subr_syscall.c revision 12702
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 1.63 1995/12/07 12:45:39 davidg Exp $ 39 */ 40 41/* 42 * 386 Trap and System call handling 43 */ 44 45#include <sys/param.h> 46#include <sys/systm.h> 47#include <sys/proc.h> 48#include <sys/acct.h> 49#include <sys/kernel.h> 50#include <sys/syscall.h> 51#include <sys/sysent.h> 52#include <sys/queue.h> 53#include <sys/vmmeter.h> 54#ifdef KTRACE 55#include <sys/ktrace.h> 56#endif 57 58#include <vm/vm.h> 59#include <vm/vm_param.h> 60#include <vm/vm_prot.h> 61#include <vm/lock.h> 62#include <vm/pmap.h> 63#include <vm/vm_kern.h> 64#include <vm/vm_map.h> 65#include <vm/vm_page.h> 66#include <vm/vm_extern.h> 67 68#include <sys/user.h> 69 70#include <machine/cpu.h> 71#include <machine/md_var.h> 72#include <machine/psl.h> 73#include <machine/reg.h> 74#include <machine/trap.h> 75#include <machine/../isa/isa_device.h> 76 77#ifdef POWERFAIL_NMI 78# include <syslog.h> 79# include <machine/clock.h> 80#endif 81 82#include "isa.h" 83#include "npx.h" 84 85extern void trap __P((struct trapframe frame)); 86extern int trapwrite __P((unsigned addr)); 87extern void syscall __P((struct trapframe frame)); 88extern void linux_syscall __P((struct trapframe frame)); 89 90static int trap_pfault __P((struct trapframe *, int)); 91static void trap_fatal __P((struct trapframe *)); 92 93extern inthand_t IDTVEC(syscall); 94 95#define MAX_TRAP_MSG 27 96static char *trap_msg[] = { 97 "", /* 0 unused */ 98 "privileged instruction fault", /* 1 T_PRIVINFLT */ 99 "", /* 2 unused */ 100 "breakpoint instruction fault", /* 3 T_BPTFLT */ 101 "", /* 4 unused */ 102 "", /* 5 unused */ 103 "arithmetic trap", /* 6 T_ARITHTRAP */ 104 "system forced exception", /* 7 T_ASTFLT */ 105 "", /* 8 unused */ 106 "general protection fault", /* 9 T_PROTFLT */ 107 "trace trap", /* 10 T_TRCTRAP */ 108 "", /* 11 unused */ 109 "page fault", /* 12 T_PAGEFLT */ 110 "", /* 13 unused */ 111 "alignment fault", /* 14 T_ALIGNFLT */ 112 "", /* 15 unused */ 113 "", /* 16 unused */ 114 "", /* 17 unused */ 115 "integer divide fault", /* 18 T_DIVIDE */ 116 "non-maskable interrupt trap", /* 19 T_NMI */ 117 "overflow trap", /* 20 T_OFLOW */ 118 "FPU bounds check fault", /* 21 T_BOUND */ 119 "FPU device not available", /* 22 T_DNA */ 120 "double fault", /* 23 T_DOUBLEFLT */ 121 "FPU operand fetch fault", /* 24 T_FPOPFLT */ 122 "invalid TSS fault", /* 25 T_TSSFLT */ 123 "segment not present fault", /* 26 T_SEGNPFLT */ 124 "stack fault", /* 27 T_STKFLT */ 125}; 126 127static void userret __P((struct proc *p, struct trapframe *frame, 128 u_quad_t oticks)); 129 130static inline void 131userret(p, frame, oticks) 132 struct proc *p; 133 struct trapframe *frame; 134 u_quad_t oticks; 135{ 136 int sig, s; 137 138 while ((sig = CURSIG(p)) != 0) 139 postsig(sig); 140 p->p_priority = p->p_usrpri; 141 if (want_resched) { 142 /* 143 * Since we are curproc, clock will normally just change 144 * our priority without moving us from one queue to another 145 * (since the running process is not on a queue.) 