[BACK]Return to trap.c CVS log [TXT][DIR] Up to [local] / sys / arch / sparc64 / sparc64

Annotation of sys/arch/sparc64/sparc64/trap.c, Revision 1.1.1.1

1.1       nbrk        1: /*     $OpenBSD: trap.c,v 1.51 2007/07/05 22:16:30 kettenis Exp $      */
                      2: /*     $NetBSD: trap.c,v 1.73 2001/08/09 01:03:01 eeh Exp $ */
                      3:
                      4: /*
                      5:  * Copyright (c) 1996
                      6:  *     The President and Fellows of Harvard College. All rights reserved.
                      7:  * Copyright (c) 1992, 1993
                      8:  *     The Regents of the University of California.  All rights reserved.
                      9:  *
                     10:  * This software was developed by the Computer Systems Engineering group
                     11:  * at Lawrence Berkeley Laboratory under DARPA contract BG 91-66 and
                     12:  * contributed to Berkeley.
                     13:  *
                     14:  * All advertising materials mentioning features or use of this software
                     15:  * must display the following acknowledgement:
                     16:  *     This product includes software developed by the University of
                     17:  *     California, Lawrence Berkeley Laboratory.
                     18:  *     This product includes software developed by Harvard University.
                     19:  *
                     20:  * Redistribution and use in source and binary forms, with or without
                     21:  * modification, are permitted provided that the following conditions
                     22:  * are met:
                     23:  * 1. Redistributions of source code must retain the above copyright
                     24:  *    notice, this list of conditions and the following disclaimer.
                     25:  * 2. Redistributions in binary form must reproduce the above copyright
                     26:  *    notice, this list of conditions and the following disclaimer in the
                     27:  *    documentation and/or other materials provided with the distribution.
                     28:  * 3. All advertising materials mentioning features or use of this software
                     29:  *    must display the following acknowledgement:
                     30:  *     This product includes software developed by the University of
                     31:  *     California, Berkeley and its contributors.
                     32:  *     This product includes software developed by Harvard University.
                     33:  * 4. Neither the name of the University nor the names of its contributors
                     34:  *    may be used to endorse or promote products derived from this software
                     35:  *    without specific prior written permission.
                     36:  *
                     37:  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
                     38:  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
                     39:  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
                     40:  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
                     41:  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
                     42:  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
                     43:  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
                     44:  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
                     45:  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
                     46:  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
                     47:  * SUCH DAMAGE.
                     48:  *
                     49:  *     @(#)trap.c      8.4 (Berkeley) 9/23/93
                     50:  */
                     51:
                     52: #include <sys/param.h>
                     53: #include <sys/systm.h>
                     54: #include <sys/proc.h>
                     55: #include <sys/signalvar.h>
                     56: #include <sys/user.h>
                     57: #include <sys/kernel.h>
                     58: #include <sys/malloc.h>
                     59: #include <sys/resource.h>
                     60: #include <sys/signal.h>
                     61: #include <sys/wait.h>
                     62: #include <sys/syscall.h>
                     63: #include <sys/syslog.h>
                     64: #ifdef KTRACE
                     65: #include <sys/ktrace.h>
                     66: #endif
                     67:
                     68: #include "systrace.h"
                     69: #include <dev/systrace.h>
                     70:
                     71: #include <uvm/uvm_extern.h>
                     72:
                     73: #include <machine/cpu.h>
                     74: #include <machine/ctlreg.h>
                     75: #include <machine/trap.h>
                     76: #include <machine/instr.h>
                     77: #include <machine/pmap.h>
                     78:
                     79: #ifdef DDB
                     80: #include <machine/db_machdep.h>
                     81: #else
                     82: #include <machine/frame.h>
                     83: #endif
                     84: #ifdef COMPAT_SVR4
                     85: #include <machine/svr4_machdep.h>
                     86: #endif
                     87: #ifdef COMPAT_SVR4_32
                     88: #include <machine/svr4_32_machdep.h>
                     89: #endif
                     90:
                     91: #include <sparc64/fpu/fpu_extern.h>
                     92: #include <sparc64/sparc64/cache.h>
                     93:
                     94: #ifndef offsetof
                     95: #define        offsetof(s, f) ((int)&((s *)0)->f)
                     96: #endif
                     97:
                     98: /* trapstats */
                     99: int trapstats = 0;
                    100: int protfix = 0;
                    101: int udmiss = 0;        /* Number of normal/nucleus data/text miss/protection faults */
                    102: int udhit = 0;
                    103: int udprot = 0;
                    104: int utmiss = 0;
                    105: int kdmiss = 0;
                    106: int kdhit = 0;
                    107: int kdprot = 0;
                    108: int ktmiss = 0;
                    109: int iveccnt = 0; /* number if normal/nucleus interrupt/interrupt vector faults */
                    110: int uintrcnt = 0;
                    111: int kiveccnt = 0;
                    112: int kintrcnt = 0;
                    113: int intristk = 0; /* interrupts when already on intrstack */
                    114: int intrpoll = 0; /* interrupts not using vector lists */
                    115: int wfill = 0;
                    116: int kwfill = 0;
                    117: int wspill = 0;
                    118: int wspillskip = 0;
                    119: int rftucnt = 0;
                    120: int rftuld = 0;
                    121: int rftudone = 0;
                    122: int rftkcnt[5] = { 0, 0, 0, 0, 0 };
                    123:
                    124: /*
                    125:  * Initial FPU state is all registers == all 1s, everything else == all 0s.
                    126:  * This makes every floating point register a signalling NaN, with sign bit
                    127:  * set, no matter how it is interpreted.  Appendix N of the Sparc V8 document
                    128:  * seems to imply that we should do this, and it does make sense.
                    129:  */
                    130: __asm(".align 64");
                    131: struct fpstate64 initfpstate = {
                    132:        { ~0, ~0, ~0, ~0, ~0, ~0, ~0, ~0, ~0, ~0, ~0, ~0, ~0, ~0, ~0, ~0,
                    133:          ~0, ~0, ~0, ~0, ~0, ~0, ~0, ~0, ~0, ~0, ~0, ~0, ~0, ~0, ~0, ~0 }
                    134: };
                    135:
                    136: /*
                    137:  * There are more than 100 trap types, but most are unused.
                    138:  *
                    139:  * Trap type 0 is taken over as an `Asynchronous System Trap'.
                    140:  * This is left-over Vax emulation crap that should be fixed.
