1 | /* $NetBSD: trap.c,v 1.87 2016/10/26 22:02:14 christos Exp $ */ |
2 | |
3 | /*- |
4 | * Copyright (c) 1998, 2000 The NetBSD Foundation, Inc. |
5 | * All rights reserved. |
6 | * |
7 | * This code is derived from software contributed to The NetBSD Foundation |
8 | * by Charles M. Hannum. |
9 | * |
10 | * Redistribution and use in source and binary forms, with or without |
11 | * modification, are permitted provided that the following conditions |
12 | * are met: |
13 | * 1. Redistributions of source code must retain the above copyright |
14 | * notice, this list of conditions and the following disclaimer. |
15 | * 2. Redistributions in binary form must reproduce the above copyright |
16 | * notice, this list of conditions and the following disclaimer in the |
17 | * documentation and/or other materials provided with the distribution. |
18 | * |
19 | * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS |
20 | * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED |
21 | * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR |
22 | * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS |
23 | * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR |
24 | * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF |
25 | * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS |
26 | * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN |
27 | * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) |
28 | * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE |
29 | * POSSIBILITY OF SUCH DAMAGE. |
30 | */ |
31 | |
32 | /*- |
33 | * Copyright (c) 1990 The Regents of the University of California. |
34 | * All rights reserved. |
35 | * |
36 | * This code is derived from software contributed to Berkeley by |
37 | * the University of Utah, and William Jolitz. |
38 | * |
39 | * Redistribution and use in source and binary forms, with or without |
40 | * modification, are permitted provided that the following conditions |
41 | * are met: |
42 | * 1. Redistributions of source code must retain the above copyright |
43 | * notice, this list of conditions and the following disclaimer. |
44 | * 2. Redistributions in binary form must reproduce the above copyright |
45 | * notice, this list of conditions and the following disclaimer in the |
46 | * documentation and/or other materials provided with the distribution. |
47 | * 3. Neither the name of the University nor the names of its contributors |
48 | * may be used to endorse or promote products derived from this software |
49 | * without specific prior written permission. |
50 | * |
51 | * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND |
52 | * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
53 | * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
54 | * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE |
55 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
56 | * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
57 | * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
58 | * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
59 | * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
60 | * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