146 * If that happened after we setrunqueue ourselves but before we 147 * mi_switch()'ed, we might not be on the queue indicated by 148 * our priority. 149 */ 150 s = splclock(); 151 setrunqueue(p); 152 p->p_stats->p_ru.ru_nivcsw++; 153 mi_switch(); 154 splx(s); 155 while ((sig = CURSIG(p)) != 0) 156 postsig(sig); 157 } 158 /* 159 * Charge system time if profiling. 160 */ 161 if (p->p_flag & P_PROFIL) { 162 u_quad_t ticks = p->p_sticks - oticks; 163 164 if (ticks) { 165#ifdef PROFTIMER 166 extern int profscale; 167 addupc(frame->tf_eip, &p->p_stats->p_prof, 168 ticks * profscale); 169#else 170 addupc(frame->tf_eip, &p->p_stats->p_prof, ticks); 171#endif 172 } 173 } 174 curpriority = p->p_priority; 175} 176 177/* 178 * Exception, fault, and trap interface to the FreeBSD kernel. 179 * This common code is called from assembly language IDT gate entry 180 * routines that prepare a suitable stack frame, and restore this 181 * frame after the exception has been processed. 182 */ 183 184void 185trap(frame) 186 struct trapframe frame; 187{ 188 struct proc *p = curproc; 189 u_quad_t sticks = 0; 190 int i = 0, ucode = 0, type, code; 191#ifdef DEBUG 192 u_long eva; 193#endif 194 195 type = frame.tf_trapno; 196 code = frame.tf_err; 197 198 if (ISPL(frame.tf_cs) == SEL_UPL) { 199 /* user trap */ 200 201 sticks = p->p_sticks; 202 p->p_md.md_regs = (int *)&frame; 203 204 switch (type) { 205 case T_PRIVINFLT: /* privileged instruction fault */ 206 ucode = type; 207 i = SIGILL; 208 break; 209 210 case T_BPTFLT: /* bpt instruction fault */ 211 case T_TRCTRAP: /* trace trap */ 212 frame.tf_eflags &= ~PSL_T; 213 i = SIGTRAP; 214 break; 215 216 case T_ARITHTRAP: /* arithmetic trap */ 217 ucode = code; 218 i = SIGFPE; 219 break; 220 221 case T_ASTFLT: /* Allow process switch */ 222 astoff(); 223 cnt.v_soft++; 224 if (p->p_flag & P_OWEUPC) { 225 addupc(frame.tf_eip, &p->p_stats->p_prof, 1); 226 p->p_flag &= ~P_OWEUPC; 227 } 228 goto out; 229 230 case T_PROTFLT: /* general protection fault */ 231 case T_SEGNPFLT: /* segment not present fault */ 232 case T_STKFLT: /* stack fault */ 233 case T_TSSFLT: /* invalid TSS fault */ 234 case T_DOUBLEFLT: /* double fault */ 235 default: 236 ucode = code + BUS_SEGM_FAULT ; 237 i = SIGBUS; 238 break; 239 240 case T_PAGEFLT: /* page fault */ 241 i = trap_pfault(&frame, TRUE); 242 if (i == -1) 243 return; 244 if (i == 0) 245 goto out; 246 247 ucode = T_PAGEFLT; 248 break; 249 250 case T_DIVIDE: /* integer divide fault */ 251 ucode = FPE_INTDIV_TRAP; 252 i = SIGFPE; 253 break; 254 255#if NISA > 0 256 case T_NMI: 257#ifdef POWERFAIL_NMI 258 goto handle_powerfail; 259#else /* !POWERFAIL_NMI */ 260#ifdef DDB 261 /* NMI can be hooked up to a pushbutton for debugging */ 262 printf ("NMI ... going to debugger\n"); 263 if (kdb_trap (type, 0, &frame)) 264 return; 265#endif /* DDB */ 266 /* machine/parity/power