subr_syscall.c revision 11163
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 1.59 1995/08/21 18:06:48 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/user.h> 49#include <sys/acct.h> 50#include <sys/kernel.h> 51#include <sys/syscall.h> 52#include <sys/sysent.h> 53#ifdef KTRACE 54#include <sys/ktrace.h> 55#endif 56 57#include <vm/vm_param.h> 58#include <vm/pmap.h> 59#include <vm/vm_kern.h> 60#include <vm/vm_map.h> 61#include <vm/vm_page.h> 62 63#include <machine/cpu.h> 64#include <machine/md_var.h> 65#include <machine/psl.h> 66#include <machine/reg.h> 67#include <machine/trap.h> 68#include <machine/../isa/isa_device.h> 69 70#ifdef POWERFAIL_NMI 71# include <syslog.h> 72# include <machine/clock.h> 73#endif 74 75#include "isa.h" 76#include "npx.h" 77 78int trap_pfault __P((struct trapframe *, int)); 79void trap_fatal __P((struct trapframe *)); 80 81extern inthand_t IDTVEC(syscall); 82#ifdef COMPAT_LINUX 83extern inthand_t IDTVEC(linux_syscall); 84#endif 85 86#define MAX_TRAP_MSG 27 87char *trap_msg[] = { 88 "", /* 0 unused */ 89 "privileged instruction fault", /* 1 T_PRIVINFLT */ 90 "", /* 2 unused */ 91 "breakpoint instruction fault", /* 3 T_BPTFLT */ 92 "", /* 4 unused */ 93 "", /* 5 unused */ 94 "arithmetic trap", /* 6 T_ARITHTRAP */ 95 "system forced exception", /* 7 T_ASTFLT */ 96 "", /* 8 unused */ 97 "general protection fault", /* 9 T_PROTFLT */ 98 "trace trap", /* 10 T_TRCTRAP */ 99 "", /* 11 unused */ 100 "page fault", /* 12 T_PAGEFLT */ 101 "", /* 13 unused */ 102 "alignment fault", /* 14 T_ALIGNFLT */ 103 "", /* 15 unused */ 104 "", /* 16 unused */ 105 "", /* 17 unused */ 106 "integer divide fault", /* 18 T_DIVIDE */ 107 "non-maskable interrupt trap", /* 19 T_NMI */ 108 "overflow trap", /* 20 T_OFLOW */ 109 "FPU bounds check fault", /* 21 T_BOUND */ 110 "FPU device not available", /* 22 T_DNA */ 111 "double fault", /* 23 T_DOUBLEFLT */ 112 "FPU operand fetch fault", /* 24 T_FPOPFLT */ 113 "invalid TSS fault", /* 25 T_TSSFLT */ 114 "segment not present fault", /* 26 T_SEGNPFLT */ 115 "stack fault", /* 27 T_STKFLT */ 116}; 117 118static inline void 119userret(p, frame, oticks) 120 struct proc *p; 121 struct trapframe *frame; 122 u_quad_t oticks; 123{ 124 int sig, s; 125 126 while ((sig = CURSIG(p)) != 0) 127 postsig(sig); 128 p->p_priority = p->p_usrpri; 129 if (want_resched) { 130 /* 131 * Since we are curproc, clock will normally just change 132 * our priority without moving us from one queue to another 133 * (since the running process is not on a queue.) 134 * If that happened after we setrunqueue ourselves but before we 135 * mi_switch()'ed, we might not be on the queue indicated by 136 * our priority. 137 */ 138 s = splclock(); 139 setrunqueue(p); 140 p->p_stats->p_ru.ru_nivcsw++; 141 mi_switch(); 142 splx(s); 143 while ((sig = CURSIG(p)) != 0) 144 postsig(sig); 145 } 146 /* 147 * Charge system time if profiling. 