame)
193 {
194 struct thread *td = curthread;
195 struct proc *p = td->td_proc;
196 int i = 0, ucode = 0, code;
197 u_int type;
198 register_t addr = 0;
199 ksiginfo_t ksi;
200
201 PCPU_INC(cnt.v_trap);
202 type = frame->tf_trapno;
203
204 #ifdef SMP
205 /* Handler for NMI IPIs used for stopping CPUs. */
206 if (type == T_NMI) {
207 if (ipi_nmi_handler() == 0)
208 goto out;
209 }
210 #endif /* SMP */
211
212 #ifdef KDB
213 if (kdb_active) {
214 kdb_reenter();
215 goto out;
216 }
217 #endif
218
219 if (type == T_RESERVED) {
220 trap_fatal(frame, 0);
221 goto out;
222 }
223
224 #ifdef HWPMC_HOOKS
225 /*
226 * CPU PMCs interrupt using an NMI. If the PMC module is
227 * active, pass the 'rip' value to the PMC module's interrupt
228 * handler. A return value of '1' from the handler means that
229 * the NMI was handled by it and we can return immediately.
230 */
231 if (type == T_NMI && pmc_intr &&
232 (*pmc_intr)(PCPU_GET(cpuid), frame))
233 goto out;
234 #endif
235
236 if (type == T_MCHK) {
237 if (!mca_intr())
238 trap_fatal(frame, 0);
239 goto out;
240 }
241
242 #ifdef KDTRACE_HOOKS
243 /*
244 * A trap can occur while DTrace executes a probe. Before
245 * executing the probe, DTrace blocks re-scheduling and sets
246 * a flag in it's per-cpu flags to indicate that it doesn't
247 * want to fault. On returning from the probe, the no-fault
248 * flag is cleared and finally re-scheduling is enabled.
249 *
250 * If the DTrace kernel module has registered a trap handler,
251 * call it and if it returns non-zero, assume that it has
252 * handled the trap and modified the trap frame so that this
253 * function can return normally.
254 */
255 if (type == T_DTRACE_PROBE || type == T_DTRACE_RET ||
256 type == T_BPTFLT) {
257 struct reg regs;
258
259 fill_frame_regs(frame, ®s);
260 if (type == T_DTRACE_PROBE &&
261 dtrace_fasttrap_probe_ptr != NULL &&
262 dtrace_fasttrap_probe_ptr(®s) == 0)
263 goto out;
264 else if (type == T_BPTFLT &&
265 dtrace_pid_probe_ptr != NULL &&
266 dtrace_pid_probe_ptr(®s) == 0)
267 goto out;
268 else if (type == T_DTRACE_RET &&
269 dtrace_return_probe_ptr != NULL &&
270 dtrace_return_probe_ptr(®s) == 0)
271 goto out;
272 }
273 if (dtrace_trap_func != NULL && (*dtrace_trap_func)(frame, type))
274 goto out;
275 #endif
276
277 if ((frame->tf_rflags & PSL_I) == 0) {
278 /*
279 * Buggy application or kernel code has disabled
280 * interrupts and then trapped. Enabling interrupts
281 * now is wrong, but it is better than running with
282 * interrupts disabled until they are accidentally
283 * enabled later.
284 */
285 if (ISPL(frame->tf_cs) == SEL_UPL)
286 uprintf(
287 "pid %ld (%s): trap %d with interrupts disabled\n",
288 (long)curproc->p_pid, curthread->td_name, type);
289 else if (type != T_NMI && type != T_BPTFLT &&
290 type != T_TRCTRAP) {
291 /*
292 * XXX not quite right, since this may be for a
293 * multiple fault in user mode.
294 */
295 printf("kernel trap %d with interrupts disabled\n",
296 type);
297
298 /*
299 * We shouldn't enable interrupts while holding a
300 * spin lock.