                    141:  *
                    142:  * Traps not supported on the spitfire are marked with `*',
                    143:  * and additions are marked with `+'
                    144:  */
                    145: static const char T[] = "*trap";
                    146: const char *trap_type[] = {
                    147:        /* non-user vectors */
                    148:        "ast",                  /* 0 */
                    149:        "power on reset",       /* 1 */
                    150:        "watchdog reset",       /* 2 */
                    151:        "externally initiated reset",/*3 */
                    152:        "software initiated reset",/* 4 */
                    153:        "RED state exception",  /* 5 */
                    154:        T, T,                   /* 6..7 */
                    155:        "instruction access exception", /* 8 */
                    156:        "*instruction MMU miss",/* 9 */
                    157:        "instruction access error",/* 0a */
                    158:        T, T, T, T, T,          /* 0b..0f */
                    159:        "illegal instruction",  /* 10 */
                    160:        "privileged opcode",    /* 11 */
                    161:        "*unimplemented LDD",   /* 12 */
                    162:        "*unimplemented STD",   /* 13 */
                    163:        T, T, T, T,             /* 14..17 */
                    164:        T, T, T, T, T, T, T, T, /* 18..1f */
                    165:        "fp disabled",          /* 20 */
                    166:        "fp exception ieee 754",/* 21 */
                    167:        "fp exception other",   /* 22 */
                    168:        "tag overflow",         /* 23 */
                    169:        "clean window",         /* 24 */
                    170:        T, T, T,                /* 25..27 -- trap continues */
                    171:        "division by zero",     /* 28 */
                    172:        "*internal processor error",/* 29 */
                    173:        T, T, T, T, T, T,       /* 2a..2f */
                    174:        "data access exception",/* 30 */
                    175:        "*data access MMU miss",/* 31 */
                    176:        "data access error",    /* 32 */
                    177:        "*data access protection",/* 33 */
                    178:        "mem address not aligned",      /* 34 */
                    179:        "LDDF mem address not aligned",/* 35 */
                    180:        "STDF mem address not aligned",/* 36 */
                    181:        "privileged action",    /* 37 */
                    182:        "LDQF mem address not aligned",/* 38 */
                    183:        "STQF mem address not aligned",/* 39 */
                    184:        T, T, T, T, T, T,       /* 3a..3f */
                    185:        "*async data error",    /* 40 */
                    186:        "level 1 int",          /* 41 */
                    187:        "level 2 int",          /* 42 */
                    188:        "level 3 int",          /* 43 */
                    189:        "level 4 int",          /* 44 */
                    190:        "level 5 int",          /* 45 */
                    191:        "level 6 int",          /* 46 */
                    192:        "level 7 int",          /* 47 */
                    193:        "level 8 int",          /* 48 */
                    194:        "level 9 int",          /* 49 */
                    195:        "level 10 int",         /* 4a */
                    196:        "level 11 int",         /* 4b */
                    197:        "level 12 int",         /* 4c */
                    198:        "level 13 int",         /* 4d */
                    199:        "level 14 int",         /* 4e */
                    200:        "level 15 int",         /* 4f */
                    201:        T, T, T, T, T, T, T, T, /* 50..57 */
                    202:        T, T, T, T, T, T, T, T, /* 58..5f */
                    203:        "+interrupt vector",    /* 60 */
                    204:        "+PA_watchpoint",       /* 61 */
                    205:        "+VA_watchpoint",       /* 62 */
                    206:        "+corrected ECC error", /* 63 */
                    207:        "+fast instruction access MMU miss",/* 64 */
                    208:        T, T, T,                /* 65..67 -- trap continues */
                    209:        "+fast data access MMU miss",/* 68 */
                    210:        T, T, T,                /* 69..6b -- trap continues */
                    211:        "+fast data access protection",/* 6c */
                    212:        T, T, T,                /* 6d..6f -- trap continues */
                    213:        T, T, T, T, T, T, T, T, /* 70..77 */
                    214:        T, T, T, T, T, T, T, T, /* 78..7f */
                    215:        "spill 0 normal",       /* 80 */
                    216:        T, T, T,                /* 81..83 -- trap continues */
                    217:        "spill 1 normal",       /* 84 */
                    218:        T, T, T,                /* 85..87 -- trap continues */
                    219:        "spill 2 normal",       /* 88 */
                    220:        T, T, T,                /* 89..8b -- trap continues */
                    221:        "spill 3 normal",       /* 8c */
                    222:        T, T, T,                /* 8d..8f -- trap continues */
                    223:        "spill 4 normal",       /* 90 */
                    224:        T, T, T,                /* 91..93 -- trap continues */
                    225:        "spill 5 normal",       /* 94 */
                    226:        T, T, T,                /* 95..97 -- trap continues */
                    227:        "spill 6 normal",       /* 98 */
                    228:        T, T, T,                /* 99..9b -- trap continues */
                    229:        "spill 7 normal",       /* 9c */
                    230:        T, T, T,                /* 9c..9f -- trap continues */
                    231:        "spill 0 other",        /* a0 */
                    232:        T, T, T,                /* a1..a3 -- trap continues */
                    233:        "spill 1 other",        /* a4 */
                    234:        T, T, T,                /* a5..a7 -- trap continues */
                    235:        "spill 2 other",        /* a8 */
                    236:        T, T, T,                /* a9..ab -- trap continues */
                    237:        "spill 3 other",        /* ac */
                    238:        T, T, T,                /* ad..af -- trap continues */
                    239:        "spill 4 other",        /* b0 */
                    240:        T, T, T,                /* b1..b3 -- trap continues */
                    241:        "spill 5 other",        /* b4 */
                    242:        T, T, T,                /* b5..b7 -- trap continues */
                    243:        "spill 6 other",        /* b8 */
                    244:        T, T, T,                /* b9..bb -- trap continues */
                    245:        "spill 7 other",        /* bc */
                    246:        T, T, T,                /* bc..bf -- trap continues */
                    247:        "fill 0 normal",        /* c0 */
                    248:        T, T, T,                /* c1..c3 -- trap continues */
                    249:        "fill 1 normal",        /* c4 */
                    250:        T, T, T,                /* c5..c7 -- trap continues */
                    251:        "fill 2 normal",        /* c8 */
                    252:        T, T, T,                /* c9..cb -- trap continues */
                    253:        "fill 3 normal",        /* cc */
                    254:        T, T, T,                /* cd..cf -- trap continues */
                    255:        "fill 4 normal",        /* d0 */
                    256:        T, T, T,                /* d1..d3 -- trap continues */
                    257:        "fill 5 normal",        /* d4 */
                    258:        T, T, T,                /* d5..d7 -- trap continues */
                    259:        "fill 6 normal",        /* d8 */
                    260:        T, T, T,                /* d9..db -- trap continues */
                    261:        "fill 7 normal",        /* dc */
                    262:        T, T, T,                /* dc..df -- trap continues */
                    263:        "fill 0 other",         /* e0 */
                    264:        T, T, T,                /* e1..e3 -- trap continues */
                    265:        "fill 1 other",         /* e4 */
                    266:        T, T, T,                /* e5..e7 -- trap continues */
                    267:        "fill 2 other",         /* e8 */
                    268:        T, T, T,                /* e9..eb -- trap continues */
                    269:        "fill 3 other",         /* ec */
                    270:        T, T, T,                /* ed..ef -- trap continues */
                    271:        "fill 4 other",         /* f0 */
                    272:        T, T, T,                /* f1..f3 -- trap continues */
                    273:        "fill 5 other",         /* f4 */
                    274:        T, T, T,                /* f5..f7 -- trap continues */
                    275:        "fill 6 other",         /* f8 */
                    276:        T, T, T,                /* f9..fb -- trap continues */
                    277:        "fill 7 other",         /* fc */
                    278:        T, T, T,                /* fc..ff -- trap continues */
                    279:
                    280:        /* user (software trap) vectors */
                    281:        "syscall",              /* 100 */
                    282:        "breakpoint",           /* 101 */
                    283:        "zero divide",          /* 102 */
                    284:        "flush windows",        /* 103 */
                    285:        "clean windows",        /* 104 */
                    286:        "range check",          /* 105 */
                    287:        "fix align",            /* 106 */
                    288:        "integer overflow",     /* 107 */
                    289:        "svr4 syscall",         /* 108 */
                    290:        "4.4 syscall",          /* 109 */
                    291:        "kgdb exec",            /* 10a */
                    292:        T, T, T, T, T,          /* 10b..10f */
                    293:        T, T, T, T, T, T, T, T, /* 11a..117 */
                    294:        T, T, T, T, T, T, T, T, /* 118..11f */
                    295:        "svr4 getcc",           /* 120 */
                    296:        "svr4 setcc",           /* 121 */
                    297:        "svr4 getpsr",          /* 122 */
                    298:        "svr4 setpsr",          /* 123 */
                    299:        "svr4 gethrtime",       /* 124 */
                    300:        "svr4 gethrvtime",      /* 125 */
                    301:        T,                      /* 126 */
                    302:        "svr4 gethrestime",     /* 127 */
                    303:        T, T, T, T, T, T, T, T, /* 128..12f */
                    304:        T, T,                   /* 130..131 */
                    305:        "get condition codes",  /* 132 */
                    306:        "set condition codes",  /* 133 */
                    307:        T, T, T, T,             /* 134..137 */
                    308:        T, T, T, T, T, T, T, T, /* 138..13f */
                    309:        T, T, T, T, T, T, T, T, /* 140..147 */
                    310:        T, T, T, T, T, T, T, T, /* 148..14f */
                    311:        T, T, T, T, T, T, T, T, /* 150..157 */
                    312:        T, T, T, T, T, T, T, T, /* 158..15f */
                    313:        T, T, T, T,             /* 160..163 */
                    314:        "SVID syscall64",       /* 164 */
                    315:        "SPARC Intl syscall64", /* 165 */
                    316:        "OS vendor spec syscall",       /* 166 */
                    317:        "HW OEM syscall",       /* 167 */
                    318:        "ret from deferred trap",       /* 168 */
                    319: };
                    320:
                    321: #define        N_TRAP_TYPES    (sizeof trap_type / sizeof *trap_type)
                    322:
                    323: static __inline void userret(struct proc *);
                    324: static __inline void share_fpu(struct proc *, struct trapframe64 *);
                    325:
                    326: void trap(struct trapframe64 *tf, unsigned type, vaddr_t pc, long tstate);
                    327: void data_access_fault(struct trapframe64 *tf, unsigned type, vaddr_t pc,
                    328:        vaddr_t va, vaddr_t sfva, u_long sfsr);
                    329: void data_access_error(struct trapframe64 *tf, unsigned type,
                    330:        vaddr_t afva, u_long afsr, vaddr_t sfva, u_long sfsr);
                    331: void text_access_fault(struct trapframe64 *tf, unsigned type,
                    332:        vaddr_t pc, u_long sfsr);
                    333: void text_access_error(struct trapframe64 *tf, unsigned type,
                    334:        vaddr_t pc, u_long sfsr, vaddr_t afva, u_long afsr);
                    335: void syscall(struct trapframe64 *, register_t code, register_t pc);
                    336:
                    337: /*
                    338:  * Define the code needed before returning to user mode, for
                    339:  * trap, mem_access_fault, and syscall.