61 | * SUCH DAMAGE. |
62 | * |
63 | * @(#)trap.c 7.4 (Berkeley) 5/13/91 |
64 | */ |
65 | |
66 | /* |
67 | * 386 Trap and System call handling |
68 | */ |
69 | |
70 | #include <sys/cdefs.h> |
71 | __KERNEL_RCSID(0, "$NetBSD: trap.c,v 1.87 2016/10/26 22:02:14 christos Exp $" ); |
72 | |
73 | #include "opt_ddb.h" |
74 | #include "opt_kgdb.h" |
75 | #include "opt_xen.h" |
76 | #include "opt_dtrace.h" |
77 | |
78 | #include <sys/param.h> |
79 | #include <sys/systm.h> |
80 | #include <sys/proc.h> |
81 | #include <sys/acct.h> |
82 | #include <sys/kauth.h> |
83 | #include <sys/kernel.h> |
84 | #include <sys/kmem.h> |
85 | #include <sys/ras.h> |
86 | #include <sys/signal.h> |
87 | #include <sys/syscall.h> |
88 | #include <sys/cpu.h> |
89 | #include <sys/ucontext.h> |
90 | |
91 | #include <uvm/uvm_extern.h> |
92 | |
93 | #include <machine/cpufunc.h> |
94 | #include <x86/fpu.h> |
95 | #include <machine/psl.h> |
96 | #include <machine/reg.h> |
97 | #include <machine/trap.h> |
98 | #include <machine/userret.h> |
99 | #include <machine/db_machdep.h> |
100 | |
101 | #include <x86/nmi.h> |
102 | |
103 | #ifndef XEN |
104 | #include "isa.h" |
105 | #endif |
106 | |
107 | #include <sys/kgdb.h> |
108 | |
109 | #ifdef KDTRACE_HOOKS |
110 | #include <sys/dtrace_bsd.h> |
111 | |
112 | /* |
113 | * This is a hook which is initialized by the dtrace module |
114 | * to handle traps which might occur during DTrace probe |
115 | * execution. |
116 | */ |
117 | dtrace_trap_func_t dtrace_trap_func = NULL; |
118 | |
119 | dtrace_doubletrap_func_t dtrace_doubletrap_func = NULL; |
120 | #endif |
121 | |
122 | void trap(struct trapframe *); |
123 | void trap_return_fault_return(struct trapframe *) __dead; |
124 | |
125 | const char * const trap_type[] = { |
126 | "privileged instruction fault" , /* 0 T_PRIVINFLT */ |
127 | "breakpoint trap" , /* 1 T_BPTFLT */ |
128 | "arithmetic trap" , /* 2 T_ARITHTRAP */ |
129 | "asynchronous system trap" , /* 3 T_ASTFLT */ |
130 | "protection fault" , /* 4 T_PROTFLT */ |
131 | "trace trap" , /* 5 T_TRCTRAP */ |
132 | "page fault" , /* 6 T_PAGEFLT */ |
133 | "alignment fault" , /* 7 T_ALIGNFLT */ |
134 | "integer divide fault" , /* 8 T_DIVIDE */ |
135 | "non-maskable interrupt" , /* 9 T_NMI */ |
136 | "overflow trap" , /* 10 T_OFLOW */ |
137 | "bounds check fault" , /* 11 T_BOUND */ |
138 | "FPU not available fault" , /* 12 T_DNA */ |
139 | "double fault" , /* 13 T_DOUBLEFLT */ |
140 | "FPU operand fetch fault" , /* 14 T_FPOPFLT */ |
141 | "invalid TSS fault" , /* 15 T_TSSFLT */ |
142 | "segment not present fault" , /* 16 T_SEGNPFLT */ |
143 | "stack fault" , /* 17 T_STKFLT */ |
144 | "machine check fault" , /* 18 T_MCA */ |
145 | "SSE FP exception" , /* 19 T_XMM */ |
146 | "reserved trap" , /* 20 T_RESERVED */ |
147 | }; |
148 | int trap_types = __arraycount(trap_type); |
149 | |
150 | #ifdef DEBUG |
151 | int trapdebug = 0; |
152 | #endif |
153 | |
154 | #define IDTVEC(name) __CONCAT(X, name) |
155 | |
156 | #ifdef TRAP_SIGDEBUG |
157 | static void frame_dump(struct trapframe *); |
158 | #endif |
159 | |
160 | static void * |
161 | onfault_handler(const struct pcb *pcb, const struct trapframe *tf) |
162 | { |