fail/"kitchen sink" faults */ 267 if (isa_nmi(code) == 0) return; 268 panic("NMI indicates hardware failure"); 269#endif /* POWERFAIL_NMI */ 270#endif /* NISA > 0 */ 271 272 case T_OFLOW: /* integer overflow fault */ 273 ucode = FPE_INTOVF_TRAP; 274 i = SIGFPE; 275 break; 276 277 case T_BOUND: /* bounds check fault */ 278 ucode = FPE_SUBRNG_TRAP; 279 i = SIGFPE; 280 break; 281 282 case T_DNA: 283#if NNPX > 0 284 /* if a transparent fault (due to context switch "late") */ 285 if (npxdna()) 286 return; 287#endif /* NNPX > 0 */ 288 289#if defined(MATH_EMULATE) || defined(GPL_MATH_EMULATE) 290 i = math_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#else /* MATH_EMULATE || GPL_MATH_EMULATE */ 299 i = SIGFPE; 300 ucode = FPE_FPU_NP_TRAP; 301#endif /* MATH_EMULATE || GPL_MATH_EMULATE */ 302 break; 303 304 case T_FPOPFLT: /* FPU operand fetch fault */ 305 ucode = T_FPOPFLT; 306 i = SIGILL; 307 break; 308 } 309 } else { 310 /* kernel trap */ 311 312 switch (type) { 313 case T_PAGEFLT: /* page fault */ 314 (void) trap_pfault(&frame, FALSE); 315 return; 316 317 case T_PROTFLT: /* general protection fault */ 318 case T_SEGNPFLT: /* segment not present fault */ 319 /* 320 * Invalid segment selectors and out of bounds 321 * %eip's and %esp's can be set up in user mode. 322 * This causes a fault in kernel mode when the 323 * kernel tries to return to user mode. We want 324 * to get this fault so that we can fix the 325 * problem here and not have to check all the 326 * selectors and pointers when the user changes 327 * them. 328 */ 329#define MAYBE_DORETI_FAULT(where, whereto) \ 330 do { \ 331 if (frame.tf_eip == (int)where) { \ 332 frame.tf_eip = (int)whereto; \ 333 return; \ 334 } \ 335 } while (0) 336 337 if (intr_nesting_level == 0) { 338 MAYBE_DORETI_FAULT(doreti_iret, 339 doreti_iret_fault); 340 MAYBE_DORETI_FAULT(doreti_popl_ds, 341 doreti_popl_ds_fault); 342 MAYBE_DORETI_FAULT(doreti_popl_es, 343 doreti_popl_es_fault); 344 } 345 if (curpcb && curpcb->pcb_onfault) { 346 frame.tf_eip = (int)curpcb->pcb_onfault; 347 return; 348 } 349 break; 350 351 case T_TSSFLT: 352 /* 353 * PSL_NT can be set in user mode and isn't cleared 354 * automatically when the kernel is entered. This 355 * causes a TSS fault when the kernel attempts to 356 * `iret' because the TSS link is uninitialized. We 357 * want to get this fault so that we can fix the 358 * problem here and not every time the kernel is 359 * entered. 360 */ 361 if (frame.tf_eflags & PSL_NT) { 362 frame.tf_eflags &= ~PSL_NT; 363 return; 364 } 365 break; 366 367 case T_TRCTRAP: /* trace trap */ 368 if (frame.tf_eip == (int)IDTVEC(syscall)) { 369 /* 370 * We've just entered system mode via the 371 * syscall lcall. Continue single stepping 372 * silently until the syscall handler has 373 * saved the flags. 374 */ 375 return; 376 } 377 if (frame.tf_eip == (int)IDTVEC(syscall) + 1) { 378 /* 379 * The syscall handler has now saved the 380 * flags. Stop single stepping it. 381 */ 382 frame.tf_eflags &= ~PSL_T; 383 return; 384 } 385 /* 386 * Fall through. 