subr_syscall.c revision 31564
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 1.118 1997/12/04 21:21:26 jmg Exp $ 39 */ 40 41/* 42 * 386 Trap and System call handling 43 */ 44 45#include "opt_cpu.h" 46#include "opt_ddb.h" 47#include "opt_ktrace.h" 48#include "opt_vm86.h" 49 50#include <sys/param.h> 51#include <sys/systm.h> 52#include <sys/proc.h> 53#include <sys/pioctl.h> 54#include <sys/kernel.h> 55#include <sys/resourcevar.h> 56#include <sys/signalvar.h> 57#include <sys/syscall.h> 58#include <sys/sysent.h> 59#include <sys/vmmeter.h> 60#ifdef KTRACE 61#include <sys/ktrace.h> 62#endif 63 64#include <vm/vm.h> 65#include <vm/vm_param.h> 66#include <vm/vm_prot.h> 67#include <sys/lock.h> 68#include <vm/pmap.h> 69#include <vm/vm_kern.h> 70#include <vm/vm_map.h> 71#include <vm/vm_page.h> 72#include <vm/vm_extern.h> 73 74#include <machine/cpu.h> 75#include <machine/ipl.h> 76#include <machine/md_var.h> 77#include <machine/pcb.h> 78#ifdef SMP 79#include <machine/smp.h> 80#endif 81#include <machine/tss.h> 82 83#include <i386/isa/intr_machdep.h> 84 85#ifdef POWERFAIL_NMI 86#include <sys/syslog.h> 87#include <machine/clock.h> 88#endif 89 90#ifdef VM86 91#include <machine/vm86.h> 92#endif 93 94#include "isa.h" 95#include "npx.h" 96 97extern struct i386tss common_tss; 98 99int (*pmath_emulate) __P((struct trapframe *)); 100 101extern void trap __P((struct trapframe frame)); 102extern int trapwrite __P((unsigned addr)); 103extern void syscall __P((struct trapframe frame)); 104 105static int trap_pfault __P((struct trapframe *, int)); 106static void trap_fatal __P((struct trapframe *)); 107void dblfault_handler __P((void)); 108 109extern inthand_t IDTVEC(syscall); 110 111#define MAX_TRAP_MSG 28 112static char *trap_msg[] = { 113 "", /* 0 unused */ 114 "privileged instruction fault", /* 1 T_PRIVINFLT */ 115 "", /* 2 unused */ 116 "breakpoint instruction fault", /* 3 T_BPTFLT */ 117 "", /* 4 unused */ 118 "", /* 5 unused */ 119 "arithmetic trap", /* 6 T_ARITHTRAP */ 120 "system forced exception", /* 7 T_ASTFLT */ 121 "", /* 8 unused */ 122 "general protection fault", /* 9 T_PROTFLT */ 123 "trace trap", /* 10 T_TRCTRAP */ 124 "", /* 11 unused */ 125 "page fault", /* 12 T_PAGEFLT */ 126 "", /* 13 unused */ 127 "alignment fault", /* 14 T_ALIGNFLT */ 128 "", /* 15 unused */ 129 "", /* 16 unused */ 130 "", /* 17 unused */ 131 "integer divide fault", /* 18 T_DIVIDE */ 132 "non-maskable interrupt trap", /* 19 T_NMI */ 133 "overflow trap", /* 20 T_OFLOW */ 134 "FPU bounds check fault", /* 21 T_BOUND */ 135 "FPU device not available", /* 22 T_DNA */ 136 "double fault", /* 23 T_DOUBLEFLT */ 137 "FPU operand fetch fault", /* 24 T_FPOPFLT */ 138 "invalid TSS fault", /* 25 T_TSSFLT */ 139 "segment not present fault", /* 26 T_SEGNPFLT */ 140 "stack fault", /* 27 T_STKFLT */ 141 "machine check trap", /* 28 T_MCHK */ 142}; 143 144static void userret __P((struct proc *p, struct trapframe *frame, 145 u_quad_t oticks)); 146 147#if defined(I586_CPU) && !defined(NO_F00F_HACK) 148extern struct gate_descriptor *t_idt; 149extern int has_f00f_bug; 150#endif 151 152static inline void 153userret(p, frame, oticks) 154 struct proc *p; 155 struct trapframe *frame; 156 u_quad_t oticks; 157{ 158 int sig, s; 159 160 while ((sig = CURSIG(p)) != 0) 