                    340:  */
                    341: static __inline void
                    342: userret(struct proc *p)
                    343: {
                    344:        int sig;
                    345:
                    346:        /* take pending signals */
                    347:        while ((sig = CURSIG(p)) != 0)
                    348:                postsig(sig);
                    349:
                    350:        curcpu()->ci_schedstate.spc_curpriority = p->p_priority = p->p_usrpri;
                    351: }
                    352:
                    353: /*
                    354:  * If someone stole the FPU while we were away, do not enable it
                    355:  * on return.  This is not done in userret() above as it must follow
                    356:  * the ktrsysret() in syscall().  Actually, it is likely that the
                    357:  * ktrsysret should occur before the call to userret.
                    358:  *
                    359:  * Oh, and don't touch the FPU bit if we're returning to the kernel.
                    360:  */
                    361: static __inline void share_fpu(p, tf)
                    362:        struct proc *p;
                    363:        struct trapframe64 *tf;
                    364: {
                    365:        if (!(tf->tf_tstate & (PSTATE_PRIV<<TSTATE_PSTATE_SHIFT)) &&
                    366:            (tf->tf_tstate & (PSTATE_PEF<<TSTATE_PSTATE_SHIFT)) && fpproc != p)
                    367:                tf->tf_tstate &= ~(PSTATE_PEF<<TSTATE_PSTATE_SHIFT);
                    368: }
                    369:
                    370: /*
                    371:  * Called from locore.s trap handling, for non-MMU-related traps.
                    372:  * (MMU-related traps go through mem_access_fault, below.)
                    373:  */
                    374: void
                    375: trap(tf, type, pc, tstate)
                    376:        struct trapframe64 *tf;
                    377:        unsigned type;
                    378:        vaddr_t pc;
                    379:        long tstate;
                    380: {
                    381:        struct proc *p;
                    382:        struct pcb *pcb;
                    383:        int pstate = (tstate>>TSTATE_PSTATE_SHIFT);
                    384:        int64_t n;
                    385:        union sigval sv;
                    386:
                    387:        sv.sival_ptr = (void *)pc;
                    388:
                    389:        /* This steps the PC over the trap. */
                    390: #define        ADVANCE (n = tf->tf_npc, tf->tf_pc = n, tf->tf_npc = n + 4)
                    391:
                    392:        uvmexp.traps++;
                    393:        /*
                    394:         * Generally, kernel traps cause a panic.  Any exceptions are
                    395:         * handled early here.
                    396:         */
                    397:        if (pstate & PSTATE_PRIV) {
                    398: #ifdef DDB
                    399:                if (type == T_BREAKPOINT) {
                    400:                        write_all_windows();
                    401:                        if (kdb_trap(type, tf)) {
                    402:                                /* ADVANCE; */
                    403:                                return;
                    404:                        }
                    405:                }
                    406:                if (type == T_PA_WATCHPT || type == T_VA_WATCHPT) {
                    407:                        if (kdb_trap(type, tf)) {
                    408:                                /* DDB must turn off watchpoints or something */
                    409:                                return;
                    410:                        }
                    411:                }
                    412: #endif
                    413:                /*
                    414:                 * The kernel needs to use FPU registers for block
                    415:                 * load/store.  If we trap in priviliged code, save
                    416:                 * the FPU state if there is any and enable the FPU.
                    417:                 *
                    418:                 * We rely on the kernel code properly enabling the FPU
                    419:                 * in %fprs, otherwise we'll hang here trying to enable
                    420:                 * the FPU.
                    421:                 */
                    422:                if (type == T_FPDISABLED) {
                    423:                        struct proc *newfpproc;
                    424:
                    425:                        if (CLKF_INTR((struct clockframe *)tf) || !curproc)
                    426:                                newfpproc = &proc0;
                    427:                        else
                    428:                                newfpproc = curproc;
                    429:
                    430:                        if (fpproc != newfpproc) {
                    431:                                if (fpproc != NULL) {
                    432:                                        /* someone else had it, maybe? */
                    433:                                        savefpstate(fpproc->p_md.md_fpstate);
                    434:                                        fpproc = NULL;
                    435:                                }
                    436:                                /* If we have an allocated fpstate, load it */
                    437:                                if (newfpproc->p_md.md_fpstate != 0) {
                    438:                                        fpproc = newfpproc;
                    439:                                        loadfpstate(fpproc->p_md.md_fpstate);
                    440:                                } else
                    441:                                        fpproc = NULL;
                    442:                        }
                    443:                        /* Enable the FPU */
                    444:                        tf->tf_tstate |= (PSTATE_PEF<<TSTATE_PSTATE_SHIFT);
                    445:                        return;
                    446:                }
                    447:                if (type != T_SPILL_N_NORM && type != T_FILL_N_NORM)
                    448:                        goto dopanic;
                    449:        }
                    450:        if ((p = curproc) == NULL)
                    451:                p = &proc0;
                    452:        pcb = &p->p_addr->u_pcb;
                    453:        p->p_md.md_tf = tf;     /* for ptrace/signals */
                    454:
                    455:        switch (type) {
                    456:
                    457:        default:
                    458:                if (type < 0x100) {
                    459:                        extern int trap_trace_dis;
                    460: dopanic:
                    461:                        trap_trace_dis = 1;
                    462:
                    463:                        panic("trap type 0x%x (%s): pc=%lx npc=%lx pstate=%b\n",
                    464:                            type, type < N_TRAP_TYPES ? trap_type[type] : T,
                    465:                            pc, (long)tf->tf_npc, pstate, PSTATE_BITS);
                    466:                        /* NOTREACHED */
                    467:                }
                    468: #if defined(COMPAT_SVR4) || defined(COMPAT_SVR4_32)
                    469: badtrap:
                    470: #endif
                    471:                trapsignal(p, SIGILL, type, ILL_ILLOPC, sv);
                    472:                break;
                    473:
                    474: #if defined(COMPAT_SVR4) || defined(COMPAT_SVR4_32)
                    475:        case T_SVR4_GETCC:
                    476:        case T_SVR4_SETCC:
                    477:        case T_SVR4_GETPSR:
                    478:        case T_SVR4_SETPSR:
                    479:        case T_SVR4_GETHRTIME:
                    480:        case T_SVR4_GETHRVTIME:
                    481:        case T_SVR4_GETHRESTIME:
                    482: #if defined(COMPAT_SVR4_32)
                    483:                if (svr4_32_trap(type, p))
                    484:                        break;
                    485: #endif
                    486: #if defined(COMPAT_SVR4)
                    487:                if (svr4_trap(type, p))
                    488:                        break;
                    489: #endif
                    490:                goto badtrap;
                    491: #endif
                    492:
                    493:        case T_AST:
                    494:                want_ast = 0;
                    495:                if (p->p_flag & P_OWEUPC) {
                    496:                        ADDUPROF(p);
                    497:                }
                    498:                if (curcpu()->ci_want_resched)
                    499:                        preempt(NULL);
                    500:                break;
                    501:
                    502:        case T_ILLINST:
                    503:        {
                    504:                union instr ins;
                    505:
                    506:                if (copyin((caddr_t)pc, &ins, sizeof(ins)) != 0) {
                    507:                        /* XXX Can this happen? */
                    508:                        trapsignal(p, SIGILL, 0, ILL_ILLOPC, sv);
