The Design and Implementation of the FreeBSD Operating System, Second Edition
Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]

FreeBSD/Linux Kernel Cross Reference
sys/i386/i386/trap.c

Version: -  FREEBSD  -  FREEBSD-13-STABLE  -  FREEBSD-13-0  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  l41  -  OPENBSD  -  linux-2.6  -  MK84  -  PLAN9  -  xnu-8792 
SearchContext: -  none  -  3  -  10 

    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  */
   39 
   40 #include <sys/cdefs.h>
   41 __FBSDID("$FreeBSD$");
   42 
   43 /*
   44  * 386 Trap and System call handling
   45  */
   46 
   47 #include "opt_clock.h"
   48 #include "opt_cpu.h"
   49 #include "opt_hwpmc_hooks.h"
   50 #include "opt_isa.h"
   51 #include "opt_kdb.h"
   52 #include "opt_kdtrace.h"
   53 #include "opt_ktrace.h"
   54 #include "opt_npx.h"
   55 #include "opt_trap.h"
   56 
   57 #include <sys/param.h>
   58 #include <sys/bus.h>
   59 #include <sys/systm.h>
   60 #include <sys/proc.h>
   61 #include <sys/pioctl.h>
   62 #include <sys/ptrace.h>
   63 #include <sys/kdb.h>
   64 #include <sys/kernel.h>
   65 #include <sys/ktr.h>
   66 #include <sys/lock.h>
   67 #include <sys/mutex.h>
   68 #include <sys/resourcevar.h>
   69 #include <sys/signalvar.h>
   70 #include <sys/syscall.h>
   71 #include <sys/sysctl.h>
   72 #include <sys/sysent.h>
   73 #include <sys/uio.h>
   74 #include <sys/vmmeter.h>
   75 #ifdef KTRACE
   76 #include <sys/ktrace.h>
   77 #endif
   78 #ifdef HWPMC_HOOKS
   79 #include <sys/pmckern.h>
   80 #endif
   81 #include <security/audit/audit.h>
   82 
   83 #include <vm/vm.h>
   84 #include <vm/vm_param.h>
   85 #include <vm/pmap.h>
   86 #include <vm/vm_kern.h>
   87 #include <vm/vm_map.h>
   88 #include <vm/vm_page.h>
   89 #include <vm/vm_extern.h>
   90 
   91 #include <machine/cpu.h>
   92 #include <machine/intr_machdep.h>
   93 #include <machine/md_var.h>
   94 #include <machine/pcb.h>
   95 #ifdef SMP
   96 #include <machine/smp.h>
   97 #endif
   98 #include <machine/tss.h>
   99 #include <machine/vm86.h>
  100 
  101 #ifdef POWERFAIL_NMI
  102 #include <sys/syslog.h>
  103 #include <machine/clock.h>
  104 #endif
  105 
  106 #ifdef KDTRACE_HOOKS
  107 #include <sys/dtrace_bsd.h>
  108 
  109 /*
  110  * This is a hook which is initialised by the dtrace module
  111  * to handle traps which might occur during DTrace probe
  112  * execution.
  113  */
  114 dtrace_trap_func_t      dtrace_trap_func;
  115 
  116 dtrace_doubletrap_func_t        dtrace_doubletrap_func;
  117 
  118 /*
  119  * This is a hook which is initialised by the systrace module
  120  * when it is loaded. This keeps the DTrace syscall provider
  121  * implementation opaque. 
  122  */
  123 systrace_probe_func_t   systrace_probe_func;
  124 #endif
  125 
  126 extern void trap(struct trapframe *frame);
  127 extern void syscall(struct trapframe *frame);
  128 
  129 static int trap_pfault(struct trapframe *, int, vm_offset_t);
  130 static void trap_fatal(struct trapframe *, vm_offset_t);
  131 void dblfault_handler(void);
  132 
  133 extern inthand_t IDTVEC(lcall_syscall);
  134 
  135 #define MAX_TRAP_MSG            30
  136 static char *trap_msg[] = {
  137         "",                                     /*  0 unused */
  138         "privileged instruction fault",         /*  1 T_PRIVINFLT */
  139         "",                                     /*  2 unused */
  140         "breakpoint instruction fault",         /*  3 T_BPTFLT */
  141         "",                                     /*  4 unused */
  142         "",                                     /*  5 unused */
  143         "arithmetic trap",                      /*  6 T_ARITHTRAP */
  144         "",                                     /*  7 unused */
  145         "",                                     /*  8 unused */
  146         "general protection fault",             /*  9 T_PROTFLT */
  147         "trace trap",                           /* 10 T_TRCTRAP */
  148         "",                                     /* 11 unused */
  149         "page fault",                           /* 12 T_PAGEFLT */
  150         "",                                     /* 13 unused */
  151         "alignment fault",                      /* 14 T_ALIGNFLT */
  152         "",                                     /* 15 unused */
  153         "",                                     /* 16 unused */
  154         "",                                     /* 17 unused */
  155         "integer divide fault",                 /* 18 T_DIVIDE */
  156         "non-maskable interrupt trap",          /* 19 T_NMI */
  157         "overflow trap",                        /* 20 T_OFLOW */
  158         "FPU bounds check fault",               /* 21 T_BOUND */
  159         "FPU device not available",             /* 22 T_DNA */
  160         "double fault",                         /* 23 T_DOUBLEFLT */
  161         "FPU operand fetch fault",              /* 24 T_FPOPFLT */
  162         "invalid TSS fault",                    /* 25 T_TSSFLT */
  163         "segment not present fault",            /* 26 T_SEGNPFLT */
  164         "stack fault",                          /* 27 T_STKFLT */
  165         "machine check trap",                   /* 28 T_MCHK */
  166         "SIMD floating-point exception",        /* 29 T_XMMFLT */
  167         "reserved (unknown) fault",             /* 30 T_RESERVED */
  168 };
  169 
  170 #if defined(I586_CPU) && !defined(NO_F00F_HACK)
  171 extern int has_f00f_bug;
  172 #endif
  173 
  174 #ifdef KDB
  175 static int kdb_on_nmi = 1;
  176 SYSCTL_INT(_machdep, OID_AUTO, kdb_on_nmi, CTLFLAG_RW,
  177         &kdb_on_nmi, 0, "Go to KDB on NMI");
  178 #endif
  179 static int panic_on_nmi = 1;
  180 SYSCTL_INT(_machdep, OID_AUTO, panic_on_nmi, CTLFLAG_RW,
  181         &panic_on_nmi, 0, "Panic on NMI");
  182 static int prot_fault_translation = 0;
  183 SYSCTL_INT(_machdep, OID_AUTO, prot_fault_translation, CTLFLAG_RW,
  184         &prot_fault_translation, 0, "Select signal to deliver on protection fault");
  185 
  186 extern char *syscallnames[];
  187 
  188 /*
  189  * Exception, fault, and trap interface to the FreeBSD kernel.
