subr_syscall.c revision 13490
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 1.70 1996/01/04 21:11:03 wollman 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)); 93extern void linux_syscall __P((struct trapframe frame)); 94 95static int trap_pfault __P((struct trapframe *, int)); 96static void trap_fatal __P((struct trapframe *)); 97void dblfault_handler __P((void)); 98 99extern inthand_t IDTVEC(syscall); 100 101#define MAX_TRAP_MSG 27 102static char *trap_msg[] = { 103 "", /* 0 unused */ 104 "privileged instruction fault", /* 1 T_PRIVINFLT */ 105 "", /* 2 unused */ 106 "breakpoint instruction fault", /* 3 T_BPTFLT */ 107 "", /* 4 unused */ 108 "", /* 5 unused */ 109 "arithmetic trap", /* 6 T_ARITHTRAP */ 110 "system forced exception", /* 7 T_ASTFLT */ 111 "", /* 8 unused */ 112 "general protection fault", /* 9 T_PROTFLT */ 113 "trace trap", /* 10 T_TRCTRAP */ 114 "", /* 11 unused */ 115 "page fault", /* 12 T_PAGEFLT */ 116 "", /* 13 unused */ 117 "alignment fault", /* 14 T_ALIGNFLT */ 118 "", /* 15 unused */ 119 "", /* 16 unused */ 120 "", /* 17 unused */ 121 "integer divide fault", /* 18 T_DIVIDE */ 122 "non-maskable interrupt trap", /* 19 T_NMI */ 123 "overflow trap", /* 20 T_OFLOW */ 124 "FPU bounds check fault", /* 21 T_BOUND */ 125 "FPU device not available", /* 22 T_DNA */ 126 "double fault", /* 23 T_DOUBLEFLT */ 127 "FPU operand fetch fault", /* 24 T_FPOPFLT */ 128 "invalid TSS fault", /* 25 T_TSSFLT */ 129 "segment not present fault", /* 26 T_SEGNPFLT */ 130 "stack fault", /* 27 T_STKFLT */ 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 p->p_priority = p->p_usrpri; 147 if (want_resched) { 148 /* 149 * Since we are curproc, clock will normally just change 150 * our priority without moving us from one queue to another 151 * (since the running process is not on a queue.) 152 * If that happened after we setrunqueue ourselves but before we 153 * mi_switch()'ed, we might not be on the queue indicated by 154 * our priority. 155 */ 156 s = splclock(); 157 setrunqueue(p); 158 p->p_stats->p_ru.ru_nivcsw++; 159 mi_switch(); 160 splx(s); 161 while ((sig = CURSIG(p)) != 0) 162 postsig(sig); 163 } 164 /* 165 * Charge system time if profiling. 166 */ 167 if (p->p_flag & P_PROFIL) { 168 u_quad_t ticks = p->p_sticks - oticks; 169 170 if (ticks) { 171#ifdef PROFTIMER 172 extern int profscale; 173 addupc(frame->tf_eip, &p->p_stats->p_prof, 174 ticks * profscale); 175#else 176 addupc(frame->tf_eip, &p->p_stats->p_prof, ticks); 177#endif 178 } 179 } 180 curpriority = p->p_priority; 181} 182 183/* 184 * Exception, fault, and trap interface to the FreeBSD kernel. 185 * This common code is called from assembly language IDT gate entry 186 * routines that prepare a suitable stack frame, and restore this 187 * frame after the exception has been processed. 