subr_trap.c revision 29675
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 * $Id: trap.c,v 1.110 1997/09/05 08:54:54 peter Exp $ 39 */ 40 41/* 42 * 386 Trap and System call handling 43 */ 44 45#include "opt_ktrace.h" 46#include "opt_ddb.h" 47 48#include <sys/param.h> 49#include <sys/systm.h> 50#include <sys/proc.h> 51#include <sys/kernel.h> 52#include <sys/syscall.h> 53#include <sys/sysent.h> 54#include <sys/vmmeter.h> 55#ifdef KTRACE 56#include <sys/ktrace.h> 57#endif 58 59#include <vm/vm.h> 60#include <vm/vm_param.h> 61#include <vm/vm_prot.h> 62#include <sys/lock.h> 63#include <vm/pmap.h> 64#include <vm/vm_kern.h> 65#include <vm/vm_map.h> 66#include <vm/vm_page.h> 67#include <vm/vm_extern.h> 68 69#include <sys/user.h> 70 71#include <machine/cpu.h> 72#include <machine/ipl.h> 73#include <machine/md_var.h> 74#include <machine/psl.h> 75#include <machine/../isa/intr_machdep.h> 76#include <machine/smp.h> 77 78#ifdef POWERFAIL_NMI 79#include <sys/syslog.h> 80#include <machine/clock.h> 81#endif 82 83#include "isa.h" 84#include "npx.h" 85 86extern struct i386tss common_tss; 87 88int (*pmath_emulate) __P((struct trapframe *)); 89 90extern void trap __P((struct trapframe frame)); 91extern int trapwrite __P((unsigned addr)); 92extern void syscall __P((struct trapframe frame)); 93 94static int trap_pfault __P((struct trapframe *, int)); 95static void trap_fatal __P((struct trapframe *)); 96void dblfault_handler __P((void)); 97 98extern inthand_t IDTVEC(syscall); 99 100#define MAX_TRAP_MSG 28 101static char *trap_msg[] = { 102 "", /* 0 unused */ 103 "privileged instruction fault", /* 1 T_PRIVINFLT */ 104 "", /* 2 unused */ 105 "breakpoint instruction fault", /* 3 T_BPTFLT */ 106 "", /* 4 unused */ 107 "", /* 5 unused */ 108 "arithmetic trap", /* 6 T_ARITHTRAP */ 109 "system forced exception", /* 7 T_ASTFLT */ 110 "", /* 8 unused */ 111 "general protection fault", /* 9 T_PROTFLT */ 112 "trace trap", /* 10 T_TRCTRAP */ 113 "", /* 11 unused */ 114 "page fault", /* 12 T_PAGEFLT */ 115 "", /* 13 unused */ 116 "alignment fault", /* 14 T_ALIGNFLT */ 117 "", /* 15 unused */ 118 "", /* 16 unused */ 119 "", /* 17 unused */ 120 "integer divide fault", /* 18 T_DIVIDE */ 121 "non-maskable interrupt trap", /* 19 T_NMI */ 122 "overflow trap", /* 20 T_OFLOW */ 123 "FPU bounds check fault", /* 21 T_BOUND */ 124 "FPU device not available", /* 22 T_DNA */ 125 "double fault", /* 23 T_DOUBLEFLT */ 126 "FPU operand fetch fault", /* 24 T_FPOPFLT */ 127 "invalid TSS fault", /* 25 T_TSSFLT */ 128 "segment not present fault", /* 26 T_SEGNPFLT */ 129 "stack fault", /* 27 T_STKFLT */ 130 "machine check trap", /* 28 T_MCHK */ 131}; 132 133static void userret __P((struct proc *p, struct trapframe *frame, 134 u_quad_t oticks)); 135 136static inline void 137userret(p, frame, oticks) 138 struct proc *p; 139 struct trapframe *frame; 140 u_quad_t oticks; 141{ 142 int sig, s; 143 144 while ((sig = CURSIG(p)) != 0) 145 postsig(sig); 146 147#if 0 148 if (!want_resched && 149 (p->p_priority <= p->p_usrpri) && 150 (p->p_rtprio.type == RTP_PRIO_NORMAL)) { 151 int newpriority; 152 p->p_estcpu += 1; 153 newpriority = PUSER + p->p_estcpu / 4 + 2 * p->p_nice; 154 newpriority = min(newpriority, MAXPRI); 155 p->p_usrpri = newpriority; 156 } 157#endif 158 159 p->p_priority = p->p_usrpri; 160 if (want_resched) { 161 /* 162 * Since we are curproc, clock will normally just change 163 * our priority without moving us from one queue to another 164 * (since the running process is not on a queue.) 