  190  * This common code is called from assembly language IDT gate entry
  191  * routines that prepare a suitable stack frame, and restore this
  192  * frame after the exception has been processed.
  193  */
  194 
  195 void
  196 trap(struct trapframe *frame)
  197 {
  198         struct thread *td = curthread;
  199         struct proc *p = td->td_proc;
  200         int i = 0, ucode = 0, code;
  201         u_int type;
  202         register_t addr = 0;
  203         vm_offset_t eva;
  204         ksiginfo_t ksi;
  205 #ifdef POWERFAIL_NMI
  206         static int lastalert = 0;
  207 #endif
  208 
  209         PCPU_INC(cnt.v_trap);
  210         type = frame->tf_trapno;
  211 
  212 #ifdef SMP
  213 #ifdef STOP_NMI
  214         /* Handler for NMI IPIs used for stopping CPUs. */
  215         if (type == T_NMI) {
  216                  if (ipi_nmi_handler() == 0)
  217                            goto out;
  218         }
  219 #endif /* STOP_NMI */
  220 #endif /* SMP */
  221 
  222 #ifdef KDB
  223         if (kdb_active) {
  224                 kdb_reenter();
  225                 goto out;
  226         }
  227 #endif
  228 
  229 #ifdef  HWPMC_HOOKS
  230         /*
  231          * CPU PMCs interrupt using an NMI so we check for that first.
  232          * If the HWPMC module is active, 'pmc_hook' will point to
  233          * the function to be called.  A return value of '1' from the
  234          * hook means that the NMI was handled by it and that we can
  235          * return immediately.
  236          */
  237         if (type == T_NMI && pmc_intr &&
  238             (*pmc_intr)(PCPU_GET(cpuid), (uintptr_t) frame->tf_eip,
  239                 TRAPF_USERMODE(frame)))
  240             goto out;
  241 #endif
  242 
  243 #ifdef KDTRACE_HOOKS
  244         /*
  245          * A trap can occur while DTrace executes a probe. Before
  246          * executing the probe, DTrace blocks re-scheduling and sets
  247          * a flag in it's per-cpu flags to indicate that it doesn't
  248          * want to fault. On returning from the the probe, the no-fault
  249          * flag is cleared and finally re-scheduling is enabled.
  250          *
  251          * If the DTrace kernel module has registered a trap handler,
  252          * call it and if it returns non-zero, assume that it has
  253          * handled the trap and modified the trap frame so that this
  254          * function can return normally.
  255          */
  256         if ((type == T_PROTFLT || type == T_PAGEFLT) &&
  257             dtrace_trap_func != NULL)
  258                 if ((*dtrace_trap_func)(frame, type))
  259                         goto out;
  260 #endif
  261 
  262         if ((frame->tf_eflags & PSL_I) == 0) {
  263                 /*
  264                  * Buggy application or kernel code has disabled
  265                  * interrupts and then trapped.  Enabling interrupts
  266                  * now is wrong, but it is better than running with
  267                  * interrupts disabled until they are accidentally
  268                  * enabled later.
  269                  */
  270                 if (ISPL(frame->tf_cs) == SEL_UPL || (frame->tf_eflags & PSL_VM))
  271                         printf(
  272                             "pid %ld (%s): trap %d with interrupts disabled\n",
  273                             (long)curproc->p_pid, curproc->p_comm, type);
  274                 else if (type != T_BPTFLT && type != T_TRCTRAP &&
  275                          frame->tf_eip != (int)cpu_switch_load_gs) {
  276                         /*
  277                          * XXX not quite right, since this may be for a
  278                          * multiple fault in user mode.
  279                          */
  280                         printf("kernel trap %d with interrupts disabled\n",
  281                             type);
  282                         /*
  283                          * Page faults need interrupts disabled until later,
  284                          * and we shouldn't enable interrupts while holding
  285                          * a spin lock or if servicing an NMI.
  286                          */
  287                         if (type != T_NMI && type != T_PAGEFLT &&
  288                             td->td_md.md_spinlock_count == 0)
  289                                 enable_intr();
  290                 }
  291         }
  292 
  293         eva = 0;
  294         code = frame->tf_err;
  295         if (type == T_PAGEFLT) {
  296                 /*
  297                  * For some Cyrix CPUs, %cr2 is clobbered by
  298                  * interrupts.  This problem is worked around by using
  299                  * an interrupt gate for the pagefault handler.  We
  300                  * are finally ready to read %cr2 and then must
  301                  * reenable interrupts.
