subr_syscall.c revision 13203
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 1.68 1995/12/19 14:47:41 davidg Exp $ 39 */ 40 41/* 42 * 386 Trap and System call handling 43 */ 44 45#include "opt_ktrace.h" 46 47#include <sys/param.h> 48#include <sys/systm.h> 49#include <sys/proc.h> 50#include <sys/acct.h> 51#include <sys/kernel.h> 52#include <sys/syscall.h> 53#include <sys/sysent.h> 54#include <sys/queue.h> 55#include <sys/vmmeter.h> 56#ifdef KTRACE 57#include <sys/ktrace.h> 58#endif 59 60#include <vm/vm.h> 61#include <vm/vm_param.h> 62#include <vm/vm_prot.h> 63#include <vm/lock.h> 64#include <vm/pmap.h> 65#include <vm/vm_kern.h> 66#include <vm/vm_map.h> 67#include <vm/vm_page.h> 68#include <vm/vm_extern.h> 69 70#include <sys/user.h> 71 72#include <machine/cpu.h> 73#include <machine/md_var.h> 74#include <machine/psl.h> 75#include <machine/reg.h> 76#include <machine/trap.h> 77#include <machine/../isa/isa_device.h> 78 79#ifdef POWERFAIL_NMI 80# include <syslog.h> 81# include <machine/clock.h> 82#endif 83 84#include "isa.h" 85#include "npx.h" 86 87int (*pmath_emulate) __P((struct trapframe *)); 88 89extern void trap __P((struct trapframe frame)); 90extern int trapwrite __P((unsigned addr)); 91extern void syscall __P((struct trapframe frame)); 92extern void linux_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 u_quad_t ticks = p->p_sticks - oticks; 168 169 if (ticks) { 170#ifdef PROFTIMER 171 extern int profscale; 172 addupc(frame->tf_eip, &p->p_stats->p_prof, 173 ticks * profscale); 174#else 175 addupc(frame->tf_eip, &p->p_stats->p_prof, ticks); 176#endif 177 } 178 } 179 curpriority = p->p_priority; 180} 181 182/* 183 * Exception, fault, and trap interface to the FreeBSD kernel. 184 * This common code is called from assembly language IDT gate entry 185 * routines that prepare a suitable stack frame, and restore this 186 * frame after the exception has been processed. 187 */ 188 189void 190trap(frame) 191 struct trapframe frame; 192{ 193 struct proc *p = curproc; 194 u_quad_t sticks = 0; 195 int i = 0, ucode = 0, type, code; 196#ifdef DEBUG 197 u_long eva; 198#endif 199 200 type = frame.tf_trapno; 201 code = frame.tf_err; 202 203 if (ISPL(frame.tf_cs) == SEL_UPL) { 204 /* user trap */ 205 206 sticks = p->p_sticks; 207 p->p_md.md_regs = (int *)&frame; 208 209 switch (type) { 210 case T_PRIVINFLT: /* privileged instruction fault */ 211 ucode = type; 212 i = SIGILL; 213 break; 214 215 case T_BPTFLT: /* bpt instruction fault */ 216 case T_TRCTRAP: /* trace trap */ 217 frame.tf_eflags &= ~PSL_T; 218 i = SIGTRAP; 219 break; 220 221 case T_ARITHTRAP: /* arithmetic trap */ 222 ucode = code; 223 i = SIGFPE; 224 break; 225 226 case T_ASTFLT: /* Allow process switch */ 227 astoff(); 228 cnt.v_soft++; 229 if (p->p_flag & P_OWEUPC) { 230 addupc(frame.tf_eip, &p->p_stats->p_prof, 1); 231 p->p_flag &= ~P_OWEUPC; 232 } 233 goto out; 234 235 case T_PROTFLT: /* general protection fault */ 236 case T_SEGNPFLT: /* segment not present fault */ 237 case T_STKFLT: /* stack fault */ 238 case T_TSSFLT: /* invalid TSS fault */ 239 case T_DOUBLEFLT: /* double fault */ 240 default: 241 ucode = code + BUS_SEGM_FAULT ; 242 i = SIGBUS; 243 break; 244 245 case T_PAGEFLT: /* page fault */ 246 i = trap_pfault(&frame, TRUE); 247 if (i == -1) 248 return; 249 if (i == 0) 250 goto out; 251 252 ucode = T_PAGEFLT; 253 break; 254 255 case T_DIVIDE: /* integer divide fault */ 256 ucode = FPE_INTDIV_TRAP; 257 i = SIGFPE; 258 break; 259 260#if NISA > 0 261 case T_NMI: 262#ifdef POWERFAIL_NMI 263 goto handle_powerfail; 264#else /* !POWERFAIL_NMI */ 265#ifdef DDB 266 /* NMI can be hooked up to a pushbutton for debugging */ 267 printf ("NMI ... going to debugger\n"); 268 if (kdb_trap (type, 0, &frame)) 269 return; 270#endif /* DDB */ 271 /* machine/parity/power fail/"kitchen sink" faults */ 272 if (isa_nmi(code) == 0) return; 273 panic("NMI indicates hardware failure"); 274#endif /* POWERFAIL_NMI */ 275#endif /* NISA > 0 */ 276 277 case T_OFLOW: /* integer overflow fault */ 278 ucode = FPE_INTOVF_TRAP; 279 i = SIGFPE; 280 break; 281 282 case T_BOUND: /* bounds check fault */ 283 ucode = FPE_SUBRNG_TRAP; 284 i = SIGFPE; 285 break; 286 287 case T_DNA: 288#if NNPX > 0 289 /* if a transparent fault (due to context switch "late") */ 290 if (npxdna()) 291 return; 292#endif /* NNPX > 0 */ 293 294 if (!pmath_emulate) { 295 i = SIGFPE; 296 ucode = FPE_FPU_NP_TRAP; 297 break; 298 } 299 i = (*pmath_emulate)(&frame); 300 if (i == 0) { 301 if (!(frame.tf_eflags & PSL_T)) 302 return; 303 frame.tf_eflags &= ~PSL_T; 304 i = SIGTRAP; 305 } 306 /* else ucode = emulator_only_knows() XXX */ 307 break; 308 309 case T_FPOPFLT: /* FPU operand fetch fault */ 310 ucode = T_FPOPFLT; 311 i = SIGILL; 312 break; 313 } 314 } else { 315 /* kernel trap */ 316 317 switch (type) { 318 case T_PAGEFLT: /* page fault */ 319 (void) trap_pfault(&frame, FALSE); 320 return; 321 322 case T_PROTFLT: /* general protection fault */ 323 case T_SEGNPFLT: /* segment not present fault */ 324 /* 325 * Invalid segment selectors and out of bounds 326 * %eip's and %esp's can be set up in user mode. 327 * This causes a fault in kernel mode when the 328 * kernel tries to return to user mode. We want 329 * to get this fault so that we can fix the 330 * problem here and not have to check all the 331 * selectors and pointers when the user changes 332 * them. 333 */ 334#define MAYBE_DORETI_FAULT(where, whereto) \ 335 do { \ 336 if (frame.tf_eip == (int)where) { \ 337 frame.tf_eip = (int)whereto; \ 338 return; \ 339 } \ 340 } while (0) 341 342 if (intr_nesting_level == 0) { 343 MAYBE_DORETI_FAULT(doreti_iret, 344 doreti_iret_fault); 345 MAYBE_DORETI_FAULT(doreti_popl_ds, 346 doreti_popl_ds_fault); 347 MAYBE_DORETI_FAULT(doreti_popl_es, 348 doreti_popl_es_fault); 349 } 350 if (curpcb && curpcb->pcb_onfault) { 351 frame.tf_eip = (int)curpcb->pcb_onfault; 352 return; 353 } 354 break; 355 356 case T_TSSFLT: 357 /* 358 * PSL_NT can be set in user mode and isn't cleared 359 * automatically when the kernel is entered. This 360 * causes a TSS fault when the kernel attempts to 361 * `iret' because the TSS link is uninitialized. We 362 * want to get this fault so that we can fix the 363 * problem here and not every time the kernel is 364 * entered. 