                    509:                        break;
                    510:                }
                    511:                if (ins.i_any.i_op == IOP_mem &&
                    512:                    (ins.i_op3.i_op3 == IOP3_LDQF ||
                    513:                     ins.i_op3.i_op3 == IOP3_STQF ||
                    514:                     ins.i_op3.i_op3 == IOP3_LDQFA ||
                    515:                     ins.i_op3.i_op3 == IOP3_STQFA)) {
                    516:                        if (emul_qf(ins.i_int, p, sv, tf))
                    517:                                ADVANCE;
                    518:                        break;
                    519:                }
                    520:                if (ins.i_any.i_op == IOP_reg &&
                    521:                    ins.i_op3.i_op3 == IOP3_POPC &&
                    522:                    ins.i_op3.i_rs1 == 0) {
                    523:                        if (emul_popc(ins.i_int, p, sv, tf))
                    524:                                ADVANCE;
                    525:                        break;
                    526:                }
                    527:                trapsignal(p, SIGILL, 0, ILL_ILLOPC, sv);       /* XXX code?? */
                    528:                break;
                    529:        }
                    530:
                    531:        case T_INST_EXCEPT:
                    532:        case T_TEXTFAULT:
                    533:        case T_PRIVINST:
                    534:                trapsignal(p, SIGILL, 0, ILL_ILLOPC, sv);       /* XXX code?? */
                    535:                break;
                    536:
                    537:        case T_FPDISABLED: {
                    538:                struct fpstate64 *fs = p->p_md.md_fpstate;
                    539:
                    540:                if (fs == NULL) {
                    541:                        /* NOTE: fpstate must be 64-bit aligned */
                    542:                        fs = malloc((sizeof *fs), M_SUBPROC, M_WAITOK);
                    543:                        *fs = initfpstate;
                    544:                        fs->fs_qsize = 0;
                    545:                        p->p_md.md_fpstate = fs;
                    546:                }
                    547:                /*
                    548:                 * If we have not found an FPU, we have to emulate it.
                    549:                 *
                    550:                 * Since All UltraSPARC CPUs have an FPU how can this happen?
                    551:                 */
                    552:                if (!foundfpu) {
                    553:                        trapsignal(p, SIGILL, 0, ILL_COPROC, sv);
                    554:                        break;
                    555:                }
                    556:
                    557:                /*
                    558:                 * We may have more FPEs stored up and/or ops queued.
                    559:                 * If they exist, handle them and get out.  Otherwise,
                    560:                 * resolve the FPU state, turn it on, and try again.
                    561:                 *
                    562:                 * Ultras should never have a FPU queue.
                    563:                 */
                    564:                if (fs->fs_qsize) {
                    565:                        printf("trap: Warning fs_qsize is %d\n",fs->fs_qsize);
                    566:                        fpu_cleanup(p, fs);
                    567:                        break;
                    568:                }
                    569:                if (fpproc != p) {              /* we do not have it */
                    570:                        if (fpproc != NULL)     /* someone else had it */
                    571:                                savefpstate(fpproc->p_md.md_fpstate);
                    572:                        loadfpstate(fs);
                    573:                        fpproc = p;             /* now we do have it */
                    574:                        uvmexp.fpswtch++;
                    575:                }
                    576:                tf->tf_tstate |= (PSTATE_PEF<<TSTATE_PSTATE_SHIFT);
                    577:                break;
                    578:        }
                    579:
                    580:        case T_LDQF_ALIGN:
                    581:        case T_STQF_ALIGN:
                    582:        {
                    583:                union instr ins;
                    584:
                    585:                if (copyin((caddr_t)pc, &ins, sizeof(ins)) != 0) {
                    586:                        /* XXX Can this happen? */
                    587:                        trapsignal(p, SIGILL, 0, ILL_ILLOPC, sv);
                    588:                        break;
                    589:                }
                    590:                if (ins.i_any.i_op == IOP_mem &&
                    591:                    (ins.i_op3.i_op3 == IOP3_LDQF ||
                    592:                     ins.i_op3.i_op3 == IOP3_STQF ||
                    593:                     ins.i_op3.i_op3 == IOP3_LDQFA ||
                    594:                     ins.i_op3.i_op3 == IOP3_STQFA)) {
                    595:                        if (emul_qf(ins.i_int, p, sv, tf))
                    596:                                ADVANCE;
                    597:                } else
                    598:                        trapsignal(p, SIGILL, 0, ILL_ILLOPC, sv);
                    599:                break;
                    600:        }
                    601:
                    602:        case T_SPILL_N_NORM:
                    603:        case T_FILL_N_NORM:
                    604:                /*
                    605:                 * We got an alignment trap in the spill/fill handler.
                    606:                 *
                    607:                 * XXX We really should generate a bus error here, but
                    608:                 * we could be on the interrupt stack, and dumping
                    609:                 * core from the interrupt stack is not a good idea.
                    610:                 * It causes random crashes.
                    611:                 */
                    612:                sigexit(p, SIGKILL);
                    613:                break;
                    614:
                    615:        case T_ALIGN:
                    616:        case T_LDDF_ALIGN:
                    617:        case T_STDF_ALIGN:
                    618:        {
                    619:                int64_t dsfsr, dsfar=0, isfsr;
                    620:
                    621:                dsfsr = ldxa(SFSR, ASI_DMMU);
                    622:                if (dsfsr & SFSR_FV)
                    623:                        dsfar = ldxa(SFAR, ASI_DMMU);
                    624:                isfsr = ldxa(SFSR, ASI_IMMU);
                    625:                /*
                    626:                 * If we're busy doing copyin/copyout continue
                    627:                 */
                    628:                if (p->p_addr && p->p_addr->u_pcb.pcb_onfault) {
                    629:                        tf->tf_pc = (vaddr_t)p->p_addr->u_pcb.pcb_onfault;
                    630:                        tf->tf_npc = tf->tf_pc + 4;
                    631:                        break;
                    632:                }
                    633:        }
                    634:
                    635:                if ((p->p_md.md_flags & MDP_FIXALIGN) != 0 &&
                    636:                    fixalign(p, tf) == 0) {
                    637:                        ADVANCE;
                    638:                        break;
                    639:                }
                    640:                /* XXX sv.sival_ptr should be the fault address! */
                    641:                trapsignal(p, SIGBUS, 0, BUS_ADRALN, sv);       /* XXX code?? */
                    642:                break;
                    643:
                    644:        case T_FP_IEEE_754:
                    645:        case T_FP_OTHER:
                    646:                /*
                    647:                 * Clean up after a floating point exception.
                    648:                 * fpu_cleanup can (and usually does) modify the
                    649:                 * state we save here, so we must `give up' the FPU
                    650:                 * chip context.  (The software and hardware states
                    651:                 * will not match once fpu_cleanup does its job, so
                    652:                 * we must not save again later.)
                    653:                 */
                    654:                if (p != fpproc)
                    655:                        panic("fpe without being the FP user");
                    656:                savefpstate(p->p_md.md_fpstate);
                    657:                fpproc = NULL;
                    658:                /* tf->tf_psr &= ~PSR_EF; */    /* share_fpu will do this */
                    659:                if (type == T_FP_OTHER && p->p_md.md_fpstate->fs_qsize == 0) {
                    660:                        /*
                    661:                         * Push the faulting instruction on the queue;
                    662:                         * we might need to emulate it.