387 */ 388 case T_BPTFLT: 389 /* 390 * If DDB is enabled, let it handle the debugger trap. 391 * Otherwise, debugger traps "can't happen". 392 */ 393#ifdef DDB 394 if (kdb_trap (type, 0, &frame)) 395 return; 396#endif 397 break; 398 399#if NISA > 0 400 case T_NMI: 401#ifdef POWERFAIL_NMI 402#ifndef TIMER_FREQ 403# define TIMER_FREQ 1193182 404#endif 405 handle_powerfail: 406 { 407 static unsigned lastalert = 0; 408 409 if(time.tv_sec - lastalert > 10) 410 { 411 log(LOG_WARNING, "NMI: power fail\n"); 412 sysbeep(TIMER_FREQ/880, hz); 413 lastalert = time.tv_sec; 414 } 415 return; 416 } 417#else /* !POWERFAIL_NMI */ 418#ifdef DDB 419 /* NMI can be hooked up to a pushbutton for debugging */ 420 printf ("NMI ... going to debugger\n"); 421 if (kdb_trap (type, 0, &frame)) 422 return; 423#endif /* DDB */ 424 /* machine/parity/power fail/"kitchen sink" faults */ 425 if (isa_nmi(code) == 0) return; 426 /* FALL THROUGH */ 427#endif /* POWERFAIL_NMI */ 428#endif /* NISA > 0 */ 429 } 430 431 trap_fatal(&frame); 432 return; 433 } 434 435 trapsignal(p, i, ucode); 436 437#ifdef DEBUG 438 eva = rcr2(); 439 if (type <= MAX_TRAP_MSG) { 440 uprintf("fatal process exception: %s", 441 trap_msg[type]); 442 if ((type == T_PAGEFLT) || (type == T_PROTFLT)) 443 uprintf(", fault VA = 0x%x", eva); 444 uprintf("\n"); 445 } 446#endif 447 448out: 449 userret(p, &frame, sticks); 450} 451 452#ifdef notyet 453/* 454 * This version doesn't allow a page fault to user space while 455 * in the kernel. The rest of the kernel needs to be made "safe" 456 * before this can be used. I think the only things remaining 457 * to be made safe are the iBCS2 code and the process tracing/ 458 * debugging code. 459 */ 460static int 461trap_pfault(frame, usermode) 462 struct trapframe *frame; 463 int usermode; 464{ 465 vm_offset_t va; 466 struct vmspace *vm = NULL; 467 vm_map_t map = 0; 468 int rv = 0; 469 vm_prot_t ftype; 470 int eva; 471 struct proc *p = curproc; 472 473 if (frame->tf_err & PGEX_W) 474 ftype = VM_PROT_READ | VM_PROT_WRITE; 475 else 476 ftype = VM_PROT_READ; 477 478 eva = rcr2(); 479 va = trunc_page((vm_offset_t)eva); 480 481 if (va < VM_MIN_KERNEL_ADDRESS) { 482 vm_offset_t v; 483 vm_page_t ptepg; 484 485 if (p == NULL || 486 (!usermode && va < VM_MAXUSER_ADDRESS && 487 (curpcb == NULL || curpcb->pcb_onfault == NULL))) { 488 trap_fatal(frame); 489 return (-1); 490 } 491 492 /* 493 * This is a fault on non-kernel virtual memory. 494 * vm is initialized above to NULL. If curproc is NULL 495 * or curproc->p_vmspace is NULL the fault is fatal. 