161 postsig(sig); 162 163#if 0 164 if (!want_resched && 165 (p->p_priority <= p->p_usrpri) && 166 (p->p_rtprio.type == RTP_PRIO_NORMAL)) { 167 int newpriority; 168 p->p_estcpu += 1; 169 newpriority = PUSER + p->p_estcpu / 4 + 2 * p->p_nice; 170 newpriority = min(newpriority, MAXPRI); 171 p->p_usrpri = newpriority; 172 } 173#endif 174 175 p->p_priority = p->p_usrpri; 176 if (want_resched) { 177 /* 178 * Since we are curproc, clock will normally just change 179 * our priority without moving us from one queue to another 180 * (since the running process is not on a queue.) 181 * If that happened after we setrunqueue ourselves but before we 182 * mi_switch()'ed, we might not be on the queue indicated by 183 * our priority. 184 */ 185 s = splhigh(); 186 setrunqueue(p); 187 p->p_stats->p_ru.ru_nivcsw++; 188 mi_switch(); 189 splx(s); 190 while ((sig = CURSIG(p)) != 0) 191 postsig(sig); 192 } 193 /* 194 * Charge system time if profiling. 195 */ 196 if (p->p_flag & P_PROFIL) 197 addupc_task(p, frame->tf_eip, 198 (u_int)(p->p_sticks - oticks) * psratio); 199 200 curpriority = p->p_priority; 201} 202 203/* 204 * Exception, fault, and trap interface to the FreeBSD kernel. 205 * This common code is called from assembly language IDT gate entry 206 * routines that prepare a suitable stack frame, and restore this 207 * frame after the exception has been processed. 208 */ 209 210void 211trap(frame) 212 struct trapframe frame; 213{ 214 struct proc *p = curproc; 215 u_quad_t sticks = 0; 216 int i = 0, ucode = 0, type, code; 217#ifdef DEBUG 218 u_long eva; 219#endif 220 221#if defined(I586_CPU) && !defined(NO_F00F_HACK) 222restart: 223#endif 224 type = frame.tf_trapno; 225 code = frame.tf_err; 226 227 if ((ISPL(frame.tf_cs) == SEL_UPL) || (frame.tf_eflags & PSL_VM)) { 228 /* user trap */ 229 230 sticks = p->p_sticks; 231 p->p_md.md_regs = &frame; 232 233 switch (type) { 234 case T_PRIVINFLT: /* privileged instruction fault */ 235 ucode = type; 236 i = SIGILL; 237 break; 238 239 case T_BPTFLT: /* bpt instruction fault */ 240 case T_TRCTRAP: /* trace trap */ 241 frame.tf_eflags &= ~PSL_T; 242 i = SIGTRAP; 243 break; 244 245 case T_ARITHTRAP: /* arithmetic trap */ 246 ucode = code; 247 i = SIGFPE; 248 break; 249 250 case T_ASTFLT: /* Allow process switch */ 251 astoff(); 252 cnt.v_soft++; 253 if (p->p_flag & P_OWEUPC) { 254 p->p_flag &= ~P_OWEUPC; 255 addupc_task(p, p->p_stats->p_prof.pr_addr, 256 p->p_stats->p_prof.pr_ticks); 257 } 258 goto out; 259 260 /* 261 * The following two traps can happen in 262 * vm86 mode, and, if so, we want to handle 263 * them specially. 264 */ 265 case T_PROTFLT: /* general protection fault */ 266 case T_STKFLT: /* stack fault */ 267#ifdef VM86 268 if (frame.tf_eflags & PSL_VM) { 269 i = vm86_emulate((struct vm86frame *)&frame); 270 if (i == 0) 271 goto out; 272 break; 273 } 274#endif /* VM86 */ 275 /* FALL THROUGH */ 276 277 case T_SEGNPFLT: /* segment not present fault */ 278 case T_TSSFLT: /* invalid TSS fault */ 279 case T_DOUBLEFLT: /* double fault */ 280 default: 281 ucode = code + BUS_SEGM_FAULT ; 282 i = SIGBUS; 283 break; 284 285 case T_PAGEFLT: /* page fault */ 286 i = trap_pfault(&frame, TRUE); 287 if (i == -1) 288 return; 289#if defined(I586_CPU) && !defined(NO_F00F_HACK) 290 if (i == -2) 291 goto restart; 292#endif 293 if (i == 0) 294 goto out; 295 296 ucode = T_PAGEFLT; 297 break; 298 299 case T_DIVIDE: /* integer divide fault */ 300 ucode = FPE_INTDIV_TRAP; 301 i = SIGFPE; 302 break; 303 304#if NISA > 0 305 case T_NMI: 306#ifdef POWERFAIL_NMI 307 goto handle_powerfail; 308#else /* !POWERFAIL_NMI */ 309#ifdef DDB 310 /* NMI can be hooked up to a pushbutton for debugging */ 311 printf ("NMI ... going to debugger\n"); 312 if (kdb_trap (type, 0, &frame)) 313 return; 314#endif /* DDB */ 315 /* machine/parity/power fail/"kitchen sink" faults */ 316 if (isa_nmi(code) == 0) return; 317 panic("NMI indicates hardware failure"); 318#endif /* POWERFAIL_NMI */ 319#endif /* NISA > 0 */ 320 321 case T_OFLOW: /* integer overflow fault */ 322 ucode = FPE_INTOVF_TRAP; 323 i = SIGFPE; 324 break; 325 326 case T_BOUND: /* bounds check fault */ 327 ucode = FPE_SUBRNG_TRAP; 328 i = SIGFPE; 329 break; 330 331 case T_DNA: 332#if NNPX > 0 333 /* if a transparent fault (due to context switch "late") */ 334 if (npxdna()) 335 return; 336#endif 337 if (!pmath_emulate) { 338 i = SIGFPE; 339 ucode = FPE_FPU_NP_TRAP; 340 break; 341 } 342 i = (*pmath_emulate)(&frame); 343 if (i == 0) { 344 if (!(frame.tf_eflags & PSL_T)) 345 return; 346 frame.tf_eflags &= ~PSL_T; 347 i = SIGTRAP; 348 } 349 /* else ucode = emulator_only_knows() XXX */ 350 break; 351 352 case T_FPOPFLT: /* FPU operand fetch fault */ 353 ucode = T_FPOPFLT; 354 i = SIGILL; 355 break; 356 } 357 } else { 358 /* kernel trap */ 359 360 switch (type) { 361 case T_PAGEFLT: /* page fault */ 362 (void) trap_pfault(&frame, FALSE); 363 return; 364 365 case T_DNA: 366#if NNPX > 0 367 /* 368 * The kernel is apparently using npx for copying. 369 * XXX this should be fatal unless the kernel has 370 * registered such use. 371 */ 372 if (npxdna()) 373 return; 374#endif 375 break; 376 377 case T_PROTFLT: /* general protection fault */ 378 case T_SEGNPFLT: /* segment not present fault */ 379 /* 380 * Invalid segment selectors and out of bounds 381 * %eip's and %esp's can be set up in user mode. 382 * This causes a fault in kernel mode when the 383 * kernel tries to return to user mode. We want 384 * to get this fault so that we can fix the 385 * problem here and not have to check all the 386 * selectors and pointers when the user changes 387 * them. 388 */ 389#define MAYBE_DORETI_FAULT(where, whereto) \ 390 do { \ 391 if (frame.tf_eip == (int)where) { \ 392 frame.tf_eip = (int)whereto; \ 393 return; \ 394 } \ 395 } while (0) 396 397 if (intr_nesting_level == 0) { 398 /* 399 * Invalid %fs's and %gs's can be created using 400 * procfs or PT_SETREGS or by invalidating the 401 * underlying LDT entry. This causes a fault 402 * in kernel mode when the kernel attempts to 403 * switch contexts. Lose the bad context 404 * (XXX) so that we can continue, and generate 405 * a signal. 