301 */
302 if (td->td_md.md_spinlock_count == 0)
303 enable_intr();
304 }
305 }
306
307 code = frame->tf_err;
308 if (type == T_PAGEFLT) {
309 /*
310 * If we get a page fault while in a critical section, then
311 * it is most likely a fatal kernel page fault. The kernel
312 * is already going to panic trying to get a sleep lock to
313 * do the VM lookup, so just consider it a fatal trap so the
314 * kernel can print out a useful trap message and even get
315 * to the debugger.
316 *
317 * If we get a page fault while holding a non-sleepable
318 * lock, then it is most likely a fatal kernel page fault.
319 * If WITNESS is enabled, then it's going to whine about
320 * bogus LORs with various VM locks, so just skip to the
321 * fatal trap handling directly.
322 */
323 if (td->td_critnest != 0 ||
324 WITNESS_CHECK(WARN_SLEEPOK | WARN_GIANTOK, NULL,
325 "Kernel page fault") != 0)
326 trap_fatal(frame, frame->tf_addr);
327 }
328
329 if (ISPL(frame->tf_cs) == SEL_UPL) {
330 /* user trap */
331
332 td->td_pticks = 0;
333 td->td_frame = frame;
334 addr = frame->tf_rip;
335 if (td->td_ucred != p->p_ucred)
336 cred_update_thread(td);
337
338 switch (type) {
339 case T_PRIVINFLT: /* privileged instruction fault */
340 i = SIGILL;
341 ucode = ILL_PRVOPC;
342 break;
343
344 case T_BPTFLT: /* bpt instruction fault */
345 case T_TRCTRAP: /* trace trap */
346 enable_intr();
347 frame->tf_rflags &= ~PSL_T;
348 i = SIGTRAP;
349 ucode = (type == T_TRCTRAP ? TRAP_TRACE : TRAP_BRKPT);
350 break;
351
352 case T_ARITHTRAP: /* arithmetic trap */
353 ucode = fputrap();
354 if (ucode == -1)
355 goto userout;
356 i = SIGFPE;
357 break;
358
359 case T_PROTFLT: /* general protection fault */
360 i = SIGBUS;
361 ucode = BUS_OBJERR;
362 break;
363 case T_STKFLT: /* stack fault */
364 case T_SEGNPFLT: /* segment not present fault */
365 i = SIGBUS;
366 ucode = BUS_ADRERR;
367 break;
368 case T_TSSFLT: /* invalid TSS fault */
369 i = SIGBUS;
370 ucode = BUS_OBJERR;
371 break;
372 case T_DOUBLEFLT: /* double fault */
373 default:
374 i = SIGBUS;
375 ucode = BUS_OBJERR;
376 break;
377
378 case T_PAGEFLT: /* page fault */
379 addr = frame->tf_addr;
380 i = trap_pfault(frame, TRUE);
381 if (i == -1)
382 goto userout;
383 if (i == 0)
384 goto user;
385
386 if (i == SIGSEGV)
387 ucode = SEGV_MAPERR;
388 else {
389 if (prot_fault_translation == 0) {
390 /*
391 * Autodetect.
392 * This check also covers the images
393 * without the ABI-tag ELF note.
394 */
395 if (SV_CURPROC_ABI() == SV_ABI_FREEBSD
396 && p->p_osrel >= P_OSREL_SIGSEGV) {
397 i = SIGSEGV;
398 ucode = SEGV_ACCERR;
399 } else {
400 i = SIGBUS;
401 ucode = BUS_PAGE_FAULT;
402 }
403 } else if (prot_fault_translation == 1) {
404 /*
405 * Always compat mode.
406 */
407 i = SIGBUS;
408 ucode = BUS_PAGE_FAULT;
409 } else {
410 /*
411 * Always SIGSEGV mode.