148 */ 149 if (p->p_flag & P_PROFIL) { 150 u_quad_t ticks = p->p_sticks - oticks; 151 152 if (ticks) { 153#ifdef PROFTIMER 154 extern int profscale; 155 addupc(frame->tf_eip, &p->p_stats->p_prof, 156 ticks * profscale); 157#else 158 addupc(frame->tf_eip, &p->p_stats->p_prof, ticks); 159#endif 160 } 161 } 162 curpriority = p->p_priority; 163} 164 165/* 166 * trap(frame): 167 * Exception, fault, and trap interface to the FreeBSD kernel. 168 * This common code is called from assembly language IDT gate entry 169 * routines that prepare a suitable stack frame, and restore this 170 * frame after the exception has been processed. 171 */ 172 173/*ARGSUSED*/ 174void 175trap(frame) 176 struct trapframe frame; 177{ 178 struct proc *p = curproc; 179 u_quad_t sticks = 0; 180 int i = 0, ucode = 0, type, code; 181#ifdef DIAGNOSTIC 182 u_long eva; 183#endif 184 185 type = frame.tf_trapno; 186 code = frame.tf_err; 187 188 if (ISPL(frame.tf_cs) == SEL_UPL) { 189 /* user trap */ 190 191 sticks = p->p_sticks; 192 p->p_md.md_regs = (int *)&frame; 193 194 switch (type) { 195 case T_PRIVINFLT: /* privileged instruction fault */ 196 ucode = type; 197 i = SIGILL; 198 break; 199 200 case T_BPTFLT: /* bpt instruction fault */ 201 case T_TRCTRAP: /* trace trap */ 202 frame.tf_eflags &= ~PSL_T; 203 i = SIGTRAP; 204 break; 205 206 case T_ARITHTRAP: /* arithmetic trap */ 207 ucode = code; 208 i = SIGFPE; 209 break; 210 211 case T_ASTFLT: /* Allow process switch */ 212 astoff(); 213 cnt.v_soft++; 214 if (p->p_flag & P_OWEUPC) { 215 addupc(frame.tf_eip, &p->p_stats->p_prof, 1); 216 p->p_flag &= ~P_OWEUPC; 217 } 218 goto out; 219 220 case T_PROTFLT: /* general protection fault */ 221 case T_SEGNPFLT: /* segment not present fault */ 222 case T_STKFLT: /* stack fault */ 223 case T_TSSFLT: /* invalid TSS fault */ 224 case T_DOUBLEFLT: /* double fault */ 225 default: 226 ucode = code + BUS_SEGM_FAULT ; 227 i = SIGBUS; 228 break; 229 230 case T_PAGEFLT: /* page fault */ 231 i = trap_pfault(&frame, TRUE); 232 if (i == -1) 233 return; 234 if (i == 0) 235 goto out; 236 237 ucode = T_PAGEFLT; 238 break; 239 240 case T_DIVIDE: /* integer divide fault */ 241 ucode = FPE_INTDIV_TRAP; 242 i = SIGFPE; 243 break; 244 245#if NISA > 0 246 case T_NMI: 247#ifdef POWERFAIL_NMI 248 goto handle_powerfail; 249#else /* !POWERFAIL_NMI */ 250#ifdef DDB 251 /* NMI can be hooked up to a pushbutton for debugging */ 252 printf ("NMI ... going to debugger\n"); 253 if (kdb_trap (type, 0, &frame)) 254 return; 255#endif /* DDB */ 256 /* machine/parity/power fail/"kitchen sink" faults */ 257 if (isa_nmi(code) == 0) return; 258 panic("NMI indicates hardware failure"); 259#endif /* POWERFAIL_NMI */ 260#endif /* NISA > 0 */ 261 262 case T_OFLOW: /* integer overflow fault */ 263 ucode = FPE_INTOVF_TRAP; 264 i = SIGFPE; 265 break; 266 267 case T_BOUND: /* bounds check fault */ 268 ucode = FPE_SUBRNG_TRAP; 269 i = SIGFPE; 270 break; 271 272 case T_DNA: 273#if NNPX > 0 274 /* if a transparent fault (due to context switch "late") */ 275 if (npxdna()) 276 return; 277#endif /* NNPX > 0 */ 278 279#if defined(MATH_EMULATE) || defined(GPL_MATH_EMULATE) 280 i = math_emulate(&frame); 281 if (i == 0) { 282 if (!