subr_syscall.c revision 17117
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 1.79 1996/06/25 20:01:57 bde 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/acct.h> 52#include <sys/kernel.h> 53#include <sys/syscall.h> 54#include <sys/sysent.h> 55#include <sys/queue.h> 56#include <sys/vmmeter.h> 57#ifdef KTRACE 58#include <sys/ktrace.h> 59#endif 60 61#include <vm/vm.h> 62#include <vm/vm_param.h> 63#include <vm/vm_prot.h> 64#include <vm/lock.h> 65#include <vm/pmap.h> 66#include <vm/vm_kern.h> 67#include <vm/vm_map.h> 68#include <vm/vm_page.h> 69#include <vm/vm_extern.h> 70 71#include <sys/user.h> 72 73#include <machine/cpu.h> 74#include <machine/md_var.h> 75#include <machine/psl.h> 76#include <machine/reg.h> 77#include <machine/trap.h> 78#include <machine/../isa/isa_device.h> 79 80#ifdef POWERFAIL_NMI 81# include <syslog.h> 82# include <machine/clock.h> 83#endif 84 85#include "isa.h" 86#include "npx.h" 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 27 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}; 131 132static void userret __P((struct proc *p, struct trapframe *frame, 133 u_quad_t oticks)); 134 135static inline void 136userret(p, frame, oticks) 137 struct proc *p; 138 struct trapframe *frame; 139 u_quad_t oticks; 140{ 141 int sig, s; 142 143 while ((sig = CURSIG(p)) != 0) 144 postsig(sig); 145 p->p_priority = p->p_usrpri; 146 if (want_resched) { 147 /* 148 * Since we are curproc, clock will normally just change 149 * our priority without moving us from one queue to another 150 * (since the running process is not on a queue.) 151 * If that happened after we setrunqueue ourselves but before we 152 * mi_switch()'ed, we might not be on the queue indicated by 153 * our priority. 154 */ 155 s = splclock(); 156 setrunqueue(p); 157 p->p_stats->p_ru.ru_nivcsw++; 158 mi_switch(); 159 splx(s); 160 while ((sig = CURSIG(p)) != 0) 161 postsig(sig); 162 } 163 /* 164 * Charge system time if profiling. 165 */ 166 if (p->p_flag & P_PROFIL) 167 addupc_task(p, frame->tf_eip, 168 (u_int)(p->p_sticks - oticks) * psratio); 169 170 curpriority = p->p_priority; 171} 172 173/* 174 * Exception, fault, and trap interface to the FreeBSD kernel. 175 * This common code is called from assembly language IDT gate entry 176 * routines that prepare a suitable stack frame, and restore this 177 * frame after the exception has been processed. 178 */ 179 180void 181trap(frame) 182 struct trapframe frame; 183{ 184 struct proc *p = curproc; 185 u_quad_t sticks = 0; 186 int i = 0, ucode = 0, type, code; 187#ifdef DEBUG 188 u_long eva; 189#endif 190 191 type = frame.tf_trapno; 192 code = frame.tf_err; 193 194 if (ISPL(frame.tf_cs) == SEL_UPL) { 195 /* user trap */ 196 197 sticks = p->p_sticks; 198 p->p_md.md_regs = (int *)&frame; 199 200 switch (type) { 201 case T_PRIVINFLT: /* privileged instruction fault */ 202 ucode = type; 203 i = SIGILL; 204 break; 205 206 case T_BPTFLT: /* bpt instruction fault */ 207 case T_TRCTRAP: /* trace trap */ 208 frame.tf_eflags &= ~PSL_T; 209 i = SIGTRAP; 210 break; 211 212 case T_ARITHTRAP: /* arithmetic trap */ 213 ucode = code; 214 i = SIGFPE; 215 break; 216 217 case T_ASTFLT: /* Allow process switch */ 218 astoff(); 219 cnt.v_soft++; 220 if (p->p_flag & P_OWEUPC) { 221 p->p_flag &= ~P_OWEUPC; 222 addupc_task(p, p->p_stats->p_prof.pr_addr, 223 p->p_stats->p_prof.pr_ticks); 224 } 225 goto out; 226 227 case T_PROTFLT: /* general protection fault */ 228 case