bsd/dev/ppc/systemcalls.c

   1 /*
   2  * Copyright (c) 2000-2005 Apple Computer, Inc. All rights reserved.
   3  *
   4  * @APPLE_LICENSE_HEADER_START@
   5  *
   6  * The contents of this file constitute Original Code as defined in and
   7  * are subject to the Apple Public Source License Version 1.1 (the
   8  * "License").  You may not use this file except in compliance with the
   9  * License.  Please obtain a copy of the License at
  10  * http://www.apple.com/publicsource and read it before using this file.
  11  *
  12  * This Original Code and all software distributed under the License are
  13  * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  14  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  15  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  16  * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT.  Please see the
  17  * License for the specific language governing rights and limitations
  18  * under the License.
  19  *
  20  * @APPLE_LICENSE_HEADER_END@
  21  */
  22
  23 #include <kern/task.h>
  24 #include <kern/thread.h>
  25 #include <kern/assert.h>
  26 #include <kern/clock.h>
  27 #include <kern/locks.h>
  28 #include <kern/sched_prim.h>
  29 #include <mach/machine/thread_status.h>
  30 #include <ppc/savearea.h>
  31
  32 #include <sys/kernel.h>
  33 #include <sys/vm.h>
  34 #include <sys/proc_internal.h>
  35 #include <sys/syscall.h>
  36 #include <sys/systm.h>
  37 #include <sys/user.h>
  38 #include <sys/errno.h>
  39 #include <sys/ktrace.h>
  40 #include <sys/kdebug.h>
  41 #include <sys/sysent.h>
  42 #include <sys/sysproto.h>
  43 #include <sys/kauth.h>
  44
  45 #include <bsm/audit_kernel.h>
  46
  47 extern void
  48 unix_syscall(struct savearea *regs);
  49 void
  50 unix_syscall_return(int error);
  51
  52 extern struct savearea *
  53 find_user_regs(
  54         thread_t act);
  55
  56 extern void enter_funnel_section(funnel_t *funnel_lock);
  57 extern void exit_funnel_section(void);
  58
  59 /*
  60  * Function:    unix_syscall
  61  *
  62  * Inputs:      regs    - pointer to Process Control Block
  63  *
  64  * Outputs:     none
  65  */
  66 void
  67 unix_syscall(struct savearea    *regs)
  68 {
  69         thread_t                        thread_act;
  70         struct uthread          *uthread;
  71         struct proc                     *proc;
  72         struct sysent           *callp;
  73         int                                     error;
  74         unsigned short          code;
  75         boolean_t                       flavor;
  76         int funnel_type;
  77         unsigned int cancel_enable;
  78
  79         flavor = (((unsigned int)regs->save_r0) == 0)? 1: 0;
  80
  81         if (flavor)
  82                 code = regs->save_r3;
  83         else
  84                 code = regs->save_r0;
  85
  86         if (kdebug_enable && (code != 180)) {
  87                 if (flavor)
  88                         KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_START,
  89                                 regs->save_r4, regs->save_r5, regs->save_r6, regs->save_r7, 0);
  90                 else
  91                         KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_START,
  92                                 regs->save_r3, regs->save_r4, regs->save_r5, regs->save_r6, 0);
  93         }
  94         thread_act = current_thread();
  95         uthread = get_bsdthread_info(thread_act);
  96
  97         if (!(uthread->uu_flag & UT_VFORK))
  98                 proc = (struct proc *)get_bsdtask_info(current_task());
  99         else
 100                 proc = current_proc();
 101
 102         /* Make sure there is a process associated with this task */
 103         if (proc == NULL) {
 104                 regs->save_r3 = (long long)EPERM;
 105                 /* set the "pc" to execute cerror routine */
 106                 regs->save_srr0 -= 4;
 107                 task_terminate_internal(current_task());
 108                 thread_exception_return();
 109                 /* NOTREACHED */
 110         }
 111
 112         /*
 113          * Delayed binding of thread credential to process credential, if we
 114          * are not running with an explicitly set thread credential.