                    663:                         */
                    664:                        copyin((caddr_t)pc, &p->p_md.md_fpstate->fs_queue[0].fq_instr, sizeof(int));
                    665:                        p->p_md.md_fpstate->fs_queue[0].fq_addr = (int *)pc;
                    666:                        p->p_md.md_fpstate->fs_qsize = 1;
                    667:                }
                    668:                ADVANCE;
                    669:                fpu_cleanup(p, p->p_md.md_fpstate);
                    670:                /* fpu_cleanup posts signals if needed */
                    671:                break;
                    672:
                    673:        case T_TAGOF:
                    674:                trapsignal(p, SIGEMT, 0, EMT_TAGOVF, sv);       /* XXX code?? */
                    675:                break;
                    676:
                    677:        case T_BREAKPOINT:
                    678:                trapsignal(p, SIGTRAP, 0, TRAP_BRKPT, sv);
                    679:                break;
                    680:
                    681:        case T_DIV0:
                    682:                ADVANCE;
                    683:                trapsignal(p, SIGFPE, 0, FPE_INTDIV, sv);
                    684:                break;
                    685:
                    686:        case T_CLEANWIN:
                    687:                uprintf("T_CLEANWIN\n");        /* XXX Should not get this */
                    688:                ADVANCE;
                    689:                break;
                    690:
                    691:        case T_FLUSHWIN:
                    692:                /* Software window flush for v8 software */
                    693:                write_all_windows();
                    694:                ADVANCE;
                    695:                break;
                    696:
                    697:        case T_RANGECHECK:
                    698:                ADVANCE;
                    699:                trapsignal(p, SIGILL, 0, ILL_ILLOPN, sv);       /* XXX code?? */
                    700:                break;
                    701:
                    702:        case T_FIXALIGN:
                    703: #ifdef DEBUG_ALIGN
                    704:                uprintf("T_FIXALIGN\n");
                    705: #endif
                    706:                /* User wants us to fix alignment faults */
                    707:                p->p_md.md_flags |= MDP_FIXALIGN;
                    708:                ADVANCE;
                    709:                break;
                    710:
                    711:        case T_INTOF:
                    712:                uprintf("T_INTOF\n");           /* XXX */
                    713:                ADVANCE;
                    714:                trapsignal(p, SIGFPE, FPE_INTOVF_TRAP, FPE_INTOVF, sv);
                    715:                break;
                    716:        }
                    717:        userret(p);
                    718:        share_fpu(p, tf);
                    719: #undef ADVANCE
                    720: }
                    721:
                    722: /*
                    723:  * Save windows from PCB into user stack, and return 0.  This is used on
                    724:  * window overflow pseudo-traps (from locore.s, just before returning to
                    725:  * user mode) and when ptrace or sendsig needs a consistent state.
                    726:  * As a side effect, rwindow_save() always sets pcb_nsaved to 0.
                    727:  *
                    728:  * If the windows cannot be saved, pcb_nsaved is restored and we return -1.
                    729:  *
                    730:  * XXXXXX This cannot work properly.  I need to re-examine this register
                    731:  * window thing entirely.
                    732:  */
                    733: int
                    734: rwindow_save(p)
                    735:        struct proc *p;
                    736: {
                    737:        struct pcb *pcb = &p->p_addr->u_pcb;
                    738:        struct rwindow64 *rw = &pcb->pcb_rw[0];
                    739:        u_int64_t rwdest;
                    740:        int i, j;
                    741:
                    742:        i = pcb->pcb_nsaved;
                    743:        if (i == 0)
                    744:                return (0);
                    745:         while (i > 0) {
                    746:                rwdest = rw[i--].rw_in[6];
                    747:                if (rwdest & 1) {
                    748:                        struct rwindow64 rwstack = rw[i];
                    749:
                    750:                        rwdest += BIAS;
                    751:                        rwstack.rw_in[7] ^= p->p_addr->u_pcb.pcb_wcookie;
                    752:                        if (copyout((caddr_t)&rwstack, (caddr_t)(u_long)rwdest,
                    753:                            sizeof(rwstack))) {
                    754:                                return (-1);
                    755:                        }
                    756:                } else {
                    757:                        struct rwindow32 rwstack;
                    758:
                    759:                        /* 32-bit window */
                    760:                        for (j = 0; j < 8; j++) {
                    761:                                rwstack.rw_local[j] = (int)rw[i].rw_local[j];
                    762:                                rwstack.rw_in[j] = (int)rw[i].rw_in[j];
                    763:                        }
                    764:                        /* Must truncate rwdest */
                    765:                        if (copyout(&rwstack, (caddr_t)(u_long)(u_int)rwdest,
                    766:                            sizeof(rwstack))) {
                    767:                                return (-1);
                    768:                        }
                    769:                }
                    770:        }
                    771:        pcb->pcb_nsaved = 0;
                    772:
                    773:        return (0);
                    774: }
                    775:
                    776: /*
                    777:  * Kill user windows (before exec) by writing back to stack or pcb
                    778:  * and then erasing any pcb tracks.  Otherwise we might try to write
                    779:  * the registers into the new process after the exec.
                    780:  */
                    781: void
                    782: pmap_unuse_final(p)
                    783:        struct proc *p;
                    784: {
                    785:
                    786:        write_user_windows();
                    787:        p->p_addr->u_pcb.pcb_nsaved = 0;
                    788: }
                    789:
                    790: /*
                    791:  * This routine handles MMU generated faults.  About half
                    792:  * of them could be recoverable through uvm_fault.
                    793:  */
                    794: void
                    795: data_access_fault(tf, type, pc, addr, sfva, sfsr)
                    796:        struct trapframe64 *tf;
                    797:        unsigned type;
                    798:        vaddr_t pc;
                    799:        vaddr_t addr;
                    800:        vaddr_t sfva;
                    801:        u_long sfsr;
                    802: {
                    803:        u_int64_t tstate;
                    804:        struct proc *p;
                    805:        struct vmspace *vm;
                    806:        vaddr_t va;
                    807:        int rv;
                    808:        vm_prot_t access_type;
                    809:        vaddr_t onfault;
                    810:        union sigval sv;
                    811:
                    812:        uvmexp.traps++;
                    813:        if ((p = curproc) == NULL)      /* safety check */
                    814:                p = &proc0;
                    815:
                    816:        tstate = tf->tf_tstate;
                    817:
                    818:        /* Find the faulting va to give to uvm_fault */
                    819:        va = trunc_page(addr);
                    820:
                    821:        /*
                    822:         * Now munch on protections.
                    823:         *
                    824:         * If it was a FAST_DATA_ACCESS_MMU_MISS we have no idea what the
                    825:         * access was since the SFSR is not set.  But we should never get
                    826:         * here from there.
                    827:         */
                    828:        if (type == T_FDMMU_MISS || (sfsr & SFSR_FV) == 0) {
                    829:                /* Punt */
                    830:                access_type = VM_PROT_READ;
                    831:        } else {
                    832:                access_type = (sfsr & SFSR_W) ? VM_PROT_READ|VM_PROT_WRITE
                    833:                        : VM_PROT_READ;
                    834:        }
                    835:        if (tstate & (PSTATE_PRIV<<TSTATE_PSTATE_SHIFT)) {
                    836: #ifdef DDB
                    837:                extern char Lfsprobe[];
                    838:                /*
                    839:                 * If this was an access that we shouldn't try to page in,
                    840:                 * resume at the fault handler without any action.
                    841:                 */
                    842:                if (p->p_addr && p->p_addr->u_pcb.pcb_onfault == Lfsprobe)
                    843:                        goto kfault;
                    844: #endif
                    845:
                    846:                /*
                    847:                 * During autoconfiguration, faults are never OK unless
                    848:                 * pcb_onfault is set.  Once running normally we must allow
                    849:                 * exec() to cause copy-on-write faults to kernel addresses.
                    850:                 */
                    851:                if (cold)
                    852:                        goto kfault;
                    853:                if (!(addr & TLB_TAG_ACCESS_CTX)) {
                    854:                        /* CTXT == NUCLEUS */
                    855:                        rv = uvm_fault(kernel_map, va, 0, access_type);
                    856:                        if (rv == 0)
                    857:                                return;
                    858:                        goto kfault;
                    859:                }
                    860:        } else
                    861:                p->p_md.md_tf = tf;
                    862:
                    863:        vm = p->p_vmspace;
                    864:        /* alas! must call the horrible vm code */
                    865:        onfault = (vaddr_t)p->p_addr->u_pcb.pcb_onfault;
                    866:        p->p_addr->u_pcb.pcb_onfault = NULL;
                    867:        rv = uvm_fault(&vm->vm_map, (vaddr_t)va, 0, access_type);
                    868:        p->p_addr->u_pcb.pcb_onfault = (void *)onfault;
                    869:
                    870:        /*
                    871:         * If this was a stack access we keep track of the maximum
                    872:         * accessed stack size.  Also, if uvm_fault gets a protection
                    873:         * failure it is due to accessing the stack region outside
                    874:         * the current limit and we need to reflect that as an access
                    875:         * error.