163 | struct onfault_table { |
164 | uintptr_t start; |
165 | uintptr_t end; |
166 | void *handler; |
167 | }; |
168 | extern const struct onfault_table onfault_table[]; |
169 | const struct onfault_table *p; |
170 | uintptr_t pc; |
171 | |
172 | if (pcb->pcb_onfault != NULL) { |
173 | return pcb->pcb_onfault; |
174 | } |
175 | |
176 | pc = tf->tf_rip; |
177 | for (p = onfault_table; p->start; p++) { |
178 | if (p->start <= pc && pc < p->end) { |
179 | return p->handler; |
180 | } |
181 | } |
182 | return NULL; |
183 | } |
184 | |
185 | static void |
186 | trap_print(const struct trapframe *frame, const lwp_t *l) |
187 | { |
188 | const int type = frame->tf_trapno; |
189 | |
190 | if (frame->tf_trapno < trap_types) { |
191 | printf("fatal %s" , trap_type[type]); |
192 | } else { |
193 | printf("unknown trap %d" , type); |
194 | } |
195 | printf(" in %s mode\n" , (type & T_USER) ? "user" : "supervisor" ); |
196 | |
197 | printf("trap type %d code %lx rip %lx cs %lx rflags %lx cr2 %lx " |
198 | "ilevel %x rsp %lx\n" , |
199 | type, frame->tf_err, (u_long)frame->tf_rip, frame->tf_cs, |
200 | frame->tf_rflags, rcr2(), curcpu()->ci_ilevel, frame->tf_rsp); |
201 | |
202 | printf("curlwp %p pid %d.%d lowest kstack %p\n" , |
203 | l, l->l_proc->p_pid, l->l_lid, KSTACK_LOWEST_ADDR(l)); |
204 | } |
205 | |
206 | /* |
207 | * trap(frame): exception, fault, and trap interface to BSD kernel. |
208 | * |
209 | * This common code is called from assembly language IDT gate entry routines |
210 | * that prepare a suitable stack frame, and restore this frame after the |
211 | * exception has been processed. Note that the effect is as if the arguments |
212 | * were passed call by reference. |
213 | * |
214 | * Note that the fpu traps (07 T_DNA, 10 T_ARITHTRAP and 13 T_XMM) |
215 | * jump directly into the code in x86/fpu.c so they get processed |
216 | * without interrupts being enabled. |
217 | */ |
218 | void |
219 | trap(struct trapframe *frame) |
220 | { |
221 | struct lwp *l = curlwp; |
222 | struct proc *p; |
223 | struct pcb *pcb; |
224 | extern char fusuintrfailure[], kcopy_fault[]; |
225 | extern char IDTVEC(oosyscall)[]; |
226 | extern char IDTVEC(osyscall)[]; |
227 | extern char IDTVEC(syscall32)[]; |
228 | #ifndef XEN |
229 | struct trapframe *vframe; |
230 | #endif |
231 | ksiginfo_t ksi; |
232 | void *onfault; |
233 | int type, error; |
234 | uint64_t cr2; |
235 | bool pfail; |
236 | |
237 | if (__predict_true(l != NULL)) { |
238 | pcb = lwp_getpcb(l); |
239 | p = l->l_proc; |
240 | } else { |
241 | /* |
242 | * this can happen eg. on break points in early on boot. |
243 | */ |
244 | pcb = NULL; |
245 | p = NULL; |
246 | } |
247 | type = frame->tf_trapno; |
248 | |
249 | #ifdef DEBUG |
250 | if (trapdebug) { |
251 | trap_print(frame, l); |
252 | } |
253 | #endif |
254 | if (type != T_NMI && !KERNELMODE(frame->tf_cs, frame->tf_rflags)) { |
255 | type |= T_USER; |
256 | l->l_md.md_regs = frame; |
257 | LWP_CACHE_CREDS(l, p); |
258 | } |
259 | |
260 | #ifdef KDTRACE_HOOKS |
261 | /* |
262 | * A trap can occur while DTrace executes a probe. Before |
263 | * executing the probe, DTrace blocks re-scheduling and sets |