496 */ 497 vm = p->p_vmspace; 498 if (vm == NULL) 499 goto nogo; 500 501 map = &vm->vm_map; 502 503 /* 504 * Keep swapout from messing with us during this 505 * critical time. 506 */ 507 ++p->p_lock; 508 509 /* 510 * Grow the stack if necessary 511 */ 512 if ((caddr_t)va > vm->vm_maxsaddr 513 && (caddr_t)va < (caddr_t)USRSTACK) { 514 if (!grow(p, va)) { 515 rv = KERN_FAILURE; 516 --p->p_lock; 517 goto nogo; 518 } 519 } 520 521 /* 522 * Check if page table is mapped, if not, 523 * fault it first 524 */ 525 v = (vm_offset_t) vtopte(va); 526 527 /* Fault the pte only if needed: */ 528 if (*((int *)vtopte(v)) == 0) 529 (void) vm_fault(map, trunc_page(v), VM_PROT_WRITE, FALSE); 530 531 pmap_use_pt( vm_map_pmap(map), va); 532 533 /* Fault in the user page: */ 534 rv = vm_fault(map, va, ftype, FALSE); 535 536 pmap_unuse_pt( vm_map_pmap(map), va); 537 538 --p->p_lock; 539 } else { 540 /* 541 * Don't allow user-mode faults in kernel address space. 542 */ 543 if (usermode) 544 goto nogo; 545 546 /* 547 * Since we know that kernel virtual address addresses 548 * always have pte pages mapped, we just have to fault 549 * the page. 550 */ 551 rv = vm_fault(kernel_map, va, ftype, FALSE); 552 } 553 554 if (rv == KERN_SUCCESS) 555 return (0); 556nogo: 557 if (!usermode) { 558 if (curpcb && curpcb->pcb_onfault) { 559 frame->tf_eip = (int)curpcb->pcb_onfault; 560 return (0); 561 } 562 trap_fatal(frame); 563 return (-1); 564 } 565 566 /* kludge to pass faulting virtual address to sendsig */ 567 frame->tf_err = eva; 568 569 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 570} 571#endif 572 573int 574trap_pfault(frame, usermode) 575 struct trapframe *frame; 576 int usermode; 577{ 578 vm_offset_t va; 579 struct vmspace *vm = NULL; 580 vm_map_t map = 0; 581 int rv = 0; 582 vm_prot_t ftype; 583 int eva; 584 struct proc *p = curproc; 585 586 eva = rcr2(); 587 va = trunc_page((vm_offset_t)eva); 588 589 if (va >= KERNBASE) { 590 /* 591 * Don't allow user-mode faults in kernel address space. 592 */ 593 if (usermode) 594 goto nogo; 595 596 map = kernel_map; 597 } else { 598 /* 599 * This is a fault on non-kernel virtual memory. 600 * vm is initialized above to NULL. If curproc is NULL 601 * or curproc->p_vmspace is NULL the fault is fatal. 602 */ 603 if (p != NULL) 604 vm = p->p_vmspace; 605 606 if (vm == NULL) 607 goto nogo; 608 609 map = &vm->vm_map; 610 } 611 612 if (frame->tf_err & PGEX_W) 613 ftype = VM_PROT_READ | VM_PROT_WRITE; 614 else 615 ftype = VM_PROT_READ; 616 617 if (map != kernel_map) { 618 vm_offset_t v; 619 620 /* 621 * Keep swapout from messing with us during this 622 * critical time. 623 */ 624 ++p->p_lock; 625 626 /* 627 * Grow the stack if necessary 628 */ 629 if ((caddr_t)va > vm->vm_maxsaddr 630 && (caddr_t)va < (caddr_t)USRSTACK) { 631 if (!grow(p, va)) { 632 rv = KERN_FAILURE; 633 --p->p_lock; 634 goto nogo; 635 } 636 } 637 638 /* 639 * Check if page table is mapped, if not, 640 * fault it first 641 */ 642 v = (vm_offset_t) vtopte(va); 643 644 /* Fault the pte only if needed: */ 645 if (*((int *)vtopte(v)) == 0) 646 (void) vm_fault(map, trunc_page(v), VM_PROT_WRITE, FALSE); 647 648 pmap_use_pt( vm_map_pmap(map), va); 649 650 /* Fault in the user page: */ 651 rv = vm_fault(map, va, ftype, FALSE); 