365 */ 366 if (frame.tf_eflags & PSL_NT) { 367 frame.tf_eflags &= ~PSL_NT; 368 return; 369 } 370 break; 371 372 case T_TRCTRAP: /* trace trap */ 373 if (frame.tf_eip == (int)IDTVEC(syscall)) { 374 /* 375 * We've just entered system mode via the 376 * syscall lcall. Continue single stepping 377 * silently until the syscall handler has 378 * saved the flags. 379 */ 380 return; 381 } 382 if (frame.tf_eip == (int)IDTVEC(syscall) + 1) { 383 /* 384 * The syscall handler has now saved the 385 * flags. Stop single stepping it. 386 */ 387 frame.tf_eflags &= ~PSL_T; 388 return; 389 } 390 /* 391 * Fall through. 392 */ 393 case T_BPTFLT: 394 /* 395 * If DDB is enabled, let it handle the debugger trap. 396 * Otherwise, debugger traps "can't happen". 397 */ 398#ifdef DDB 399 if (kdb_trap (type, 0, &frame)) 400 return; 401#endif 402 break; 403 404#if NISA > 0 405 case T_NMI: 406#ifdef POWERFAIL_NMI 407#ifndef TIMER_FREQ 408# define TIMER_FREQ 1193182 409#endif 410 handle_powerfail: 411 { 412 static unsigned lastalert = 0; 413 414 if(time.tv_sec - lastalert > 10) 415 { 416 log(LOG_WARNING, "NMI: power fail\n"); 417 sysbeep(TIMER_FREQ/880, hz); 418 lastalert = time.tv_sec; 419 } 420 return; 421 } 422#else /* !POWERFAIL_NMI */ 423#ifdef DDB 424 /* NMI can be hooked up to a pushbutton for debugging */ 425 printf ("NMI ... going to debugger\n"); 426 if (kdb_trap (type, 0, &frame)) 427 return; 428#endif /* DDB */ 429 /* machine/parity/power fail/"kitchen sink" faults */ 430 if (isa_nmi(code) == 0) return; 431 /* FALL THROUGH */ 432#endif /* POWERFAIL_NMI */ 433#endif /* NISA > 0 */ 434 } 435 436 trap_fatal(&frame); 437 return; 438 } 439 440 trapsignal(p, i, ucode); 441 442#ifdef DEBUG 443 eva = rcr2(); 444 if (type <= MAX_TRAP_MSG) { 445 uprintf("fatal process exception: %s", 446 trap_msg[type]); 447 if ((type == T_PAGEFLT) || (type == T_PROTFLT)) 448 uprintf(", fault VA = 0x%x", eva); 449 uprintf("\n"); 450 } 451#endif 452 453out: 454 userret(p, &frame, sticks); 455} 456 457#ifdef notyet 458/* 459 * This version doesn't allow a page fault to user space while 460 * in the kernel. The rest of the kernel needs to be made "safe" 461 * before this can be used. I think the only things remaining 462 * to be made safe are the iBCS2 code and the process tracing/ 463 * debugging code. 464 */ 465static int 466trap_pfault(frame, usermode) 467 struct trapframe *frame; 468 int usermode; 469{ 470 vm_offset_t va; 471 struct vmspace *vm = NULL; 472 vm_map_t map = 0; 473 int rv = 0; 474 vm_prot_t ftype; 475 int eva; 476 struct proc *p = curproc; 477 478 if (frame->tf_err & PGEX_W) 479 ftype = VM_PROT_READ | VM_PROT_WRITE; 480 else 481 ftype = VM_PROT_READ; 482 483 eva = rcr2(); 484 va = trunc_page((vm_offset_t)eva); 485 486 if (va < VM_MIN_KERNEL_ADDRESS) { 487 vm_offset_t v; 488 vm_page_t ptepg; 489 490 if (p == NULL || 491 (!usermode && va < VM_MAXUSER_ADDRESS && 492 (curpcb == NULL || curpcb->pcb_onfault == NULL))) { 493 trap_fatal(frame); 494 return (-1); 495 } 496 497 /* 498 * This is a fault on non-kernel virtual memory. 499 * vm is initialized above to NULL. If curproc is NULL 500 * or curproc->p_vmspace is NULL the fault is fatal. 