1 /*        $NetBSD: trap.c,v 1.129 2023/10/05 19:41:03 ad Exp $        */
2 
3 /*
4  * Copyright (c) 1998, 2000, 2017 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, and by Maxime Villard.
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 #include <sys/cdefs.h>
67 __KERNEL_RCSID(0, "$NetBSD: trap.c,v 1.129 2023/10/05 19:41:03 ad Exp $");
68 
69 #include "opt_ddb.h"
70 #include "opt_kgdb.h"
71 #include "opt_xen.h"
72 #include "opt_dtrace.h"
73 
74 #include <sys/param.h>
75 #include <sys/systm.h>
76 #include <sys/proc.h>
77 #include <sys/acct.h>
78 #include <sys/kauth.h>
79 #include <sys/kernel.h>
80 #include <sys/kmem.h>
81 #include <sys/ras.h>
82 #include <sys/signal.h>
83 #include <sys/syscall.h>
84 #include <sys/cpu.h>
85 #include <sys/ucontext.h>
86 #include <sys/module_hook.h>
87 #include <sys/compat_stub.h>
88 
89 #include <uvm/uvm_extern.h>
90 
91 #include <machine/cpufunc.h>
92 #include <x86/fpu.h>
93 #include <x86/dbregs.h>
94 #include <machine/psl.h>
95 #include <machine/reg.h>
96 #include <machine/trap.h>
97 #include <machine/userret.h>
98 #include <machine/db_machdep.h>
99 
100 #include <x86/nmi.h>
101 
102 #ifndef XENPV
103 #include "isa.h"
104 #endif
105 
106 #include <sys/kgdb.h>
107 
108 #ifdef KDTRACE_HOOKS
109 #include <sys/dtrace_bsd.h>
110 /*
111  * This is a hook which is initialized by the dtrace module to handle traps
112  * which might occur during DTrace probe execution.
113  */
114 dtrace_trap_func_t dtrace_trap_func = NULL;
115 dtrace_doubletrap_func_t dtrace_doubletrap_func = NULL;
116 #endif
117 
118 /*
119  * Module hook for amd64_oosyscall
120  */
121 struct amd64_oosyscall_hook_t amd64_oosyscall_hook;
122 
123 void nmitrap(struct trapframe *);
124 void doubletrap(struct trapframe *);
125 void trap(struct trapframe *);
126 
127 const char * const trap_type[] = {
128           "privileged instruction fault",                   /*  0 T_PRIVINFLT */
129           "breakpoint trap",                      /*  1 T_BPTFLT */
130           "arithmetic trap",                      /*  2 T_ARITHTRAP */
131           "asynchronous system trap",             /*  3 T_ASTFLT */
132           "protection fault",                     /*  4 T_PROTFLT */
133           "trace trap",                                     /*  5 T_TRCTRAP */
134           "page fault",                                     /*  6 T_PAGEFLT */
135           "alignment fault",                      /*  7 T_ALIGNFLT */
136           "integer divide fault",                           /*  8 T_DIVIDE */
137           "non-maskable interrupt",               /*  9 T_NMI */
138           "overflow trap",                        /* 10 T_OFLOW */
139           "bounds check fault",                             /* 11 T_BOUND */
140           "FPU not available fault",              /* 12 T_DNA */
141           "double fault",                                   /* 13 T_DOUBLEFLT */
142           "FPU operand fetch fault",              /* 14 T_FPOPFLT */
143           "invalid TSS fault",                              /* 15 T_TSSFLT */
144           "segment not present fault",            /* 16 T_SEGNPFLT */
145           "stack fault",                                    /* 17 T_STKFLT */
146           "machine check fault",                            /* 18 T_MCA */
147           "SSE FP exception",                     /* 19 T_XMM */
148           "reserved trap",                        /* 20 T_RESERVED */
149 };
150 int trap_types = __arraycount(trap_type);
151 
152 #ifdef TRAP_SIGDEBUG
153 static void sigdebug(const struct trapframe *, const ksiginfo_t *, int);
154 #define SIGDEBUG(a, b, c) sigdebug(a, b, c)
155 #else
156 #define SIGDEBUG(a, b, c)
157 #endif
158 
159 static void
onfault_restore(struct trapframe * frame,void * onfault,int error)160 onfault_restore(struct trapframe *frame, void *onfault, int error)
161 {
162           frame->tf_rip = (uintptr_t)onfault;
163           frame->tf_rax = error;