652 653 pmap_unuse_pt( vm_map_pmap(map), va); 654 655 --p->p_lock; 656 } else { 657 /* 658 * Since we know that kernel virtual address addresses 659 * always have pte pages mapped, we just have to fault 660 * the page. 661 */ 662 rv = vm_fault(map, va, ftype, FALSE); 663 } 664 665 if (rv == KERN_SUCCESS) 666 return (0); 667nogo: 668 if (!usermode) { 669 if (curpcb && curpcb->pcb_onfault) { 670 frame->tf_eip = (int)curpcb->pcb_onfault; 671 return (0); 672 } 673 trap_fatal(frame); 674 return (-1); 675 } 676 677 /* kludge to pass faulting virtual address to sendsig */ 678 frame->tf_err = eva; 679 680 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 681} 682 683static void 684trap_fatal(frame) 685 struct trapframe *frame; 686{ 687 int code, type, eva; 688 struct soft_segment_descriptor softseg; 689 690 code = frame->tf_err; 691 type = frame->tf_trapno; 692 eva = rcr2(); 693 sdtossd(&gdt[IDXSEL(frame->tf_cs & 0xffff)].sd, &softseg); 694 695 if (type <= MAX_TRAP_MSG) 696 printf("\n\nFatal trap %d: %s while in %s mode\n", 697 type, trap_msg[type], 698 ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel"); 699 if (type == T_PAGEFLT) { 700 printf("fault virtual address = 0x%x\n", eva); 701 printf("fault code = %s %s, %s\n", 702 code & PGEX_U ? "user" : "supervisor", 703 code & PGEX_W ? "write" : "read", 704 code & PGEX_P ? "protection violation" : "page not present"); 705 } 706 printf("instruction pointer = 0x%x:0x%x\n", frame->tf_cs & 0xffff, frame->tf_eip); 707 printf("code segment = base 0x%x, limit 0x%x, type 0x%x\n", 708 softseg.ssd_base, softseg.ssd_limit, softseg.ssd_type); 709 printf(" = DPL %d, pres %d, def32 %d, gran %d\n", 710 softseg.ssd_dpl, softseg.ssd_p, softseg.ssd_def32, softseg.ssd_gran); 711 printf("processor eflags = "); 712 if (frame->tf_eflags & PSL_T) 713 printf("trace/trap, "); 714 if (frame->tf_eflags & PSL_I) 715 printf("interrupt enabled, "); 716 if (frame->tf_eflags & PSL_NT) 717 printf("nested task, "); 718 if (frame->tf_eflags & PSL_RF) 719 printf("resume, "); 720 if (frame->tf_eflags & PSL_VM) 721 printf("vm86, "); 722 printf("IOPL = %d\n", (frame->tf_eflags & PSL_IOPL) >> 12); 723 printf("current process = "); 724 if (curproc) { 725 printf("%lu (%s)\n", 726 (u_long)curproc->p_pid, curproc->p_comm ? 727 curproc->p_comm : ""); 728 } else { 729 printf("Idle\n"); 730 } 731 printf("interrupt mask = "); 732 if ((cpl & net_imask) == net_imask) 733 printf("net "); 734 if ((cpl & tty_imask) == tty_imask) 735 printf("tty "); 736 if ((cpl & bio_imask) == bio_imask) 737 printf("bio "); 738 if (cpl == 0) 739 printf("none"); 740 printf("\n"); 741 742#ifdef KDB 743 if (kdb_trap(&psl)) 744 return; 745#endif 746#ifdef DDB 747 if (kdb_trap (type, 0, frame)) 748 return; 749#endif 750 if (type <= MAX_TRAP_MSG) 751 panic(trap_msg[type]); 752 else 753 panic("unknown/reserved trap"); 754} 755 756/* 757 * Compensate for 386 brain damage (missing URKR). 