 115          */
 116         if (uthread->uu_ucred != proc->p_ucred &&
 117             (uthread->uu_flag & UT_SETUID) == 0) {
 118                 kauth_cred_t old = uthread->uu_ucred;
 119                 proc_lock(proc);
 120                 uthread->uu_ucred = proc->p_ucred;
 121                 kauth_cred_ref(uthread->uu_ucred);
 122                 proc_unlock(proc);
 123                 if (old != NOCRED)
 124                         kauth_cred_rele(old);
 125         }
 126
 127         callp = (code >= nsysent) ? &sysent[63] : &sysent[code];
 128
 129         if (callp->sy_narg != 0) {
 130                 void            *regsp;
 131                 sy_munge_t      *mungerp;
 132
 133                 if (IS_64BIT_PROCESS(proc)) {
 134                         /* XXX Turn 64 bit unsafe calls into nosys() */
 135                         if (callp->sy_funnel & UNSAFE_64BIT) {
 136                                 callp = &sysent[63];
 137                                 goto unsafe;
 138                         }
 139                         mungerp = callp->sy_arg_munge64;
 140                 }
 141                 else {
 142                         mungerp = callp->sy_arg_munge32;
 143                 }
 144                 if ( !flavor) {
 145                         regsp = (void *) &regs->save_r3;
 146                 } else {
 147                         /* indirect system call consumes an argument so only 7 are supported */
 148                         if (callp->sy_narg > 7) {
 149                                 callp = &sysent[63];
 150                                 goto unsafe;
 151                         }
 152                         regsp = (void *) &regs->save_r4;
 153                 }
 154                 /* call syscall argument munger to copy in arguments (see xnu/bsd/dev/ppc/munge.s) */
 155                 (*mungerp)(regsp, (void *) &uthread->uu_arg[0]);
 156         }
 157
 158 unsafe:
 159         cancel_enable = callp->sy_cancel;
 160
 161         if (cancel_enable == _SYSCALL_CANCEL_NONE) {
 162                         uthread->uu_flag |= UT_NOTCANCELPT;
 163         } else {
 164                 if((uthread->uu_flag & (UT_CANCELDISABLE | UT_CANCEL | UT_CANCELED)) == UT_CANCEL) {
 165                         if (cancel_enable == _SYSCALL_CANCEL_PRE) {
 166                                         /* system call cancelled; return to handle cancellation */
 167                                         regs->save_r3 = (long long)EINTR;
 168                                         thread_exception_return();
 169                                         /* NOTREACHED */
 170                         } else {
 171                         thread_abort_safely(thread_act);
 172                         }
 173                 }
 174         }
 175
 176         funnel_type = (int)(callp->sy_funnel & FUNNEL_MASK);
 177         if (funnel_type == KERNEL_FUNNEL)
 178                  enter_funnel_section(kernel_flock);
 179
 180         uthread->uu_rval[0] = 0;
 181
 182         /*
 183          * r4 is volatile, if we set it to regs->save_r4 here the child
 184          * will have parents r4 after execve
 185          */
 186         uthread->uu_rval[1] = 0;
 187
 188         error = 0;
 189
 190         /*
 191          * PPC runtime calls cerror after every unix system call, so
 192          * assume no error and adjust the "pc" to skip this call.
 193          * It will be set back to the cerror call if an error is detected.