165 * If that happened after we setrunqueue ourselves but before we 166 * mi_switch()'ed, we might not be on the queue indicated by 167 * our priority. 168 */ 169 s = splhigh(); 170 setrunqueue(p); 171 p->p_stats->p_ru.ru_nivcsw++; 172 mi_switch(); 173 splx(s); 174 while ((sig = CURSIG(p)) != 0) 175 postsig(sig); 176 } 177 /* 178 * Charge system time if profiling. 179 */ 180 if (p->p_flag & P_PROFIL) 181 addupc_task(p, frame->tf_eip, 182 (u_int)(p->p_sticks - oticks) * psratio); 183 184 curpriority = p->p_priority; 185} 186 187/* 188 * Exception, fault, and trap interface to the FreeBSD kernel. 189 * This common code is called from assembly language IDT gate entry 190 * routines that prepare a suitable stack frame, and restore this 191 * frame after the exception has been processed. 192 */ 193 194void 195trap(frame) 196 struct trapframe frame; 197{ 198 struct proc *p = curproc; 199 u_quad_t sticks = 0; 200 int i = 0, ucode = 0, type, code; 201#ifdef DEBUG 202 u_long eva; 203#endif 204 205 type = frame.tf_trapno; 206 code = frame.tf_err; 207 208 if ((ISPL(frame.tf_cs) == SEL_UPL) || (frame.tf_eflags & PSL_VM)) { 209 /* user trap */ 210 211 sticks = p->p_sticks; 212 p->p_md.md_regs = &frame; 213 214 switch (type) { 215 case T_PRIVINFLT: /* privileged instruction fault */ 216 ucode = type; 217 i = SIGILL; 218 break; 219 220 case T_BPTFLT: /* bpt instruction fault */ 221 case T_TRCTRAP: /* trace trap */ 222 frame.tf_eflags &= ~PSL_T; 223 i = SIGTRAP; 224 break; 225 226 case T_ARITHTRAP: /* arithmetic trap */ 227 ucode = code; 228 i = SIGFPE; 229 break; 230 231 case T_ASTFLT: /* Allow process switch */ 232 astoff(); 233 cnt.v_soft++; 234 if (p->p_flag & P_OWEUPC) { 235 p->p_flag &= ~P_OWEUPC; 236 addupc_task(p, p->p_stats->p_prof.pr_addr, 237 p->p_stats->p_prof.pr_ticks); 238 } 239 goto out; 240 241 /* 242 * The following two traps can happen in 243 * vm86 mode, and, if so, we want to handle 244 * them specially. 245 */ 246 case T_PROTFLT: /* general protection fault */ 247 case T_STKFLT: /* stack fault */ 248#ifdef VM86 249 if (frame.tf_eflags & PSL_VM) { 250 i = vm86_emulate((struct vm86frame *)&frame); 251 if (i == 0) 252 goto out; 253 break; 254 } 255#endif /* VM86 */ 256 /* FALL THROUGH */ 257 258 case T_SEGNPFLT: /* segment not present fault */ 259 case T_TSSFLT: /* invalid TSS fault */ 260 case T_DOUBLEFLT: /* double fault */ 261 default: 262 ucode = code + BUS_SEGM_FAULT ; 263 i = SIGBUS; 264 break; 265 266 case T_PAGEFLT: /* page fault */ 267 i = trap_pfault(&frame, TRUE); 268 if (i == -1) 269 return; 270 if (i == 0) 271 goto out; 272 273 ucode = T_PAGEFLT; 274 break; 275 276 case T_DIVIDE: /* integer divide fault */ 277 ucode = FPE_INTDIV_TRAP; 