758 * This is a little simpler than the pagefault handler in trap() because 759 * it the page tables have already been faulted in and high addresses 760 * are thrown out early for other reasons. 761 */ 762int trapwrite(addr) 763 unsigned addr; 764{ 765 struct proc *p; 766 vm_offset_t va, v; 767 struct vmspace *vm; 768 int rv; 769 770 va = trunc_page((vm_offset_t)addr); 771 /* 772 * XXX - MAX is END. Changed > to >= for temp. fix. 773 */ 774 if (va >= VM_MAXUSER_ADDRESS) 775 return (1); 776 777 p = curproc; 778 vm = p->p_vmspace; 779 780 ++p->p_lock; 781 782 if ((caddr_t)va >= vm->vm_maxsaddr 783 && (caddr_t)va < (caddr_t)USRSTACK) { 784 if (!grow(p, va)) { 785 --p->p_lock; 786 return (1); 787 } 788 } 789 790 v = trunc_page(vtopte(va)); 791 792 /* 793 * wire the pte page 794 */ 795 if (va < USRSTACK) { 796 vm_map_pageable(&vm->vm_map, v, round_page(v+1), FALSE); 797 } 798 799 /* 800 * fault the data page 801 */ 802 rv = vm_fault(&vm->vm_map, va, VM_PROT_READ|VM_PROT_WRITE, FALSE); 803 804 /* 805 * unwire the pte page 806 */ 807 if (va < USRSTACK) { 808 vm_map_pageable(&vm->vm_map, v, round_page(v+1), TRUE); 809 } 810 811 --p->p_lock; 812 813 if (rv != KERN_SUCCESS) 814 return 1; 815 816 return (0); 817} 818 819/* 820 * System call request from POSIX system call gate interface to kernel. 821 * Like trap(), argument is call by reference. 822 */ 823void 824syscall(frame) 825 struct trapframe frame; 826{ 827 caddr_t params; 828 int i; 829 struct sysent *callp; 830 struct proc *p = curproc; 831 u_quad_t sticks; 832 int error; 833 int args[8], rval[2]; 834 u_int code; 835 836 sticks = p->p_sticks; 837 if (ISPL(frame.tf_cs) != SEL_UPL) 838 panic("syscall"); 839 840 p->p_md.md_regs = (int *)&frame; 841 params = (caddr_t)frame.tf_esp + sizeof(int); 842 code = frame.tf_eax; 843 /* 844 * Need to check if this is a 32 bit or 64 bit syscall. 845 */ 846 if (code == SYS_syscall) { 847 /* 848 * Code is first argument, followed by actual args. 849 */ 850 code = fuword(params); 851 params += sizeof(int); 852 } else if (code == SYS___syscall) { 853 /* 854 * Like syscall, but code is a quad, so as to maintain 855 * quad alignment for the rest of the arguments. 856 */ 857 code = fuword(params); 858 params += sizeof(quad_t); 859 } 860 861 if (p->p_sysent->sv_mask) 862 code &= p->p_sysent->sv_mask; 863 864 if (code >= p->p_sysent->sv_size) 865 callp = &p->p_sysent->sv_table[0]; 866 else 867 callp = &p->p_sysent->sv_table[code]; 868 869 if ((i = callp->sy_narg * sizeof(int)) && 870 (error = copyin(params, (caddr_t)args, (u_int)i))) { 871#ifdef KTRACE 872 if (KTRPOINT(p, KTR_SYSCALL)) 873 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 874#endif 875 goto bad; 876 } 877#ifdef KTRACE 878 if (KTRPOINT(p, KTR_SYSCALL)) 879 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 880#endif 881 rval[0] = 0; 882 rval[1] = frame.tf_edx; 883 884 error = (*callp->sy_call)(p, args, rval); 885 886 switch (error) { 887 888 case 0: 889 /* 890 * Reinitialize proc pointer `p' as it may be different 891 * if this is a child returning from fork syscall. 