  302                  *
  303                  * If we get a page fault while in a critical section, then
  304                  * it is most likely a fatal kernel page fault.  The kernel
  305                  * is already going to panic trying to get a sleep lock to
  306                  * do the VM lookup, so just consider it a fatal trap so the
  307                  * kernel can print out a useful trap message and even get
  308                  * to the debugger.
  309                  *
  310                  * If we get a page fault while holding a non-sleepable
  311                  * lock, then it is most likely a fatal kernel page fault.
  312                  * If WITNESS is enabled, then it's going to whine about
  313                  * bogus LORs with various VM locks, so just skip to the
  314                  * fatal trap handling directly.
  315                  */
  316                 eva = rcr2();
  317                 if (td->td_critnest != 0 ||
  318                     WITNESS_CHECK(WARN_SLEEPOK | WARN_GIANTOK, NULL,
  319                     "Kernel page fault") != 0)
  320                         trap_fatal(frame, eva);
  321                 else
  322                         enable_intr();
  323         }
  324 
  325         if ((ISPL(frame->tf_cs) == SEL_UPL) ||
  326             ((frame->tf_eflags & PSL_VM) && 
  327                 !(PCPU_GET(curpcb)->pcb_flags & PCB_VM86CALL))) {
  328                 /* user trap */
  329 
  330                 td->td_pticks = 0;
  331                 td->td_frame = frame;
  332                 addr = frame->tf_eip;
  333                 if (td->td_ucred != p->p_ucred) 
  334                         cred_update_thread(td);
  335 
  336                 switch (type) {
  337                 case T_PRIVINFLT:       /* privileged instruction fault */
  338                         i = SIGILL;
  339                         ucode = ILL_PRVOPC;
  340                         break;
  341 
  342                 case T_BPTFLT:          /* bpt instruction fault */
  343                 case T_TRCTRAP:         /* trace trap */
  344                         enable_intr();
  345                         frame->tf_eflags &= ~PSL_T;
  346                         i = SIGTRAP;
  347                         ucode = (type == T_TRCTRAP ? TRAP_TRACE : TRAP_BRKPT);
  348                         break;
  349 
  350                 case T_ARITHTRAP:       /* arithmetic trap */
  351 #ifdef DEV_NPX
  352                         ucode = npxtrap();
  353                         if (ucode == -1)
  354                                 goto userout;
  355 #else
  356                         ucode = 0;
  357 #endif
  358                         i = SIGFPE;
  359                         break;
  360 
  361                         /*
  362                          * The following two traps can happen in
  363                          * vm86 mode, and, if so, we want to handle
  364                          * them specially.
  365                          */
  366                 case T_PROTFLT:         /* general protection fault */
  367                 case T_STKFLT:          /* stack fault */
  368                         if (frame->tf_eflags & PSL_VM) {
  369                                 i = vm86_emulate((struct vm86frame *)frame);
  370                                 if (i == 0)
  371                                         goto user;
  372                                 break;
  373                         }
  374                         i = SIGBUS;
  375                         ucode = (type == T_PROTFLT) ? BUS_OBJERR : BUS_ADRERR;
  376                         break;
  377                 case T_SEGNPFLT:        /* segment not present fault */
  378                         i = SIGBUS;
  379                         ucode = BUS_ADRERR;
  380                         break;
  381                 case T_TSSFLT:          /* invalid TSS fault */
  382                         i = SIGBUS;
  383                         ucode = BUS_OBJERR;
  384                         break;
  385                 case T_DOUBLEFLT:       /* double fault */
  386                 default:
  387                         i = SIGBUS;
  388                         ucode = BUS_OBJERR;
  389                         break;
  390 
  391                 case T_PAGEFLT:         /* page fault */
  392 #ifdef KSE
  393                         if (td->td_pflags & TDP_SA)
  394                                 thread_user_enter(td);
  395 #endif
  396 
  397                         i = trap_pfault(frame, TRUE, eva);
  398 #if defined(I586_CPU) && !defined(NO_F00F_HACK)
  399                         if (i == -2) {
  400                                 /*
  401                                  * The f00f hack workaround has triggered, so
  402                                  * treat the fault as an illegal instruction 
  403                                  * (T_PRIVINFLT) instead of a page fault.
  404                                  */
  405                                 type = frame->tf_trapno = T_PRIVINFLT;
  406 
  407                                 /* Proceed as in that case. */
  408                                 ucode = ILL_PRVOPC;
  409                                 i = SIGILL;
  410                                 break;
  411                         }
  412 #endif
  413                         if (i == -1)
  414                                 goto userout;
  415                         if (i == 0)
  416                                 goto user;
  417 
  418                         if (i == SIGSEGV)
  419                                 ucode = SEGV_MAPERR;
  420                         else {
  421                                 if (prot_fault_translation == 0) {
  422                                         /*
  423                                          * Autodetect.
  424                                          * This check also covers the images
  425                                          * without the ABI-tag ELF note.
  426                                          */
  427                                         if (p->p_osrel >= 700004) {
  428                                                 i = SIGSEGV;
  429                                                 ucode = SEGV_ACCERR;
  430                                         } else {
  431                                                 i = SIGBUS;
  432                                                 ucode = BUS_PAGE_FAULT;
  433                                         }
  434                                 } else if (prot_fault_translation == 1) {
  435                                         /*
  436                                          * Always compat mode.
  437                                          */
  438                                         i = SIGBUS;
  439                                         ucode = BUS_PAGE_FAULT;
  440                                 } else {
  441                                         /*
  442                                          * Always SIGSEGV mode.