278 i = SIGFPE; 279 break; 280 281#if NISA > 0 282 case T_NMI: 283#ifdef POWERFAIL_NMI 284 goto handle_powerfail; 285#else /* !POWERFAIL_NMI */ 286#ifdef DDB 287 /* NMI can be hooked up to a pushbutton for debugging */ 288 printf ("NMI ... going to debugger\n"); 289 if (kdb_trap (type, 0, &frame)) 290 return; 291#endif /* DDB */ 292 /* machine/parity/power fail/"kitchen sink" faults */ 293 if (isa_nmi(code) == 0) return; 294 panic("NMI indicates hardware failure"); 295#endif /* POWERFAIL_NMI */ 296#endif /* NISA > 0 */ 297 298 case T_OFLOW: /* integer overflow fault */ 299 ucode = FPE_INTOVF_TRAP; 300 i = SIGFPE; 301 break; 302 303 case T_BOUND: /* bounds check fault */ 304 ucode = FPE_SUBRNG_TRAP; 305 i = SIGFPE; 306 break; 307 308 case T_DNA: 309#if NNPX > 0 310 /* if a transparent fault (due to context switch "late") */ 311 if (npxdna()) 312 return; 313#endif 314 if (!pmath_emulate) { 315 i = SIGFPE; 316 ucode = FPE_FPU_NP_TRAP; 317 break; 318 } 319 i = (*pmath_emulate)(&frame); 320 if (i == 0) { 321 if (!(frame.tf_eflags & PSL_T)) 322 return; 323 frame.tf_eflags &= ~PSL_T; 324 i = SIGTRAP; 325 } 326 /* else ucode = emulator_only_knows() XXX */ 327 break; 328 329 case T_FPOPFLT: /* FPU operand fetch fault */ 330 ucode = T_FPOPFLT; 331 i = SIGILL; 332 break; 333 } 334 } else { 335 /* kernel trap */ 336 337 switch (type) { 338 case T_PAGEFLT: /* page fault */ 339 (void) trap_pfault(&frame, FALSE); 340 return; 341 342 case T_DNA: 343#if NNPX > 0 344 /* 345 * The kernel is apparently using npx for copying. 346 * XXX this should be fatal unless the kernel has 347 * registered such use. 348 */ 349 if (npxdna()) 350 return; 351#endif 352 break; 353 354 case T_PROTFLT: /* general protection fault */ 355 case T_SEGNPFLT: /* segment not present fault */ 356 /* 357 * Invalid segment selectors and out of bounds 358 * %eip's and %esp's can be set up in user mode. 359 * This causes a fault in kernel mode when the 360 * kernel tries to return to user mode. We want 361 * to get this fault so that we can fix the 362 * problem here and not have to check all the 363 * selectors and pointers when the user changes 364 * them. 365 */ 366#define MAYBE_DORETI_FAULT(where, whereto) \ 367 do { \ 368 if (frame.tf_eip == (int)where) { \ 369 frame.tf_eip = (int)whereto; \ 370 return; \ 371 } \ 372 } while (0) 373 374 if (intr_nesting_level == 0) { 375 /* 376 * Invalid %fs's and %gs's can be created using 377 * procfs or PT_SETREGS or by invalidating the 378 * underlying LDT entry. This causes a fault 379 * in kernel mode when the kernel attempts to 380 * switch contexts. Lose the bad context 381 * (XXX) so that we can continue, and generate 382 * a signal. 