406 */ 407 if (frame.tf_eip == (int)cpu_switch_load_fs) { 408 curpcb->pcb_fs = 0; 409 psignal(p, SIGBUS); 410 return; 411 } 412 if (frame.tf_eip == (int)cpu_switch_load_gs) { 413 curpcb->pcb_gs = 0; 414 psignal(p, SIGBUS); 415 return; 416 } 417 MAYBE_DORETI_FAULT(doreti_iret, 418 doreti_iret_fault); 419 MAYBE_DORETI_FAULT(doreti_popl_ds, 420 doreti_popl_ds_fault); 421 MAYBE_DORETI_FAULT(doreti_popl_es, 422 doreti_popl_es_fault); 423 if (curpcb && curpcb->pcb_onfault) { 424 frame.tf_eip = (int)curpcb->pcb_onfault; 425 return; 426 } 427 } 428 break; 429 430 case T_TSSFLT: 431 /* 432 * PSL_NT can be set in user mode and isn't cleared 433 * automatically when the kernel is entered. This 434 * causes a TSS fault when the kernel attempts to 435 * `iret' because the TSS link is uninitialized. We 436 * want to get this fault so that we can fix the 437 * problem here and not every time the kernel is 438 * entered. 439 */ 440 if (frame.tf_eflags & PSL_NT) { 441 frame.tf_eflags &= ~PSL_NT; 442 return; 443 } 444 break; 445 446 case T_TRCTRAP: /* trace trap */ 447 if (frame.tf_eip == (int)IDTVEC(syscall)) { 448 /* 449 * We've just entered system mode via the 450 * syscall lcall. Continue single stepping 451 * silently until the syscall handler has 452 * saved the flags. 453 */ 454 return; 455 } 456 if (frame.tf_eip == (int)IDTVEC(syscall) + 1) { 457 /* 458 * The syscall handler has now saved the 459 * flags. Stop single stepping it. 460 */ 461 frame.tf_eflags &= ~PSL_T; 462 return; 463 } 464 /* 465 * Fall through. 466 */ 467 case T_BPTFLT: 468 /* 469 * If DDB is enabled, let it handle the debugger trap. 470 * Otherwise, debugger traps "can't happen". 471 */ 472#ifdef DDB 473 if (kdb_trap (type, 0, &frame)) 474 return; 475#endif 476 break; 477 478#if NISA > 0 479 case T_NMI: 480#ifdef POWERFAIL_NMI 481#ifndef TIMER_FREQ 482# define TIMER_FREQ 1193182 483#endif 484 handle_powerfail: 485 { 486 static unsigned lastalert = 0; 487 488 if(time.tv_sec - lastalert > 10) 489 { 490 log(LOG_WARNING, "NMI: power fail\n"); 491 sysbeep(TIMER_FREQ/880, hz); 492 lastalert = time.tv_sec; 493 } 494 return; 495 } 496#else /* !POWERFAIL_NMI */ 497#ifdef DDB 498 /* NMI can be hooked up to a pushbutton for debugging */ 499 printf ("NMI ... going to debugger\n"); 500 if (kdb_trap (type, 0, &frame)) 501 return; 502#endif /* DDB */ 503 /* machine/parity/power fail/"kitchen sink" faults */ 504 if (isa_nmi(code) == 0) return; 505 /* FALL THROUGH */ 506#endif /* POWERFAIL_NMI */ 507#endif /* NISA > 0 */ 508 } 509 510 trap_fatal(&frame); 511 return; 512 } 513 514 trapsignal(p, i, ucode); 515 516#ifdef DEBUG 517 eva = rcr2(); 518 if (type <= MAX_TRAP_MSG) { 519 uprintf("fatal process exception: %s", 520 trap_msg[type]); 521 if ((type == T_PAGEFLT) || (type == T_PROTFLT)) 522 uprintf(", fault VA = 0x%x", eva); 523 uprintf("\n"); 524 } 525#endif 526 527out: 528 userret(p, &frame, sticks); 529} 530 531#ifdef notyet 532/* 533 * This version doesn't allow a page fault to user space while 534 * in the kernel. The rest of the kernel needs to be made "safe" 535 * before this can be used. I think the only things remaining 536 * to be made safe are the iBCS2 code and the process tracing/ 537 * debugging code. 538 */ 539static int 540trap_pfault(frame, usermode) 541 struct trapframe *frame; 542 int usermode; 543{ 544 vm_offset_t va; 545 struct vmspace *vm = NULL; 546 vm_map_t map = 0; 547 int rv = 0; 548 vm_prot_t ftype; 549 int eva; 550 struct proc *p = curproc; 551 552 if (frame->tf_err & PGEX_W) 553 ftype = VM_PROT_READ | VM_PROT_WRITE; 554 else 555 ftype = VM_PROT_READ; 556 557 eva = rcr2(); 558 va = trunc_page((vm_offset_t)eva); 559 560 if (va < VM_MIN_KERNEL_ADDRESS) { 561 vm_offset_t v; 562 vm_page_t mpte; 563 564 if (p == NULL || 565 (!usermode && va < VM_MAXUSER_ADDRESS && 566 (intr_nesting_level != 0 || curpcb == NULL || 567 curpcb->pcb_onfault == NULL))) { 568 trap_fatal(frame); 569 return (-1); 570 } 571 572 /* 573 * This is a fault on non-kernel virtual memory. 