412 */
413 i = SIGSEGV;
414 ucode = SEGV_ACCERR;
415 }
416 }
417 break;
418
419 case T_DIVIDE: /* integer divide fault */
420 ucode = FPE_INTDIV;
421 i = SIGFPE;
422 break;
423
424 #ifdef DEV_ISA
425 case T_NMI:
426 /* machine/parity/power fail/"kitchen sink" faults */
427 if (isa_nmi(code) == 0) {
428 #ifdef KDB
429 /*
430 * NMI can be hooked up to a pushbutton
431 * for debugging.
432 */
433 if (kdb_on_nmi) {
434 printf ("NMI ... going to debugger\n");
435 kdb_trap(type, 0, frame);
436 }
437 #endif /* KDB */
438 goto userout;
439 } else if (panic_on_nmi)
440 panic("NMI indicates hardware failure");
441 break;
442 #endif /* DEV_ISA */
443
444 case T_OFLOW: /* integer overflow fault */
445 ucode = FPE_INTOVF;
446 i = SIGFPE;
447 break;
448
449 case T_BOUND: /* bounds check fault */
450 ucode = FPE_FLTSUB;
451 i = SIGFPE;
452 break;
453
454 case T_DNA:
455 /* transparent fault (due to context switch "late") */
456 KASSERT(PCB_USER_FPU(td->td_pcb),
457 ("kernel FPU ctx has leaked"));
458 fpudna();
459 goto userout;
460
461 case T_FPOPFLT: /* FPU operand fetch fault */
462 ucode = ILL_COPROC;
463 i = SIGILL;
464 break;
465
466 case T_XMMFLT: /* SIMD floating-point exception */
467 ucode = 0; /* XXX */
468 i = SIGFPE;
469 break;
470 }
471 } else {
472 /* kernel trap */
473
474 KASSERT(cold || td->td_ucred != NULL,
475 ("kernel trap doesn't have ucred"));
476 switch (type) {
477 case T_PAGEFLT: /* page fault */
478 (void) trap_pfault(frame, FALSE);
479 goto out;
480
481 case T_DNA:
482 KASSERT(!PCB_USER_FPU(td->td_pcb),
483 ("Unregistered use of FPU in kernel"));
484 fpudna();
485 goto out;
486
487 case T_ARITHTRAP: /* arithmetic trap */
488 case T_XMMFLT: /* SIMD floating-point exception */
489 case T_FPOPFLT: /* FPU operand fetch fault */
490 /*
491 * XXXKIB for now disable any FPU traps in kernel
492 * handler registration seems to be overkill
493 */
494 trap_fatal(frame, 0);
495 goto out;
496
497 case T_STKFLT: /* stack fault */
498 break;
499
500 case T_PROTFLT: /* general protection fault */
501 case T_SEGNPFLT: /* segment not present fault */
502 if (td->td_intr_nesting_level != 0)
503 break;
504
505 /*
506 * Invalid segment selectors and out of bounds
507 * %rip's and %rsp's can be set up in user mode.
508 * This causes a fault in kernel mode when the
509 * kernel tries to return to user mode. We want
510 * to get this fault so that we can fix the
511 * problem here and not have to check all the
512 * selectors and pointers when the user changes
513 * them.