T_SEGNPFLT: /* segment not present fault */ 229 case T_STKFLT: /* stack fault */ 230 case T_TSSFLT: /* invalid TSS fault */ 231 case T_DOUBLEFLT: /* double fault */ 232 default: 233 ucode = code + BUS_SEGM_FAULT ; 234 i = SIGBUS; 235 break; 236 237 case T_PAGEFLT: /* page fault */ 238 i = trap_pfault(&frame, TRUE); 239 if (i == -1) 240 return; 241 if (i == 0) 242 goto out; 243 244 ucode = T_PAGEFLT; 245 break; 246 247 case T_DIVIDE: /* integer divide fault */ 248 ucode = FPE_INTDIV_TRAP; 249 i = SIGFPE; 250 break; 251 252#if NISA > 0 253 case T_NMI: 254#ifdef POWERFAIL_NMI 255 goto handle_powerfail; 256#else /* !POWERFAIL_NMI */ 257#ifdef DDB 258 /* NMI can be hooked up to a pushbutton for debugging */ 259 printf ("NMI ... going to debugger\n"); 260 if (kdb_trap (type, 0, &frame)) 261 return; 262#endif /* DDB */ 263 /* machine/parity/power fail/"kitchen sink" faults */ 264 if (isa_nmi(code) == 0) return; 265 panic("NMI indicates hardware failure"); 266#endif /* POWERFAIL_NMI */ 267#endif /* NISA > 0 */ 268 269 case T_OFLOW: /* integer overflow fault */ 270 ucode = FPE_INTOVF_TRAP; 271 i = SIGFPE; 272 break; 273 274 case T_BOUND: /* bounds check fault */ 275 ucode = FPE_SUBRNG_TRAP; 276 i = SIGFPE; 277 break; 278 279 case T_DNA: 280#if NNPX > 0 281 /* if a transparent fault (due to context switch "late") */ 282 if (npxdna()) 283 return; 284#endif 285 if (!pmath_emulate) { 286 i = SIGFPE; 287 ucode = FPE_FPU_NP_TRAP; 288 break; 289 } 290 i = (*pmath_emulate)(&frame); 291 if (i == 0) { 292 if (!(frame.tf_eflags & PSL_T)) 293 return; 294 frame.tf_eflags &= ~PSL_T; 295 i = SIGTRAP; 296 } 297 /* else ucode = emulator_only_knows() XXX */ 298 break; 299 300 case T_FPOPFLT: /* FPU operand fetch fault */ 301 ucode = T_FPOPFLT; 302 i = SIGILL; 303 break; 304 } 305 } else { 306 /* kernel trap */ 307 308 switch (type) { 309 case T_PAGEFLT: /* page fault */ 310 (void) trap_pfault(&frame, FALSE); 311 return; 312 313 case T_DNA: 314#if NNPX > 0 315 /* 316 * The kernel is apparently using npx for copying. 317 * XXX this should be fatal unless the kernel has 318 * registered such use. 319 */ 320 if (npxdna()) 321 return; 322#endif 323 break; 324 325 case T_PROTFLT: /* general protection fault */ 326 case T_SEGNPFLT: /* segment not present fault */ 327 /* 328 * Invalid segment selectors and out of bounds 329 * %eip's and %esp's can be set up in user mode. 330 * This causes a fault in kernel mode when the 331 * kernel tries to return to user mode. We want 332 * to get this fault so that we can fix the 333 * problem here and not have to check all the 334 * selectors and pointers when the user changes 335 * them. 336 */ 337#define MAYBE_DORETI_FAULT(where, whereto) \ 338 do { \ 339 if (frame.tf_eip == (int)where) { \ 340 frame.tf_eip = (int)whereto; \ 341 return; \ 342 } \ 343 } while (0) 344 345 if (intr_nesting_level == 0) { 346 MAYBE_DORETI_FAULT(doreti_iret, 347 doreti_iret_fault); 348 MAYBE_DORETI_FAULT(doreti_popl_ds, 349 doreti_popl_ds_fault); 350 MAYBE_DORETI_FAULT(doreti_popl_es, 351 doreti_popl_es_fault); 352 } 353 if (curpcb && curpcb->pcb_onfault) { 354 frame.tf_eip = (int)curpcb->pcb_onfault; 355 return; 356 } 357 break; 358 359 case T_TSSFLT: 360 /* 361 * PSL_NT can be set in user mode and isn't cleared 362 * automatically when the kernel is entered. This 363 * causes a TSS fault when the kernel attempts to 364 * `iret' because the TSS link is uninitialized. We 365 * want to get this fault so that we can fix the 366 * problem here and not every time the kernel is 367 * entered. 