164 }
165 
166 static void *
onfault_handler(const struct pcb * pcb,const struct trapframe * tf)167 onfault_handler(const struct pcb *pcb, const struct trapframe *tf)
168 {
169           struct onfault_table {
170                     uintptr_t start;
171                     uintptr_t end;
172                     void *handler;
173           };
174           extern const struct onfault_table onfault_table[];
175           const struct onfault_table *p;
176           uintptr_t pc;
177 
178           if (pcb->pcb_onfault != NULL) {
179                     return pcb->pcb_onfault;
180           }
181 
182           pc = tf->tf_rip;
183           for (p = onfault_table; p->start; p++) {
184                     if (p->start <= pc && pc < p->end) {
185                               return p->handler;
186                     }
187           }
188           return NULL;
189 }
190 
191 static void
trap_print(const struct trapframe * frame,const lwp_t * l)192 trap_print(const struct trapframe *frame, const lwp_t *l)
193 {
194           const int type = frame->tf_trapno;
195 
196           if (frame->tf_trapno < trap_types) {
197                     printf("fatal %s", trap_type[type]);
198           } else {
199                     printf("unknown trap %d", type);
200           }
201           printf(" in %s mode\n", (type & T_USER) ? "user" : "supervisor");
202 
203           printf("trap type %d code %#lx rip %#lx cs %#lx rflags %#lx cr2 %#lx "
204               "ilevel %#x rsp %#lx\n",
205               type, frame->tf_err, (u_long)frame->tf_rip, frame->tf_cs,
206               frame->tf_rflags, rcr2(), curcpu()->ci_ilevel, frame->tf_rsp);
207 
208           printf("curlwp %p pid %d.%d lowest kstack %p\n",
209               l, l->l_proc->p_pid, l->l_lid, KSTACK_LOWEST_ADDR(l));
210 }
211 
212 void
nmitrap(struct trapframe * frame)213 nmitrap(struct trapframe *frame)
214 {
215           const int type = T_NMI;
216 
217           if (nmi_dispatch(frame))
218                     return;
219           /* NMI can be hooked up to a pushbutton for debugging */
220           if (kgdb_trap(type, frame))
221                     return;
222           if (kdb_trap(type, 0, frame))
223                     return;
224           /* machine/parity/power fail/"kitchen sink" faults */
225 
226           x86_nmi();
227 }
228 
229 void
doubletrap(struct trapframe * frame)230 doubletrap(struct trapframe *frame)
231 {
232           const int type = T_DOUBLEFLT;
233           struct lwp *l = curlwp;
234 
235           trap_print(frame, l);
236 
237           if (kdb_trap(type, 0, frame))
238                     return;
239           if (kgdb_trap(type, frame))
240                     return;
241 
242           panic("double fault");
243 }
244 
245 /*
246  * trap(frame): exception, fault, and trap interface to BSD kernel.
247  *
248  * This common code is called from assembly language IDT gate entry routines
249  * that prepare a suitable stack frame, and restore this frame after the
250  * exception has been processed. Note that the effect is as if the arguments
251  * were passed call by reference.
252  *
253  * Note that the fpu traps (07 T_DNA, 10 T_ARITHTRAP and 13 T_XMM)
254  * jump directly into the code in x86/fpu.c so they get processed
255  * without interrupts being enabled.
256  */
257 void
trap(struct trapframe * frame)258 trap(struct trapframe *frame)
259 {
260           struct lwp *l = curlwp;
261           struct proc *p;
262           struct pcb *pcb;
263           extern char kcopy_fault[];
264           ksiginfo_t ksi;
265           void *onfault;
266           int type, error;
267           uint64_t cr2;
268           bool pfail;
269 
270           if (__predict_true(l != NULL)) {
271                     pcb = lwp_getpcb(l);
272                     p = l->l_proc;
273           } else {
274                     /*
275                      * This can happen eg on break points in early on boot.