574 * vm is initialized above to NULL. If curproc is NULL 575 * or curproc->p_vmspace is NULL the fault is fatal. 576 */ 577 vm = p->p_vmspace; 578 if (vm == NULL) 579 goto nogo; 580 581 map = &vm->vm_map; 582 583 /* 584 * Keep swapout from messing with us during this 585 * critical time. 586 */ 587 ++p->p_lock; 588 589 /* 590 * Grow the stack if necessary 591 */ 592 if ((caddr_t)va > vm->vm_maxsaddr 593 && (caddr_t)va < (caddr_t)USRSTACK) { 594 if (!grow(p, va)) { 595 rv = KERN_FAILURE; 596 --p->p_lock; 597 goto nogo; 598 } 599 } 600 601 /* Fault in the user page: */ 602 rv = vm_fault(map, va, ftype, 603 (ftype & VM_PROT_WRITE) ? VM_FAULT_DIRTY : 0); 604 605 --p->p_lock; 606 } else { 607 /* 608 * Don't allow user-mode faults in kernel address space. 609 */ 610 if (usermode) 611 goto nogo; 612 613 /* 614 * Since we know that kernel virtual address addresses 615 * always have pte pages mapped, we just have to fault 616 * the page. 617 */ 618 rv = vm_fault(kernel_map, va, ftype, FALSE); 619 } 620 621 if (rv == KERN_SUCCESS) 622 return (0); 623nogo: 624 if (!usermode) { 625 if (intr_nesting_level == 0 && curpcb && curpcb->pcb_onfault) { 626 frame->tf_eip = (int)curpcb->pcb_onfault; 627 return (0); 628 } 629 trap_fatal(frame); 630 return (-1); 631 } 632 633 /* kludge to pass faulting virtual address to sendsig */ 634 frame->tf_err = eva; 635 636 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 637} 638#endif 639 640int 641trap_pfault(frame, usermode) 642 struct trapframe *frame; 643 int usermode; 644{ 645 vm_offset_t va; 646 struct vmspace *vm = NULL; 647 vm_map_t map = 0; 648 int rv = 0; 649 vm_prot_t ftype; 650 int eva; 651 struct proc *p = curproc; 652 653 eva = rcr2(); 654 va = trunc_page((vm_offset_t)eva); 655 656 if (va >= KERNBASE) { 657 /* 658 * Don't allow user-mode faults in kernel address space. 659 * An exception: if the faulting address is the invalid 660 * instruction entry in the IDT, then the Intel Pentium 661 * F00F bug workaround was triggered, and we need to 662 * treat it is as an illegal instruction, and not a page 663 * fault. 664 */ 665#if defined(I586_CPU) && !defined(NO_F00F_HACK) 666 if ((eva == (unsigned int)&t_idt[6]) && has_f00f_bug) { 667 frame->tf_trapno = T_PRIVINFLT; 668 return -2; 669 } 670#endif 671 if (usermode) 672 goto nogo; 673 674 map = kernel_map; 675 } else { 676 /* 677 * This is a fault on non-kernel virtual memory. 678 * vm is initialized above to NULL. If curproc is NULL 679 * or curproc->p_vmspace is NULL the fault is fatal. 