264 | * a flag in its per-cpu flags to indicate that it doesn't |
265 | * want to fault. On returning from the the probe, the no-fault |
266 | * flag is cleared and finally re-scheduling is enabled. |
267 | * |
268 | * If the DTrace kernel module has registered a trap handler, |
269 | * call it and if it returns non-zero, assume that it has |
270 | * handled the trap and modified the trap frame so that this |
271 | * function can return normally. |
272 | */ |
273 | if ((type == T_PROTFLT || type == T_PAGEFLT) && |
274 | dtrace_trap_func != NULL) { |
275 | if ((*dtrace_trap_func)(frame, type)) { |
276 | return; |
277 | } |
278 | } |
279 | #endif |
280 | |
281 | switch (type) { |
282 | |
283 | default: |
284 | we_re_toast: |
285 | trap_print(frame, l); |
286 | |
287 | if (kdb_trap(type, 0, frame)) |
288 | return; |
289 | if (kgdb_trap(type, frame)) |
290 | return; |
291 | /* |
292 | * If this is a breakpoint, don't panic if we're not connected. |
293 | */ |
294 | if (type == T_BPTFLT && kgdb_disconnected()) { |
295 | printf("kgdb: ignored %s\n" , trap_type[type]); |
296 | return; |
297 | } |
298 | panic("trap" ); |
299 | /*NOTREACHED*/ |
300 | |
301 | case T_PROTFLT: |
302 | case T_SEGNPFLT: |
303 | case T_ALIGNFLT: |
304 | case T_TSSFLT: |
305 | if (p == NULL) |
306 | goto we_re_toast; |
307 | /* Check for copyin/copyout fault. */ |
308 | onfault = onfault_handler(pcb, frame); |
309 | if (onfault != NULL) { |
310 | copyefault: |
311 | error = EFAULT; |
312 | copyfault: |
313 | frame->tf_rip = (uintptr_t)onfault; |
314 | frame->tf_rax = error; |
315 | return; |
316 | } |
317 | |
318 | /* |
319 | * Check for failure during return to user mode. |
320 | * This can happen loading invalid values into the segment |
321 | * registers, or during the 'iret' itself. |
322 | * |
323 | * We do this by looking at the instruction we faulted on. |
324 | * The specific instructions we recognize only happen when |
325 | * returning from a trap, syscall, or interrupt. |
326 | */ |
327 | |
328 | kernelfault: |
329 | #ifdef XEN |
330 | /* |
331 | * XXX: there has to be an equivalent 'problem' |
332 | * but I (dsl) don't know exactly what happens! |
333 | * For now panic the kernel. |
334 | */ |
335 | goto we_re_toast; |
336 | #else |
337 | KSI_INIT_TRAP(&ksi); |
338 | ksi.ksi_signo = SIGSEGV; |
339 | ksi.ksi_code = SEGV_ACCERR; |
340 | ksi.ksi_trap = type; |
341 | |
342 | /* Get %rsp value before fault - there may be a pad word |
343 | * below the trap frame. */ |
344 | vframe = (void *)frame->tf_rsp; |
345 | if (frame->tf_rip == 0) { |
346 | /* |
347 | * Assume that if we jumped to null we |
348 | * probably did it via a null function |
349 | * pointer, so print the return address. |
350 | */ |
351 | printf("kernel jumped to null; return addr was %p\n" , |
352 | *(void **)frame->tf_rsp); |
353 | goto we_re_toast; |
354 | } |
355 | switch (*(uint16_t *)frame->tf_rip) { |
356 | case 0xcf48: /* iretq */ |
357 | /* |
358 | * The 'iretq' instruction faulted, so we have the |
359 | * 'user' registers saved after the kernel |
360 | * %rip:%cs:%fl:%rsp:%ss of the iret, and below that |
361 | * the user %rip:%cs:%fl:%rsp:%ss the 'iret' was |
362 | * processing. |