                    876:         */
                    877:        if ((caddr_t)va >= vm->vm_maxsaddr) {
                    878:                if (rv == 0)
                    879:                        uvm_grow(p, va);
                    880:                else if (rv == EACCES)
                    881:                        rv = EFAULT;
                    882:        }
                    883:        if (rv != 0) {
                    884:                /*
                    885:                 * Pagein failed.  If doing copyin/out, return to onfault
                    886:                 * address.  Any other page fault in kernel, die; if user
                    887:                 * fault, deliver SIGSEGV.
                    888:                 */
                    889:                if (tstate & (PSTATE_PRIV<<TSTATE_PSTATE_SHIFT)) {
                    890: kfault:
                    891:                        onfault = p->p_addr ?
                    892:                            (long)p->p_addr->u_pcb.pcb_onfault : 0;
                    893:                        if (!onfault) {
                    894:                                extern int trap_trace_dis;
                    895:                                trap_trace_dis = 1; /* Disable traptrace for printf */
                    896:                                (void) splhigh();
                    897:                                panic("kernel data fault: pc=%lx addr=%lx\n",
                    898:                                    pc, addr);
                    899:                                /* NOTREACHED */
                    900:                        }
                    901:                        tf->tf_pc = onfault;
                    902:                        tf->tf_npc = onfault + 4;
                    903:                        return;
                    904:                }
                    905:
                    906:                if (type == T_FDMMU_MISS || (sfsr & SFSR_FV) == 0)
                    907:                        sv.sival_ptr = (void *)va;
                    908:                else
                    909:                        sv.sival_ptr = (void *)sfva;
                    910:
                    911:                if (rv == ENOMEM) {
                    912:                        printf("UVM: pid %d (%s), uid %u killed: out of swap\n",
                    913:                               p->p_pid, p->p_comm,
                    914:                               p->p_cred && p->p_ucred ?
                    915:                               p->p_ucred->cr_uid : -1);
                    916:                        trapsignal(p, SIGKILL, access_type, SEGV_MAPERR, sv);
                    917:                } else {
                    918:                        trapsignal(p, SIGSEGV, access_type, SEGV_MAPERR, sv);
                    919:                }
                    920:        }
                    921:        if ((tstate & TSTATE_PRIV) == 0) {
                    922:                userret(p);
                    923:                share_fpu(p, tf);
                    924:        }
                    925: }
                    926:
                    927: /*
                    928:  * This routine handles deferred errors caused by the memory
                    929:  * or I/O bus subsystems.  Most of these are fatal, and even
                    930:  * if they are not, recovery is painful.  Also, the TPC and
                    931:  * TNPC values are probably not valid if we're not doing a
                    932:  * special PEEK/POKE code sequence.
                    933:  */
                    934: void
                    935: data_access_error(tf, type, afva, afsr, sfva, sfsr)
                    936:        struct trapframe64 *tf;
                    937:        unsigned type;
                    938:        vaddr_t sfva;
                    939:        u_long sfsr;
                    940:        vaddr_t afva;
                    941:        u_long afsr;
                    942: {
                    943:        u_long pc;
                    944:        u_int64_t tstate;
                    945:        struct proc *p;
                    946:        vaddr_t onfault;
                    947:        union sigval sv;
                    948:
                    949:        uvmexp.traps++;
                    950:        if ((p = curproc) == NULL)      /* safety check */
                    951:                p = &proc0;
                    952:
                    953:        pc = tf->tf_pc;
                    954:        tstate = tf->tf_tstate;
                    955:
                    956:        sv.sival_ptr = (void *)pc;
                    957:
                    958:        onfault = p->p_addr ? (long)p->p_addr->u_pcb.pcb_onfault : 0;
                    959:        printf("data error type %x sfsr=%lx sfva=%lx afsr=%lx afva=%lx tf=%p\n",
                    960:                type, sfsr, sfva, afsr, afva, tf);
                    961:
                    962:        if (afsr == 0) {
                    963:                printf("data_access_error: no fault\n");
                    964:                goto out;       /* No fault. Why were we called? */
                    965:        }
                    966:
                    967:        if (tstate & (PSTATE_PRIV<<TSTATE_PSTATE_SHIFT)) {
                    968:
                    969:                if (!onfault) {
                    970:                        extern int trap_trace_dis;
                    971:
                    972:                        trap_trace_dis = 1; /* Disable traptrace for printf */
                    973:                        (void) splhigh();
                    974:                        panic("data fault: pc=%lx addr=%lx sfsr=%b\n",
                    975:                                (u_long)pc, (long)sfva, sfsr, SFSR_BITS);
                    976:                        /* NOTREACHED */
                    977:                }
                    978:
                    979:                /*
                    980:                 * If this was a priviliged error but not a probe, we
                    981:                 * cannot recover, so panic.
                    982:                 */
                    983:                if (afsr & ASFR_PRIV) {
                    984:                        panic("Privileged Async Fault: AFAR %p AFSR %lx\n%b",
                    985:                                (void *)afva, afsr, afsr, AFSR_BITS);
                    986:                        /* NOTREACHED */
                    987:                }
                    988:                tf->tf_pc = onfault;
                    989:                tf->tf_npc = onfault + 4;
                    990:                return;
                    991:        }
                    992:
                    993:        trapsignal(p, SIGSEGV, VM_PROT_READ|VM_PROT_WRITE, SEGV_MAPERR, sv);
                    994: out:
                    995:        if ((tstate & TSTATE_PRIV) == 0) {
                    996:                userret(p);
                    997:                share_fpu(p, tf);
                    998:        }
                    999: }
                   1000:
                   1001: /*
                   1002:  * This routine handles MMU generated faults.  About half
                   1003:  * of them could be recoverable through uvm_fault.
                   1004:  */
                   1005: void
                   1006: text_access_fault(tf, type, pc, sfsr)
                   1007:        unsigned type;
                   1008:        vaddr_t pc;
                   1009:        struct trapframe64 *tf;
                   1010:        u_long sfsr;
                   1011: {
                   1012:        u_int64_t tstate;
                   1013:        struct proc *p;
                   1014:        struct vmspace *vm;
                   1015:        vaddr_t va;
                   1016:        int rv;
                   1017:        vm_prot_t access_type;
                   1018:        union sigval sv;
                   1019:
                   1020:        sv.sival_ptr = (void *)pc;
                   1021:
                   1022:        uvmexp.traps++;
                   1023:        if ((p = curproc) == NULL)      /* safety check */
                   1024:                panic("text_access_fault: no curproc");
                   1025:
                   1026:        tstate = tf->tf_tstate;
                   1027:
                   1028:        va = trunc_page(pc);
                   1029:
                   1030:        /* Now munch on protections... */
                   1031:
                   1032:        access_type = VM_PROT_EXECUTE;
                   1033:        if (tstate & (PSTATE_PRIV<<TSTATE_PSTATE_SHIFT)) {
                   1034:                extern int trap_trace_dis;
                   1035:                trap_trace_dis = 1; /* Disable traptrace for printf */
                   1036:                (void) splhigh();
                   1037:                panic("kernel text_access_fault: pc=%lx va=%lx\n", pc, va);
                   1038:                /* NOTREACHED */
                   1039:        } else
                   1040:                p->p_md.md_tf = tf;
                   1041:
                   1042:        vm = p->p_vmspace;
                   1043:        /* alas! must call the horrible vm code */
                   1044:        rv = uvm_fault(&vm->vm_map, va, 0, access_type);
                   1045:
                   1046:        /*
                   1047:         * If this was a stack access we keep track of the maximum
                   1048:         * accessed stack size.  Also, if uvm_fault gets a protection
                   1049:         * failure it is due to accessing the stack region outside
                   1050:         * the current limit and we need to reflect that as an access
                   1051:         * error.