501 */ 502 vm = p->p_vmspace; 503 if (vm == NULL) 504 goto nogo; 505 506 map = &vm->vm_map; 507 508 /* 509 * Keep swapout from messing with us during this 510 * critical time. 511 */ 512 ++p->p_lock; 513 514 /* 515 * Grow the stack if necessary 516 */ 517 if ((caddr_t)va > vm->vm_maxsaddr 518 && (caddr_t)va < (caddr_t)USRSTACK) { 519 if (!grow(p, va)) { 520 rv = KERN_FAILURE; 521 --p->p_lock; 522 goto nogo; 523 } 524 } 525 526 /* 527 * Check if page table is mapped, if not, 528 * fault it first 529 */ 530 v = (vm_offset_t) vtopte(va); 531 532 /* Fault the pte only if needed: */ 533 if (*((int *)vtopte(v)) == 0) 534 (void) vm_fault(map, trunc_page(v), VM_PROT_WRITE, FALSE); 535 536 pmap_use_pt( vm_map_pmap(map), va); 537 538 /* Fault in the user page: */ 539 rv = vm_fault(map, va, ftype, FALSE); 540 541 pmap_unuse_pt( vm_map_pmap(map), va); 542 543 --p->p_lock; 544 } else { 545 /* 546 * Don't allow user-mode faults in kernel address space. 547 */ 548 if (usermode) 549 goto nogo; 550 551 /* 552 * Since we know that kernel virtual address addresses 553 * always have pte pages mapped, we just have to fault 554 * the page. 555 */ 556 rv = vm_fault(kernel_map, va, ftype, FALSE); 557 } 558 559 if (rv == KERN_SUCCESS) 560 return (0); 561nogo: 562 if (!usermode) { 563 if (curpcb && curpcb->pcb_onfault) { 564 frame->tf_eip = (int)curpcb->pcb_onfault; 565 return (0); 566 } 567 trap_fatal(frame); 568 return (-1); 569 } 570 571 /* kludge to pass faulting virtual address to sendsig */ 572 frame->tf_err = eva; 573 574 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 575} 576#endif 577 578int 579trap_pfault(frame, usermode) 580 struct trapframe *frame; 581 int usermode; 582{ 583 vm_offset_t va; 584 struct vmspace *vm = NULL; 585 vm_map_t map = 0; 586 int rv = 0; 587 vm_prot_t ftype; 588 int eva; 589 struct proc *p = curproc; 590 591 eva = rcr2(); 592 va = trunc_page((vm_offset_t)eva); 593 594 if (va >= KERNBASE) { 595 /* 596 * Don't allow user-mode faults in kernel address space. 597 */ 598 if (usermode) 599 goto nogo; 600 601 map = kernel_map; 602 } else { 603 /* 604 * This is a fault on non-kernel virtual memory. 605 * vm is initialized above to NULL. If curproc is NULL 606 * or curproc->p_vmspace is NULL the fault is fatal. 607 */ 608 if (p != NULL) 609 vm = p->p_vmspace; 610 611 if (vm == NULL) 612 goto nogo; 613 614 map = &vm->vm_map; 615 } 616 617 if (frame->tf_err & PGEX_W) 618 ftype = VM_PROT_READ | VM_PROT_WRITE; 619 else 620 ftype = VM_PROT_READ; 621 622 if (map != kernel_map) { 623 vm_offset_t v; 624 625 /* 626 * Keep swapout from messing with us during this 627 * critical time. 628 */ 629 ++p->p_lock; 630 631 /* 632 * Grow the stack if necessary 633 */ 634 if ((caddr_t)va > vm->vm_maxsaddr 635 && (caddr_t)va < (caddr_t)USRSTACK) { 636 if (!grow(p, va)) { 637 rv = KERN_FAILURE; 638 --p->p_lock; 639 goto nogo; 640 } 641 } 642 643 /* 644 * Check if page table is mapped, if not, 645 * fault it first 646 */ 647 v = (vm_offset_t) vtopte(va); 648 649 /* Fault the pte only if needed: */ 650 if (*((int *)vtopte(v)) == 0) 651 (void) vm_fault(map, trunc_page(v), VM_PROT_WRITE, FALSE); 652 653 pmap_use_pt( vm_map_pmap(map), va); 654 655 /* Fault in the user page: */ 656 rv = vm_fault(map, va, ftype, FALSE); 657 658 pmap_unuse_pt( vm_map_pmap(map), va); 659 660 --p->p_lock; 661 } else { 662 /* 663 * Since we know that kernel virtual address addresses 664 * always have pte pages mapped, we just