363 | * We must copy the user register back over the |
364 | * kernel fault frame to generate a normal stack |
365 | * frame (eg for sending a SIGSEGV). |
366 | */ |
367 | vframe = (void *)((char *)vframe |
368 | - offsetof(struct trapframe, tf_rip)); |
369 | memmove(vframe, frame, |
370 | offsetof(struct trapframe, tf_rip)); |
371 | /* Set the faulting address to the user %eip */ |
372 | ksi.ksi_addr = (void *)vframe->tf_rip; |
373 | break; |
374 | case 0x848e: /* mov 0xa8(%rsp),%es (8e 84 24 a8 00 00 00) */ |
375 | case 0x9c8e: /* mov 0xb0(%rsp),%ds (8e 9c 24 b0 00 00 00) */ |
376 | /* |
377 | * We faulted loading one of the user segment registers. |
378 | * The stack frame containing the user registers is |
379 | * still valid and pointed to by tf_rsp. |
380 | * Maybe we should check the iretq follows. |
381 | */ |
382 | if (KERNELMODE(vframe->tf_cs, vframe->tf_eflags)) |
383 | goto we_re_toast; |
384 | /* There is no valid address for the fault */ |
385 | break; |
386 | |
387 | default: |
388 | goto we_re_toast; |
389 | } |
390 | |
391 | /* XXX: worry about on-stack trampolines for nested |
392 | * handlers?? */ |
393 | /* Save outer frame for any signal return */ |
394 | l->l_md.md_regs = vframe; |
395 | (*p->p_emul->e_trapsignal)(l, &ksi); |
396 | /* Return to user by reloading the user frame */ |
397 | trap_return_fault_return(vframe); |
398 | /* NOTREACHED */ |
399 | #endif |
400 | |
401 | case T_PROTFLT|T_USER: /* protection fault */ |
402 | case T_TSSFLT|T_USER: |
403 | case T_SEGNPFLT|T_USER: |
404 | case T_STKFLT|T_USER: |
405 | case T_ALIGNFLT|T_USER: |
406 | #ifdef TRAP_SIGDEBUG |
407 | printf("pid %d.%d (%s): BUS/SEGV (%x) at rip %lx addr %lx\n" , |
408 | p->p_pid, l->l_lid, p->p_comm, type, frame->tf_rip, rcr2()); |
409 | frame_dump(frame); |
410 | #endif |
411 | KSI_INIT_TRAP(&ksi); |
412 | ksi.ksi_trap = type & ~T_USER; |
413 | ksi.ksi_addr = (void *)rcr2(); |
414 | switch (type) { |
415 | case T_SEGNPFLT|T_USER: |
416 | case T_STKFLT|T_USER: |
417 | ksi.ksi_signo = SIGBUS; |
418 | ksi.ksi_code = BUS_ADRERR; |
419 | break; |
420 | case T_TSSFLT|T_USER: |
421 | ksi.ksi_signo = SIGBUS; |
422 | ksi.ksi_code = BUS_OBJERR; |
423 | break; |
424 | case T_ALIGNFLT|T_USER: |
425 | ksi.ksi_signo = SIGBUS; |
426 | ksi.ksi_code = BUS_ADRALN; |
427 | break; |
428 | case T_PROTFLT|T_USER: |
429 | ksi.ksi_signo = SIGSEGV; |
430 | ksi.ksi_code = SEGV_ACCERR; |
431 | break; |
432 | default: |
433 | KASSERT(0); |
434 | break; |
435 | } |
436 | goto trapsignal; |
437 | |
438 | case T_PRIVINFLT|T_USER: /* privileged instruction fault */ |
439 | case T_FPOPFLT|T_USER: /* coprocessor operand fault */ |
440 | #ifdef TRAP_SIGDEBUG |
441 | printf("pid %d.%d (%s): ILL at rip %lx addr %lx\n" , |
442 | p->p_pid, l->l_lid, p->p_comm, frame->tf_rip, rcr2()); |
443 | frame_dump(frame); |
444 | #endif |
445 | KSI_INIT_TRAP(&ksi); |
446 | ksi.ksi_signo = SIGILL; |
447 | ksi.ksi_trap = type & ~T_USER; |
448 | ksi.ksi_addr = (void *) frame->tf_rip; |
449 | switch (type) { |
450 | case T_PRIVINFLT|T_USER: |
451 | ksi.ksi_code = ILL_PRVOPC; |
452 | break; |
453 | case T_FPOPFLT|T_USER: |
454 | ksi.ksi_code = ILL_COPROC; |
455 | break; |
456 | default: |