276                      */
277                     pcb = NULL;
278                     p = NULL;
279           }
280           type = frame->tf_trapno;
281 
282           if (!KERNELMODE(frame->tf_cs)) {
283                     type |= T_USER;
284                     l->l_md.md_regs = frame;
285           }
286 
287 #ifdef KDTRACE_HOOKS
288           /*
289            * A trap can occur while DTrace executes a probe. Before
290            * executing the probe, DTrace blocks re-scheduling and sets
291            * a flag in its per-cpu flags to indicate that it doesn't
292            * want to fault. On returning from the probe, the no-fault
293            * flag is cleared and finally re-scheduling is enabled.
294            *
295            * If the DTrace kernel module has registered a trap handler,
296            * call it and if it returns non-zero, assume that it has
297            * handled the trap and modified the trap frame so that this
298            * function can return normally.
299            */
300           if ((type == T_PROTFLT || type == T_PAGEFLT) &&
301               dtrace_trap_func != NULL) {
302                     if ((*dtrace_trap_func)(frame, type)) {
303                               return;
304                     }
305           }
306 #endif
307 
308           switch (type) {
309 
310           default:
311           we_re_toast:
312                     trap_print(frame, l);
313 
314                     if (kdb_trap(type, 0, frame))
315                               return;
316                     if (kgdb_trap(type, frame))
317                               return;
318                     /*
319                      * If this is a breakpoint, don't panic if we're not connected.
320                      */
321                     if (type == T_BPTFLT && kgdb_disconnected()) {
322                               printf("kgdb: ignored %s\n", trap_type[type]);
323                               return;
324                     }
325                     panic("trap");
326                     /*NOTREACHED*/
327 
328           case T_PROTFLT:
329           case T_SEGNPFLT:
330           case T_ALIGNFLT:
331           case T_STKFLT:
332           case T_TSSFLT:
333                     if (p == NULL)
334                               goto we_re_toast;
335 
336                     /* Check for copyin/copyout fault. */
337                     onfault = onfault_handler(pcb, frame);
338                     if (onfault != NULL) {
339                               onfault_restore(frame, onfault, EFAULT);
340                               return;
341                     }
342 
343                     goto we_re_toast;
344 
345           case T_PROTFLT|T_USER:                  /* protection fault */
346           {         int hook_ret;
347 
348                     MODULE_HOOK_CALL(amd64_oosyscall_hook, (p, frame),
349                               ENOSYS, hook_ret);
350                     if (hook_ret == 0) {
351                               /* Do the syscall */
352                               p->p_md.md_syscall(frame);
353                               goto out;
354                     }
355           }
356                     /* FALLTHROUGH */
357           case T_TSSFLT|T_USER:
358           case T_SEGNPFLT|T_USER:
359           case T_STKFLT|T_USER:
360           case T_ALIGNFLT|T_USER:
361                     KSI_INIT_TRAP(&ksi);
362                     ksi.ksi_trap = type & ~T_USER;
363                     ksi.ksi_addr = (void *)frame->tf_rip;
364                     switch (type) {
365                     case T_SEGNPFLT|T_USER:
366                     case T_STKFLT|T_USER:
367                               ksi.ksi_signo = SIGBUS;
368                               ksi.ksi_code = BUS_ADRERR;
369                               break;
370                     case T_TSSFLT|T_USER:
371                               ksi.ksi_signo = SIGBUS;
372                               ksi.ksi_code = BUS_OBJERR;
373                               break;
374                     case T_ALIGNFLT|T_USER:
375                               ksi.ksi_signo = SIGBUS;
376                               ksi.ksi_code = BUS_ADRALN;
377                               break;
378                     case T_PROTFLT|T_USER:
379                               ksi.ksi_signo = SIGSEGV;
380                               ksi.ksi_code = SEGV_ACCERR;
381                               break;
382                     default:
383                               KASSERT(0);
384                               break;
385                     }
386                     goto trapsignal;
387 
388           case T_PRIVINFLT|T_USER:      /* privileged instruction fault */
389           case T_FPOPFLT|T_USER:                  /* coprocessor operand fault */
390                     KSI_INIT_TRAP(&ksi);
391                     ksi.ksi_signo = SIGILL;
392                     ksi.ksi_trap = type & ~T_USER;
393                     ksi.ksi_addr = (void *) frame->tf_rip;
394                     switch (type) {
395                     case T_PRIVINFLT|T_USER:
396                               ksi.ksi_code = ILL_PRVOPC;
397                               break;
398                     case T_FPOPFLT|T_USER:
399                               ksi.ksi_code = ILL_COPROC;
400                               break;
401                     default:
402                               KASSERT(0);
403                               break;
404                     }
405                     goto trapsignal;
406 
407           case T_ASTFLT|T_USER:
408                     /* Allow process switch. */