188 */ 189 190void 191trap(frame) 192 struct trapframe frame; 193{ 194 struct proc *p = curproc; 195 u_quad_t sticks = 0; 196 int i = 0, ucode = 0, type, code; 197#ifdef DEBUG 198 u_long eva; 199#endif 200 201 type = frame.tf_trapno; 202 code = frame.tf_err; 203 204 if (ISPL(frame.tf_cs) == SEL_UPL) { 205 /* user trap */ 206 207 sticks = p->p_sticks; 208 p->p_md.md_regs = (int *)&frame; 209 210 switch (type) { 211 case T_PRIVINFLT: /* privileged instruction fault */ 212 ucode = type; 213 i = SIGILL; 214 break; 215 216 case T_BPTFLT: /* bpt instruction fault */ 217 case T_TRCTRAP: /* trace trap */ 218 frame.tf_eflags &= ~PSL_T; 219 i = SIGTRAP; 220 break; 221 222 case T_ARITHTRAP: /* arithmetic trap */ 223 ucode = code; 224 i = SIGFPE; 225 break; 226 227 case T_ASTFLT: /* Allow process switch */ 228 astoff(); 229 cnt.v_soft++; 230 if (p->p_flag & P_OWEUPC) { 231 addupc(frame.tf_eip, &p->p_stats->p_prof, 1); 232 p->p_flag &= ~P_OWEUPC; 233 } 234 goto out; 235 236 case T_PROTFLT: /* general protection fault */ 237 case T_SEGNPFLT: /* segment not present fault */ 238 case T_STKFLT: /* stack fault */ 239 case T_TSSFLT: /* invalid TSS fault */ 240 case T_DOUBLEFLT: /* double fault */ 241 default: 242 ucode = code + BUS_SEGM_FAULT ; 243 i = SIGBUS; 244 break; 245 246 case T_PAGEFLT: /* page fault */ 247 i = trap_pfault(&frame, TRUE); 248 if (i == -1) 249 return; 250 if (i == 0) 251 goto out; 252 253 ucode = T_PAGEFLT; 254 break; 255 256 case T_DIVIDE: /* integer divide fault */ 257 ucode = FPE_INTDIV_TRAP; 258 i = SIGFPE; 259 break; 260 261#if NISA > 0 262 case T_NMI: 263#ifdef POWERFAIL_NMI 264 goto handle_powerfail; 265#else /* !POWERFAIL_NMI */ 266#ifdef DDB 267 /* NMI can be hooked up to a pushbutton for debugging */ 268 printf ("NMI ... going to debugger\n"); 269 if (kdb_trap (type, 0, &frame)) 270 return; 271#endif /* DDB */ 272 /* machine/parity/power fail/"kitchen sink" faults */ 273 if (isa_nmi(code) == 0) return; 274 panic("NMI indicates hardware failure"); 275#endif /* POWERFAIL_NMI */ 276#endif /* NISA > 0 */ 277 278 case T_OFLOW: /* integer overflow fault */ 279 ucode = FPE_INTOVF_TRAP; 280 i = SIGFPE; 281 break; 282 283 case T_BOUND: /* bounds check fault */ 284 ucode = FPE_SUBRNG_TRAP; 285 i = SIGFPE; 286 break; 287 288 case T_DNA: 289#if NNPX > 0 290 /* if a transparent fault (due to context switch "late") */ 291 if (npxdna()) 292 return; 293#endif /* NNPX > 0 */ 294 295 if (!pmath_emulate) { 296 i = SIGFPE; 297 ucode = FPE_FPU_NP_TRAP; 298 break; 299 } 300 i = (*pmath_emulate)(&frame); 301 if (i == 0) { 302 if (!(frame.tf_eflags & PSL_T)) 303 return; 304 frame.tf_eflags &= ~PSL_T; 305 i = SIGTRAP; 306 } 307 /* else ucode = emulator_only_knows() XXX */ 308 break; 309 310 case T_FPOPFLT: /* FPU operand fetch fault */ 311 ucode = T_FPOPFLT; 312 i = SIGILL; 313 break; 314 } 315 } else { 316 /* kernel trap */ 317 318 switch (type) { 319 case T_PAGEFLT: /* page fault */ 320 (void) trap_pfault(&frame, FALSE); 321 return; 322 323 case T_PROTFLT: /* general protection fault */ 324 case