(frame.tf_eflags & PSL_T)) 283 return; 284 frame.tf_eflags &= ~PSL_T; 285 i = SIGTRAP; 286 } 287 /* else ucode = emulator_only_knows() XXX */ 288#else /* MATH_EMULATE || GPL_MATH_EMULATE */ 289 i = SIGFPE; 290 ucode = FPE_FPU_NP_TRAP; 291#endif /* MATH_EMULATE || GPL_MATH_EMULATE */ 292 break; 293 294 case T_FPOPFLT: /* FPU operand fetch fault */ 295 ucode = T_FPOPFLT; 296 i = SIGILL; 297 break; 298 } 299 } else { 300 /* kernel trap */ 301 302 switch (type) { 303 case T_PAGEFLT: /* page fault */ 304 (void) trap_pfault(&frame, FALSE); 305 return; 306 307 case T_PROTFLT: /* general protection fault */ 308 case T_SEGNPFLT: /* segment not present fault */ 309 /* 310 * Invalid segment selectors and out of bounds 311 * %eip's and %esp's can be set up in user mode. 312 * This causes a fault in kernel mode when the 313 * kernel tries to return to user mode. We want 314 * to get this fault so that we can fix the 315 * problem here and not have to check all the 316 * selectors and pointers when the user changes 317 * them. 318 */ 319#define MAYBE_DORETI_FAULT(where, whereto) \ 320 do { \ 321 if (frame.tf_eip == (int)where) { \ 322 frame.tf_eip = (int)whereto; \ 323 return; \ 324 } \ 325 } while (0) 326 327 if (intr_nesting_level == 0) { 328 MAYBE_DORETI_FAULT(doreti_iret, 329 doreti_iret_fault); 330 MAYBE_DORETI_FAULT(doreti_popl_ds, 331 doreti_popl_ds_fault); 332 MAYBE_DORETI_FAULT(doreti_popl_es, 333 doreti_popl_es_fault); 334 } 335 if (curpcb && curpcb->pcb_onfault) { 336 frame.tf_eip = (int)curpcb->pcb_onfault; 337 return; 338 } 339 break; 340 341 case T_TSSFLT: 342 /* 343 * PSL_NT can be set in user mode and isn't cleared 344 * automatically when the kernel is entered. This 345 * causes a TSS fault when the kernel attempts to 346 * `iret' because the TSS link is uninitialized. We 347 * want to get this fault so that we can fix the 348 * problem here and not every time the kernel is 349 * entered. 350 */ 351 if (frame.tf_eflags & PSL_NT) { 352 frame.tf_eflags &= ~PSL_NT; 353 return; 354 } 355 break; 356 357#ifdef DDB 358 case T_BPTFLT: 359#ifndef DDB_NO_LCALLS 360 case T_TRCTRAP: 361#endif 362 if (kdb_trap (type, 0, &frame)) 363 return; 364 break; 365#endif 366#if !defined (DDB) || defined (DDB_NO_LCALLS) 367 case T_TRCTRAP: /* trace trap -- someone single stepping lcall's */ 368 /* Q: how do we turn it on again? */ 369#ifdef COMPAT_LINUX 370 if (frame.tf_eip != (int) IDTVEC(syscall) && 371 frame.tf_eip != (int) IDTVEC(linux_syscall)) 372#else 373 if (frame.tf_eip != IDTVEC(syscall)) 374#endif 375 frame.tf_eflags &= ~PSL_T; 376 return; 377#endif 378 379#if NISA > 0 380 case T_NMI: 381#ifdef POWERFAIL_NMI 382#ifndef TIMER_FREQ 383# define TIMER_FREQ 1193182 