  443                                          */
  444                                         i = SIGSEGV;
  445                                         ucode = SEGV_ACCERR;
  446                                 }
  447                         }
  448                         addr = eva;
  449                         break;
  450 
  451                 case T_DIVIDE:          /* integer divide fault */
  452                         ucode = FPE_INTDIV;
  453                         i = SIGFPE;
  454                         break;
  455 
  456 #ifdef DEV_ISA
  457                 case T_NMI:
  458 #ifdef POWERFAIL_NMI
  459 #ifndef TIMER_FREQ
  460 #  define TIMER_FREQ 1193182
  461 #endif
  462                         mtx_lock(&Giant);
  463                         if (time_second - lastalert > 10) {
  464                                 log(LOG_WARNING, "NMI: power fail\n");
  465                                 sysbeep(TIMER_FREQ/880, hz);
  466                                 lastalert = time_second;
  467                         }
  468                         mtx_unlock(&Giant);
  469                         goto userout;
  470 #else /* !POWERFAIL_NMI */
  471                         /* machine/parity/power fail/"kitchen sink" faults */
  472                         /* XXX Giant */
  473                         if (isa_nmi(code) == 0) {
  474 #ifdef KDB
  475                                 /*
  476                                  * NMI can be hooked up to a pushbutton
  477                                  * for debugging.
  478                                  */
  479                                 if (kdb_on_nmi) {
  480                                         printf ("NMI ... going to debugger\n");
  481                                         kdb_trap(type, 0, frame);
  482                                 }
  483 #endif /* KDB */
  484                                 goto userout;
  485                         } else if (panic_on_nmi)
  486                                 panic("NMI indicates hardware failure");
  487                         break;
  488 #endif /* POWERFAIL_NMI */
  489 #endif /* DEV_ISA */
  490 
  491                 case T_OFLOW:           /* integer overflow fault */
  492                         ucode = FPE_INTOVF;
  493                         i = SIGFPE;
  494                         break;
  495 
  496                 case T_BOUND:           /* bounds check fault */
  497                         ucode = FPE_FLTSUB;
  498                         i = SIGFPE;
  499                         break;
  500 
  501                 case T_DNA:
  502 #ifdef DEV_NPX
  503                         /* transparent fault (due to context switch "late") */
  504                         if (npxdna())
  505                                 goto userout;
  506 #endif
  507                         printf("pid %d killed due to lack of floating point\n",
  508                                 p->p_pid);
  509                         i = SIGKILL;
  510                         ucode = 0;
  511                         break;
  512 
  513                 case T_FPOPFLT:         /* FPU operand fetch fault */
  514                         ucode = ILL_COPROC;
  515                         i = SIGILL;
  516                         break;
  517 
  518                 case T_XMMFLT:          /* SIMD floating-point exception */
  519                         ucode = 0; /* XXX */
  520                         i = SIGFPE;
  521                         break;
  522                 }
  523         } else {
  524                 /* kernel trap */
  525 
  526                 KASSERT(cold || td->td_ucred != NULL,
  527                     ("kernel trap doesn't have ucred"));
  528                 switch (type) {
  529                 case T_PAGEFLT:                 /* page fault */
  530                         (void) trap_pfault(frame, FALSE, eva);
  531                         goto out;
  532 
  533                 case T_DNA:
  534 #ifdef DEV_NPX
  535                         /*
  536                          * The kernel is apparently using npx for copying.
  537                          * XXX this should be fatal unless the kernel has
  538                          * registered such use.
  539                          */
  540                         if (npxdna())
  541                                 goto out;
  542 #endif
  543                         break;
  544 
  545                         /*
  546                          * The following two traps can happen in
  547                          * vm86 mode, and, if so, we want to handle
  548                          * them specially.
  549                          */
  550                 case T_PROTFLT:         /* general protection fault */
  551                 case T_STKFLT:          /* stack fault */
  552                         if (frame->tf_eflags & PSL_VM) {
  553                                 i = vm86_emulate((struct vm86frame *)frame);
  554                                 if (i != 0)
  555                                         /*
  556                                          * returns to original process
  557                                          */
  558                                         vm86_trap((struct vm86frame *)frame);
  559                                 goto out;
  560                         }
  561                         if (type == T_STKFLT)
  562                                 break;
  563 
  564                         /* FALL THROUGH */
  565 
  566                 case T_SEGNPFLT:        /* segment not present fault */
  567                         if (PCPU_GET(curpcb)->pcb_flags & PCB_VM86CALL)
  568                                 break;
  569 
  570                         /*
  571                          * Invalid %fs's and %gs's can be created using
  572                          * procfs or PT_SETREGS or by invalidating the
  573                          * underlying LDT entry.  This causes a fault
  574                          * in kernel mode when the kernel attempts to
  575                          * switch contexts.  Lose the bad context
  576                          * (XXX) so that we can continue, and generate
  577                          * a signal.
  578                          */
  579                         if (frame->tf_eip == (int)cpu_switch_load_gs) {
  580                                 PCPU_GET(curpcb)->pcb_gs = 0;
  581 #if 0                           
  582                                 PROC_LOCK(p);
  583                                 psignal(p, SIGBUS);
  584                                 PROC_UNLOCK(p);
  585 #endif                          
  586                                 goto out;
  587                         }
  588 
  589                         if (td->td_intr_nesting_level != 0)
  590                                 break;
  591 
  592                         /*
  593                          * Invalid segment selectors and out of bounds
  594                          * %eip's and %esp's can be set up in user mode.
  595                          * This causes a fault in kernel mode when the
  596                          * kernel tries to return to user mode.  We want
  597                          * to get this fault so that we can fix the
  598                          * problem here and not have to check all the
  599                          * selectors and pointers when the user changes
  600                          * them.