457 | KASSERT(0); |
458 | break; |
459 | } |
460 | goto trapsignal; |
461 | |
462 | case T_ASTFLT|T_USER: |
463 | /* Allow process switch. */ |
464 | //curcpu()->ci_data.cpu_nast++; |
465 | if (l->l_pflag & LP_OWEUPC) { |
466 | l->l_pflag &= ~LP_OWEUPC; |
467 | ADDUPROF(l); |
468 | } |
469 | /* Allow a forced task switch. */ |
470 | if (curcpu()->ci_want_resched) { |
471 | preempt(); |
472 | } |
473 | goto out; |
474 | |
475 | case T_BOUND|T_USER: |
476 | case T_OFLOW|T_USER: |
477 | case T_DIVIDE|T_USER: |
478 | KSI_INIT_TRAP(&ksi); |
479 | ksi.ksi_signo = SIGFPE; |
480 | ksi.ksi_trap = type & ~T_USER; |
481 | ksi.ksi_addr = (void *)frame->tf_rip; |
482 | switch (type) { |
483 | case T_BOUND|T_USER: |
484 | ksi.ksi_code = FPE_FLTSUB; |
485 | break; |
486 | case T_OFLOW|T_USER: |
487 | ksi.ksi_code = FPE_INTOVF; |
488 | break; |
489 | case T_DIVIDE|T_USER: |
490 | ksi.ksi_code = FPE_INTDIV; |
491 | break; |
492 | default: |
493 | #ifdef DIAGNOSTIC |
494 | panic("unhandled type %x\n" , type); |
495 | #endif |
496 | break; |
497 | } |
498 | goto trapsignal; |
499 | |
500 | case T_PAGEFLT: |
501 | /* Allow page faults in kernel mode. */ |
502 | if (__predict_false(l == NULL)) |
503 | goto we_re_toast; |
504 | |
505 | /* |
506 | * fusuintrfailure is used by [fs]uswintr() to prevent |
507 | * page faulting from inside the profiling interrupt. |
508 | */ |
509 | onfault = pcb->pcb_onfault; |
510 | if (onfault == fusuintrfailure) { |
511 | goto copyefault; |
512 | } |
513 | if (cpu_intr_p() || (l->l_pflag & LP_INTR) != 0) { |
514 | goto we_re_toast; |
515 | } |
516 | |
517 | cr2 = rcr2(); |
518 | |
519 | if (frame->tf_err & PGEX_X) { |
520 | /* SMEP might have brought us here */ |
521 | if (cr2 > VM_MIN_ADDRESS && cr2 <= VM_MAXUSER_ADDRESS) |
522 | panic("prevented execution of %p (SMEP)" , |
523 | (void *)cr2); |
524 | } |
525 | |
526 | goto faultcommon; |
527 | |
528 | case T_PAGEFLT|T_USER: { /* page fault */ |
529 | register vaddr_t va; |
530 | register struct vmspace *vm; |
531 | register struct vm_map *map; |
532 | vm_prot_t ftype; |
533 | extern struct vm_map *kernel_map; |
534 | |
535 | cr2 = rcr2(); |
536 | if (p->p_emul->e_usertrap != NULL && |
537 | (*p->p_emul->e_usertrap)(l, cr2, frame) != 0) |
538 | return; |
539 | faultcommon: |
540 | vm = p->p_vmspace; |
541 | if (__predict_false(vm == NULL)) { |
542 | goto we_re_toast; |
543 | } |
544 | pcb->pcb_cr2 = cr2; |
545 | va = trunc_page((vaddr_t)cr2); |
546 | /* |
547 | * It is only a kernel address space fault iff: |
548 | * 1. (type & T_USER) == 0 and |
549 | * 2. pcb_onfault not set or |
550 | * 3. pcb_onfault set but supervisor space fault |
551 | * The last can occur during an exec() copyin where the |
552 | * argument space is lazy-allocated. |
553 | */ |
554 | if (type == T_PAGEFLT && va >= VM_MIN_KERNEL_ADDRESS) |
555 | map = kernel_map; |
556 | else |
557 | map = &vm->vm_map; |
558 | if (frame->tf_err & PGEX_W) |
559 | ftype = VM_PROT_WRITE; |
560 | else if (frame->tf_err & PGEX_X) |
561 | ftype = VM_PROT_EXECUTE; |
562 | else |
563 | ftype = VM_PROT_READ; |
564 | |
565 | #ifdef DIAGNOSTIC |
566 | if (map == kernel_map && va == 0) { |