T_SEGNPFLT: /* segment not present fault */ 325 /* 326 * Invalid segment selectors and out of bounds 327 * %eip's and %esp's can be set up in user mode. 328 * This causes a fault in kernel mode when the 329 * kernel tries to return to user mode. We want 330 * to get this fault so that we can fix the 331 * problem here and not have to check all the 332 * selectors and pointers when the user changes 333 * them. 334 */ 335#define MAYBE_DORETI_FAULT(where, whereto) \ 336 do { \ 337 if (frame.tf_eip == (int)where) { \ 338 frame.tf_eip = (int)whereto; \ 339 return; \ 340 } \ 341 } while (0) 342 343 if (intr_nesting_level == 0) { 344 MAYBE_DORETI_FAULT(doreti_iret, 345 doreti_iret_fault); 346 MAYBE_DORETI_FAULT(doreti_popl_ds, 347 doreti_popl_ds_fault); 348 MAYBE_DORETI_FAULT(doreti_popl_es, 349 doreti_popl_es_fault); 350 } 351 if (curpcb && curpcb->pcb_onfault) { 352 frame.tf_eip = (int)curpcb->pcb_onfault; 353 return; 354 } 355 break; 356 357 case T_TSSFLT: 358 /* 359 * PSL_NT can be set in user mode and isn't cleared 360 * automatically when the kernel is entered. This 361 * causes a TSS fault when the kernel attempts to 362 * `iret' because the TSS link is uninitialized. We 363 * want to get this fault so that we can fix the 364 * problem here and not every time the kernel is 365 * entered. 366 */ 367 if (frame.tf_eflags & PSL_NT) { 368 frame.tf_eflags &= ~PSL_NT; 369 return; 370 } 371 break; 372 373 case T_TRCTRAP: /* trace trap */ 374 if (frame.tf_eip == (int)IDTVEC(syscall)) { 375 /* 376 * We've just entered system mode via the 377 * syscall lcall. Continue single stepping 378 * silently until the syscall handler has 379 * saved the flags. 380 */ 381 return; 382 } 383 if (frame.tf_eip == (int)IDTVEC(syscall) + 1) { 384 /* 385 * The syscall handler has now saved the 386 * flags. Stop single stepping it. 387 */ 388 frame.tf_eflags &= ~PSL_T; 389 return; 390 } 391 /* 392 * Fall through. 393 */ 394 case T_BPTFLT: 395 /* 396 * If DDB is enabled, let it handle the debugger trap. 397 * Otherwise, debugger traps "can't happen". 398 */ 399#ifdef DDB 400 if (kdb_trap (type, 0, &frame)) 401 return; 402#endif 403 break; 404 405#if NISA > 0 406 case T_NMI: 407#ifdef POWERFAIL_NMI 408#ifndef TIMER_FREQ 409# define TIMER_FREQ 1193182 410#endif 411 handle_powerfail: 412 { 413 static unsigned lastalert = 0; 414 415 if(time.tv_sec - lastalert > 10) 416 { 417 log(LOG_WARNING, "NMI: power fail\n"); 418 sysbeep(TIMER_FREQ/880, hz); 419 lastalert = time.tv_sec; 420 } 421 return; 422 } 423#else /* !POWERFAIL_NMI */ 424#ifdef DDB 425 /* NMI can be hooked up to a pushbutton for debugging */ 426 printf ("NMI ... going to debugger\n"); 427 if (kdb_trap (type, 0, &frame)) 428 return; 429#endif /* DDB */ 430 /* machine/parity/power fail/"kitchen sink" faults */ 431 if (isa_nmi(code) == 0) return; 432 /* FALL THROUGH */ 433#endif /* POWERFAIL_NMI */ 434#endif /* NISA > 0 */ 435 } 436 437 trap_fatal(&frame); 