514 */
515 if (frame->tf_rip == (long)doreti_iret) {
516 frame->tf_rip = (long)doreti_iret_fault;
517 goto out;
518 }
519 if (frame->tf_rip == (long)ld_ds) {
520 frame->tf_rip = (long)ds_load_fault;
521 goto out;
522 }
523 if (frame->tf_rip == (long)ld_es) {
524 frame->tf_rip = (long)es_load_fault;
525 goto out;
526 }
527 if (frame->tf_rip == (long)ld_fs) {
528 frame->tf_rip = (long)fs_load_fault;
529 goto out;
530 }
531 if (frame->tf_rip == (long)ld_gs) {
532 frame->tf_rip = (long)gs_load_fault;
533 goto out;
534 }
535 if (frame->tf_rip == (long)ld_gsbase) {
536 frame->tf_rip = (long)gsbase_load_fault;
537 goto out;
538 }
539 if (frame->tf_rip == (long)ld_fsbase) {
540 frame->tf_rip = (long)fsbase_load_fault;
541 goto out;
542 }
543 if (PCPU_GET(curpcb)->pcb_onfault != NULL) {
544 frame->tf_rip =
545 (long)PCPU_GET(curpcb)->pcb_onfault;
546 goto out;
547 }
548 break;
549
550 case T_TSSFLT:
551 /*
552 * PSL_NT can be set in user mode and isn't cleared
553 * automatically when the kernel is entered. This
554 * causes a TSS fault when the kernel attempts to
555 * `iret' because the TSS link is uninitialized. We
556 * want to get this fault so that we can fix the
557 * problem here and not every time the kernel is
558 * entered.
559 */
560 if (frame->tf_rflags & PSL_NT) {
561 frame->tf_rflags &= ~PSL_NT;
562 goto out;
563 }
564 break;
565
566 case T_TRCTRAP: /* trace trap */
567 /*
568 * Ignore debug register trace traps due to
569 * accesses in the user's address space, which
570 * can happen under several conditions such as
571 * if a user sets a watchpoint on a buffer and
572 * then passes that buffer to a system call.
573 * We still want to get TRCTRAPS for addresses
574 * in kernel space because that is useful when
575 * debugging the kernel.
576 */
577 if (user_dbreg_trap()) {
578 /*
579 * Reset breakpoint bits because the
580 * processor doesn't
581 */
582 /* XXX check upper bits here */
583 load_dr6(rdr6() & 0xfffffff0);
584 goto out;
585 }
586 /*
587 * FALLTHROUGH (TRCTRAP kernel mode, kernel address)
588 */
589 case T_BPTFLT:
590 /*
591 * If KDB is enabled, let it handle the debugger trap.
592 * Otherwise, debugger traps "can't happen".
593 */
594 #ifdef KDB
595 if (kdb_trap(type, 0, frame))
596 goto out;
597 #endif
598 break;
599
600 #ifdef DEV_ISA
601 case T_NMI:
602 /* machine/parity/power fail/"kitchen sink" faults */
603 if (isa_nmi(code) == 0) {
604 #ifdef KDB
605 /*
606 * NMI can be hooked up to a pushbutton
607 * for debugging.
608 */
609 if (kdb_on_nmi) {
610 printf ("NMI ... going to debugger\n");
611 kdb_trap(type, 0, frame);
612 }
613 #endif /* KDB */
614 goto out;
615 } else if (panic_on_nmi == 0)
616 goto out;
617 /* FALLTHROUGH */
618 #endif /* DEV_ISA */
619 }
620
621 trap_fatal(frame, 0);
622 goto out;
623 }
624
625 /* Translate fault for emulators (e.g. Linux) */
626 if (*p->p_sysent->sv_transtrap)
627 i = (*p->p_sysent->sv_transtrap)(i, type);
628
629 ksiginfo_init_trap(&ksi);
630 ksi.ksi_signo = i;
631 ksi.ksi_code = ucode;
632 ksi.ksi_trapno = type;
633 ksi.ksi_addr = (void *)addr;
634 trapsignal(td, &ksi);
635
636 user:
637 userret(td, frame);
638 mtx_assert(&Giant, MA_NOTOWNED);
639 KASSERT(PCB_USER_FPU(td->td_pcb),
640 ("Return from trap with kernel FPU ctx leaked"));
641 userout:
642 out:
643 return;
644 }
645
646 static int
647 trap_pfault(frame, usermode)
648 struct trapframe *frame;
649 int usermode;
650 {
651 vm_offset_t va;
652 struct vmspace *vm = NULL;
653 vm_map_t map;
654 int rv = 0;
655 vm_prot_t ftype;
656 struct thread *td = curthread;
657 struct proc *p = td->td_proc;
658 vm_offset_t eva = frame->tf_addr;
659
660 va = trunc_page(eva);
661 if (va >= VM_MIN_KERNEL_ADDRESS) {
662 /*
663 * Don't allow user-mode faults in kernel address space.