384#endif 385 handle_powerfail: 386 { 387 static unsigned lastalert = 0; 388 389 if(time.tv_sec - lastalert > 10) 390 { 391 log(LOG_WARNING, "NMI: power fail\n"); 392 sysbeep(TIMER_FREQ/880, hz); 393 lastalert = time.tv_sec; 394 } 395 return; 396 } 397#else /* !POWERFAIL_NMI */ 398#ifdef DDB 399 /* NMI can be hooked up to a pushbutton for debugging */ 400 printf ("NMI ... going to debugger\n"); 401 if (kdb_trap (type, 0, &frame)) 402 return; 403#endif /* DDB */ 404 /* machine/parity/power fail/"kitchen sink" faults */ 405 if (isa_nmi(code) == 0) return; 406 /* FALL THROUGH */ 407#endif /* POWERFAIL_NMI */ 408#endif /* NISA > 0 */ 409 } 410 411 trap_fatal(&frame); 412 return; 413 } 414 415 trapsignal(p, i, ucode); 416 417#ifdef DEBUG 418 eva = rcr2(); 419 if (type <= MAX_TRAP_MSG) { 420 uprintf("fatal process exception: %s", 421 trap_msg[type]); 422 if ((type == T_PAGEFLT) || (type == T_PROTFLT)) 423 uprintf(", fault VA = 0x%x", eva); 424 uprintf("\n"); 425 } 426#endif 427 428out: 429 userret(p, &frame, sticks); 430} 431 432#ifdef notyet 433/* 434 * This version doesn't allow a page fault to user space while 435 * in the kernel. The rest of the kernel needs to be made "safe" 436 * before this can be used. I think the only things remaining 437 * to be made safe are the iBCS2 code and the process tracing/ 438 * debugging code. 439 */ 440int 441trap_pfault(frame, usermode) 442 struct trapframe *frame; 443 int usermode; 444{ 445 vm_offset_t va; 446 struct vmspace *vm = NULL; 447 vm_map_t map = 0; 448 int rv = 0; 449 vm_prot_t ftype; 450 int eva; 451 struct proc *p = curproc; 452 453 if (frame->tf_err & PGEX_W) 454 ftype = VM_PROT_READ | VM_PROT_WRITE; 455 else 456 ftype = VM_PROT_READ; 457 458 eva = rcr2(); 459 va = trunc_page((vm_offset_t)eva); 460 461 if (va < VM_MIN_KERNEL_ADDRESS) { 462 vm_offset_t v; 463 vm_page_t ptepg; 464 465 if (p == NULL || 466 (!usermode && va < VM_MAXUSER_ADDRESS && 467 (curpcb == NULL || curpcb->pcb_onfault == NULL))) { 468 trap_fatal(frame); 469 return (-1); 470 } 471 472 /* 473 * This is a fault on non-kernel virtual memory. 474 * vm is initialized above to NULL. If curproc is NULL 475 * or curproc->p_vmspace is NULL the fault is fatal. 476 */ 477 vm = p->p_vmspace; 478 if (vm == NULL) 479 goto nogo; 480 481 map = &vm->vm_map; 482 483 /* 484 * Keep swapout from messing with us during this 485 * critical time. 486 */ 487 ++p->p_lock; 488 489 /* 490 * Grow the stack if necessary 491 */ 492 if ((caddr_t)va > vm->vm_maxsaddr 493 && (caddr_t)va < (caddr_t)USRSTACK) { 494 if (!grow(p, va)) { 495 rv = KERN_FAILURE; 496 --p->p_lock; 497 goto nogo; 498 } 499 } 500 501 /* 502 * Check if page table is mapped, if not, 503 * fault it first 504 */ 505 v = (vm_offset_t) vtopte(va); 506 507 /* Fault the pte only if needed: */ 508 if (*((int *)vtopte(v)) == 0) 509 (void) vm_fault(map, trunc_page(v), VM_PROT_WRITE, FALSE); 510 511 pmap_use_pt( vm_map_pmap(map), va); 512 513 /* Fault in the user page: */ 514 rv = vm_fault(map, va, ftype, FALSE); 515 516 pmap_unuse_pt( vm_map_pmap(map), va); 517 518 --p->p_lock; 519 } else { 520 /* 521 * Don't allow user-mode faults in kernel address space. 