have to fault 665 * the page. 666 */ 667 rv = vm_fault(map, va, ftype, FALSE); 668 } 669 670 if (rv == KERN_SUCCESS) 671 return (0); 672nogo: 673 if (!usermode) { 674 if (curpcb && curpcb->pcb_onfault) { 675 frame->tf_eip = (int)curpcb->pcb_onfault; 676 return (0); 677 } 678 trap_fatal(frame); 679 return (-1); 680 } 681 682 /* kludge to pass faulting virtual address to sendsig */ 683 frame->tf_err = eva; 684 685 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 686} 687 688static void 689trap_fatal(frame) 690 struct trapframe *frame; 691{ 692 int code, type, eva; 693 struct soft_segment_descriptor softseg; 694 695 code = frame->tf_err; 696 type = frame->tf_trapno; 697 eva = rcr2(); 698 sdtossd(&gdt[IDXSEL(frame->tf_cs & 0xffff)].sd, &softseg); 699 700 if (type <= MAX_TRAP_MSG) 701 printf("\n\nFatal trap %d: %s while in %s mode\n", 702 type, trap_msg[type], 703 ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel"); 704 if (type == T_PAGEFLT) { 705 printf("fault virtual address = 0x%x\n", eva); 706 printf("fault code = %s %s, %s\n", 707 code & PGEX_U ? "user" : "supervisor", 708 code & PGEX_W ? "write" : "read", 709 code & PGEX_P ? "protection violation" : "page not present"); 710 } 711 printf("instruction pointer = 0x%x:0x%x\n", frame->tf_cs & 0xffff, frame->tf_eip); 712 printf("code segment = base 0x%x, limit 0x%x, type 0x%x\n", 713 softseg.ssd_base, softseg.ssd_limit, softseg.ssd_type); 714 printf(" = DPL %d, pres %d, def32 %d, gran %d\n", 715 softseg.ssd_dpl, softseg.ssd_p, softseg.ssd_def32, softseg.ssd_gran); 716 printf("processor eflags = "); 717 if (frame->tf_eflags & PSL_T) 718 printf("trace/trap, "); 719 if (frame->tf_eflags & PSL_I) 720 printf("interrupt enabled, "); 721 if (frame->tf_eflags & PSL_NT) 722 printf("nested task, "); 723 if (frame->tf_eflags & PSL_RF) 724 printf("resume, "); 725 if (frame->tf_eflags & PSL_VM) 726 printf("vm86, "); 727 printf("IOPL = %d\n", (frame->tf_eflags & PSL_IOPL) >> 12); 728 printf("current process = "); 729 if (curproc) { 730 printf("%lu (%s)\n", 731 (u_long)curproc->p_pid, curproc->p_comm ? 732 curproc->p_comm : ""); 733 } else { 734 printf("Idle\n"); 735 } 736 printf("interrupt mask = "); 737 if ((cpl & net_imask) == net_imask) 738 printf("net "); 739 if ((cpl & tty_imask) == tty_imask) 740 printf("tty "); 741 if ((cpl & bio_imask) == bio_imask) 742 printf("bio "); 743 if (cpl == 0) 744 printf("none"); 745 printf("\n"); 746 747#ifdef KDB 748 if (kdb_trap(&psl)) 749 return; 750#endif 751#ifdef DDB 752 if (kdb_trap (type, 0, frame)) 753 return; 754#endif 755 if (type <= MAX_TRAP_MSG) 756 panic(trap_msg[type]); 757 else 758 panic("unknown/reserved trap"); 759} 760 761/* 762 * Double fault handler. Called when a fault occurs while writing 763 * a frame for a trap/exception onto the stack. This usually occurs 764 * when the stack overflows (such is the case with infinite recursion, 765 * for example). 