664 */
665 if (usermode)
666 goto nogo;
667
668 map = kernel_map;
669 } else {
670 /*
671 * This is a fault on non-kernel virtual memory.
672 * vm is initialized above to NULL. If curproc is NULL
673 * or curproc->p_vmspace is NULL the fault is fatal.
674 */
675 if (p != NULL)
676 vm = p->p_vmspace;
677
678 if (vm == NULL)
679 goto nogo;
680
681 map = &vm->vm_map;
682
683 /*
684 * When accessing a usermode address, kernel must be
685 * ready to accept the page fault, and provide a
686 * handling routine. Since accessing the address
687 * without the handler is a bug, do not try to handle
688 * it normally, and panic immediately.
689 */
690 if (!usermode && (td->td_intr_nesting_level != 0 ||
691 PCPU_GET(curpcb)->pcb_onfault == NULL)) {
692 trap_fatal(frame, eva);
693 return (-1);
694 }
695 }
696
697 /*
698 * PGEX_I is defined only if the execute disable bit capability is
699 * supported and enabled.
700 */
701 if (frame->tf_err & PGEX_W)
702 ftype = VM_PROT_WRITE;
703 else if ((frame->tf_err & PGEX_I) && pg_nx != 0)
704 ftype = VM_PROT_EXECUTE;
705 else
706 ftype = VM_PROT_READ;
707
708 if (map != kernel_map) {
709 /*
710 * Keep swapout from messing with us during this
711 * critical time.
712 */
713 PROC_LOCK(p);
714 ++p->p_lock;
715 PROC_UNLOCK(p);
716
717 /* Fault in the user page: */
718 rv = vm_fault(map, va, ftype,
719 (ftype & VM_PROT_WRITE) ? VM_FAULT_DIRTY
720 : VM_FAULT_NORMAL);
721
722 PROC_LOCK(p);
723 --p->p_lock;
724 PROC_UNLOCK(p);
725 } else {
726 /*
727 * Don't have to worry about process locking or stacks in the
728 * kernel.
729 */
730 rv = vm_fault(map, va, ftype, VM_FAULT_NORMAL);
731 }
732 if (rv == KERN_SUCCESS)
733 return (0);
734 nogo:
735 if (!usermode) {
736 if (td->td_intr_nesting_level == 0 &&
737 PCPU_GET(curpcb)->pcb_onfault != NULL) {
738 frame->tf_rip = (long)PCPU_GET(curpcb)->pcb_onfault;
739 return (0);
740 }
741 trap_fatal(frame, eva);
742 return (-1);
743 }
744
745 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV);
746 }
747
748 static void
749 trap_fatal(frame, eva)
750 struct trapframe *frame;
751 vm_offset_t eva;
752 {
753 int code, ss;
754 u_int type;
755 long esp;
756 struct soft_segment_descriptor softseg;
757 char *msg;
758
759 code = frame->tf_err;
760 type = frame->tf_trapno;
761 sdtossd(&gdt[NGDT * PCPU_GET(cpuid) + IDXSEL(frame->tf_cs & 0xffff)],
762 &softseg);
763
764 if (type <= MAX_TRAP_MSG)
765 msg = trap_msg[type];