522 */ 523 if (usermode) 524 goto nogo; 525 526 /* 527 * Since we know that kernel virtual address addresses 528 * always have pte pages mapped, we just have to fault 529 * the page. 530 */ 531 rv = vm_fault(kernel_map, va, ftype, FALSE); 532 } 533 534 if (rv == KERN_SUCCESS) 535 return (0); 536nogo: 537 if (!usermode) { 538 if (curpcb && curpcb->pcb_onfault) { 539 frame->tf_eip = (int)curpcb->pcb_onfault; 540 return (0); 541 } 542 trap_fatal(frame); 543 return (-1); 544 } 545 546 /* kludge to pass faulting virtual address to sendsig */ 547 frame->tf_err = eva; 548 549 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 550} 551#endif 552 553int 554trap_pfault(frame, usermode) 555 struct trapframe *frame; 556 int usermode; 557{ 558 vm_offset_t va; 559 struct vmspace *vm = NULL; 560 vm_map_t map = 0; 561 int rv = 0; 562 vm_prot_t ftype; 563 int eva; 564 struct proc *p = curproc; 565 566 eva = rcr2(); 567 va = trunc_page((vm_offset_t)eva); 568 569 if (va >= KERNBASE) { 570 /* 571 * Don't allow user-mode faults in kernel address space. 572 */ 573 if (usermode) 574 goto nogo; 575 576 map = kernel_map; 577 } else { 578 /* 579 * This is a fault on non-kernel virtual memory. 580 * vm is initialized above to NULL. If curproc is NULL 581 * or curproc->p_vmspace is NULL the fault is fatal. 582 */ 583 if (p != NULL) 584 vm = p->p_vmspace; 585 586 if (vm == NULL) 587 goto nogo; 588 589 map = &vm->vm_map; 590 } 591 592 if (frame->tf_err & PGEX_W) 593 ftype = VM_PROT_READ | VM_PROT_WRITE; 594 else 595 ftype = VM_PROT_READ; 596 597 if (map != kernel_map) { 598 vm_offset_t v; 599 vm_page_t ptepg; 600 601 /* 602 * Keep swapout from messing with us during this 603 * critical time. 604 */ 605 ++p->p_lock; 606 607 /* 608 * Grow the stack if necessary 609 */ 610 if ((caddr_t)va > vm->vm_maxsaddr 611 && (caddr_t)va < (caddr_t)USRSTACK) { 612 if (!grow(p, va)) { 613 rv = KERN_FAILURE; 614 --p->p_lock; 615 goto nogo; 616 } 617 } 618 619 /* 620 * Check if page table is mapped, if not, 621 * fault it first 622 */ 623 v = (vm_offset_t) vtopte(va); 624 625 /* Fault the pte only if needed: */ 626 if (*((int *)vtopte(v)) == 0) 627 (void) vm_fault(map, trunc_page(v), VM_PROT_WRITE, FALSE); 628 629 pmap_use_pt( vm_map_pmap(map), va); 630 631 /* Fault in the user page: */ 632 rv = vm_fault(map, va, ftype, FALSE); 633 634 pmap_unuse_pt( vm_map_pmap(map), va); 635 636 --p->p_lock; 637 } else { 638 /* 639 * Since we know that kernel virtual address addresses 640 * always have pte pages mapped, we just have to fault 641 * the page. 642 */ 643 rv = vm_fault(map, va, ftype, FALSE); 644 } 645 646 if (rv == KERN_SUCCESS) 647 return (0); 648nogo: 649 if (!usermode) { 650 if (curpcb && curpcb->pcb_onfault) { 651 frame->tf_eip = (int)curpcb->pcb_onfault; 652 return (0); 653 } 654 trap_fatal(frame); 655 return (-1); 656 } 657 658 /* kludge to pass faulting virtual address to sendsig */ 659 frame->tf_err = eva; 660 661 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV); 662} 663 664void 665trap_fatal(frame) 666 struct trapframe *frame; 667{ 668 int code, type, eva; 669 struct soft_segment_descriptor softseg; 670 671 code = frame->tf_err; 672 type = frame->tf_trapno; 673 eva = rcr2(); 674 sdtossd(&gdt[IDXSEL(frame->tf_cs & 0xffff)].sd, &softseg); 675 676 if (type <= MAX_TRAP_MSG) 677 printf("\n\nFatal trap %d: %s while in %s mode\n", 678 type, trap_msg[type], 679 ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel"); 680 if (type == T_PAGEFLT) { 681 printf("fault virtual address = 0x%x\n", eva); 682 printf("fault code = %s %s, %s\n", 683 code & PGEX_U ? "user" : "supervisor", 684 code & PGEX_W ? "write" : "read", 685 code & PGEX_P ? "protection violation" : "page not present"); 686 } 687 printf("instruction pointer = 0x%x:0x%x\n", frame->tf_cs & 0xffff, frame->tf_eip); 688 printf("code segment = base 0x%x, limit 0x%x, type 0x%x\n", 689 softseg.ssd_base, softseg.ssd_limit, softseg.ssd_type); 690 printf(" = DPL %d, pres %d, def32 %d, gran %d\n", 691 softseg.ssd_dpl, softseg.ssd_p, softseg.ssd_def32, softseg.ssd_gran); 692 printf("processor eflags = "); 693 if (frame->tf_eflags & PSL_T) 694 printf("trace/trap, "); 695 if (frame->tf_eflags & PSL_I) 696 printf("interrupt enabled, "); 697 if (frame->tf_eflags & PSL_NT) 698 printf("nested task, "); 699 if (frame->tf_eflags & PSL_RF) 700 printf("resume, "); 701 if (frame->tf_eflags & PSL_VM) 702 printf("vm86, "); 703 printf("IOPL = %d\n", (frame->tf_eflags & PSL_IOPL) >> 12); 704 printf("current process = "); 705 if (curproc) { 706 printf("%lu (%s)\n", 707 (u_long)curproc->p_pid, curproc->p_comm ? 708 curproc->p_comm : ""); 709 } else { 710 printf("Idle\n"); 711 } 712 printf("interrupt mask = "); 713 if ((cpl & net_imask) == net_imask) 714 printf("net "); 715 if ((cpl & tty_imask) == tty_imask) 716 printf("tty "); 717 if ((cpl & bio_imask) == bio_imask) 718 printf("bio "); 719 if (cpl == 0) 720 printf("none"); 721 printf("\n"); 722 723#ifdef KDB 724 if (kdb_trap(&psl)) 725 return; 726#endif 727#ifdef DDB 728 if (kdb_trap (type, 0, frame)) 729 return; 730#endif 731 if (type <= MAX_TRAP_MSG) 732 panic(trap_msg[type]); 733 else 734 panic("unknown/reserved trap"); 735} 736 737/* 738 * Compensate for 386 brain damage (missing URKR). 739 * This is a little simpler than the pagefault handler in trap() because 740 * it the page tables have already been faulted in and high addresses 741 * are thrown out early for other reasons. 