766 * 767 * XXX Note that the current PTD gets replaced by IdlePTD when the 768 * task switch occurs. This means that the stack that was active at 769 * the time of the double fault is not available at <kstack> unless 770 * the machine was idle when the double fault occurred. The downside 771 * of this is that "trace <ebp>" in ddb won't work. 772 */ 773void 774dblfault_handler() 775{ 776 struct pcb *pcb = curpcb; 777 778 if (pcb != NULL) { 779 printf("\nFatal double fault:\n"); 780 printf("eip = 0x%x\n", pcb->pcb_tss.tss_eip); 781 printf("esp = 0x%x\n", pcb->pcb_tss.tss_esp); 782 printf("ebp = 0x%x\n", pcb->pcb_tss.tss_ebp); 783 } 784 785 panic("double fault"); 786} 787 788/* 789 * Compensate for 386 brain damage (missing URKR). 790 * This is a little simpler than the pagefault handler in trap() because 791 * it the page tables have already been faulted in and high addresses 792 * are thrown out early for other reasons. 793 */ 794int trapwrite(addr) 795 unsigned addr; 796{ 797 struct proc *p; 798 vm_offset_t va, v; 799 struct vmspace *vm; 800 int rv; 801 802 va = trunc_page((vm_offset_t)addr); 803 /* 804 * XXX - MAX is END. Changed > to >= for temp. fix. 805 */ 806 if (va >= VM_MAXUSER_ADDRESS) 807 return (1); 808 809 p = curproc; 810 vm = p->p_vmspace; 811 812 ++p->p_lock; 813 814 if ((caddr_t)va >= vm->vm_maxsaddr 815 && (caddr_t)va < (caddr_t)USRSTACK) { 816 if (!grow(p, va)) { 817 --p->p_lock; 818 return (1); 819 } 820 } 821 822 v = trunc_page(vtopte(va)); 823 824 /* 825 * wire the pte page 826 */ 827 if (va < USRSTACK) { 828 vm_map_pageable(&vm->vm_map, v, round_page(v+1), FALSE); 829 } 830 831 /* 832 * fault the data page 833 */ 834 rv = vm_fault(&vm->vm_map, va, VM_PROT_READ|VM_PROT_WRITE, FALSE); 835 836 /* 837 * unwire the pte page 838 */ 839 if (va < USRSTACK) { 840 vm_map_pageable(&vm->vm_map, v, round_page(v+1), TRUE); 841 } 842 843 --p->p_lock; 844 845 if (rv != KERN_SUCCESS) 846 return 1; 847 848 return (0); 849} 850 851/* 852 * System call request from POSIX system call gate interface to kernel. 853 * Like trap(), argument is call by reference. 854 */ 855void 856syscall(frame) 857 struct trapframe frame; 858{ 859 caddr_t params; 860 int i; 861 struct sysent *callp; 862 struct proc *p = curproc; 863 u_quad_t sticks; 864 int error; 865 int args[8], rval[2]; 866 u_int code; 867 868 sticks = p->p_sticks; 869 if (ISPL(frame.tf_cs) != SEL_UPL) 870 panic("syscall"); 871 872 p->p_md.md_regs = (int *)&frame; 873 params = (caddr_t)frame.tf_esp + sizeof(int); 874 code = frame.tf_eax; 875 /* 876 * Need to check if this is a 32 bit or 64 bit syscall. 877 */ 878 if (code == SYS_syscall) { 879 /* 880 * Code is first argument, followed by actual args. 881 */ 882 code = fuword(params); 883 params += sizeof(int); 884 } else if (code == SYS___syscall) { 885 /* 886 * Like syscall, but code is a quad, so as to maintain 887 * quad alignment for the rest of the arguments. 888 */ 889 code = fuword(params); 890 params += sizeof(quad_t); 891 } 892 893 if (p->p_sysent->sv_mask) 894 code &= p->p_sysent->sv_mask; 895 896 if (code >= p->p_sysent->sv_size) 897 callp = &p->p_sysent->sv_table[0]; 898 else 899 callp = &p->p_sysent->sv_table[code]; 900 901 if ((i = callp->sy_narg * sizeof(int)) && 902 (error = copyin(params, (caddr_t)args, (u_int)i))) { 903#ifdef KTRACE 904 if (KTRPOINT(p, KTR_SYSCALL)) 905 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 906#endif 907 goto bad; 908 } 909#ifdef KTRACE 910 if (KTRPOINT(p, KTR_SYSCALL)) 911 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 912#endif 913 rval[0] = 0; 914 rval[1] = frame.tf_edx; 915 916 error = (*callp->sy_call)(p, args, rval); 917 918 switch (error) { 919 920 case 0: 921 /* 922 * Reinitialize proc pointer `p' as it may be different 923 * if this is a child returning from fork syscall. 