438 return; 439 } 440 441 trapsignal(p, i, ucode); 442 443#ifdef DEBUG 444 eva = rcr2(); 445 if (type <= MAX_TRAP_MSG) { 446 uprintf("fatal process exception: %s", 447 trap_msg[type]); 448 if ((type == T_PAGEFLT) || (type == T_PROTFLT)) 449 uprintf(", fault VA = 0x%x", eva); 450 uprintf("\n"); 451 } 452#endif 453 454out: 455 userret(p, &frame, sticks); 456} 457 458#ifdef notyet 459/* 460 * This version doesn't allow a page fault to user space while 461 * in the kernel. The rest of the kernel needs to be made "safe" 462 * before this can be used. I think the only things remaining 463 * to be made safe are the iBCS2 code and the process tracing/ 464 * debugging code. 465 */ 466static int 467trap_pfault(frame, usermode) 468 struct trapframe *frame; 469 int usermode; 470{ 471 vm_offset_t va; 472 struct vmspace *vm = NULL; 473 vm_map_t map = 0; 474 int rv = 0; 475 vm_prot_t ftype; 476 int eva; 477 struct proc *p = curproc; 478 479 if (frame->tf_err & PGEX_W) 480 ftype = VM_PROT_READ | VM_PROT_WRITE; 481 else 482 ftype = VM_PROT_READ; 483 484 eva = rcr2(); 485 va = trunc_page((vm_offset_t)eva); 486 487 if (va < VM_MIN_KERNEL_ADDRESS) { 488 vm_offset_t v; 489 vm_page_t ptepg; 490 491 if (p == NULL || 492 (!usermode && va < VM_MAXUSER_ADDRESS && 493 (curpcb == NULL || curpcb->pcb_onfault == NULL))) { 494 trap_fatal(frame); 495 return (-1); 496 } 497 498 /* 499 * This is a fault on non-kernel virtual memory. 500 * vm is initialized above to NULL. If curproc is NULL 501 * or curproc->p_vmspace is NULL the fault is fatal. 502 */ 503 vm = p->p_vmspace; 504 if (vm == NULL) 505 goto nogo; 506 507 map = &vm->vm_map; 508 509 /* 510 * Keep swapout from messing with us during this 511 * critical time. 512 */ 513 ++p->p_lock; 514 515 /* 516 * Grow the stack if necessary 517 */ 518 if ((caddr_t)va > vm->vm_maxsaddr 519 && (caddr_t)va < (caddr_t)USRSTACK) { 520 if (!grow(p, va)) { 521 rv = KERN_FAILURE; 522 --p->p_lock; 523 goto nogo; 524 } 525 } 526 527 /* 528 * Check if page table is mapped, if not, 529 * fault it first 530 */ 531 v = (vm_offset_t) vtopte(va); 532 533 /* Fault the pte only if needed: */ 534 if (*((int *)vtopte(v)) == 0) 535 (void) vm_fault(map, trunc_page(v), VM_PROT_WRITE, FALSE); 536 537 pmap_use_pt( vm_map_pmap(map), va); 538 539 /* Fault in the user page: */ 540 rv = vm_fault(map, va, ftype, FALSE); 541 542 pmap_unuse_pt( vm_map_pmap(map), va); 543 544 --p->p_lock; 545 } else { 546 /* 547 * Don't allow user-mode faults in kernel address space. 