368 */ 369 if (frame.tf_eflags & PSL_NT) { 370 frame.tf_eflags &= ~PSL_NT; 371 return; 372 } 373 break; 374 375 case T_TRCTRAP: /* trace trap */ 376 if (frame.tf_eip == (int)IDTVEC(syscall)) { 377 /* 378 * We've just entered system mode via the 379 * syscall lcall. Continue single stepping 380 * silently until the syscall handler has 381 * saved the flags. 382 */ 383 return; 384 } 385 if (frame.tf_eip == (int)IDTVEC(syscall) + 1) { 386 /* 387 * The syscall handler has now saved the 388 * flags. Stop single stepping it. 389 */ 390 frame.tf_eflags &= ~PSL_T; 391 return; 392 } 393 /* 394 * Fall through. 395 */ 396 case T_BPTFLT: 397 /* 398 * If DDB is enabled, let it handle the debugger trap. 399 * Otherwise, debugger traps "can't happen". 400 */ 401#ifdef DDB 402 if (kdb_trap (type, 0, &frame)) 403 return; 404#endif 405 break; 406 407#if NISA > 0 408 case T_NMI: 409#ifdef POWERFAIL_NMI 410#ifndef TIMER_FREQ 411# define TIMER_FREQ 1193182 412#endif 413 handle_powerfail: 414 { 415 static unsigned lastalert = 0; 416 417 if(time.tv_sec - lastalert > 10) 418 { 419 log(LOG_WARNING, "NMI: power fail\n"); 420 sysbeep(TIMER_FREQ/880, hz); 421 lastalert = time.tv_sec; 422 } 423 return; 424 } 425#else /* !POWERFAIL_NMI */ 426#ifdef DDB 427 /* NMI can be hooked up to a pushbutton for debugging */ 428 printf ("NMI ... going to debugger\n"); 429 if (kdb_trap (type, 0, &frame)) 430 return; 431#endif /* DDB */ 432 /* machine/parity/power fail/"kitchen sink" faults */ 433 if (isa_nmi(code) == 0) return; 434 /* FALL THROUGH */ 435#endif /* POWERFAIL_NMI */ 436#endif /* NISA > 0 */ 437 } 438 439 trap_fatal(&frame); 440 return; 441 } 442 443 trapsignal(p, i, ucode); 444 445#ifdef DEBUG 446 eva = rcr2(); 447 if (type <= MAX_TRAP_MSG) { 448 uprintf("fatal process exception: %s", 449 trap_msg[type]); 450 if ((type == T_PAGEFLT) || (type == T_PROTFLT)) 451 uprintf(", fault VA = 0x%x", eva); 452 uprintf("\n"); 453 } 454#endif 455 456out: 457 userret(p, &frame, sticks); 458} 459 460#ifdef notyet 461/* 462 * This version doesn't allow a page fault to user space while 463 * in the kernel. The rest of the kernel needs to be made "safe" 464 * before this can be used. I think the only things remaining 465 * to be made safe are the iBCS2 code and the process tracing/ 466 * debugging code. 467 */ 468static int 469trap_pfault(frame, usermode) 470 struct trapframe *frame; 471 int usermode; 472{ 473 vm_offset_t va; 474 struct vmspace *vm = NULL; 475 vm_map_t map = 0; 476 int rv = 0; 477 vm_prot_t ftype; 478 int eva; 479 struct proc *p = curproc; 480 481 if (frame->tf_err & PGEX_W) 482 ftype = VM_PROT_READ | VM_PROT_WRITE; 483 else 484 ftype = VM_PROT_READ; 485 486 eva = rcr2(); 487 va = trunc_page((vm_offset_t)eva); 488 489 if (va < VM_MIN_KERNEL_ADDRESS) { 490 vm_offset_t v; 491 vm_page_t mpte; 492 493 if (p == NULL || 494 (!usermode && va < VM_MAXUSER_ADDRESS && 495 (curpcb == NULL || curpcb->pcb_onfault == NULL))) { 496 trap_fatal(frame); 497 return (-1); 498 } 499 500 /* 501 * This is a fault on non-kernel virtual memory. 502 * vm is initialized above to NULL. If curproc is NULL 503 * or curproc->p_vmspace is NULL the fault is fatal. 