680 */ 681 if (p != NULL) 682 vm = p->p_vmspace; 683 684 if (vm == NULL) 685 goto nogo; 686 687 map = &vm->vm_map; 688 } 689 690 if (frame->tf_err & PGEX_W) 691 ftype = VM_PROT_READ | VM_PROT_WRITE; 692 else 693 ftype = VM_PROT_READ; 694 695 if (map != kernel_map) { 696 /* 697 * Keep swapout from messing with us during this 698 * critical time. 699 */ 700 ++p->p_lock; 701 702 /* 703 * Grow the stack if necessary 704 */ 705 if ((caddr_t)va > vm->vm_maxsaddr 706 && (caddr_t)va < (caddr_t)USRSTACK) { 707 if (!grow(p, va)) { 708 rv = KERN_FAILURE; 709 --p->p_lock; 710 goto nogo; 711 } 712 } 713 714 /* Fault in the user page: */ 715 rv = vm_fault(map, va, ftype, 716 (ftype & VM_PROT_WRITE) ? VM_FAULT_DIRTY : 0); 717 718 --p->p_lock; 719 } else { 720 /* 721 * Don't have to worry about process locking or stacks in the kernel. 722 */ 723 rv = vm_fault(map, va, ftype, FALSE); 724 } 725 726 if (rv == KERN_SUCCESS) 727 return (0); 728nogo: 729 if (!usermode) { 730 if (intr_nesting_level == 0 && curpcb && curpcb->pcb_onfault) { 731 frame->tf_eip = (int)curpcb->pcb_onfault; 732 return (0); 733 } 734 trap_fatal(frame); 735 return (-1); 736 } 737 738 /* kludge to pass faulting virtual address to sendsig */ 739 frame->tf_err = eva; 740 741 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 742} 743 744static void 745trap_fatal(frame) 746 struct trapframe *frame; 747{ 748 int code, type, eva, ss, esp; 749 struct soft_segment_descriptor softseg; 750 751 code = frame->tf_err; 752 type = frame->tf_trapno; 753 eva = rcr2(); 754 sdtossd(&gdt[IDXSEL(frame->tf_cs & 0xffff)].sd, &softseg); 755 756 if (type <= MAX_TRAP_MSG) 757 printf("\n\nFatal trap %d: %s while in %s mode\n", 758 type, trap_msg[type], 759 frame->tf_eflags & PSL_VM ? "vm86" : 760 ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel"); 761#ifdef SMP 762 /* three seperate prints in case of a trap on an unmapped page */ 763 printf("mp_lock = %08x; ", mp_lock); 764 printf("cpuid = %d; ", cpuid); 765 printf("lapic.id = %08x\n", lapic.id); 766#endif 767 if (type == T_PAGEFLT) { 768 printf("fault virtual address = 0x%x\n", eva); 769 printf("fault code = %s %s, %s\n", 770 code & PGEX_U ? "user" : "supervisor", 771 code & PGEX_W ? "write" : "read", 772 code & PGEX_P ? "protection violation" : "page not present"); 773 } 774 printf("instruction pointer = 0x%x:0x%x\n", 775 frame->tf_cs & 0xffff, frame->tf_eip); 776 if ((ISPL(frame->tf_cs) == SEL_UPL) || (frame->tf_eflags & PSL_VM)) { 777 ss = frame->tf_ss & 0xffff; 778 esp = frame->tf_esp; 779 } else { 780 ss = GSEL(GDATA_SEL, SEL_KPL); 781 esp = (int)&frame->tf_esp; 782 } 783 printf("stack pointer = 0x%x:0x%x\n", ss, esp); 784 printf("frame pointer = 0x%x:0x%x\n", ss, frame->tf_ebp); 785 printf("code segment = base 0x%x, limit 0x%x, type 0x%x\n", 786 softseg.ssd_base, softseg.ssd_limit, softseg.ssd_type); 787 printf(" = DPL %d, pres %d, def32 %d, gran %d\n", 788 softseg.ssd_dpl, softseg.ssd_p, softseg.ssd_def32, 789 softseg.ssd_gran); 790 printf("processor eflags = "); 791 if (frame->tf_eflags & PSL_T) 792 printf("trace trap, "); 793 if (frame->tf_eflags & PSL_I) 794 printf("interrupt enabled, "); 795 if (frame->tf_eflags & PSL_NT) 796 printf("nested task, "); 797 if (frame->tf_eflags & PSL_RF) 798 printf("resume, "); 799 if (frame->tf_eflags & PSL_VM) 800 printf("vm86, "); 801 printf("IOPL = %d\n", (frame->tf_eflags & PSL_IOPL) >> 12); 802 printf("current process = "); 803 if (curproc) { 804 printf("%lu (%s)\n", 805 (u_long)curproc->p_pid, curproc->p_comm ? 