  601                          */
  602                         if (frame->tf_eip == (int)doreti_iret) {
  603                                 frame->tf_eip = (int)doreti_iret_fault;
  604                                 goto out;
  605                         }
  606                         if (frame->tf_eip == (int)doreti_popl_ds) {
  607                                 frame->tf_eip = (int)doreti_popl_ds_fault;
  608                                 goto out;
  609                         }
  610                         if (frame->tf_eip == (int)doreti_popl_es) {
  611                                 frame->tf_eip = (int)doreti_popl_es_fault;
  612                                 goto out;
  613                         }
  614                         if (frame->tf_eip == (int)doreti_popl_fs) {
  615                                 frame->tf_eip = (int)doreti_popl_fs_fault;
  616                                 goto out;
  617                         }
  618                         if (PCPU_GET(curpcb)->pcb_onfault != NULL) {
  619                                 frame->tf_eip =
  620                                     (int)PCPU_GET(curpcb)->pcb_onfault;
  621                                 goto out;
  622                         }
  623                         break;
  624 
  625                 case T_TSSFLT:
  626                         /*
  627                          * PSL_NT can be set in user mode and isn't cleared
  628                          * automatically when the kernel is entered.  This
  629                          * causes a TSS fault when the kernel attempts to
  630                          * `iret' because the TSS link is uninitialized.  We
  631                          * want to get this fault so that we can fix the
  632                          * problem here and not every time the kernel is
  633                          * entered.
  634                          */
  635                         if (frame->tf_eflags & PSL_NT) {
  636                                 frame->tf_eflags &= ~PSL_NT;
  637                                 goto out;
  638                         }
  639                         break;
  640 
  641                 case T_TRCTRAP:  /* trace trap */
  642                         if (frame->tf_eip == (int)IDTVEC(lcall_syscall)) {
  643                                 /*
  644                                  * We've just entered system mode via the
  645                                  * syscall lcall.  Continue single stepping
  646                                  * silently until the syscall handler has
  647                                  * saved the flags.
  648                                  */
  649                                 goto out;
  650                         }
  651                         if (frame->tf_eip == (int)IDTVEC(lcall_syscall) + 1) {
  652                                 /*
  653                                  * The syscall handler has now saved the
  654                                  * flags.  Stop single stepping it.
  655                                  */
  656                                 frame->tf_eflags &= ~PSL_T;
  657                                 goto out;
  658                         }
  659                         /*
  660                          * Ignore debug register trace traps due to
  661                          * accesses in the user's address space, which
  662                          * can happen under several conditions such as
  663                          * if a user sets a watchpoint on a buffer and
  664                          * then passes that buffer to a system call.
  665                          * We still want to get TRCTRAPS for addresses
  666                          * in kernel space because that is useful when
  667                          * debugging the kernel.
  668                          */
  669                         /* XXX Giant */
  670                         if (user_dbreg_trap() && 
  671                            !(PCPU_GET(curpcb)->pcb_flags & PCB_VM86CALL)) {
  672                                 /*
  673                                  * Reset breakpoint bits because the
  674                                  * processor doesn't
  675                                  */
  676                                 load_dr6(rdr6() & 0xfffffff0);
  677                                 goto out;
  678                         }
  679                         /*
  680                          * FALLTHROUGH (TRCTRAP kernel mode, kernel address)
  681                          */
  682                 case T_BPTFLT:
  683                         /*
  684                          * If KDB is enabled, let it handle the debugger trap.
  685                          * Otherwise, debugger traps "can't happen".
  686                          */
  687 #ifdef KDB
  688                         if (kdb_trap(type, 0, frame))
  689                                 goto out;
  690 #endif
  691                         break;
  692 
  693 #ifdef DEV_ISA
  694                 case T_NMI:
  695 #ifdef POWERFAIL_NMI
  696                         mtx_lock(&Giant);
  697                         if (time_second - lastalert > 10) {
  698                                 log(LOG_WARNING, "NMI: power fail\n");
  699                                 sysbeep(TIMER_FREQ/880, hz);
  700                                 lastalert = time_second;
  701                         }
  702                         mtx_unlock(&Giant);
  703                         goto out;
  704 #else /* !POWERFAIL_NMI */
  705                         /* XXX Giant */
  706                         /* machine/parity/power fail/"kitchen sink" faults */
  707                         if (isa_nmi(code) == 0) {
  708 #ifdef KDB
  709                                 /*
  710                                  * NMI can be hooked up to a pushbutton
  711                                  * for debugging.
  712                                  */
  713                                 if (kdb_on_nmi) {
  714                                         printf ("NMI ... going to debugger\n");
  715                                         kdb_trap(type, 0, frame);
  716                                 }
  717 #endif /* KDB */
  718                                 goto out;
  719                         } else if (panic_on_nmi == 0)
  720                                 goto out;
  721                         /* FALLTHROUGH */
  722 #endif /* POWERFAIL_NMI */
  723 #endif /* DEV_ISA */
  724                 }
  725 
  726                 trap_fatal(frame, eva);
  727                 goto out;
  728         }
  729 
  730         /* Translate fault for emulators (e.g. Linux) */
  731         if (*p->p_sysent->sv_transtrap)
  732                 i = (*p->p_sysent->sv_transtrap)(i, type);
  733 
  734         ksiginfo_init_trap(&ksi);
  735         ksi.ksi_signo = i;
  736         ksi.ksi_code = ucode;
  737         ksi.ksi_addr = (void *)addr;
  738         ksi.ksi_trapno = type;
  739         trapsignal(td, &ksi);
  740 
  741 #ifdef DEBUG
  742         if (type <= MAX_TRAP_MSG) {
  743                 uprintf("fatal process exception: %s",
  744                         trap_msg[type]);
  745                 if ((type == T_PAGEFLT) || (type == T_PROTFLT))
  746                         uprintf(", fault VA = 0x%lx", (u_long)eva);
  747                 uprintf("\n");
  748         }
  749 #endif
  750 
  751 user:
  752         userret(td, frame);
  753         mtx_assert(&Giant, MA_NOTOWNED);
  754 userout:
  755 out:
  756         return;
  757 }
  758 
  759 static int
  760 trap_pfault(frame, usermode, eva)
  761         struct trapframe *frame;
  762         int usermode;
  763         vm_offset_t eva;
  764 {
  765         vm_offset_t va;
  766         struct vmspace *vm = NULL;
  767         vm_map_t map;
  768         int rv = 0;
  769         vm_prot_t ftype;
  770         struct thread *td = curthread;
  771         struct proc *p = td->td_proc;
  772 
  773         va = trunc_page(eva);
  774         if (va >= KERNBASE) {
  775                 /*
  776                  * Don't allow user-mode faults in kernel address space.