504 */ 505 vm = p->p_vmspace; 506 if (vm == NULL) 507 goto nogo; 508 509 map = &vm->vm_map; 510 511 /* 512 * Keep swapout from messing with us during this 513 * critical time. 514 */ 515 ++p->p_lock; 516 517 /* 518 * Grow the stack if necessary 519 */ 520 if ((caddr_t)va > vm->vm_maxsaddr 521 && (caddr_t)va < (caddr_t)USRSTACK) { 522 if (!grow(p, va)) { 523 rv = KERN_FAILURE; 524 --p->p_lock; 525 goto nogo; 526 } 527 } 528 529 /* Fault in the user page: */ 530 rv = vm_fault(map, va, ftype, FALSE); 531 532 --p->p_lock; 533 } else { 534 /* 535 * Don't allow user-mode faults in kernel address space. 536 */ 537 if (usermode) 538 goto nogo; 539 540 /* 541 * Since we know that kernel virtual address addresses 542 * always have pte pages mapped, we just have to fault 543 * the page. 544 */ 545 rv = vm_fault(kernel_map, va, ftype, FALSE); 546 } 547 548 if (rv == KERN_SUCCESS) 549 return (0); 550nogo: 551 if (!usermode) { 552 if (curpcb && curpcb->pcb_onfault) { 553 frame->tf_eip = (int)curpcb->pcb_onfault; 554 return (0); 555 } 556 trap_fatal(frame); 557 return (-1); 558 } 559 560 /* kludge to pass faulting virtual address to sendsig */ 561 frame->tf_err = eva; 562 563 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 564} 565#endif 566 567int 568trap_pfault(frame, usermode) 569 struct trapframe *frame; 570 int usermode; 571{ 572 vm_offset_t va; 573 struct vmspace *vm = NULL; 574 vm_map_t map = 0; 575 int rv = 0; 576 vm_prot_t ftype; 577 int eva; 578 struct proc *p = curproc; 579 580 eva = rcr2(); 581 va = trunc_page((vm_offset_t)eva); 582 583 if (va >= KERNBASE) { 584 /* 585 * Don't allow user-mode faults in kernel address space. 586 */ 587 if (usermode) 588 goto nogo; 589 590 map = kernel_map; 591 } else { 592 /* 593 * This is a fault on non-kernel virtual memory. 594 * vm is initialized above to NULL. If curproc is NULL 595 * or curproc->p_vmspace is NULL the fault is fatal. 596 */ 597 if (p != NULL) 598 vm = p->p_vmspace; 599 600 if (vm == NULL) 601 goto nogo; 602 603 map = &vm->vm_map; 604 } 605 606 if (frame->tf_err & PGEX_W) 607 ftype = VM_PROT_READ | VM_PROT_WRITE; 608 else 609 ftype = VM_PROT_READ; 610 611 if (map != kernel_map) { 612 /* 613 * Keep swapout from messing with us during this 614 * critical time. 615 */ 616 ++p->p_lock; 617 618 /* 619 * Grow the stack if necessary 620 */ 621 if ((caddr_t)va > vm->vm_maxsaddr 622 && (caddr_t)va < (caddr_t)USRSTACK) { 623 if (!grow(p, va)) { 624 rv = KERN_FAILURE; 625 --p->p_lock; 626 goto nogo; 627 } 628 } 629 630 /* Fault in the user page: */ 631 rv = vm_fault(map, va, ftype, FALSE); 632 633 --p->p_lock; 634 } else { 635 /* 636 * Since we know that kernel virtual address addresses 637 * always have pte pages mapped, we just have to fault 638 * the page. 639 */ 640 rv = vm_fault(map, va, ftype, FALSE); 641 } 642 643 if (rv == KERN_SUCCESS) 644 return (0); 645nogo: 646 if (!usermode) { 647 if (curpcb && curpcb->pcb_onfault) { 648 frame->tf_eip = (int)curpcb->pcb_onfault; 649 return (0); 650 } 651 trap_fatal(frame); 652 return (-1); 653 } 654 655 /* kludge to pass faulting virtual address to sendsig */ 656 frame->tf_err = eva; 657 658 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 659} 660 661static void 662trap_fatal(frame) 663 struct trapframe *frame; 664{ 665 int code, type, eva, ss, esp; 666 struct soft_segment_descriptor softseg; 667 668 code = frame->tf_err; 669 type = frame->tf_trapno; 670 eva = rcr2(); 671 sdtossd(&gdt[IDXSEL(frame->tf_cs & 0xffff)].sd, &softseg); 672 673 if (type <= MAX_TRAP_MSG) 674 printf("\n\nFatal trap %d: %s while in %s mode\n", 675 type, trap_msg[type], 676 ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel"); 677 if (type == T_PAGEFLT) { 678 printf("fault virtual address = 0x%x\n", eva); 679 printf("fault code = %s %s, %s\n", 680 code & PGEX_U ? "user" : "supervisor", 681 code & PGEX_W ? "write" : "read", 682 code & PGEX_P ? "protection violation" : "page not present"); 683 } 684 printf("instruction pointer = 0x%x:0x%x\n", 685 frame->tf_cs & 0xffff, frame->tf_eip); 686 if (ISPL(frame->tf_cs) == SEL_UPL) { 687 ss = frame->tf_ss & 0xffff; 688 esp = frame->tf_esp; 689 } else { 690 ss = GSEL(GDATA_SEL, SEL_KPL); 691 esp = (int)&frame->tf_esp; 692 } 693 printf("stack pointer = 0x%x:0x%x\n", ss, esp); 694 printf("frame pointer = 0x%x:0x%x\n", ss, frame->tf_ebp); 695 printf("code segment = base 0x%x, limit 0x%x, type 0x%x\n", 696 softseg.ssd_base, softseg.ssd_limit, softseg.ssd_type); 697 printf(" = DPL %d, pres %d, def32 %d, gran %d\n", 698 softseg.ssd_dpl, softseg.ssd_p, softseg.ssd_def32, 699 softseg.ssd_gran); 700 printf("processor eflags = "); 701 if (frame->tf_eflags & PSL_T) 702 printf("trace trap, "); 703 if (frame->tf_eflags & PSL_I) 704 printf("interrupt enabled, "); 705 if (frame->tf_eflags & PSL_NT) 706 printf("nested task, "); 707 if (frame->tf_eflags & PSL_RF) 708 printf("resume, "); 709 if (frame->tf_eflags & PSL_VM) 710 printf("vm86, "); 711 printf("IOPL = %d\n", (frame->tf_eflags & PSL_IOPL) >> 12); 712 printf("current process = "); 713 if (curproc) { 714 printf("%lu (%s)\n", 715 (u_long)curproc->p_pid, curproc->p_comm ? 716 curproc->p_comm : ""); 717 } else { 718 printf("Idle\n"); 719 } 720 printf("interrupt mask = "); 721 if ((cpl & net_imask) == net_imask) 722 printf("net "); 723 if ((cpl & tty_imask) == tty_imask) 724 printf("tty "); 725 if ((cpl & bio_imask) == bio_imask) 726 printf("bio "); 727 if (cpl == 0) 728 printf("none"); 729 printf("\n"); 730 731#ifdef KDB 732 if (kdb_trap(&psl)) 733 return; 734#endif 735#ifdef DDB 736 if (kdb_trap (type, 0, frame)) 737 return; 738#endif 739 if (type <= MAX_TRAP_MSG) 740 panic(trap_msg[type]); 741 else 742 panic("unknown/reserved trap"); 743} 744 745/* 746 * Double fault handler. Called when a fault occurs while writing 747 * a frame for a trap/exception onto the stack. This usually occurs 748 * when the stack overflows (such is the case with infinite recursion, 749 * for example). 750 * 751 * XXX Note that the current PTD gets replaced by IdlePTD when the 752 * task switch occurs. This means that the stack that was active at 753 * the time of the double fault is not available at <kstack> unless 754 * the machine was idle when the double fault occurred. The downside 755 * of this is that "trace <ebp>" in ddb won't work. 756 */ 757void 758dblfault_handler() 759{ 760 struct pcb *pcb = curpcb; 761 762 if (pcb != NULL) { 763 printf("\nFatal double fault:\n"); 764 printf("eip = 0x%x\n", pcb->pcb_tss.tss_eip); 765 printf("esp = 0x%x\n", pcb->pcb_tss.tss_esp); 766 printf("ebp = 0x%x\n", pcb->pcb_tss.tss_ebp); 767 } 768 769 panic("double fault"); 770} 771 772/* 773 * Compensate for 386 brain damage (missing URKR). 774 * This is a little simpler than the pagefault handler in trap() because 775 * it the page tables have already been faulted in and high addresses 776 * are thrown out early for other reasons. 