 194          */
 195         regs->save_srr0 += 4;
 196
 197         if (KTRPOINT(proc, KTR_SYSCALL))
 198                 ktrsyscall(proc, code, callp->sy_narg, uthread->uu_arg);
 199
 200 #ifdef JOE_DEBUG
 201         uthread->uu_iocount = 0;
 202         uthread->uu_vpindex = 0;
 203 #endif
 204         AUDIT_SYSCALL_ENTER(code, proc, uthread);
 205         error = (*(callp->sy_call))(proc, (void *)uthread->uu_arg, &(uthread->uu_rval[0]));
 206         AUDIT_SYSCALL_EXIT(error, proc, uthread);
 207
 208 #ifdef JOE_DEBUG
 209         if (uthread->uu_iocount)
 210                 joe_debug("system call returned with uu_iocount != 0");
 211 #endif
 212         regs = find_user_regs(thread_act);
 213
 214         if (error == ERESTART) {
 215                 regs->save_srr0 -= 8;
 216         } else if (error != EJUSTRETURN) {
 217                 if (error) {
 218                         regs->save_r3 = (long long)error;
 219                         /* set the "pc" to execute cerror routine */
 220                         regs->save_srr0 -= 4;
 221                 } else { /* (not error) */
 222                         switch (callp->sy_return_type) {
 223                         case _SYSCALL_RET_INT_T:
 224                                 regs->save_r3 = uthread->uu_rval[0];
 225                                 regs->save_r4 = uthread->uu_rval[1];
 226                                 break;
 227                         case _SYSCALL_RET_UINT_T:
 228                                 regs->save_r3 = ((u_int)uthread->uu_rval[0]);
 229                                 regs->save_r4 = ((u_int)uthread->uu_rval[1]);
 230                                 break;
 231                         case _SYSCALL_RET_OFF_T:
 232                                 /* off_t returns 64 bits split across two registers for 32 bit */
 233                                 /* process and in one register for 64 bit process */
 234                                 if (IS_64BIT_PROCESS(proc)) {
 235                                         u_int64_t       *retp = (u_int64_t *)&uthread->uu_rval[0];
 236                                         regs->save_r3 = *retp;
 237                                         regs->save_r4 = 0;
 238                                 }
 239                                 else {
 240                                         regs->save_r3 = uthread->uu_rval[0];
 241                                         regs->save_r4 = uthread->uu_rval[1];
 242                                 }
 243                                 break;
 244                         case _SYSCALL_RET_ADDR_T:
 245                         case _SYSCALL_RET_SIZE_T:
 246                         case _SYSCALL_RET_SSIZE_T:
 247                                 /* the variable length return types (user_addr_t, user_ssize_t,
 248                                  * and user_size_t) are always the largest possible size in the
 249                                  * kernel (we use uu_rval[0] and [1] as one 64 bit value).
 250                                  */
 251                                 {
 252                                         user_addr_t *retp = (user_addr_t *)&uthread->uu_rval[0];
 253                                         regs->save_r3 = *retp;
 254                                         regs->save_r4 = 0;
 255                                 }
 256                                 break;
 257                         case _SYSCALL_RET_NONE:
 258                                 break;
 259                         default:
 260                                 panic("unix_syscall: unknown return type");
 261                                 break;
 262                         }
 263                 }
 264         }
 265         /* else  (error == EJUSTRETURN) { nothing } */
 266
 267
 268         if (KTRPOINT(proc, KTR_SYSRET)) {
 269                 switch(callp->sy_return_type) {
 270                 case _SYSCALL_RET_ADDR_T:
 271                 case _SYSCALL_RET_SIZE_T:
 272                 case _SYSCALL_RET_SSIZE_T:
 273                         /*
 274                          * Trace the value of the least significant bits,
 275                          * until we can revise the ktrace API safely.