  777                  * An exception:  if the faulting address is the invalid
  778                  * instruction entry in the IDT, then the Intel Pentium
  779                  * F00F bug workaround was triggered, and we need to
  780                  * treat it is as an illegal instruction, and not a page
  781                  * fault.
  782                  */
  783 #if defined(I586_CPU) && !defined(NO_F00F_HACK)
  784                 if ((eva == (unsigned int)&idt[6]) && has_f00f_bug)
  785                         return -2;
  786 #endif
  787                 if (usermode)
  788                         goto nogo;
  789 
  790                 map = kernel_map;
  791         } else {
  792                 /*
  793                  * This is a fault on non-kernel virtual memory.
  794                  * vm is initialized above to NULL. If curproc is NULL
  795                  * or curproc->p_vmspace is NULL the fault is fatal.
  796                  */
  797                 if (p != NULL)
  798                         vm = p->p_vmspace;
  799 
  800                 if (vm == NULL)
  801                         goto nogo;
  802 
  803                 map = &vm->vm_map;
  804         }
  805 
  806         /*
  807          * PGEX_I is defined only if the execute disable bit capability is
  808          * supported and enabled.
  809          */
  810         if (frame->tf_err & PGEX_W)
  811                 ftype = VM_PROT_WRITE;
  812 #ifdef PAE
  813         else if ((frame->tf_err & PGEX_I) && pg_nx != 0)
  814                 ftype = VM_PROT_EXECUTE;
  815 #endif
  816         else
  817                 ftype = VM_PROT_READ;
  818 
  819         if (map != kernel_map) {
  820                 /*
  821                  * Keep swapout from messing with us during this
  822                  *      critical time.
  823                  */
  824                 PROC_LOCK(p);
  825                 ++p->p_lock;
  826                 PROC_UNLOCK(p);
  827 
  828                 /* Fault in the user page: */
  829                 rv = vm_fault(map, va, ftype,
  830                               (ftype & VM_PROT_WRITE) ? VM_FAULT_DIRTY
  831                                                       : VM_FAULT_NORMAL);
  832 
  833                 PROC_LOCK(p);
  834                 --p->p_lock;
  835                 PROC_UNLOCK(p);
  836         } else {
  837                 /*
  838                  * Don't have to worry about process locking or stacks in the
  839                  * kernel.
  840                  */
  841                 rv = vm_fault(map, va, ftype, VM_FAULT_NORMAL);
  842         }
  843         if (rv == KERN_SUCCESS)
  844                 return (0);
  845 nogo:
  846         if (!usermode) {
  847                 if (td->td_intr_nesting_level == 0 &&
  848                     PCPU_GET(curpcb)->pcb_onfault != NULL) {
  849                         frame->tf_eip = (int)PCPU_GET(curpcb)->pcb_onfault;
  850                         return (0);
  851                 }
  852                 trap_fatal(frame, eva);
  853                 return (-1);
  854         }
  855 
  856         return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV);
  857 }
  858 
  859 static void
  860 trap_fatal(frame, eva)
  861         struct trapframe *frame;
  862         vm_offset_t eva;
  863 {
  864         int code, ss, esp;
  865         u_int type;
  866         struct soft_segment_descriptor softseg;
  867         char *msg;
  868 
  869         code = frame->tf_err;
  870         type = frame->tf_trapno;
  871         sdtossd(&gdt[IDXSEL(frame->tf_cs & 0xffff)].sd, &softseg);
  872 
  873         if (type <= MAX_TRAP_MSG)
  874                 msg = trap_msg[type];
  875         else
  876                 msg = "UNKNOWN";
  877         printf("\n\nFatal trap %d: %s while in %s mode\n", type, msg,
  878             frame->tf_eflags & PSL_VM ? "vm86" :
  879             ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel");
  880 #ifdef SMP
  881         /* two separate prints in case of a trap on an unmapped page */
  882         printf("cpuid = %d; ", PCPU_GET(cpuid));
  883         printf("apic id = %02x\n", PCPU_GET(apic_id));
  884 #endif
  885         if (type == T_PAGEFLT) {
  886                 printf("fault virtual address   = 0x%x\n", eva);
  887                 printf("fault code              = %s %s, %s\n",
  888                         code & PGEX_U ? "user" : "supervisor",
  889                         code & PGEX_W ? "write" : "read",
  890                         code & PGEX_P ? "protection violation" : "page not present");
  891         }
  892         printf("instruction pointer     = 0x%x:0x%x\n",
  893                frame->tf_cs & 0xffff, frame->tf_eip);
  894         if ((ISPL(frame->tf_cs) == SEL_UPL) || (frame->tf_eflags & PSL_VM)) {
  895                 ss = frame->tf_ss & 0xffff;
  896                 esp = frame->tf_esp;
  897         } else {
  898                 ss = GSEL(GDATA_SEL, SEL_KPL);
  899                 esp = (int)&frame->tf_esp;
  900         }
  901         printf("stack pointer           = 0x%x:0x%x\n", ss, esp);
  902         printf("frame pointer           = 0x%x:0x%x\n", ss, frame->tf_ebp);
  903         printf("code segment            = base 0x%x, limit 0x%x, type 0x%x\n",