924 */ 925 p = curproc; 926 frame.tf_eax = rval[0]; 927 frame.tf_edx = rval[1]; 928 frame.tf_eflags &= ~PSL_C; 929 break; 930 931 case ERESTART: 932 /* 933 * Reconstruct pc, assuming lcall $X,y is 7 bytes. 934 */ 935 frame.tf_eip -= 7; 936 break; 937 938 case EJUSTRETURN: 939 break; 940 941 default: 942bad: 943 if (p->p_sysent->sv_errsize) 944 if (error >= p->p_sysent->sv_errsize) 945 error = -1; /* XXX */ 946 else 947 error = p->p_sysent->sv_errtbl[error]; 948 frame.tf_eax = error; 949 frame.tf_eflags |= PSL_C; 950 break; 951 } 952 953 if (frame.tf_eflags & PSL_T) { 954 /* Traced syscall. */ 955 frame.tf_eflags &= ~PSL_T; 956 trapsignal(p, SIGTRAP, 0); 957 } 958 959 userret(p, &frame, sticks); 960 961#ifdef KTRACE 962 if (KTRPOINT(p, KTR_SYSRET)) 963 ktrsysret(p->p_tracep, code, error, rval[0]); 964#endif 965} 966 967#if defined(COMPAT_LINUX) || defined(LINUX) 968void 969linux_syscall(frame) 970 struct trapframe frame; 971{ 972 struct proc *p = curproc; 973 struct sysent *callp; 974 u_quad_t sticks; 975 int error; 976 int rval[2]; 977 u_int code; 978 struct linux_syscall_args { 979 int arg1; 980 int arg2; 981 int arg3; 982 int arg4; 983 int arg5; 984 } args; 985 986 args.arg1 = frame.tf_ebx; 987 args.arg2 = frame.tf_ecx; 988 args.arg3 = frame.tf_edx; 989 args.arg4 = frame.tf_esi; 990 args.arg5 = frame.tf_edi; 991 992 sticks = p->p_sticks; 993 if (ISPL(frame.tf_cs) != SEL_UPL) 994 panic("linux syscall"); 995 996 p->p_md.md_regs = (int *)&frame; 997 code = frame.tf_eax; 998 999 if (p->p_sysent->sv_mask) 1000 code &= p->p_sysent->sv_mask; 1001 1002 if (code >= p->p_sysent->sv_size) 1003 callp = &p->p_sysent->sv_table[0]; 1004 else 1005 callp = &p->p_sysent->sv_table[code]; 1006 1007#ifdef KTRACE 1008 if (KTRPOINT(p, KTR_SYSCALL)) 1009 ktrsyscall(p->p_tracep, code, callp->sy_narg, (int *)&args); 1010#endif 1011 1012 rval[0] = 0; 1013 1014 error = (*callp->sy_call)(p, &args, rval); 1015 1016 switch (error) { 1017 1018 case 0: 1019 /* 1020 * Reinitialize proc pointer `p' as it may be different 1021 * if this is a child returning from fork syscall. 1022 */ 1023 p = curproc; 1024 frame.tf_eax = rval[0]; 1025 frame.tf_eflags &= ~PSL_C; 1026 break; 1027 1028 case ERESTART: 1029 /* Reconstruct pc, subtract size of int 0x80 */ 1030 frame.tf_eip -= 2; 1031 break; 1032 1033 case EJUSTRETURN: 1034 break; 1035 1036 default: 1037 if (p->p_sysent->sv_errsize) 1038 if (error >= p->p_sysent->sv_errsize) 1039 error = -1; /* XXX */ 1040 else 1041 error = p->p_sysent->sv_errtbl[error]; 1042 frame.tf_eax = -error; 1043 frame.tf_eflags |= PSL_C; 1044 break; 1045 } 1046 1047 if (frame.tf_eflags & PSL_T) { 1048 /* Traced syscall. */ 1049 frame.tf_eflags &= ~PSL_T; 1050 trapsignal(p, SIGTRAP, 0); 1051 } 1052 1053 userret(p, &frame, sticks); 1054 1055#ifdef KTRACE 1056 if (KTRPOINT(p, KTR_SYSRET)) 1057 ktrsysret(p->p_tracep, code, error, rval[0]); 1058#endif 1059} 1060#endif /* COMPAT_LINUX || LINUX */ 1061