                   1052:         */
                   1053:        if ((caddr_t)va >= vm->vm_maxsaddr) {
                   1054:                if (rv == 0)
                   1055:                        uvm_grow(p, va);
                   1056:                else if (rv == EACCES)
                   1057:                        rv = EFAULT;
                   1058:        }
                   1059:        if (rv != 0) {
                   1060:                /*
                   1061:                 * Pagein failed. Any other page fault in kernel, die; if user
                   1062:                 * fault, deliver SIGSEGV.
                   1063:                 */
                   1064:                if (tstate & TSTATE_PRIV) {
                   1065:                        extern int trap_trace_dis;
                   1066:                        trap_trace_dis = 1; /* Disable traptrace for printf */
                   1067:                        (void) splhigh();
                   1068:                        panic("kernel text fault: pc=%llx\n", (unsigned long long)pc);
                   1069:                        /* NOTREACHED */
                   1070:                }
                   1071:                trapsignal(p, SIGSEGV, access_type, SEGV_MAPERR, sv);
                   1072:        }
                   1073:        if ((tstate & TSTATE_PRIV) == 0) {
                   1074:                userret(p);
                   1075:                share_fpu(p, tf);
                   1076:        }
                   1077: }
                   1078:
                   1079:
                   1080: /*
                   1081:  * This routine handles deferred errors caused by the memory
                   1082:  * or I/O bus subsystems.  Most of these are fatal, and even
                   1083:  * if they are not, recovery is painful.  Also, the TPC and
                   1084:  * TNPC values are probably not valid if we're not doing a
                   1085:  * special PEEK/POKE code sequence.
                   1086:  */
                   1087: void
                   1088: text_access_error(tf, type, pc, sfsr, afva, afsr)
                   1089:        struct trapframe64 *tf;
                   1090:        unsigned type;
                   1091:        vaddr_t pc;
                   1092:        u_long sfsr;
                   1093:        vaddr_t afva;
                   1094:        u_long afsr;
                   1095: {
                   1096:        int64_t tstate;
                   1097:        struct proc *p;
                   1098:        struct vmspace *vm;
                   1099:        vaddr_t va;
                   1100:        int rv;
                   1101:        vm_prot_t access_type;
                   1102:        union sigval sv;
                   1103:
                   1104:        sv.sival_ptr = (void *)pc;
                   1105:        uvmexp.traps++;
                   1106:        if ((p = curproc) == NULL)      /* safety check */
                   1107:                p = &proc0;
                   1108:
                   1109:        tstate = tf->tf_tstate;
                   1110:
                   1111:        if ((afsr) != 0) {
                   1112:                extern int trap_trace_dis;
                   1113:
                   1114:                trap_trace_dis++; /* Disable traptrace for printf */
                   1115:                printf("text_access_error: memory error...\n");
                   1116:                printf("text memory error type %d sfsr=%lx sfva=%lx afsr=%lx afva=%lx tf=%p\n",
                   1117:                       type, sfsr, pc, afsr, afva, tf);
                   1118:                trap_trace_dis--; /* Reenable traptrace for printf */
                   1119:
                   1120:                if (tstate & (PSTATE_PRIV<<TSTATE_PSTATE_SHIFT))
                   1121:                        panic("text_access_error: kernel memory error");
                   1122:
                   1123:                /* User fault -- Berr */
                   1124:                trapsignal(p, SIGBUS, 0, BUS_ADRALN, sv);
                   1125:        }
                   1126:
                   1127:        if ((sfsr & SFSR_FV) == 0 || (sfsr & SFSR_FT) == 0)
                   1128:                goto out;       /* No fault. Why were we called? */
                   1129:
                   1130:        va = trunc_page(pc);
                   1131:
                   1132:        /* Now munch on protections... */
                   1133:        access_type = VM_PROT_EXECUTE;
                   1134:        if (tstate & (PSTATE_PRIV<<TSTATE_PSTATE_SHIFT)) {
                   1135:                extern int trap_trace_dis;
                   1136:                trap_trace_dis = 1; /* Disable traptrace for printf */
                   1137:                (void) splhigh();
                   1138:                panic("kernel text error: pc=%lx sfsr=%b\n", pc, sfsr, SFSR_BITS);
                   1139:                /* NOTREACHED */
                   1140:        } else
                   1141:                p->p_md.md_tf = tf;
                   1142:
                   1143:        vm = p->p_vmspace;
                   1144:        /* alas! must call the horrible vm code */
                   1145:        rv = uvm_fault(&vm->vm_map, va, 0, access_type);
                   1146:
                   1147:        /*
                   1148:         * If this was a stack access we keep track of the maximum
                   1149:         * accessed stack size.  Also, if uvm_fault gets a protection
                   1150:         * failure it is due to accessing the stack region outside
                   1151:         * the current limit and we need to reflect that as an access
                   1152:         * error.
                   1153:         */
                   1154:        if ((caddr_t)va >= vm->vm_maxsaddr) {
                   1155:                if (rv == 0)
                   1156:                        uvm_grow(p, va);
                   1157:                else if (rv == EACCES)
                   1158:                        rv = EFAULT;
                   1159:        }
                   1160:        if (rv != 0) {
                   1161:                /*
                   1162:                 * Pagein failed.  If doing copyin/out, return to onfault
                   1163:                 * address.  Any other page fault in kernel, die; if user
                   1164:                 * fault, deliver SIGSEGV.
                   1165:                 */
                   1166:                if (tstate & TSTATE_PRIV) {
                   1167:                        extern int trap_trace_dis;
                   1168:                        trap_trace_dis = 1; /* Disable traptrace for printf */
                   1169:                        (void) splhigh();
                   1170:                        panic("kernel text error: pc=%lx sfsr=%b\n", pc,
                   1171:                            sfsr, SFSR_BITS);
                   1172:                        /* NOTREACHED */
                   1173:                }
                   1174:                trapsignal(p, SIGSEGV, access_type, SEGV_MAPERR, sv);
                   1175:        }
                   1176: out:
                   1177:        if ((tstate & TSTATE_PRIV) == 0) {
                   1178:                userret(p);
                   1179:                share_fpu(p, tf);
                   1180:        }
                   1181: }
                   1182:
                   1183: /*
                   1184:  * System calls.  `pc' is just a copy of tf->tf_pc.
                   1185:  *
                   1186:  * Note that the things labelled `out' registers in the trapframe were the
                   1187:  * `in' registers within the syscall trap code (because of the automatic
                   1188:  * `save' effect of each trap).  They are, however, the %o registers of the
                   1189:  * thing that made the system call, and are named that way here.
                   1190:  *
                   1191:  * 32-bit system calls on a 64-bit system are a problem.  Each system call
                   1192:  * argument is stored in the smaller of the argument's true size or a
                   1193:  * `register_t'.  Now on a 64-bit machine all normal types can be stored in a
                   1194:  * `register_t'.  (The only exceptions would be 128-bit `quad's or 128-bit
                   1195:  * extended precision floating point values, which we don't support.)  For
                   1196:  * 32-bit syscalls, 64-bit integers like `off_t's, double precision floating
                   1197:  * point values, and several other types cannot fit in a 32-bit `register_t'.
                   1198:  * These will require reading in two `register_t' values for one argument.
                   1199:  *
                   1200:  * In order to calculate the true size of the arguments and therefore whether
                   1201:  * any argument needs to be split into two slots, the system call args
                   1202:  * structure needs to be built with the appropriately sized register_t.
                   1203:  * Otherwise the emul needs to do some magic to split oversized arguments.
                   1204:  *
                   1205:  * We can handle most this stuff for normal syscalls by using either a 32-bit
                   1206:  * or 64-bit array of `register_t' arguments.  Unfortunately ktrace always
                   1207:  * expects arguments to be `register_t's, so it loses badly.  What's worse,
                   1208:  * ktrace may need to do size translations to massage the argument array
                   1209:  * appropriately according to the emulation that is doing the ktrace.