  904                softseg.ssd_base, softseg.ssd_limit, softseg.ssd_type);
  905         printf("                        = DPL %d, pres %d, def32 %d, gran %d\n",
  906                softseg.ssd_dpl, softseg.ssd_p, softseg.ssd_def32,
  907                softseg.ssd_gran);
  908         printf("processor eflags        = ");
  909         if (frame->tf_eflags & PSL_T)
  910                 printf("trace trap, ");
  911         if (frame->tf_eflags & PSL_I)
  912                 printf("interrupt enabled, ");
  913         if (frame->tf_eflags & PSL_NT)
  914                 printf("nested task, ");
  915         if (frame->tf_eflags & PSL_RF)
  916                 printf("resume, ");
  917         if (frame->tf_eflags & PSL_VM)
  918                 printf("vm86, ");
  919         printf("IOPL = %d\n", (frame->tf_eflags & PSL_IOPL) >> 12);
  920         printf("current process         = ");
  921         if (curproc) {
  922                 printf("%lu (%s)\n", (u_long)curproc->p_pid, curproc->p_comm);
  923         } else {
  924                 printf("Idle\n");
  925         }
  926 
  927 #ifdef KDB
  928         if (debugger_on_panic || kdb_active) {
  929                 frame->tf_err = eva;    /* smuggle fault address to ddb */
  930                 if (kdb_trap(type, 0, frame)) {
  931                         frame->tf_err = code;   /* restore error code */
  932                         return;
  933                 }
  934                 frame->tf_err = code;           /* restore error code */
  935         }
  936 #endif
  937         printf("trap number             = %d\n", type);
  938         if (type <= MAX_TRAP_MSG)
  939                 panic("%s", trap_msg[type]);
  940         else
  941                 panic("unknown/reserved trap");
  942 }
  943 
  944 /*
  945  * Double fault handler. Called when a fault occurs while writing
  946  * a frame for a trap/exception onto the stack. This usually occurs
  947  * when the stack overflows (such is the case with infinite recursion,
  948  * for example).
  949  *
  950  * XXX Note that the current PTD gets replaced by IdlePTD when the
  951  * task switch occurs. This means that the stack that was active at
  952  * the time of the double fault is not available at <kstack> unless
  953  * the machine was idle when the double fault occurred. The downside
  954  * of this is that "trace <ebp>" in ddb won't work.
  955  */
  956 void
  957 dblfault_handler()
  958 {
  959 #ifdef KDTRACE_HOOKS
  960         if (dtrace_doubletrap_func != NULL)
  961                 (*dtrace_doubletrap_func)();
  962 #endif
  963         printf("\nFatal double fault:\n");
  964         printf("eip = 0x%x\n", PCPU_GET(common_tss.tss_eip));
  965         printf("esp = 0x%x\n", PCPU_GET(common_tss.tss_esp));
  966         printf("ebp = 0x%x\n", PCPU_GET(common_tss.tss_ebp));
  967 #ifdef SMP
  968         /* two separate prints in case of a trap on an unmapped page */
  969         printf("cpuid = %d; ", PCPU_GET(cpuid));
  970         printf("apic id = %02x\n", PCPU_GET(apic_id));
  971 #endif
  972         panic("double fault");
  973 }
  974 
  975 /*
  976  *      syscall -       system call request C handler
  977  *
  978  *      A system call is essentially treated as a trap.
  979  */
  980 void
  981 syscall(struct trapframe *frame)
  982 {
  983         caddr_t params;
  984         struct sysent *callp;
  985         struct thread *td = curthread;
  986         struct proc *p = td->td_proc;
  987         register_t orig_tf_eflags;
  988         int error;
  989         int narg;
  990         int args[8];
  991         u_int code;
  992         ksiginfo_t ksi;
  993 
  994         PCPU_INC(cnt.v_syscall);
  995 
  996 #ifdef DIAGNOSTIC
  997         if (ISPL(frame->tf_cs) != SEL_UPL) {
  998                 panic("syscall");
  999                 /* NOT REACHED */
 1000         }
 1001 #endif
 1002 
 1003         td->td_pticks = 0;
 1004         td->td_frame = frame;
 1005         if (td->td_ucred != p->p_ucred) 
 1006                 cred_update_thread(td);
 1007 #ifdef KSE
 1008         if (p->p_flag & P_SA)
 1009                 thread_user_enter(td);
 1010 #endif
 1011         params = (caddr_t)frame->tf_esp + sizeof(int);
 1012         code = frame->tf_eax;
 1013         orig_tf_eflags = frame->tf_eflags;
 1014 
 1015         if (p->p_sysent->sv_prepsyscall) {
 1016                 /*
 1017                  * The prep code is MP aware.
 1018                  */
 1019                 (*p->p_sysent->sv_prepsyscall)(frame, args, &code, &params);
 1020         } else {
 1021                 /*
 1022                  * Need to check if this is a 32 bit or 64 bit syscall.
 1023                  * fuword is MP aware.
 1024                  */
 1025                 if (code == SYS_syscall) {
 1026                         /*
 1027                          * Code is first argument, followed by actual args.
 1028                          */
 1029                         code = fuword(params);
 1030                         params += sizeof(int);
 1031                 } else if (code == SYS___syscall) {
 1032                         /*
 1033                          * Like syscall, but code is a quad, so as to maintain
 1034                          * quad alignment for the rest of the arguments.