548 */ 549 if (usermode) 550 goto nogo; 551 552 /* 553 * Since we know that kernel virtual address addresses 554 * always have pte pages mapped, we just have to fault 555 * the page. 556 */ 557 rv = vm_fault(kernel_map, va, ftype, FALSE); 558 } 559 560 if (rv == KERN_SUCCESS) 561 return (0); 562nogo: 563 if (!usermode) { 564 if (curpcb && curpcb->pcb_onfault) { 565 frame->tf_eip = (int)curpcb->pcb_onfault; 566 return (0); 567 } 568 trap_fatal(frame); 569 return (-1); 570 } 571 572 /* kludge to pass faulting virtual address to sendsig */ 573 frame->tf_err = eva; 574 575 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 576} 577#endif 578 579int 580trap_pfault(frame, usermode) 581 struct trapframe *frame; 582 int usermode; 583{ 584 vm_offset_t va; 585 struct vmspace *vm = NULL; 586 vm_map_t map = 0; 587 int rv = 0; 588 vm_prot_t ftype; 589 int eva; 590 struct proc *p = curproc; 591 592 eva = rcr2(); 593 va = trunc_page((vm_offset_t)eva); 594 595 if (va >= KERNBASE) { 596 /* 597 * Don't allow user-mode faults in kernel address space. 598 */ 599 if (usermode) 600 goto nogo; 601 602 map = kernel_map; 603 } else { 604 /* 605 * This is a fault on non-kernel virtual memory. 606 * vm is initialized above to NULL. If curproc is NULL 607 * or curproc->p_vmspace is NULL the fault is fatal. 608 */ 609 if (p != NULL) 610 vm = p->p_vmspace; 611 612 if (vm == NULL) 613 goto nogo; 614 615 map = &vm->vm_map; 616 } 617 618 if (frame->tf_err & PGEX_W) 619 ftype = VM_PROT_READ | VM_PROT_WRITE; 620 else 621 ftype = VM_PROT_READ; 622 623 if (map != kernel_map) { 624 vm_offset_t v; 625 626 /* 627 * Keep swapout from messing with us during this 628 * critical time. 629 */ 630 ++p->p_lock; 631 632 /* 633 * Grow the stack if necessary 634 */ 635 if ((caddr_t)va > vm->vm_maxsaddr 636 && (caddr_t)va < (caddr_t)USRSTACK) { 637 if (!grow(p, va)) { 638 rv = KERN_FAILURE; 639 --p->p_lock; 640 goto nogo; 641 } 642 } 643 644 /* 645 * Check if page table is mapped, if not, 646 * fault it first 647 */ 648 v = (vm_offset_t) vtopte(va); 649 650 /* Fault the pte only if needed: */ 651 if (*((int *)vtopte(v)) == 0) 652 (void) vm_fault(map, 653 trunc_page(v), VM_PROT_WRITE, FALSE); 654 655 pmap_use_pt( vm_map_pmap(map), va); 656 657 /* Fault in the user page: */ 658 rv = vm_fault(map, va, ftype, FALSE); 659 660 pmap_unuse_pt( vm_map_pmap(map), va); 661 662 --p->p_lock; 663 } else { 664 /* 665 * Since we know that kernel virtual address addresses 666 * always have pte pages mapped, we just have to fault 667 * the page. 668 */ 669 rv = vm_fault(map, va, ftype, FALSE); 670 } 671 672 if (rv == KERN_SUCCESS) 673 return (0); 674nogo: 675 if (!usermode) { 676 if (curpcb && curpcb->pcb_onfault) { 677 frame->tf_eip = (int)curpcb->pcb_onfault; 678 return (0); 679 } 680 trap_fatal(frame); 681 return (-1); 682 } 683 684 /* kludge to pass faulting virtual address to sendsig */ 685 frame->tf_err = eva; 686 687 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 688} 689 690static void 691trap_fatal(frame) 692 struct trapframe *frame; 693{ 694 int code, type, eva; 695 struct soft_segment_descriptor softseg; 696 697 code = frame->tf_err; 698 type = frame->tf_trapno; 699 eva = rcr2(); 700 sdtossd(&gdt[IDXSEL(frame->tf_cs & 0xffff)].sd, &softseg); 701 702 if (type <= MAX_TRAP_MSG) 703 printf("\n\nFatal trap %d: %s while in %s mode\n", 704 type, trap_msg[type], 