766 else
767 msg = "UNKNOWN";
768 printf("\n\nFatal trap %d: %s while in %s mode\n", type, msg,
769 ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel");
770 #ifdef SMP
771 /* two separate prints in case of a trap on an unmapped page */
772 printf("cpuid = %d; ", PCPU_GET(cpuid));
773 printf("apic id = %02x\n", PCPU_GET(apic_id));
774 #endif
775 if (type == T_PAGEFLT) {
776 printf("fault virtual address = 0x%lx\n", eva);
777 printf("fault code = %s %s %s, %s\n",
778 code & PGEX_U ? "user" : "supervisor",
779 code & PGEX_W ? "write" : "read",
780 code & PGEX_I ? "instruction" : "data",
781 code & PGEX_P ? "protection violation" : "page not present");
782 }
783 printf("instruction pointer = 0x%lx:0x%lx\n",
784 frame->tf_cs & 0xffff, frame->tf_rip);
785 if (ISPL(frame->tf_cs) == SEL_UPL) {
786 ss = frame->tf_ss & 0xffff;
787 esp = frame->tf_rsp;
788 } else {
789 ss = GSEL(GDATA_SEL, SEL_KPL);
790 esp = (long)&frame->tf_rsp;
791 }
792 printf("stack pointer = 0x%x:0x%lx\n", ss, esp);
793 printf("frame pointer = 0x%x:0x%lx\n", ss, frame->tf_rbp);
794 printf("code segment = base 0x%lx, limit 0x%lx, type 0x%x\n",
795 softseg.ssd_base, softseg.ssd_limit, softseg.ssd_type);
796 printf(" = DPL %d, pres %d, long %d, def32 %d, gran %d\n",
797 softseg.ssd_dpl, softseg.ssd_p, softseg.ssd_long, softseg.ssd_def32,
798 softseg.ssd_gran);
799 printf("processor eflags = ");
800 if (frame->tf_rflags & PSL_T)
801 printf("trace trap, ");
802 if (frame->tf_rflags & PSL_I)
803 printf("interrupt enabled, ");
804 if (frame->tf_rflags & PSL_NT)
805 printf("nested task, ");
806 if (frame->tf_rflags & PSL_RF)
807 printf("resume, ");
808 printf("IOPL = %ld\n", (frame->tf_rflags & PSL_IOPL) >> 12);
809 printf("current process = ");
810 if (curproc) {
811 printf("%lu (%s)\n",
812 (u_long)curproc->p_pid, curthread->td_name ?
813 curthread->td_name : "");
814 } else {
815 printf("Idle\n");
816 }
817
818 #ifdef KDB
819 if (debugger_on_panic || kdb_active)
820 if (kdb_trap(type, 0, frame))
821 return;
822 #endif
823 printf("trap number = %d\n", type);
824 if (type <= MAX_TRAP_MSG)
825 panic("%s", trap_msg[type]);
826 else
827 panic("unknown/reserved trap");
828 }
829
830 /*
831 * Double fault handler. Called when a fault occurs while writing
832 * a frame for a trap/exception onto the stack. This usually occurs
833 * when the stack overflows (such is the case with infinite recursion,
834 * for example).