409                     //curcpu()->ci_data.cpu_nast++;
410                     if (l->l_pflag & LP_OWEUPC) {
411                               l->l_pflag &= ~LP_OWEUPC;
412                               ADDUPROF(l);
413                     }
414                     goto out;
415 
416           case T_BOUND|T_USER:
417           case T_OFLOW|T_USER:
418           case T_DIVIDE|T_USER:
419                     KSI_INIT_TRAP(&ksi);
420                     ksi.ksi_signo = SIGFPE;
421                     ksi.ksi_trap = type & ~T_USER;
422                     ksi.ksi_addr = (void *)frame->tf_rip;
423                     switch (type) {
424                     case T_BOUND|T_USER:
425                               ksi.ksi_code = FPE_FLTSUB;
426                               break;
427                     case T_OFLOW|T_USER:
428                               ksi.ksi_code = FPE_INTOVF;
429                               break;
430                     case T_DIVIDE|T_USER:
431                               ksi.ksi_code = FPE_INTDIV;
432                               break;
433                     default:
434                               KASSERT(0);
435                               break;
436                     }
437                     goto trapsignal;
438 
439           case T_PAGEFLT:
440                     /* Allow page faults in kernel mode. */
441                     if (__predict_false(l == NULL))
442                               goto we_re_toast;
443 
444                     onfault = pcb->pcb_onfault;
445 
446                     if (cpu_intr_p() || (l->l_pflag & LP_INTR) != 0) {
447                               goto we_re_toast;
448                     }
449 
450                     cr2 = rcr2();
451 
452                     if (frame->tf_err & PGEX_I) {
453                               /* SMEP might have brought us here */
454                               if (cr2 < VM_MAXUSER_ADDRESS) {
455                                         printf("prevented execution of %p (SMEP)\n",
456                                             (void *)cr2);
457                                         goto we_re_toast;
458                               }
459                     }
460 
461                     if ((frame->tf_err & PGEX_P) &&
462                         cr2 < VM_MAXUSER_ADDRESS) {
463                               /* SMAP might have brought us here */
464                               if (onfault_handler(pcb, frame) == NULL) {
465                                         printf("prevented access to %p (SMAP)\n",
466                                             (void *)cr2);
467                                         goto we_re_toast;
468                               }
469                     }
470 
471                     goto pagefltcommon;
472 
473           case T_PAGEFLT|T_USER: {
474                     register vaddr_t va;
475                     register struct vmspace *vm;
476                     register struct vm_map *map;
477                     vm_prot_t ftype;
478                     extern struct vm_map *kernel_map;
479 
480                     cr2 = rcr2();
481                     if (p->p_emul->e_usertrap != NULL &&
482                         (*p->p_emul->e_usertrap)(l, cr2, frame) != 0)
483                               return;
484 pagefltcommon:
485                     vm = p->p_vmspace;
486                     if (__predict_false(vm == NULL)) {
487                               goto we_re_toast;
488                     }
489                     pcb->pcb_cr2 = cr2;
490                     va = trunc_page((vaddr_t)cr2);
491                     /*
492                      * It is only a kernel address space fault iff:
493                      *        1. (type & T_USER) == 0  and
494                      *        2. pcb_onfault not set or
495                      *        3. pcb_onfault set but supervisor space fault
496                      * The last can occur during an exec() copyin where the
497                      * argument space is lazy-allocated.
498                      */
499                     if (type == T_PAGEFLT && va >= VM_MIN_KERNEL_ADDRESS)
500                               map = kernel_map;
501                     else
502                               map = &vm->vm_map;
503                     if (frame->tf_err & PGEX_W)
504                               ftype = VM_PROT_WRITE;
505                     else if (frame->tf_err & PGEX_I)
506                               ftype = VM_PROT_EXECUTE;
507                     else
508                               ftype = VM_PROT_READ;
509 
510 #ifdef DIAGNOSTIC
511                     if (map == kernel_map && va == 0) {
512                               printf("trap: bad kernel access at %lx\n", va);
513                               goto we_re_toast;
514                     }
515 #endif
516                     /* Fault the original page in. */
517                     onfault = pcb->pcb_onfault;
518                     pcb->pcb_onfault = NULL;
519                     error = uvm_fault(map, va, ftype);
520                     pcb->pcb_onfault = onfault;
521                     if (error == 0) {
522                               if (map != kernel_map && (void *)va >= vm->vm_maxsaddr)
523                                         uvm_grow(p, va);
524 
525                               pfail = false;
526                               while (type == T_PAGEFLT) {
527                                         /*
528                                          * we need to switch pmap now if we're in
529                                          * the middle of copyin/out.