705 ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel"); 706 if (type == T_PAGEFLT) { 707 printf("fault virtual address = 0x%x\n", eva); 708 printf("fault code = %s %s, %s\n", 709 code & PGEX_U ? "user" : "supervisor", 710 code & PGEX_W ? "write" : "read", 711 code & PGEX_P ? "protection violation" : "page not present"); 712 } 713 printf("instruction pointer = 0x%x:0x%x\n", frame->tf_cs & 0xffff, frame->tf_eip); 714 printf("code segment = base 0x%x, limit 0x%x, type 0x%x\n", 715 softseg.ssd_base, softseg.ssd_limit, softseg.ssd_type); 716 printf(" = DPL %d, pres %d, def32 %d, gran %d\n", 717 softseg.ssd_dpl, softseg.ssd_p, softseg.ssd_def32, softseg.ssd_gran); 718 printf("processor eflags = "); 719 if (frame->tf_eflags & PSL_T) 720 printf("trace/trap, "); 721 if (frame->tf_eflags & PSL_I) 722 printf("interrupt enabled, "); 723 if (frame->tf_eflags & PSL_NT) 724 printf("nested task, "); 725 if (frame->tf_eflags & PSL_RF) 726 printf("resume, "); 727 if (frame->tf_eflags & PSL_VM) 728 printf("vm86, "); 729 printf("IOPL = %d\n", (frame->tf_eflags & PSL_IOPL) >> 12); 730 printf("current process = "); 731 if (curproc) { 732 printf("%lu (%s)\n", 733 (u_long)curproc->p_pid, curproc->p_comm ? 734 curproc->p_comm : ""); 735 } else { 736 printf("Idle\n"); 737 } 738 printf("interrupt mask = "); 739 if ((cpl & net_imask) == net_imask) 740 printf("net "); 741 if ((cpl & tty_imask) == tty_imask) 742 printf("tty "); 743 if ((cpl & bio_imask) == bio_imask) 744 printf("bio "); 745 if (cpl == 0) 746 printf("none"); 747 printf("\n"); 748 749#ifdef KDB 750 if (kdb_trap(&psl)) 751 return; 752#endif 753#ifdef DDB 754 if (kdb_trap (type, 0, frame)) 755 return; 756#endif 757 if (type <= MAX_TRAP_MSG) 758 panic(trap_msg[type]); 759 else 760 panic("unknown/reserved trap"); 761} 762 763/* 764 * Double fault handler. Called when a fault occurs while writing 765 * a frame for a trap/exception onto the stack. This usually occurs 766 * when the stack overflows (such is the case with infinite recursion, 767 * for example). 768 * 769 * XXX Note that the current PTD gets replaced by IdlePTD when the 770 * task switch occurs. This means that the stack that was active at 771 * the time of the double fault is not available at <kstack> unless 772 * the machine was idle when the double fault occurred. The downside 773 * of this is that "trace <ebp>" in ddb won't work. 774 */ 775void 776dblfault_handler() 777{ 778 struct pcb *pcb = curpcb; 779 780 if (pcb != NULL) { 781 printf("\nFatal double fault:\n"); 782 printf("eip = 0x%x\n", pcb->pcb_tss.tss_eip); 783 printf("esp = 0x%x\n", pcb->pcb_tss.tss_esp); 784 printf("ebp = 0x%x\n", pcb->pcb_tss.tss_ebp); 785 } 786 787 panic("double fault"); 788} 789 790/* 791 * Compensate for 386 brain damage (missing URKR). 792 * This is a little simpler than the pagefault handler in trap() because 793 * it the page tables have already been faulted in and high addresses 794 * are thrown out early for other reasons. 