742 */ 743int trapwrite(addr) 744 unsigned addr; 745{ 746 struct proc *p; 747 vm_offset_t va, v; 748 struct vmspace *vm; 749 int rv; 750 751 va = trunc_page((vm_offset_t)addr); 752 /* 753 * XXX - MAX is END. Changed > to >= for temp. fix. 754 */ 755 if (va >= VM_MAXUSER_ADDRESS) 756 return (1); 757 758 p = curproc; 759 vm = p->p_vmspace; 760 761 ++p->p_lock; 762 763 if ((caddr_t)va >= vm->vm_maxsaddr 764 && (caddr_t)va < (caddr_t)USRSTACK) { 765 if (!grow(p, va)) { 766 --p->p_lock; 767 return (1); 768 } 769 } 770 771 v = trunc_page(vtopte(va)); 772 773 /* 774 * wire the pte page 775 */ 776 if (va < USRSTACK) { 777 vm_map_pageable(&vm->vm_map, v, round_page(v+1), FALSE); 778 } 779 780 /* 781 * fault the data page 782 */ 783 rv = vm_fault(&vm->vm_map, va, VM_PROT_READ|VM_PROT_WRITE, FALSE); 784 785 /* 786 * unwire the pte page 787 */ 788 if (va < USRSTACK) { 789 vm_map_pageable(&vm->vm_map, v, round_page(v+1), TRUE); 790 } 791 792 --p->p_lock; 793 794 if (rv != KERN_SUCCESS) 795 return 1; 796 797 return (0); 798} 799 800/* 801 * syscall(frame): 802 * System call request from POSIX system call gate interface to kernel. 803 * Like trap(), argument is call by reference. 804 */ 805/*ARGSUSED*/ 806void 807syscall(frame) 808 struct trapframe frame; 809{ 810 caddr_t params; 811 int i; 812 struct sysent *callp; 813 struct proc *p = curproc; 814 u_quad_t sticks; 815 int error; 816 int args[8], rval[2]; 817 u_int code; 818 819 sticks = p->p_sticks; 820 if (ISPL(frame.tf_cs) != SEL_UPL) 821 panic("syscall"); 822 823 p->p_md.md_regs = (int *)&frame; 824 params = (caddr_t)frame.tf_esp + sizeof(int); 825 code = frame.tf_eax; 826 /* 827 * Need to check if this is a 32 bit or 64 bit syscall. 828 */ 829 if (code == SYS_syscall) { 830 /* 831 * Code is first argument, followed by actual args. 832 */ 833 code = fuword(params); 834 params += sizeof(int); 835 } else if (code == SYS___syscall) { 836 /* 837 * Like syscall, but code is a quad, so as to maintain 838 * quad alignment for the rest of the arguments. 839 */ 840 code = fuword(params); 841 params += sizeof(quad_t); 842 } 843 844 if (p->p_sysent->sv_mask) 845 code &= p->p_sysent->sv_mask; 846 847 if (code >= p->p_sysent->sv_size) 848 callp = &p->p_sysent->sv_table[0]; 849 else 850 callp = &p->p_sysent->sv_table[code]; 851 852 if ((i = callp->sy_narg * sizeof(int)) && 853 (error = copyin(params, (caddr_t)args, (u_int)i))) { 854#ifdef KTRACE 855 if (KTRPOINT(p, KTR_SYSCALL)) 856 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 857#endif 858 goto bad; 859 } 860#ifdef KTRACE 861 if (KTRPOINT(p, KTR_SYSCALL)) 862 ktrsyscall(p->p_tracep, code, callp->sy_narg, args); 863#endif 864 rval[0] = 0; 865 rval[1] = frame.tf_edx; 866 867 error = (*callp->sy_call)(p, args, rval); 868 869 switch (error) { 870 871 case 0: 872 /* 873 * Reinitialize proc pointer `p' as it may be different 874 * if this is a child returning from fork syscall. 875 */ 876 p = curproc; 877 frame.tf_eax = rval[0]; 878 frame.tf_edx = rval[1]; 879 frame.tf_eflags &= ~PSL_C; /* carry bit */ 880 break; 881 882 case ERESTART: 883 /* 884 * Reconstruct pc, assuming lcall $X,y is 7 bytes. 