383 */ 384 if (frame.tf_eip == (int)cpu_switch_load_fs) { 385 curpcb->pcb_fs = 0; 386 psignal(p, SIGBUS); 387 return; 388 } 389 if (frame.tf_eip == (int)cpu_switch_load_gs) { 390 curpcb->pcb_gs = 0; 391 psignal(p, SIGBUS); 392 return; 393 } 394 MAYBE_DORETI_FAULT(doreti_iret, 395 doreti_iret_fault); 396 MAYBE_DORETI_FAULT(doreti_popl_ds, 397 doreti_popl_ds_fault); 398 MAYBE_DORETI_FAULT(doreti_popl_es, 399 doreti_popl_es_fault); 400 if (curpcb && curpcb->pcb_onfault) { 401 frame.tf_eip = (int)curpcb->pcb_onfault; 402 return; 403 } 404 } 405 break; 406 407 case T_TSSFLT: 408 /* 409 * PSL_NT can be set in user mode and isn't cleared 410 * automatically when the kernel is entered. This 411 * causes a TSS fault when the kernel attempts to 412 * `iret' because the TSS link is uninitialized. We 413 * want to get this fault so that we can fix the 414 * problem here and not every time the kernel is 415 * entered. 416 */ 417 if (frame.tf_eflags & PSL_NT) { 418 frame.tf_eflags &= ~PSL_NT; 419 return; 420 } 421 break; 422 423 case T_TRCTRAP: /* trace trap */ 424 if (frame.tf_eip == (int)IDTVEC(syscall)) { 425 /* 426 * We've just entered system mode via the 427 * syscall lcall. Continue single stepping 428 * silently until the syscall handler has 429 * saved the flags. 430 */ 431 return; 432 } 433 if (frame.tf_eip == (int)IDTVEC(syscall) + 1) { 434 /* 435 * The syscall handler has now saved the 436 * flags. Stop single stepping it. 437 */ 438 frame.tf_eflags &= ~PSL_T; 439 return; 440 } 441 /* 442 * Fall through. 443 */ 444 case T_BPTFLT: 445 /* 446 * If DDB is enabled, let it handle the debugger trap. 447 * Otherwise, debugger traps "can't happen". 448 */ 449#ifdef DDB 450 if (kdb_trap (type, 0, &frame)) 451 return; 452#endif 453 break; 454 455#if NISA > 0 456 case T_NMI: 457#ifdef POWERFAIL_NMI 458#ifndef TIMER_FREQ 459# define TIMER_FREQ 1193182 460#endif 461 handle_powerfail: 462 { 463 static unsigned lastalert = 0; 464 465 if(time.tv_sec - lastalert > 10) 466 { 467 log(LOG_WARNING, "NMI: power fail\n"); 468 sysbeep(TIMER_FREQ/880, hz); 469 lastalert = time.tv_sec; 470 } 471 return; 472 } 473#else /* !POWERFAIL_NMI */ 474#ifdef DDB 475 /* NMI can be hooked up to a pushbutton for debugging */ 476 printf ("NMI ... going to debugger\n"); 477 if (kdb_trap (type, 0, &frame)) 478 return; 479#endif /* DDB */ 480 /* machine/parity/power fail/"kitchen sink" faults */ 481 if (isa_nmi(code) == 0) return; 482 /* FALL THROUGH */ 483#endif /* POWERFAIL_NMI */ 484#endif /* NISA > 0 */ 485 } 486 487 trap_fatal(&frame); 488 return; 489 } 490 491 trapsignal(p, i, ucode); 492 493#ifdef DEBUG 494 eva = rcr2(); 495 if (type <= MAX_TRAP_MSG) { 496 uprintf("fatal process exception: %s", 497 trap_msg[type]); 498 if ((type == T_PAGEFLT) || (type == T_PROTFLT)) 499 uprintf(", fault VA = 0x%x", eva); 500 uprintf("\n"); 501 } 502#endif 503 504out: 505 userret(p, &frame, sticks); 506} 507 508#ifdef notyet 509/* 510 * This version doesn't allow a page fault to user space while 511 * in the kernel. The rest of the kernel needs to be made "safe" 512 * before this can be used. I think the only things remaining 513 * to be made safe are the iBCS2 code and the process tracing/ 514 * debugging code. 515 */ 516static int 517trap_pfault(frame, usermode) 518 struct trapframe *frame; 519 int usermode; 520{ 521 vm_offset_t va; 522 struct vmspace *vm = NULL; 523 vm_map_t map = 0; 524 int rv = 0; 525 vm_prot_t ftype; 526 int eva; 527 struct proc *p = curproc; 528 529 if (frame->tf_err & PGEX_W) 530 ftype = VM_PROT_READ | VM_PROT_WRITE; 531 else 532 ftype = VM_PROT_READ; 533 534 eva = rcr2(); 535 va = trunc_page((vm_offset_t)eva); 536 537 if (va < VM_MIN_KERNEL_ADDRESS) { 538 vm_offset_t v; 539 vm_page_t mpte; 540 541 if (p == NULL || 542 (!usermode && va < VM_MAXUSER_ADDRESS && 543 (intr_nesting_level != 0 || curpcb == NULL || 544 curpcb->pcb_onfault == NULL))) { 545 trap_fatal(frame); 546 return (-1); 547 } 548 549 /* 550 * This is a fault on non-kernel virtual memory. 551 * vm is initialized above to NULL. If curproc is NULL 552 * or curproc->p_vmspace is NULL the fault is fatal. 553 */ 554 vm = p->p_vmspace; 555 if (vm == NULL) 556 goto nogo; 557 558 map = &vm->vm_map; 559 560 /* 561 * Keep swapout from messing with us during this 562 * critical time. 563 */ 564 ++p->p_lock; 565 566 /* 567 * Grow the stack if necessary 568 */ 569 if ((caddr_t)va > vm->vm_maxsaddr 570 && (caddr_t)va < (caddr_t)USRSTACK) { 571 if (!grow(p, va)) { 572 rv = KERN_FAILURE; 573 --p->p_lock; 574 goto nogo; 575 } 576 } 577 578 /* Fault in the user page: */ 579 rv = vm_fault(map, va, ftype, 580 (ftype & VM_PROT_WRITE) ? VM_FAULT_DIRTY : 0); 581 582 --p->p_lock; 583 } else { 584 /* 585 * Don't allow user-mode faults in kernel address space. 586 */ 587 if (usermode) 588 goto nogo; 589 590 /* 591 * Since we know that kernel virtual address addresses 592 * always have pte pages mapped, we just have to fault 593 * the page. 594 */ 595 rv = vm_fault(kernel_map, va, ftype, FALSE); 596 } 597 598 if (rv == KERN_SUCCESS) 599 return (0); 600nogo: 601 if (!usermode) { 602 if (intr_nesting_level == 0 && curpcb && curpcb->pcb_onfault) { 603 frame->tf_eip = (int)curpcb->pcb_onfault; 604 return (0); 605 } 606 trap_fatal(frame); 607 return (-1); 608 } 609 610 /* kludge to pass faulting virtual address to sendsig */ 611 frame->tf_err = eva; 612 613 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 614} 615#endif 616 617int 618trap_pfault(frame, usermode) 619 struct trapframe *frame; 620 int usermode; 621{ 622 vm_offset_t va; 623 struct vmspace *vm = NULL; 624 vm_map_t map = 0; 625 int rv = 0; 626 vm_prot_t ftype; 627 int eva; 628 struct proc *p = curproc; 629 630 eva = rcr2(); 631 va = trunc_page((vm_offset_t)eva); 632 633 if (va >= KERNBASE) { 634 /* 635 * Don't allow user-mode faults in kernel address space. 636 */ 637 if (usermode) 638 goto nogo; 639 640 map = kernel_map; 641 } else { 642 /* 643 * This is a fault on non-kernel virtual memory. 644 * vm is initialized above to NULL. If curproc is NULL 645 * or curproc->p_vmspace is NULL the fault is fatal. 646 */ 647 if (p != NULL) 648 vm = p->p_vmspace; 649 650 if (vm == NULL) 651 goto nogo; 652 653 map = &vm->vm_map; 654 } 655 656 if (frame->tf_err & PGEX_W) 657 ftype = VM_PROT_READ | VM_PROT_WRITE; 658 else 659 ftype = VM_PROT_READ; 660 661 if (map != kernel_map) { 662 /* 663 * Keep swapout from messing with us during this 664 * critical time. 665 */ 666 ++p->p_lock; 667 668 /* 669 * Grow the stack if necessary 670 */ 671 if ((caddr_t)va > vm->vm_maxsaddr 672 && (caddr_t)va < (caddr_t)USRSTACK) { 673 if (!grow(p, va)) { 674 rv = KERN_FAILURE; 675 --p->p_lock; 676 goto nogo; 677 } 678 } 679 680 /* Fault in the user page: */ 681 rv = vm_fault(map, va, ftype, 682 (ftype & VM_PROT_WRITE) ? VM_FAULT_DIRTY : 0); 683 684 --p->p_lock; 685 } else { 686 /* 687 * Don't have to worry about process locking or stacks in the kernel. 