795 */ 796int trapwrite(addr) 797 unsigned addr; 798{ 799 struct proc *p; 800 vm_offset_t va, v; 801 struct vmspace *vm; 802 int rv; 803 804 va = trunc_page((vm_offset_t)addr); 805 /* 806 * XXX - MAX is END. Changed > to >= for temp. fix. 807 */ 808 if (va >= VM_MAXUSER_ADDRESS) 809 return (1); 810 811 p = curproc; 812 vm = p->p_vmspace; 813 814 ++p->p_lock; 815 816 if ((caddr_t)va >= vm->vm_maxsaddr 817 && (caddr_t)va < (caddr_t)USRSTACK) { 818 if (!grow(p, va)) { 819 --p->p_lock; 820 return (1); 821 } 822 } 823 824 v = trunc_page(vtopte(va)); 825 826 /* 827 * wire the pte page 828 */ 829 if (va < USRSTACK) { 830 vm_map_pageable(&vm->vm_map, v, round_page(v+1), FALSE); 831 } 832 833 /* 834 * fault the data page 835 */ 836 rv = vm_fault(&vm->vm_map, va, VM_PROT_READ|VM_PROT_WRITE, FALSE); 837 838 /* 839 * unwire the pte page 840 */ 841 if (va < USRSTACK) { 842 vm_map_pageable(&vm->vm_map, v, round_page(v+1), TRUE); 843 } 844 845 --p->p_lock; 846 847 if (rv != KERN_SUCCESS) 848 return 1; 849 850 return (0); 851} 852 853/* 854 * System call request from POSIX system call gate interface to kernel. 855 * Like trap(), argument is call by reference. 856 */ 857void 858syscall(frame) 859 struct trapframe frame; 860{ 861 caddr_t params; 862 int i; 863 struct sysent *callp; 864 struct proc *p = curproc; 865 u_quad_t sticks; 866 int error; 867 int args[8], rval[2]; 868 u_int code; 869 870 sticks = p->p_sticks; 871 if (ISPL(frame.tf_cs) != SEL_UPL) 872 panic("syscall"); 873 874 p->p_md.md_regs = (int *)&frame; 875 params = (caddr_t)frame.tf_esp + sizeof(int); 876 code = frame.tf_eax; 877 /* 878 * Need to check if this is a 32 bit or 64 bit syscall. 879 */ 880 if (code == SYS_syscall) { 881 /* 882 * Code is first argument, followed by actual args. 883 */ 884 code = fuword(params); 885 params += sizeof(int); 886 } else if (code == SYS___syscall) { 887 /* 888 * Like syscall, but code is a quad, so as to maintain 889 * quad alignment for the rest of the arguments. 890 */ 891 code = fuword(params); 892 params += sizeof(quad_t); 893 } 894 895 if (p->p_sysent->sv_mask) 896 code &= p->p_sysent->sv_mask; 897 898 if (code >= p->p_sysent->sv_size) 899 callp = &p->p_sysent->sv_table[0]; 900 else 901 callp = &p->p_sysent->sv_table[code]; 902 903 if ((i = callp->sy_narg * sizeof(int)) && 904 (error = copyin(params, (caddr_t)args, (u_int)i))) { 905#ifdef KTRACE 906 if (KTRPOINT(p, KTR_SYSCALL)) 907 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 908#endif 909 goto bad; 910 } 911#ifdef KTRACE 912 if (KTRPOINT(p, KTR_SYSCALL)) 913 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 914#endif 915 rval[0] = 0; 916 rval[1] = frame.tf_edx; 917 918 error = (*callp->sy_call)(p, args, rval); 919 920 switch (error) { 921 922 case 0: 923 /* 924 * Reinitialize proc pointer `p' as it may be different 925 * if this is a child returning from fork syscall. 