892 */ 893 p = curproc; 894 frame.tf_eax = rval[0]; 895 frame.tf_edx = rval[1]; 896 frame.tf_eflags &= ~PSL_C; 897 break; 898 899 case ERESTART: 900 /* 901 * Reconstruct pc, assuming lcall $X,y is 7 bytes. 902 */ 903 frame.tf_eip -= 7; 904 break; 905 906 case EJUSTRETURN: 907 break; 908 909 default: 910bad: 911 if (p->p_sysent->sv_errsize) 912 if (error >= p->p_sysent->sv_errsize) 913 error = -1; /* XXX */ 914 else 915 error = p->p_sysent->sv_errtbl[error]; 916 frame.tf_eax = error; 917 frame.tf_eflags |= PSL_C; 918 break; 919 } 920 921 if (frame.tf_eflags & PSL_T) { 922 /* Traced syscall. */ 923 frame.tf_eflags &= ~PSL_T; 924 trapsignal(p, SIGTRAP, 0); 925 } 926 927 userret(p, &frame, sticks); 928 929#ifdef KTRACE 930 if (KTRPOINT(p, KTR_SYSRET)) 931 ktrsysret(p->p_tracep, code, error, rval[0]); 932#endif 933} 934 935#ifdef COMPAT_LINUX 936void 937linux_syscall(frame) 938 struct trapframe frame; 939{ 940 struct proc *p = curproc; 941 struct sysent *callp; 942 u_quad_t sticks; 943 int error; 944 int rval[2]; 945 u_int code; 946 struct linux_syscall_args { 947 int arg1; 948 int arg2; 949 int arg3; 950 int arg4; 951 int arg5; 952 } args; 953 954 args.arg1 = frame.tf_ebx; 955 args.arg2 = frame.tf_ecx; 956 args.arg3 = frame.tf_edx; 957 args.arg4 = frame.tf_esi; 958 args.arg5 = frame.tf_edi; 959 960 sticks = p->p_sticks; 961 if (ISPL(frame.tf_cs) != SEL_UPL) 962 panic("linux syscall"); 963 964 p->p_md.md_regs = (int *)&frame; 965 code = frame.tf_eax; 966 967 if (p->p_sysent->sv_mask) 968 code &= p->p_sysent->sv_mask; 969 970 if (code >= p->p_sysent->sv_size) 971 callp = &p->p_sysent->sv_table[0]; 972 else 973 callp = &p->p_sysent->sv_table[code]; 974 975#ifdef KTRACE 976 if (KTRPOINT(p, KTR_SYSCALL)) 977 ktrsyscall(p->p_tracep, code, callp->sy_narg, (int *)&args); 978#endif 979 980 rval[0] = 0; 981 982 error = (*callp->sy_call)(p, &args, rval); 983 984 switch (error) { 985 986 case 0: 987 /* 988 * Reinitialize proc pointer `p' as it may be different 989 * if this is a child returning from fork syscall. 990 */ 991 p = curproc; 992 frame.tf_eax = rval[0]; 993 frame.tf_eflags &= ~PSL_C; 994 break; 995 996 case ERESTART: 997 /* Reconstruct pc, subtract size of int 0x80 */ 998 frame.tf_eip -= 2; 999 break; 1000 1001 case EJUSTRETURN: 1002 break; 1003 1004 default: 1005 if (p->p_sysent->sv_errsize) 1006 if (error >= p->p_sysent->sv_errsize) 1007 error = -1; /* XXX */ 1008 else 1009 error = p->p_sysent->sv_errtbl[error]; 1010 frame.tf_eax = -error; 1011 frame.tf_eflags |= PSL_C; 1012 break; 1013 } 1014 1015 if (frame.tf_eflags & PSL_T) { 1016 /* Traced syscall. */ 1017 frame.tf_eflags &= ~PSL_T; 1018 trapsignal(p, SIGTRAP, 0); 1019 } 1020 1021 userret(p, &frame, sticks); 1022 1023#ifdef KTRACE 1024 if (KTRPOINT(p, KTR_SYSRET)) 1025 ktrsysret(p->p_tracep, code, error, rval[0]); 1026#endif 1027} 1028#endif /* COMPAT_LINUX */ 1029