688 */ 689 rv = vm_fault(map, va, ftype, FALSE); 690 } 691 692 if (rv == KERN_SUCCESS) 693 return (0); 694nogo: 695 if (!usermode) { 696 if (intr_nesting_level == 0 && curpcb && curpcb->pcb_onfault) { 697 frame->tf_eip = (int)curpcb->pcb_onfault; 698 return (0); 699 } 700 trap_fatal(frame); 701 return (-1); 702 } 703 704 /* kludge to pass faulting virtual address to sendsig */ 705 frame->tf_err = eva; 706 707 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 708} 709 710static void 711trap_fatal(frame) 712 struct trapframe *frame; 713{ 714 int code, type, eva, ss, esp; 715 struct soft_segment_descriptor softseg; 716 717 code = frame->tf_err; 718 type = frame->tf_trapno; 719 eva = rcr2(); 720 sdtossd(&gdt[IDXSEL(frame->tf_cs & 0xffff)].sd, &softseg); 721 722 if (type <= MAX_TRAP_MSG) 723 printf("\n\nFatal trap %d: %s while in %s mode\n", 724 type, trap_msg[type], 725 frame->tf_eflags & PSL_VM ? "vm86" : 726 ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel"); 727#ifdef SMP 728 /* three seperate prints in case of a trap on an unmapped page */ 729 printf("mp_lock = %08x; ", mp_lock); 730 printf("cpuid = %d; ", cpuid); 731 printf("lapic.id = %08x\n", lapic.id); 732#endif 733 if (type == T_PAGEFLT) { 734 printf("fault virtual address = 0x%x\n", eva); 735 printf("fault code = %s %s, %s\n", 736 code & PGEX_U ? "user" : "supervisor", 737 code & PGEX_W ? "write" : "read", 738 code & PGEX_P ? "protection violation" : "page not present"); 739 } 740 printf("instruction pointer = 0x%x:0x%x\n", 741 frame->tf_cs & 0xffff, frame->tf_eip); 742 if ((ISPL(frame->tf_cs) == SEL_UPL) || (frame->tf_eflags & PSL_VM)) { 743 ss = frame->tf_ss & 0xffff; 744 esp = frame->tf_esp; 745 } else { 746 ss = GSEL(GDATA_SEL, SEL_KPL); 747 esp = (int)&frame->tf_esp; 748 } 749 printf("stack pointer = 0x%x:0x%x\n", ss, esp); 750 printf("frame pointer = 0x%x:0x%x\n", ss, frame->tf_ebp); 751 printf("code segment = base 0x%x, limit 0x%x, type 0x%x\n", 752 softseg.ssd_base, softseg.ssd_limit, softseg.ssd_type); 753 printf(" = DPL %d, pres %d, def32 %d, gran %d\n", 754 softseg.ssd_dpl, softseg.ssd_p, softseg.ssd_def32, 755 softseg.ssd_gran); 756 printf("processor eflags = "); 757 if (frame->tf_eflags & PSL_T) 758 printf("trace trap, "); 759 if (frame->tf_eflags & PSL_I) 760 printf("interrupt enabled, "); 761 if (frame->tf_eflags & PSL_NT) 762 printf("nested task, "); 763 if (frame->tf_eflags & PSL_RF) 764 printf("resume, "); 765 if (frame->tf_eflags & PSL_VM) 766 printf("vm86, "); 767 printf("IOPL = %d\n", (frame->tf_eflags & PSL_IOPL) >> 12); 768 printf("current process = "); 769 if (curproc) { 770 printf("%lu (%s)\n", 771 (u_long)curproc->p_pid, curproc->p_comm ? 