806 curproc->p_comm : ""); 807 } else { 808 printf("Idle\n"); 809 } 810 printf("interrupt mask = "); 811 if ((cpl & net_imask) == net_imask) 812 printf("net "); 813 if ((cpl & tty_imask) == tty_imask) 814 printf("tty "); 815 if ((cpl & bio_imask) == bio_imask) 816 printf("bio "); 817 if ((cpl & cam_imask) == cam_imask) 818 printf("cam "); 819 if (cpl == 0) 820 printf("none"); 821#ifdef SMP 822/** 823 * XXX FIXME: 824 * we probably SHOULD have stopped the other CPUs before now! 825 * another CPU COULD have been touching cpl at this moment... 826 */ 827 printf(" <- SMP: XXX"); 828#endif 829 printf("\n"); 830 831#ifdef KDB 832 if (kdb_trap(&psl)) 833 return; 834#endif 835#ifdef DDB 836 if (kdb_trap (type, 0, frame)) 837 return; 838#endif 839 printf("trap number = %d\n", type); 840 if (type <= MAX_TRAP_MSG) 841 panic(trap_msg[type]); 842 else 843 panic("unknown/reserved trap"); 844} 845 846/* 847 * Double fault handler. Called when a fault occurs while writing 848 * a frame for a trap/exception onto the stack. This usually occurs 849 * when the stack overflows (such is the case with infinite recursion, 850 * for example). 851 * 852 * XXX Note that the current PTD gets replaced by IdlePTD when the 853 * task switch occurs. This means that the stack that was active at 854 * the time of the double fault is not available at <kstack> unless 855 * the machine was idle when the double fault occurred. The downside 856 * of this is that "trace <ebp>" in ddb won't work. 857 */ 858void 859dblfault_handler() 860{ 861 printf("\nFatal double fault:\n"); 862 printf("eip = 0x%x\n", common_tss.tss_eip); 863 printf("esp = 0x%x\n", common_tss.tss_esp); 864 printf("ebp = 0x%x\n", common_tss.tss_ebp); 865#ifdef SMP 866 /* three seperate prints in case of a trap on an unmapped page */ 867 printf("mp_lock = %08x; ", mp_lock); 868 printf("cpuid = %d; ", cpuid); 869 printf("lapic.id = %08x\n", lapic.id); 870#endif 871 panic("double fault"); 872} 873 874/* 875 * Compensate for 386 brain damage (missing URKR). 876 * This is a little simpler than the pagefault handler in trap() because 877 * it the page tables have already been faulted in and high addresses 878 * are thrown out early for other reasons. 879 */ 880int trapwrite(addr) 881 unsigned addr; 882{ 883 struct proc *p; 884 vm_offset_t va; 885 struct vmspace *vm; 886 int rv; 887 888 va = trunc_page((vm_offset_t)addr); 889 /* 890 * XXX - MAX is END. Changed > to >= for temp. fix. 891 */ 892 if (va >= VM_MAXUSER_ADDRESS) 893 return (1); 894 895 p = curproc; 896 vm = p->p_vmspace; 897 898 ++p->p_lock; 899 900 if ((caddr_t)va >= vm->vm_maxsaddr 901 && (caddr_t)va < (caddr_t)USRSTACK) { 902 if (!grow(p, va)) { 903 --p->p_lock; 904 return (1); 905 } 906 } 907 908 /* 909 * fault the data page 910 */ 911 rv = vm_fault(&vm->vm_map, va, VM_PROT_READ|VM_PROT_WRITE, VM_FAULT_DIRTY); 912 913 --p->p_lock; 914 915 if (rv != KERN_SUCCESS) 916 return 1; 917 918 return (0); 919} 920 921/* 922 * System call request from POSIX system call gate interface to kernel. 923 * Like trap(), argument is call by reference. 924 */ 925void 926syscall(frame) 927 struct trapframe frame; 928{ 929 caddr_t params; 930 int i; 931 struct sysent *callp; 932 struct proc *p = curproc; 933 u_quad_t sticks; 934 int error; 935 int args[8]; 936 u_int code; 937 938#ifdef DIAGNOSTIC 939 if (ISPL(frame.tf_cs) != SEL_UPL) 940 panic("syscall"); 941#endif 942 sticks = p->p_sticks; 943 p->p_md.md_regs = &frame; 944 params = (caddr_t)frame.tf_esp + sizeof(int); 945 code = frame.tf_eax; 946 if (p->p_sysent->sv_prepsyscall) { 947 (*p->p_sysent->sv_prepsyscall)(&frame, args, &code, ¶ms); 948 } else { 949 /* 950 * Need to check if this is a 32 bit or 64 bit syscall. 