777 */ 778int trapwrite(addr) 779 unsigned addr; 780{ 781 struct proc *p; 782 vm_offset_t va, v; 783 struct vmspace *vm; 784 int rv; 785 786 va = trunc_page((vm_offset_t)addr); 787 /* 788 * XXX - MAX is END. Changed > to >= for temp. fix. 789 */ 790 if (va >= VM_MAXUSER_ADDRESS) 791 return (1); 792 793 p = curproc; 794 vm = p->p_vmspace; 795 796 ++p->p_lock; 797 798 if ((caddr_t)va >= vm->vm_maxsaddr 799 && (caddr_t)va < (caddr_t)USRSTACK) { 800 if (!grow(p, va)) { 801 --p->p_lock; 802 return (1); 803 } 804 } 805 806 v = trunc_page(vtopte(va)); 807 808 /* 809 * fault the data page 810 */ 811 rv = vm_fault(&vm->vm_map, va, VM_PROT_READ|VM_PROT_WRITE, FALSE); 812 813 --p->p_lock; 814 815 if (rv != KERN_SUCCESS) 816 return 1; 817 818 return (0); 819} 820 821/* 822 * System call request from POSIX system call gate interface to kernel. 823 * Like trap(), argument is call by reference. 824 */ 825void 826syscall(frame) 827 struct trapframe frame; 828{ 829 caddr_t params; 830 int i; 831 struct sysent *callp; 832 struct proc *p = curproc; 833 u_quad_t sticks; 834 int error; 835 int args[8], rval[2]; 836 u_int code; 837 838 sticks = p->p_sticks; 839 if (ISPL(frame.tf_cs) != SEL_UPL) 840 panic("syscall"); 841 842 p->p_md.md_regs = (int *)&frame; 843 params = (caddr_t)frame.tf_esp + sizeof(int); 844 code = frame.tf_eax; 845 if (p->p_sysent->sv_prepsyscall) { 846 (*p->p_sysent->sv_prepsyscall)(&frame, args, &code, ¶ms); 847 } else { 848 /* 849 * Need to check if this is a 32 bit or 64 bit syscall. 850 */ 851 if (code == SYS_syscall) { 852 /* 853 * Code is first argument, followed by actual args. 854 */ 855 code = fuword(params); 856 params += sizeof(int); 857 } else if (code == SYS___syscall) { 858 /* 859 * Like syscall, but code is a quad, so as to maintain 860 * quad alignment for the rest of the arguments. 861 */ 862 code = fuword(params); 863 params += sizeof(quad_t); 864 } 865 } 866 867 if (p->p_sysent->sv_mask) 868 code &= p->p_sysent->sv_mask; 869 870 if (code >= p->p_sysent->sv_size) 871 callp = &p->p_sysent->sv_table[0]; 872 else 873 callp = &p->p_sysent->sv_table[code]; 874 875 if (params && (i = callp->sy_narg * sizeof(int)) && 876 (error = copyin(params, (caddr_t)args, (u_int)i))) { 877#ifdef KTRACE 878 if (KTRPOINT(p, KTR_SYSCALL)) 879 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 880#endif 881 goto bad; 882 } 883#ifdef KTRACE 884 if (KTRPOINT(p, KTR_SYSCALL)) 885 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 886#endif 887 rval[0] = 0; 888 rval[1] = frame.tf_edx; 889 890 error = (*callp->sy_call)(p, args, rval); 891 892 switch (error) { 893 894 case 0: 895 /* 896 * Reinitialize proc pointer `p' as it may be different 897 * if this is a child returning from fork syscall. 898 */ 899 p = curproc; 900 frame.tf_eax = rval[0]; 901 frame.tf_edx = rval[1]; 902 frame.tf_eflags &= ~PSL_C; 903 break; 904 905 case ERESTART: 906 /* 907 * Reconstruct pc, assuming lcall $X,y is 7 bytes, 908 * int 0x80 is 2 bytes. We saved this in tf_err. 909 */ 910 frame.tf_eip -= frame.tf_err; 911 break; 912 913 case EJUSTRETURN: 914 break; 915 916 default: 917bad: 918 if (p->p_sysent->sv_errsize) 919 if (error >= p->p_sysent->sv_errsize) 920 error = -1; /* XXX */ 921 else 922 error = p->p_sysent->sv_errtbl[error]; 923 frame.tf_eax = error; 924 frame.tf_eflags |= PSL_C; 925 break; 926 } 927 928 if (frame.tf_eflags & PSL_T) { 929 /* Traced syscall. */ 930 frame.tf_eflags &= ~PSL_T; 931 trapsignal(p, SIGTRAP, 0); 932 } 933 934 userret(p, &frame, sticks); 935 936#ifdef KTRACE 937 if (KTRPOINT(p, KTR_SYSRET)) 938 ktrsysret(p->p_tracep, code, error, rval[0]); 939#endif 940} 941