926 */ 927 p = curproc; 928 frame.tf_eax = rval[0]; 929 frame.tf_edx = rval[1]; 930 frame.tf_eflags &= ~PSL_C; 931 break; 932 933 case ERESTART: 934 /* 935 * Reconstruct pc, assuming lcall $X,y is 7 bytes. 936 */ 937 frame.tf_eip -= 7; 938 break; 939 940 case EJUSTRETURN: 941 break; 942 943 default: 944bad: 945 if (p->p_sysent->sv_errsize) 946 if (error >= p->p_sysent->sv_errsize) 947 error = -1; /* XXX */ 948 else 949 error = p->p_sysent->sv_errtbl[error]; 950 frame.tf_eax = error; 951 frame.tf_eflags |= PSL_C; 952 break; 953 } 954 955 if (frame.tf_eflags & PSL_T) { 956 /* Traced syscall. */ 957 frame.tf_eflags &= ~PSL_T; 958 trapsignal(p, SIGTRAP, 0); 959 } 960 961 userret(p, &frame, sticks); 962 963#ifdef KTRACE 964 if (KTRPOINT(p, KTR_SYSRET)) 965 ktrsysret(p->p_tracep, code, error, rval[0]); 966#endif 967} 968 969#if defined(COMPAT_LINUX) || defined(LINUX) 970void 971linux_syscall(frame) 972 struct trapframe frame; 973{ 974 struct proc *p = curproc; 975 struct sysent *callp; 976 u_quad_t sticks; 977 int error; 978 int rval[2]; 979 u_int code; 980 struct linux_syscall_args { 981 int arg1; 982 int arg2; 983 int arg3; 984 int arg4; 985 int arg5; 986 } args; 987 988 args.arg1 = frame.tf_ebx; 989 args.arg2 = frame.tf_ecx; 990 args.arg3 = frame.tf_edx; 991 args.arg4 = frame.tf_esi; 992 args.arg5 = frame.tf_edi; 993 994 sticks = p->p_sticks; 995 if (ISPL(frame.tf_cs) != SEL_UPL) 996 panic("linux syscall"); 997 998 p->p_md.md_regs = (int *)&frame; 999 code = frame.tf_eax; 1000 1001 if (p->p_sysent->sv_mask) 1002 code &= p->p_sysent->sv_mask; 1003 1004 if (code >= p->p_sysent->sv_size) 1005 callp = &p->p_sysent->sv_table[0]; 1006 else 1007 callp = &p->p_sysent->sv_table[code]; 1008 1009#ifdef KTRACE 1010 if (KTRPOINT(p, KTR_SYSCALL)) 1011 ktrsyscall(p->p_tracep, code, callp->sy_narg, (int *)&args); 1012#endif 1013 1014 rval[0] = 0; 1015 1016 error = (*callp->sy_call)(p, &args, rval); 1017 1018 switch (error) { 1019 1020 case 0: 1021 /* 1022 * Reinitialize proc pointer `p' as it may be different 1023 * if this is a child returning from fork syscall. 1024 */ 1025 p = curproc; 1026 frame.tf_eax = rval[0]; 1027 frame.tf_eflags &= ~PSL_C; 1028 break; 1029 1030 case ERESTART: 1031 /* Reconstruct pc, subtract size of int 0x80 */ 1032 frame.tf_eip -= 2; 1033 break; 1034 1035 case EJUSTRETURN: 1036 break; 1037 1038 default: 1039 if (p->p_sysent->sv_errsize) 1040 if (error >= p->p_sysent->sv_errsize) 1041 error = -1; /* XXX */ 1042 else 1043 error = p->p_sysent->sv_errtbl[error]; 1044 frame.tf_eax = -error; 1045 frame.tf_eflags |= PSL_C; 1046 break; 1047 } 1048 1049 if (frame.tf_eflags & PSL_T) { 1050 /* Traced syscall. */ 1051 frame.tf_eflags &= ~PSL_T; 1052 trapsignal(p, SIGTRAP, 0); 1053 } 1054 1055 userret(p, &frame, sticks); 1056 1057#ifdef KTRACE 1058 if (KTRPOINT(p, KTR_SYSRET)) 1059 ktrsysret(p->p_tracep, code, error, rval[0]); 1060#endif 1061} 1062#endif /* COMPAT_LINUX || LINUX */ 1063