885 */ 886 frame.tf_eip -= 7; 887 break; 888 889 case EJUSTRETURN: 890 break; 891 892 default: 893bad: 894 if (p->p_sysent->sv_errsize) 895 if (error >= p->p_sysent->sv_errsize) 896 error = -1; /* XXX */ 897 else 898 error = p->p_sysent->sv_errtbl[error]; 899 frame.tf_eax = error; 900 frame.tf_eflags |= PSL_C; /* carry bit */ 901 break; 902 } 903 904#if 1 905 if (frame.tf_eflags & PSL_T) { 906 /* traced syscall, raise sig */ 907 frame.tf_eflags &= ~PSL_T; 908 if (ISPL(frame.tf_cs) == SEL_UPL) { 909 trapsignal(p, SIGTRAP, 0); 910 } 911 } 912#endif 913 userret(p, &frame, sticks); 914 915#ifdef KTRACE 916 if (KTRPOINT(p, KTR_SYSRET)) 917 ktrsysret(p->p_tracep, code, error, rval[0]); 918#endif 919} 920 921#ifdef COMPAT_LINUX 922/* 923 * linux_syscall(frame): 924 */ 925/*ARGSUSED*/ 926void 927linux_syscall(frame) 928 struct trapframe frame; 929{ 930 int i; 931 struct proc *p = curproc; 932 struct sysent *callp; 933 u_quad_t sticks; 934 int error; 935 int rval[2]; 936 u_int code; 937 struct linux_syscall_args { 938 int arg1; 939 int arg2; 940 int arg3; 941 int arg4; 942 int arg5; 943 } args; 944 945 args.arg1 = frame.tf_ebx; 946 args.arg2 = frame.tf_ecx; 947 args.arg3 = frame.tf_edx; 948 args.arg4 = frame.tf_esi; 949 args.arg5 = frame.tf_edi; 950 951 sticks = p->p_sticks; 952 if (ISPL(frame.tf_cs) != SEL_UPL) 953 panic("linux syscall"); 954 955 p->p_md.md_regs = (int *)&frame; 956 code = frame.tf_eax; 957 958 if (p->p_sysent->sv_mask) 959 code &= p->p_sysent->sv_mask; 960 961 if (code >= p->p_sysent->sv_size) 962 callp = &p->p_sysent->sv_table[0]; 963 else 964 callp = &p->p_sysent->sv_table[code]; 965 966#ifdef KTRACE 967 if (KTRPOINT(p, KTR_SYSCALL)) 968 ktrsyscall(p->p_tracep, code, callp->sy_narg, (int *)&args); 969#endif 970 971 rval[0] = 0; 972 973 error = (*callp->sy_call)(p, &args, rval); 974 975 switch (error) { 976 977 case 0: 978 /* 979 * Reinitialize proc pointer `p' as it may be different 980 * if this is a child returning from fork syscall. 981 */ 982 p = curproc; 983 frame.tf_eax = rval[0]; 984 frame.tf_eflags &= ~PSL_C; /* carry bit */ 985 break; 986 987 case ERESTART: 988 /* Reconstruct pc, subtract size of int 0x80 */ 989 frame.tf_eip -= 2; 990 break; 991 992 case EJUSTRETURN: 993 break; 994 995 default: 996 if (p->p_sysent->sv_errsize) 997 if (error >= p->p_sysent->sv_errsize) 998 error = -1; /* XXX */ 999 else 1000 error = p->p_sysent->sv_errtbl[error]; 1001 frame.tf_eax = -error; 1002 frame.tf_eflags |= PSL_C; /* carry bit */ 1003 break; 1004 } 1005 1006#if 1 1007 if (frame.tf_eflags & PSL_T) { 1008 /* traced syscall, raise sig */ 1009 frame.tf_eflags &= ~PSL_T; 1010 if (ISPL(frame.tf_cs) == SEL_UPL) { 1011 trapsignal(p, SIGTRAP, 0); 1012 } 1013 } 1014#endif 1015 userret(p, &frame, sticks); 1016 1017#ifdef KTRACE 1018 if (KTRPOINT(p, KTR_SYSRET)) 1019 ktrsysret(p->p_tracep, code, error, rval[0]); 1020#endif 1021} 1022#endif /* COMPAT_LINUX */ 1023