567 | printf("trap: bad kernel access at %lx\n" , va); |
568 | goto we_re_toast; |
569 | } |
570 | #endif |
571 | /* Fault the original page in. */ |
572 | onfault = pcb->pcb_onfault; |
573 | pcb->pcb_onfault = NULL; |
574 | error = uvm_fault(map, va, ftype); |
575 | pcb->pcb_onfault = onfault; |
576 | if (error == 0) { |
577 | if (map != kernel_map && (void *)va >= vm->vm_maxsaddr) |
578 | uvm_grow(p, va); |
579 | |
580 | pfail = false; |
581 | while (type == T_PAGEFLT) { |
582 | /* |
583 | * we need to switch pmap now if we're in |
584 | * the middle of copyin/out. |
585 | * |
586 | * but we don't need to do so for kcopy as |
587 | * it never touch userspace. |
588 | */ |
589 | kpreempt_disable(); |
590 | if (curcpu()->ci_want_pmapload) { |
591 | onfault = onfault_handler(pcb, frame); |
592 | if (onfault != kcopy_fault) { |
593 | pmap_load(); |
594 | } |
595 | } |
596 | /* |
597 | * We need to keep the pmap loaded and |
598 | * so avoid being preempted until back |
599 | * into the copy functions. Disable |
600 | * interrupts at the hardware level before |
601 | * re-enabling preemption. Interrupts |
602 | * will be re-enabled by 'iret' when |
603 | * returning back out of the trap stub. |
604 | * They'll only be re-enabled when the |
605 | * program counter is once again in |
606 | * the copy functions, and so visible |
607 | * to cpu_kpreempt_exit(). |
608 | */ |
609 | #ifndef XEN |
610 | x86_disable_intr(); |
611 | #endif |
612 | l->l_nopreempt--; |
613 | if (l->l_nopreempt > 0 || !l->l_dopreempt || |
614 | pfail) { |
615 | return; |
616 | } |
617 | #ifndef XEN |
618 | x86_enable_intr(); |
619 | #endif |
620 | /* |
621 | * If preemption fails for some reason, |
622 | * don't retry it. The conditions won't |
623 | * change under our nose. |
624 | */ |
625 | pfail = kpreempt(0); |
626 | } |
627 | goto out; |
628 | } |
629 | |
630 | if (type == T_PAGEFLT) { |
631 | onfault = onfault_handler(pcb, frame); |
632 | if (onfault != NULL) |
633 | goto copyfault; |
634 | printf("uvm_fault(%p, 0x%lx, %d) -> %x\n" , |
635 | map, va, ftype, error); |
636 | goto kernelfault; |
637 | } |
638 | |
639 | KSI_INIT_TRAP(&ksi); |
640 | ksi.ksi_trap = type & ~T_USER; |
641 | ksi.ksi_addr = (void *)cr2; |
642 | switch (error) { |
643 | case EINVAL: |
644 | ksi.ksi_signo = SIGBUS; |
645 | ksi.ksi_code = BUS_ADRERR; |
646 | break; |
647 | case EACCES: |
648 | ksi.ksi_signo = SIGSEGV; |
649 | ksi.ksi_code = SEGV_ACCERR; |
650 | error = EFAULT; |
651 | break; |
652 | case ENOMEM: |
653 | ksi.ksi_signo = SIGKILL; |
654 | printf("UVM: pid %d.%d (%s), uid %d killed: " |
655 | "out of swap\n" , p->p_pid, l->l_lid, p->p_comm, |
656 | l->l_cred ? kauth_cred_geteuid(l->l_cred) : -1); |
657 | break; |
658 | default: |
659 | ksi.ksi_signo = SIGSEGV; |
660 | ksi.ksi_code = SEGV_MAPERR; |
661 | break; |
662 | } |
663 | |
664 | #ifdef TRAP_SIGDEBUG |
665 | printf("pid %d.%d (%s): signal %d at rip %#lx addr %#lx " |
666 | "error %d trap %d cr2 %p\n" , p->p_pid, l->l_lid, p->p_comm, |
667 | ksi.ksi_signo, frame->tf_rip, va, error, ksi.ksi_trap, |
668 | ksi.ksi_addr); |
669 | frame_dump(frame); |
670 | #endif |
671 | (*p->p_emul->e_trapsignal)(l, &ksi); |
672 | break; |
673 | } |
674 | |