                   1210:  *
                   1211:  */
                   1212: void
                   1213: syscall(tf, code, pc)
                   1214:        register_t code;
                   1215:        struct trapframe64 *tf;
                   1216:        register_t pc;
                   1217: {
                   1218:        int i, nsys, nap;
                   1219:        int64_t *ap;
                   1220:        const struct sysent *callp;
                   1221:        struct proc *p;
                   1222:        int error = 0, new;
                   1223:        register_t args[8];
                   1224:        register_t rval[2];
                   1225:
                   1226:        uvmexp.syscalls++;
                   1227:        p = curproc;
                   1228: #ifdef DIAGNOSTIC
                   1229:        if (tf->tf_tstate & TSTATE_PRIV)
                   1230:                panic("syscall from kernel");
                   1231:        if (curpcb != &p->p_addr->u_pcb)
                   1232:                panic("syscall: cpcb/ppcb mismatch");
                   1233:        if (tf != (struct trapframe64 *)((caddr_t)curpcb + USPACE) - 1)
                   1234:                panic("syscall: trapframe");
                   1235: #endif
                   1236:        p->p_md.md_tf = tf;
                   1237:        new = code & (SYSCALL_G7RFLAG | SYSCALL_G2RFLAG);
                   1238:        code &= ~(SYSCALL_G7RFLAG | SYSCALL_G2RFLAG);
                   1239:
                   1240:        callp = p->p_emul->e_sysent;
                   1241:        nsys = p->p_emul->e_nsysent;
                   1242:
                   1243:        /*
                   1244:         * The first six system call arguments are in the six %o registers.
                   1245:         * Any arguments beyond that are in the `argument extension' area
                   1246:         * of the user's stack frame (see <machine/frame.h>).
                   1247:         *
                   1248:         * Check for ``special'' codes that alter this, namely syscall and
                   1249:         * __syscall.  The latter takes a quad syscall number, so that other
                   1250:         * arguments are at their natural alignments.  Adjust the number
                   1251:         * of ``easy'' arguments as appropriate; we will copy the hard
                   1252:         * ones later as needed.
                   1253:         */
                   1254:        ap = &tf->tf_out[0];
                   1255:        nap = 6;
                   1256:
                   1257:        switch (code) {
                   1258:        case SYS_syscall:
                   1259:                code = *ap++;
                   1260:                nap--;
                   1261:                break;
                   1262:        case SYS___syscall:
                   1263:                if (code < nsys &&
                   1264:                    callp[code].sy_call != callp[p->p_emul->e_nosys].sy_call)
                   1265:                        break; /* valid system call */
                   1266:                if (tf->tf_out[6] & 1L) {
                   1267:                        /* longs *are* quadwords */
                   1268:                        code = ap[0];
                   1269:                        ap += 1;
                   1270:                        nap -= 1;
                   1271:                } else {
                   1272:                        code = ap[_QUAD_LOWWORD];
                   1273:                        ap += 2;
                   1274:                        nap -= 2;
                   1275:                }
                   1276:                break;
                   1277:        }
                   1278:
                   1279:        if (code < 0 || code >= nsys)
                   1280:                callp += p->p_emul->e_nosys;
                   1281:        else if (tf->tf_out[6] & 1L) {
                   1282:                register_t *argp;
                   1283:
                   1284:                callp += code;
                   1285:                i = callp->sy_narg; /* Why divide? */
                   1286:                if (i > nap) {  /* usually false */
                   1287:                        if (i > 8)
                   1288:                                panic("syscall nargs");
                   1289:                        /* Read the whole block in */
                   1290:                        error = copyin((caddr_t)(u_long)tf->tf_out[6] + BIAS +
                   1291:                                       offsetof(struct frame64, fr_argx),
                   1292:                                       (caddr_t)&args[nap], (i - nap) * sizeof(register_t));
                   1293:                        i = nap;
                   1294:                }
                   1295:                /* It should be faster to do <=6 longword copies than call bcopy */
                   1296:                for (argp = args; i--;)
                   1297:                        *argp++ = *ap++;
                   1298:
                   1299: #ifdef KTRACE
                   1300:                if (KTRPOINT(p, KTR_SYSCALL))
                   1301:                        ktrsyscall(p, code,
                   1302:                                   callp->sy_argsize, args);
                   1303: #endif
                   1304:                if (error)
                   1305:                        goto bad;
                   1306:        } else {
                   1307:                error = EFAULT;
                   1308:                goto bad;
                   1309:        }
                   1310: #ifdef SYSCALL_DEBUG
                   1311:        scdebug_call(p, code, args);
                   1312: #endif
                   1313:        rval[0] = 0;
                   1314:        rval[1] = tf->tf_out[1];
                   1315: #if NSYSTRACE > 0
                   1316:        if (ISSET(p->p_flag, P_SYSTRACE))
                   1317:                error = systrace_redirect(code, p, args, rval);
                   1318:        else
                   1319: #endif
                   1320:                error = (*callp->sy_call)(p, args, rval);
                   1321:
                   1322:        switch (error) {
                   1323:                vaddr_t dest;
                   1324:        case 0:
                   1325:                /* Note: fork() does not return here in the child */
                   1326:                tf->tf_out[0] = rval[0];
                   1327:                tf->tf_out[1] = rval[1];
                   1328:                if (new) {
                   1329:                        /* jmp %g2 (or %g7, deprecated) on success */
                   1330:                        dest = tf->tf_global[new & SYSCALL_G2RFLAG ? 2 : 7];
                   1331:                        if (dest & 3) {
                   1332:                                error = EINVAL;
                   1333:                                goto bad;
                   1334:                        }
                   1335:                } else {
                   1336:                        /* old system call convention: clear C on success */
                   1337:                        tf->tf_tstate &= ~(((int64_t)(ICC_C|XCC_C))<<TSTATE_CCR_SHIFT); /* success */
                   1338:                        dest = tf->tf_npc;
                   1339:                }
                   1340:                tf->tf_pc = dest;
                   1341:                tf->tf_npc = dest + 4;
                   1342:                break;
                   1343:
                   1344:        case ERESTART:
                   1345:        case EJUSTRETURN:
                   1346:                /* nothing to do */
                   1347:                break;
                   1348:
                   1349:        default:
                   1350:        bad:
                   1351:                if (p->p_emul->e_errno)
                   1352:                        error = p->p_emul->e_errno[error];
                   1353:                tf->tf_out[0] = error;
                   1354:                tf->tf_tstate |= (((int64_t)(ICC_C|XCC_C))<<TSTATE_CCR_SHIFT);  /* fail */
                   1355:                dest = tf->tf_npc;
                   1356:                tf->tf_pc = dest;
                   1357:                tf->tf_npc = dest + 4;
                   1358:                break;
                   1359:        }
                   1360:
                   1361: #ifdef SYSCALL_DEBUG
                   1362:        scdebug_ret(p, code, error, rval);
                   1363: #endif
                   1364:        userret(p);
                   1365: #ifdef KTRACE
                   1366:        if (KTRPOINT(p, KTR_SYSRET))
                   1367:                ktrsysret(p, code, error, rval[0]);
                   1368: #endif
                   1369:        share_fpu(p, tf);
                   1370: }
                   1371:
                   1372: /*
                   1373:  * Process the tail end of a fork() for the child.
                   1374:  */
                   1375: void
                   1376: child_return(arg)
                   1377:        void *arg;
                   1378: {
                   1379:        struct proc *p = (struct proc *)arg;
                   1380:        struct trapframe64 *tf = p->p_md.md_tf;
                   1381:
                   1382:        /*
                   1383:         * Return values in the frame set by cpu_fork().
                   1384:         */
                   1385:        tf->tf_out[0] = 0;
                   1386:        tf->tf_out[1] = 0;
                   1387:        tf->tf_tstate &= ~(((int64_t)(ICC_C|XCC_C))<<TSTATE_CCR_SHIFT);
                   1388:
                   1389:        userret(p);
                   1390: #ifdef KTRACE
                   1391:        if (KTRPOINT(p, KTR_SYSRET))
                   1392:                ktrsysret(p,
                   1393:                    (p->p_flag & P_PPWAIT) ? SYS_vfork : SYS_fork, 0, 0);
                   1394: #endif
                   1395: }
                   1396:

CVSweb