835 */
836 void
837 dblfault_handler(struct trapframe *frame)
838 {
839 #ifdef KDTRACE_HOOKS
840 if (dtrace_doubletrap_func != NULL)
841 (*dtrace_doubletrap_func)();
842 #endif
843 printf("\nFatal double fault\n");
844 printf("rip = 0x%lx\n", frame->tf_rip);
845 printf("rsp = 0x%lx\n", frame->tf_rsp);
846 printf("rbp = 0x%lx\n", frame->tf_rbp);
847 #ifdef SMP
848 /* two separate prints in case of a trap on an unmapped page */
849 printf("cpuid = %d; ", PCPU_GET(cpuid));
850 printf("apic id = %02x\n", PCPU_GET(apic_id));
851 #endif
852 panic("double fault");
853 }
854
855 int
856 cpu_fetch_syscall_args(struct thread *td, struct syscall_args *sa)
857 {
858 struct proc *p;
859 struct trapframe *frame;
860 register_t *argp;
861 caddr_t params;
862 int reg, regcnt, error;
863
864 p = td->td_proc;
865 frame = td->td_frame;
866 reg = 0;
867 regcnt = 6;
868
869 params = (caddr_t)frame->tf_rsp + sizeof(register_t);
870 sa->code = frame->tf_rax;
871
872 if (sa->code == SYS_syscall || sa->code == SYS___syscall) {
873 sa->code = frame->tf_rdi;
874 reg++;
875 regcnt--;
876 }
877 if (p->p_sysent->sv_mask)
878 sa->code &= p->p_sysent->sv_mask;
879
880 if (sa->code >= p->p_sysent->sv_size)
881 sa->callp = &p->p_sysent->sv_table[0];
882 else
883 sa->callp = &p->p_sysent->sv_table[sa->code];
884
885 sa->narg = sa->callp->sy_narg;
886 KASSERT(sa->narg <= sizeof(sa->args) / sizeof(sa->args[0]),
887 ("Too many syscall arguments!"));
888 error = 0;
889 argp = &frame->tf_rdi;
890 argp += reg;
891 bcopy(argp, sa->args, sizeof(sa->args[0]) * regcnt);
892 if (sa->narg > regcnt) {
893 KASSERT(params != NULL, ("copyin args with no params!"));
894 error = copyin(params, &sa->args[regcnt],
895 (sa->narg - regcnt) * sizeof(sa->args[0]));
896 }
897
898 if (error == 0) {
899 td->td_retval[0] = 0;
900 td->td_retval[1] = frame->tf_rdx;
901 }
902
903 return (error);
904 }
905
906 #include "../../kern/subr_syscall.c"
907
908 /*
909 * syscall - system call request C handler
910 *
911 * A system call is essentially treated as a trap.
912 */
913 void
914 amd64_syscall(struct thread *td, int traced)
915 {
916 struct syscall_args sa;
917 int error;
918 ksiginfo_t ksi;
919
920 #ifdef DIAGNOSTIC
921 if (ISPL(td->td_frame->tf_cs) != SEL_UPL) {
922 panic("syscall");
923 /* NOT REACHED */
924 }
925 #endif
926 error = syscallenter(td, &sa);
927
928 /*
929 * Traced syscall.
930 */
931 if (__predict_false(traced)) {
932 td->td_frame->tf_rflags &= ~PSL_T;
933 ksiginfo_init_trap(&ksi);
934 ksi.ksi_signo = SIGTRAP;
935 ksi.ksi_code = TRAP_TRACE;
936 ksi.ksi_addr = (void *)td->td_frame->tf_rip;
937 trapsignal(td, &ksi);
938 }
939
940 KASSERT(PCB_USER_FPU(td->td_pcb),
941 ("System call %s returing with kernel FPU ctx leaked",
942 syscallname(td->td_proc, sa.code)));
943 KASSERT(td->td_pcb->pcb_save == &td->td_pcb->pcb_user_save,
944 ("System call %s returning with mangled pcb_save",
945 syscallname(td->td_proc, sa.code)));
946
947 syscallret(td, error, &sa);
948
949 /*
950 * If the user-supplied value of %rip is not a canonical
951 * address, then some CPUs will trigger a ring 0 #GP during
952 * the sysret instruction. However, the fault handler would
953 * execute with the user's %gs and %rsp in ring 0 which would
954 * not be safe. Instead, preemptively kill the thread with a
955 * SIGBUS.
956 */
957 if (td->td_frame->tf_rip >= VM_MAXUSER_ADDRESS) {
958 ksiginfo_init_trap(&ksi);
959 ksi.ksi_signo = SIGBUS;
960 ksi.ksi_code = BUS_OBJERR;
961 ksi.ksi_trapno = T_PROTFLT;
962 ksi.ksi_addr = (void *)td->td_frame->tf_rip;
963 trapsignal(td, &ksi);
964 }
965 }
Cache object: 9de28fa56a3eeb1622cbdadb7bcf8a22
|