530                                          *
531                                          * but we don't need to do so for kcopy as
532                                          * it never touch userspace.
533                                          */
534                                         kpreempt_disable();
535                                         if (curcpu()->ci_want_pmapload) {
536                                                   onfault = onfault_handler(pcb, frame);
537                                                   if (onfault != kcopy_fault) {
538                                                             pmap_load();
539                                                   }
540                                         }
541                                         /*
542                                          * We need to keep the pmap loaded and
543                                          * so avoid being preempted until back
544                                          * into the copy functions.  Disable
545                                          * interrupts at the hardware level before
546                                          * re-enabling preemption.  Interrupts
547                                          * will be re-enabled by 'iret' when
548                                          * returning back out of the trap stub.
549                                          * They'll only be re-enabled when the
550                                          * program counter is once again in
551                                          * the copy functions, and so visible
552                                          * to cpu_kpreempt_exit().
553                                          */
554 #ifndef XENPV
555                                         x86_disable_intr();
556 #endif
557                                         l->l_nopreempt--;
558                                         if (l->l_nopreempt > 0 || !l->l_dopreempt ||
559                                             pfail) {
560                                                   return;
561                                         }
562 #ifndef XENPV
563                                         x86_enable_intr();
564 #endif
565                                         /*
566                                          * If preemption fails for some reason,
567                                          * don't retry it.  The conditions won't
568                                          * change under our nose.
569                                          */
570                                         pfail = kpreempt(0);
571                               }
572                               goto out;
573                     }
574 
575                     if (type == T_PAGEFLT) {
576                               onfault = onfault_handler(pcb, frame);
577                               if (onfault != NULL) {
578                                         onfault_restore(frame, onfault, error);
579                                         return;
580                               }
581 
582                               printf("uvm_fault(%p, 0x%lx, %d) -> %x\n",
583                                   map, va, ftype, error);
584                               goto we_re_toast;
585                     }
586 
587                     KSI_INIT_TRAP(&ksi);
588                     ksi.ksi_trap = type & ~T_USER;
589                     ksi.ksi_addr = (void *)cr2;
590                     switch (error) {
591                     case EINVAL:
592                               ksi.ksi_signo = SIGBUS;
593                               ksi.ksi_code = BUS_ADRERR;
594                               break;
595                     case EACCES:
596                               ksi.ksi_signo = SIGSEGV;
597                               ksi.ksi_code = SEGV_ACCERR;
598                               error = EFAULT;
599                               break;
600                     case ENOMEM:
601                               ksi.ksi_signo = SIGKILL;
602                               printf("UVM: pid %d.%d (%s), uid %d killed: "
603                                   "out of swap\n", p->p_pid, l->l_lid, p->p_comm,
604                                   l->l_cred ?  kauth_cred_geteuid(l->l_cred) : -1);
605                               break;
606                     default:
607                               ksi.ksi_signo = SIGSEGV;
608                               ksi.ksi_code = SEGV_MAPERR;
609                               break;
610                     }
611 
612                     SIGDEBUG(frame, &ksi, error);
613                     (*p->p_emul->e_trapsignal)(l, &ksi);
614                     break;
615           }
616 
617           case T_TRCTRAP:
618                     /*
619                      * Ignore debug register trace traps due to
620                      * accesses in the user's address space, which
621                      * can happen under several conditions such as
622                      * if a user sets a watchpoint on a buffer and
623                      * then passes that buffer to a system call.
624                      * We still want to get TRCTRAPS for addresses
625                      * in kernel space because that is useful when
626                      * debugging the kernel.
627                      */
628                     if (x86_dbregs_user_trap())
629                               break;
630 
631                     goto we_re_toast;
632 
633           case T_BPTFLT|T_USER:                   /* bpt instruction fault */
634           case T_TRCTRAP|T_USER:                  /* trace trap */
635                     /*
636                      * Don't go single-stepping into a RAS.