772 curproc->p_comm : ""); 773 } else { 774 printf("Idle\n"); 775 } 776 printf("interrupt mask = "); 777 if ((cpl & net_imask) == net_imask) 778 printf("net "); 779 if ((cpl & tty_imask) == tty_imask) 780 printf("tty "); 781 if ((cpl & bio_imask) == bio_imask) 782 printf("bio "); 783 if ((cpl & cam_imask) == cam_imask) 784 printf("cam "); 785 if (cpl == 0) 786 printf("none"); 787#ifdef SMP 788/** 789 * XXX FIXME: 790 * we probably SHOULD have stopped the other CPUs before now! 791 * another CPU COULD have been touching cpl at this moment... 792 */ 793 printf(" <- SMP: XXX"); 794#endif 795 printf("\n"); 796 797#ifdef KDB 798 if (kdb_trap(&psl)) 799 return; 800#endif 801#ifdef DDB 802 if (kdb_trap (type, 0, frame)) 803 return; 804#endif 805 printf("trap number = %d\n", type); 806 if (type <= MAX_TRAP_MSG) 807 panic(trap_msg[type]); 808 else 809 panic("unknown/reserved trap"); 810} 811 812/* 813 * Double fault handler. Called when a fault occurs while writing 814 * a frame for a trap/exception onto the stack. This usually occurs 815 * when the stack overflows (such is the case with infinite recursion, 816 * for example). 817 * 818 * XXX Note that the current PTD gets replaced by IdlePTD when the 819 * task switch occurs. This means that the stack that was active at 820 * the time of the double fault is not available at <kstack> unless 821 * the machine was idle when the double fault occurred. The downside 822 * of this is that "trace <ebp>" in ddb won't work. 823 */ 824void 825dblfault_handler() 826{ 827 printf("\nFatal double fault:\n"); 828 printf("eip = 0x%x\n", common_tss.tss_eip); 829 printf("esp = 0x%x\n", common_tss.tss_esp); 830 printf("ebp = 0x%x\n", common_tss.tss_ebp); 831#ifdef SMP 832 /* three seperate prints in case of a trap on an unmapped page */ 833 printf("mp_lock = %08x; ", mp_lock); 834 printf("cpuid = %d; ", cpuid); 835 printf("lapic.id = %08x\n", lapic.id); 836#endif 837 panic("double fault"); 838} 839 840/* 841 * Compensate for 386 brain damage (missing URKR). 842 * This is a little simpler than the pagefault handler in trap() because 843 * it the page tables have already been faulted in and high addresses 844 * are thrown out early for other reasons. 845 */ 846int trapwrite(addr) 847 unsigned addr; 848{ 849 struct proc *p; 850 vm_offset_t va; 851 struct vmspace *vm; 852 int rv; 853 854 va = trunc_page((vm_offset_t)addr); 855 /* 856 * XXX - MAX is END. Changed > to >= for temp. fix. 857 */ 858 if (va >= VM_MAXUSER_ADDRESS) 859 return (1); 860 861 p = curproc; 862 vm = p->p_vmspace; 863 864 ++p->p_lock; 865 866 if ((caddr_t)va >= vm->vm_maxsaddr 867 && (caddr_t)va < (caddr_t)USRSTACK) { 868 if (!grow(p, va)) { 869 --p->p_lock; 870 return (1); 871 } 872 } 873 874 /* 875 * fault the data page 876 */ 877 rv = vm_fault(&vm->vm_map, va, VM_PROT_READ|VM_PROT_WRITE, VM_FAULT_DIRTY); 878 879 --p->p_lock; 880 881 if (rv != KERN_SUCCESS) 882 return 1; 883 884 return (0); 885} 886 887/* 888 * System call request from POSIX system call gate interface to kernel. 889 * Like trap(), argument is call by reference. 