 1035                          */
 1036                         code = fuword(params);
 1037                         params += sizeof(quad_t);
 1038                 }
 1039         }
 1040 
 1041         if (p->p_sysent->sv_mask)
 1042                 code &= p->p_sysent->sv_mask;
 1043 
 1044         if (code >= p->p_sysent->sv_size)
 1045                 callp = &p->p_sysent->sv_table[0];
 1046         else
 1047                 callp = &p->p_sysent->sv_table[code];
 1048 
 1049         narg = callp->sy_narg;
 1050 
 1051         /*
 1052          * copyin and the ktrsyscall()/ktrsysret() code is MP-aware
 1053          */
 1054         if (params != NULL && narg != 0)
 1055                 error = copyin(params, (caddr_t)args,
 1056                     (u_int)(narg * sizeof(int)));
 1057         else
 1058                 error = 0;
 1059                 
 1060 #ifdef KTRACE
 1061         if (KTRPOINT(td, KTR_SYSCALL))
 1062                 ktrsyscall(code, narg, args);
 1063 #endif
 1064 
 1065         CTR4(KTR_SYSC, "syscall enter thread %p pid %d proc %s code %d", td,
 1066             td->td_proc->p_pid, td->td_proc->p_comm, code);
 1067 
 1068         td->td_syscalls++;
 1069 
 1070         if (error == 0) {
 1071                 td->td_retval[0] = 0;
 1072                 td->td_retval[1] = frame->tf_edx;
 1073 
 1074                 STOPEVENT(p, S_SCE, narg);
 1075 
 1076                 PTRACESTOP_SC(p, td, S_PT_SCE);
 1077 
 1078 #ifdef KDTRACE_HOOKS
 1079                 /*
 1080                  * If the systrace module has registered it's probe
 1081                  * callback and if there is a probe active for the
 1082                  * syscall 'entry', process the probe.
 1083                  */
 1084                 if (systrace_probe_func != NULL && callp->sy_entry != 0)
 1085                         (*systrace_probe_func)(callp->sy_entry, code, callp,
 1086                             args);
 1087 #endif
 1088 
 1089                 AUDIT_SYSCALL_ENTER(code, td);
 1090                 error = (*callp->sy_call)(td, args);
 1091                 AUDIT_SYSCALL_EXIT(error, td);
 1092 
 1093                 /* Save the latest error return value. */
 1094                 td->td_errno = error;
 1095 
 1096 #ifdef KDTRACE_HOOKS
 1097                 /*
 1098                  * If the systrace module has registered it's probe
 1099                  * callback and if there is a probe active for the
 1100                  * syscall 'return', process the probe.
 1101                  */
 1102                 if (systrace_probe_func != NULL && callp->sy_return != 0)
 1103                         (*systrace_probe_func)(callp->sy_return, code, callp,
 1104                             args);
 1105 #endif
 1106         }
 1107 
 1108         switch (error) {
 1109         case 0:
 1110                 frame->tf_eax = td->td_retval[0];
 1111                 frame->tf_edx = td->td_retval[1];
 1112                 frame->tf_eflags &= ~PSL_C;
 1113                 break;
 1114 
 1115         case ERESTART:
 1116                 /*
 1117                  * Reconstruct pc, assuming lcall $X,y is 7 bytes,
 1118                  * int 0x80 is 2 bytes. We saved this in tf_err.
 1119                  */
 1120                 frame->tf_eip -= frame->tf_err;
 1121                 break;
 1122 
 1123         case EJUSTRETURN:
 1124                 break;
 1125 
 1126         default:
 1127                 if (p->p_sysent->sv_errsize) {
 1128                         if (error >= p->p_sysent->sv_errsize)
 1129                                 error = -1;     /* XXX */
 1130                         else
 1131                                 error = p->p_sysent->sv_errtbl[error];
 1132                 }
 1133                 frame->tf_eax = error;
 1134                 frame->tf_eflags |= PSL_C;
 1135                 break;
 1136         }
 1137 
 1138         /*
 1139          * Traced syscall.
 1140          */
 1141         if ((orig_tf_eflags & PSL_T) && !(orig_tf_eflags & PSL_VM)) {
 1142                 frame->tf_eflags &= ~PSL_T;
 1143                 ksiginfo_init_trap(&ksi);
 1144                 ksi.ksi_signo = SIGTRAP;
 1145                 ksi.ksi_code = TRAP_TRACE;
 1146                 ksi.ksi_addr = (void *)frame->tf_eip;
 1147                 trapsignal(td, &ksi);
 1148         }
 1149 
 1150         /*
 1151          * Check for misbehavior.
 1152          */
 1153         WITNESS_WARN(WARN_PANIC, NULL, "System call %s returning",
 1154             (code >= 0 && code < SYS_MAXSYSCALL) ? syscallnames[code] : "???");
 1155         KASSERT(td->td_critnest == 0,
 1156             ("System call %s returning in a critical section",
 1157             (code >= 0 && code < SYS_MAXSYSCALL) ? syscallnames[code] : "???"));
 1158         KASSERT(td->td_locks == 0,
 1159             ("System call %s returning with %d locks held",
 1160             (code >= 0 && code < SYS_MAXSYSCALL) ? syscallnames[code] : "???",
 1161             td->td_locks));
 1162 
 1163         /*
 1164          * Handle reschedule and other end-of-syscall issues
 1165          */
 1166         userret(td, frame);
 1167 
 1168         CTR4(KTR_SYSC, "syscall exit thread %p pid %d proc %s code %d", td,
 1169             td->td_proc->p_pid, td->td_proc->p_comm, code);
 1170 
 1171 #ifdef KTRACE
 1172         if (KTRPOINT(td, KTR_SYSRET))
 1173                 ktrsysret(code, error, td->td_retval[0]);
 1174 #endif
 1175 
 1176         /*
 1177          * This works because errno is findable through the
 1178          * register set.  If we ever support an emulation where this
 1179          * is not the case, this code will need to be revisited.
 1180          */
 1181         STOPEVENT(p, S_SCX, code);
 1182 
 1183         PTRACESTOP_SC(p, td, S_PT_SCX);
 1184 }
 1185 

Cache object: 5a70adf03ed58b7af36f5e5338de2112


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]


This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.