637                      */
638                     if (p->p_raslist == NULL ||
639                         (ras_lookup(p, (void *)frame->tf_rip) == (void *)-1)) {
640                               KSI_INIT_TRAP(&ksi);
641                               ksi.ksi_signo = SIGTRAP;
642                               ksi.ksi_trap = type & ~T_USER;
643                               if (x86_dbregs_user_trap()) {
644                                         x86_dbregs_store_dr6(l);
645                                         ksi.ksi_code = TRAP_DBREG;
646                               } else if (type == (T_BPTFLT|T_USER))
647                                         ksi.ksi_code = TRAP_BRKPT;
648                               else
649                                         ksi.ksi_code = TRAP_TRACE;
650                               (*p->p_emul->e_trapsignal)(l, &ksi);
651                     }
652                     break;
653           }
654 
655           if ((type & T_USER) == 0)
656                     return;
657 out:
658           userret(l);
659           return;
660 trapsignal:
661           SIGDEBUG(frame, &ksi, 0);
662           (*p->p_emul->e_trapsignal)(l, &ksi);
663           userret(l);
664 }
665 
666 /*
667  * startlwp: start of a new LWP.
668  */
669 void
startlwp(void * arg)670 startlwp(void *arg)
671 {
672           ucontext_t *uc = arg;
673           lwp_t *l = curlwp;
674           int error __diagused;
675 
676           error = cpu_setmcontext(l, &uc->uc_mcontext, uc->uc_flags);
677           KASSERT(error == 0);
678 
679           kmem_free(uc, sizeof(ucontext_t));
680           userret(l);
681 }
682 
683 #ifdef TRAP_SIGDEBUG
684 static void
frame_dump(const struct trapframe * tf,struct pcb * pcb)685 frame_dump(const struct trapframe *tf, struct pcb *pcb)
686 {
687 
688           printf("trapframe %p\n", tf);
689           printf("rip %#018lx  rsp %#018lx  rfl %#018lx\n",
690               tf->tf_rip, tf->tf_rsp, tf->tf_rflags);
691           printf("rdi %#018lx  rsi %#018lx  rdx %#018lx\n",
692               tf->tf_rdi, tf->tf_rsi, tf->tf_rdx);
693           printf("rcx %#018lx  r8  %#018lx  r9  %#018lx\n",
694               tf->tf_rcx, tf->tf_r8, tf->tf_r9);
695           printf("r10 %#018lx  r11 %#018lx  r12 %#018lx\n",
696               tf->tf_r10, tf->tf_r11, tf->tf_r12);
697           printf("r13 %#018lx  r14 %#018lx  r15 %#018lx\n",
698               tf->tf_r13, tf->tf_r14, tf->tf_r15);
699           printf("rbp %#018lx  rbx %#018lx  rax %#018lx\n",
700               tf->tf_rbp, tf->tf_rbx, tf->tf_rax);
701           printf("cs %#04lx  ds %#04lx  es %#04lx  "
702               "fs %#04lx  gs %#04lx  ss %#04lx\n",
703               tf->tf_cs & 0xffff, tf->tf_ds & 0xffff, tf->tf_es & 0xffff,
704               tf->tf_fs & 0xffff, tf->tf_gs & 0xffff, tf->tf_ss & 0xffff);
705           printf("fsbase %#018lx gsbase %#018lx\n", pcb->pcb_fs, pcb->pcb_gs);
706           printf("\n");
707           hexdump(printf, "Stack dump", tf, 256);
708 }
709 
710 static void
sigdebug(const struct trapframe * tf,const ksiginfo_t * ksi,int e)711 sigdebug(const struct trapframe *tf, const ksiginfo_t *ksi, int e)
712 {
713           struct lwp *l = curlwp;
714           struct proc *p = l->l_proc;
715 
716           printf("pid %d.%d (%s): signal %d code=%d (trap %#lx) "
717               "@rip %#lx addr %#lx error=%d\n",
718               p->p_pid, l->l_lid, p->p_comm, ksi->ksi_signo, ksi->ksi_code,
719               tf->tf_trapno, tf->tf_rip, rcr2(), e);
720           frame_dump(tf, lwp_getpcb(l));
721 }
722 #endif
723