890 */ 891void 892syscall(frame) 893 struct trapframe frame; 894{ 895 caddr_t params; 896 int i; 897 struct sysent *callp; 898 struct proc *p = curproc; 899 u_quad_t sticks; 900 int error; 901 int args[8], rval[2]; 902 u_int code; 903 904 sticks = p->p_sticks; 905 if (ISPL(frame.tf_cs) != SEL_UPL) 906 panic("syscall"); 907 908 p->p_md.md_regs = &frame; 909 params = (caddr_t)frame.tf_esp + sizeof(int); 910 code = frame.tf_eax; 911 if (p->p_sysent->sv_prepsyscall) { 912 (*p->p_sysent->sv_prepsyscall)(&frame, args, &code, ¶ms); 913 } else { 914 /* 915 * Need to check if this is a 32 bit or 64 bit syscall. 916 */ 917 if (code == SYS_syscall) { 918 /* 919 * Code is first argument, followed by actual args. 920 */ 921 code = fuword(params); 922 params += sizeof(int); 923 } else if (code == SYS___syscall) { 924 /* 925 * Like syscall, but code is a quad, so as to maintain 926 * quad alignment for the rest of the arguments. 927 */ 928 code = fuword(params); 929 params += sizeof(quad_t); 930 } 931 } 932 933 if (p->p_sysent->sv_mask) 934 code &= p->p_sysent->sv_mask; 935 936 if (code >= p->p_sysent->sv_size) 937 callp = &p->p_sysent->sv_table[0]; 938 else 939 callp = &p->p_sysent->sv_table[code]; 940 941 if (params && (i = callp->sy_narg * sizeof(int)) && 942 (error = copyin(params, (caddr_t)args, (u_int)i))) { 943#ifdef KTRACE 944 if (KTRPOINT(p, KTR_SYSCALL)) 945 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 946#endif 947 goto bad; 948 } 949#ifdef KTRACE 950 if (KTRPOINT(p, KTR_SYSCALL)) 951 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 952#endif 953 rval[0] = 0; 954 rval[1] = frame.tf_edx; 955 956 error = (*callp->sy_call)(p, args, rval); 957 958 switch (error) { 959 960 case 0: 961 /* 962 * Reinitialize proc pointer `p' as it may be different 963 * if this is a child returning from fork syscall. 964 */ 965 p = curproc; 966 frame.tf_eax = rval[0]; 967 frame.tf_edx = rval[1]; 968 frame.tf_eflags &= ~PSL_C; 969 break; 970 971 case ERESTART: 972 /* 973 * Reconstruct pc, assuming lcall $X,y is 7 bytes, 974 * int 0x80 is 2 bytes. We saved this in tf_err. 975 */ 976 frame.tf_eip -= frame.tf_err; 977 break; 978 979 case EJUSTRETURN: 980 break; 981 982 default: 983bad: 984 if (p->p_sysent->sv_errsize) 985 if (error >= p->p_sysent->sv_errsize) 986 error = -1; /* XXX */ 987 else 988 error = p->p_sysent->sv_errtbl[error]; 989 frame.tf_eax = error; 990 frame.tf_eflags |= PSL_C; 991 break; 992 } 993 994 if ((frame.tf_eflags & PSL_T) && !(frame.tf_eflags & PSL_VM)) { 995 /* Traced syscall. */ 996 frame.tf_eflags &= ~PSL_T; 997 trapsignal(p, SIGTRAP, 0); 998 } 999 1000 userret(p, &frame, sticks); 1001 1002#ifdef KTRACE 1003 if (KTRPOINT(p, KTR_SYSRET)) 1004 ktrsysret(p->p_tracep, code, error, rval[0]); 1005#endif 1006} 1007 1008/* 1009 * Simplified back end of syscall(), used when returning from fork() 1010 * directly into user mode. 1011 */ 1012void 1013fork_return(p, frame) 1014 struct proc *p; 1015 struct trapframe frame; 1016{ 1017 frame.tf_eax = 0; /* Child returns zero */ 1018 frame.tf_eflags &= ~PSL_C; /* success */ 1019 frame.tf_edx = 1; 1020 1021 userret(p, &frame, 0); 1022#ifdef KTRACE 1023 if (KTRPOINT(p, KTR_SYSRET)) 1024 ktrsysret(p->p_tracep, SYS_fork, 0, 0); 1025#endif 1026} 1027