675 | case T_TRCTRAP: |
676 | /* Check whether they single-stepped into a lcall. */ |
677 | if (frame->tf_rip == (uint64_t)IDTVEC(oosyscall) || |
678 | frame->tf_rip == (uint64_t)IDTVEC(osyscall) || |
679 | frame->tf_rip == (uint64_t)IDTVEC(syscall32)) { |
680 | frame->tf_rflags &= ~PSL_T; |
681 | return; |
682 | } |
683 | goto we_re_toast; |
684 | |
685 | case T_BPTFLT|T_USER: /* bpt instruction fault */ |
686 | case T_TRCTRAP|T_USER: /* trace trap */ |
687 | /* |
688 | * Don't go single-stepping into a RAS. |
689 | */ |
690 | |
691 | if (p->p_raslist == NULL || |
692 | (ras_lookup(p, (void *)frame->tf_rip) == (void *)-1)) { |
693 | KSI_INIT_TRAP(&ksi); |
694 | ksi.ksi_signo = SIGTRAP; |
695 | ksi.ksi_trap = type & ~T_USER; |
696 | if (type == (T_BPTFLT|T_USER)) |
697 | ksi.ksi_code = TRAP_BRKPT; |
698 | else |
699 | ksi.ksi_code = TRAP_TRACE; |
700 | (*p->p_emul->e_trapsignal)(l, &ksi); |
701 | } |
702 | break; |
703 | |
704 | case T_NMI: |
705 | if (nmi_dispatch(frame)) |
706 | return; |
707 | /* NMI can be hooked up to a pushbutton for debugging */ |
708 | if (kgdb_trap(type, frame)) |
709 | return; |
710 | if (kdb_trap(type, 0, frame)) |
711 | return; |
712 | /* machine/parity/power fail/"kitchen sink" faults */ |
713 | |
714 | x86_nmi(); |
715 | return; |
716 | } |
717 | |
718 | if ((type & T_USER) == 0) |
719 | return; |
720 | out: |
721 | userret(l); |
722 | return; |
723 | trapsignal: |
724 | (*p->p_emul->e_trapsignal)(l, &ksi); |
725 | userret(l); |
726 | } |
727 | |
728 | /* |
729 | * startlwp: start of a new LWP. |
730 | */ |
731 | void |
732 | startlwp(void *arg) |
733 | { |
734 | ucontext_t *uc = arg; |
735 | lwp_t *l = curlwp; |
736 | int error __diagused; |
737 | |
738 | error = cpu_setmcontext(l, &uc->uc_mcontext, uc->uc_flags); |
739 | KASSERT(error == 0); |
740 | |
741 | kmem_free(uc, sizeof(ucontext_t)); |
742 | userret(l); |
743 | } |
744 | |
745 | #ifdef TRAP_SIGDEBUG |
746 | static void |
747 | frame_dump(struct trapframe *tf) |
748 | { |
749 | int i; |
750 | unsigned long *p; |
751 | |
752 | printf("rip %p rsp %p rfl %p\n" , |
753 | (void *)tf->tf_rip, (void *)tf->tf_rsp, (void *)tf->tf_rflags); |
754 | printf("rdi %p rsi %p rdx %p\n" , |
755 | (void *)tf->tf_rdi, (void *)tf->tf_rsi, (void *)tf->tf_rdx); |
756 | printf("rcx %p r8 %p r9 %p\n" , |
757 | (void *)tf->tf_rcx, (void *)tf->tf_r8, (void *)tf->tf_r9); |
758 | printf("r10 %p r11 %p r12 %p\n" , |
759 | (void *)tf->tf_r10, (void *)tf->tf_r11, (void *)tf->tf_r12); |
760 | printf("r13 %p r14 %p r15 %p\n" , |
761 | (void *)tf->tf_r13, (void *)tf->tf_r14, (void *)tf->tf_r15); |
762 | printf("rbp %p rbx %p rax %p\n" , |
763 | (void *)tf->tf_rbp, (void *)tf->tf_rbx, (void *)tf->tf_rax); |
764 | printf("cs %lx ds %lx es %lx fs %lx gs %lx ss %lx\n" , |
765 | tf->tf_cs & 0xffff, tf->tf_ds & 0xffff, tf->tf_es & 0xffff, |
766 | tf->tf_fs & 0xffff, tf->tf_gs & 0xffff, tf->tf_ss & 0xffff); |
767 | |
768 | printf("\n" ); |
769 | printf("Stack dump:\n" ); |
770 | for (i = 0, p = (unsigned long *) tf; i < 20; i ++, p += 4) |
771 | printf(" 0x%.16lx 0x%.16lx 0x%.16lx 0x%.16lx\n" , *p, p[1], p[2], p[3]); |
772 | printf("\n" ); |
773 | } |
774 | #endif |
775 | |