951 */ 952 if (code == SYS_syscall) { 953 /* 954 * Code is first argument, followed by actual args. 955 */ 956 code = fuword(params); 957 params += sizeof(int); 958 } else if (code == SYS___syscall) { 959 /* 960 * Like syscall, but code is a quad, so as to maintain 961 * quad alignment for the rest of the arguments. 962 */ 963 code = fuword(params); 964 params += sizeof(quad_t); 965 } 966 } 967 968 if (p->p_sysent->sv_mask) 969 code &= p->p_sysent->sv_mask; 970 971 if (code >= p->p_sysent->sv_size) 972 callp = &p->p_sysent->sv_table[0]; 973 else 974 callp = &p->p_sysent->sv_table[code]; 975 976 if (params && (i = callp->sy_narg * sizeof(int)) && 977 (error = copyin(params, (caddr_t)args, (u_int)i))) { 978#ifdef KTRACE 979 if (KTRPOINT(p, KTR_SYSCALL)) 980 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 981#endif 982 goto bad; 983 } 984#ifdef KTRACE 985 if (KTRPOINT(p, KTR_SYSCALL)) 986 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 987#endif 988 p->p_retval[0] = 0; 989 p->p_retval[1] = frame.tf_edx; 990 991 STOPEVENT(p, S_SCE, callp->sy_narg); 992 993 error = (*callp->sy_call)(p, args); 994 995 switch (error) { 996 997 case 0: 998 /* 999 * Reinitialize proc pointer `p' as it may be different 1000 * if this is a child returning from fork syscall. 1001 */ 1002 p = curproc; 1003 frame.tf_eax = p->p_retval[0]; 1004 frame.tf_edx = p->p_retval[1]; 1005 frame.tf_eflags &= ~PSL_C; 1006 break; 1007 1008 case ERESTART: 1009 /* 1010 * Reconstruct pc, assuming lcall $X,y is 7 bytes, 1011 * int 0x80 is 2 bytes. We saved this in tf_err. 1012 */ 1013 frame.tf_eip -= frame.tf_err; 1014 break; 1015 1016 case EJUSTRETURN: 1017 break; 1018 1019 default: 1020bad: 1021 if (p->p_sysent->sv_errsize) 1022 if (error >= p->p_sysent->sv_errsize) 1023 error = -1; /* XXX */ 1024 else 1025 error = p->p_sysent->sv_errtbl[error]; 1026 frame.tf_eax = error; 1027 frame.tf_eflags |= PSL_C; 1028 break; 1029 } 1030 1031 if ((frame.tf_eflags & PSL_T) && !(frame.tf_eflags & PSL_VM)) { 1032 /* Traced syscall. */ 1033 frame.tf_eflags &= ~PSL_T; 1034 trapsignal(p, SIGTRAP, 0); 1035 } 1036 1037 userret(p, &frame, sticks); 1038 1039#ifdef KTRACE 1040 if (KTRPOINT(p, KTR_SYSRET)) 1041 ktrsysret(p->p_tracep, code, error, p->p_retval[0]); 1042#endif 1043 1044 /* 1045 * This works because errno is findable through the 1046 * register set. If we ever support an emulation where this 1047 * is not the case, this code will need to be revisited. 1048 */ 1049 STOPEVENT(p, S_SCX, code); 1050 1051} 1052 1053/* 1054 * Simplified back end of syscall(), used when returning from fork() 1055 * directly into user mode. 1056 */ 1057void 1058fork_return(p, frame) 1059 struct proc *p; 1060 struct trapframe frame; 1061{ 1062 frame.tf_eax = 0; /* Child returns zero */ 1063 frame.tf_eflags &= ~PSL_C; /* success */ 1064 frame.tf_edx = 1; 1065 1066 userret(p, &frame, 0); 1067#ifdef KTRACE 1068 if (KTRPOINT(p, KTR_SYSRET)) 1069 ktrsysret(p->p_tracep, SYS_fork, 0, 0); 1070#endif 1071} 1072