 276                          */
 277                         ktrsysret(proc, code, error, uthread->uu_rval[1]);
 278                         break;
 279                 default:
 280                         ktrsysret(proc, code, error, uthread->uu_rval[0]);
 281                         break;
 282                 }
 283         }
 284
 285         if (cancel_enable == _SYSCALL_CANCEL_NONE)
 286                 uthread->uu_flag &= ~UT_NOTCANCELPT;
 287
 288         exit_funnel_section();
 289
 290         if (uthread->uu_lowpri_delay) {
 291                 /*
 292                  * task is marked as a low priority I/O type
 293                  * and the I/O we issued while in this system call
 294                  * collided with normal I/O operations... we'll
 295                  * delay in order to mitigate the impact of this
 296                  * task on the normal operation of the system
 297                  */
 298                 IOSleep(uthread->uu_lowpri_delay);
 299                 uthread->uu_lowpri_delay = 0;
 300         }
 301         if (kdebug_enable && (code != 180)) {
 302
 303                 if (callp->sy_return_type == _SYSCALL_RET_SSIZE_T)
 304                         KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_END,
 305                                               error, uthread->uu_rval[1], 0, 0, 0);
 306                 else
 307                         KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_END,
 308                                               error, uthread->uu_rval[0], uthread->uu_rval[1], 0, 0);
 309         }
 310
 311         thread_exception_return();
 312         /* NOTREACHED */
 313 }
 314
 315 void
 316 unix_syscall_return(int error)
 317 {
 318         thread_t                                        thread_act;
 319         struct uthread                          *uthread;
 320         struct proc                                     *proc;
 321         struct savearea                         *regs;
 322         unsigned short                          code;
 323         struct sysent                           *callp;
 324         unsigned int cancel_enable;
 325
 326         thread_act = current_thread();
 327         proc = current_proc();
 328         uthread = get_bsdthread_info(thread_act);
 329
 330         regs = find_user_regs(thread_act);
 331
 332         if (regs->save_r0 != 0)
 333                 code = regs->save_r0;
 334         else
 335                 code = regs->save_r3;
 336
 337         callp = (code >= nsysent) ? &sysent[63] : &sysent[code];
 338
 339         /*
 340          * Get index into sysent table
 341          */
 342         if (error == ERESTART) {
 343                 regs->save_srr0 -= 8;
 344         } else if (error != EJUSTRETURN) {
 345                 if (error) {
 346                         regs->save_r3 = (long long)error;
 347                         /* set the "pc" to execute cerror routine */
 348                         regs->save_srr0 -= 4;
 349                 } else { /* (not error) */
 350                         switch (callp->sy_return_type) {
 351                         case _SYSCALL_RET_INT_T:
 352                                 regs->save_r3 = uthread->uu_rval[0];
 353                                 regs->save_r4 = uthread->uu_rval[1];
 354                                 break;
 355                         case _SYSCALL_RET_UINT_T:
 356                                 regs->save_r3 = ((u_int)uthread->uu_rval[0]);
 357                                 regs->save_r4 = ((u_int)uthread->uu_rval[1]);
 358                                 break;
 359                         case _SYSCALL_RET_OFF_T:
 360                                 /* off_t returns 64 bits split across two registers for 32 bit */
 361                                 /* process and in one register for 64 bit process */
 362                                 if (IS_64BIT_PROCESS(proc)) {
 363                                         u_int64_t       *retp = (u_int64_t *)&uthread->uu_rval[0];
 364                                         regs->save_r3 = *retp;
 365                                 }
 366                                 else {
 367                                         regs->save_r3 = uthread->uu_rval[0];
 368                                         regs->save_r4 = uthread->uu_rval[1];
 369                                 }
 370                                 break;
 371                         case _SYSCALL_RET_ADDR_T:
 372                         case _SYSCALL_RET_SIZE_T:
 373                         case _SYSCALL_RET_SSIZE_T:
 374                                 /* the variable length return types (user_addr_t, user_ssize_t,
 375                                  * and user_size_t) are always the largest possible size in the
 376                                  * kernel (we use uu_rval[0] and [1] as one 64 bit value).
 377                                  */
 378                                 {
 379                                         u_int64_t       *retp = (u_int64_t *)&uthread->uu_rval[0];
 380                                         regs->save_r3 = *retp;
 381                                 }
 382                                 break;
 383                         case _SYSCALL_RET_NONE:
 384                                 break;
 385                         default:
 386                                 panic("unix_syscall: unknown return type");
 387                                 break;
 388                         }
 389                 }
 390         }
 391         /* else  (error == EJUSTRETURN) { nothing } */
 392
 393         if (KTRPOINT(proc, KTR_SYSRET)) {
 394                 switch(callp->sy_return_type) {
 395                 case _SYSCALL_RET_ADDR_T:
 396                 case _SYSCALL_RET_SIZE_T:
 397                 case _SYSCALL_RET_SSIZE_T:
 398                         /*
 399                          * Trace the value of the least significant bits,
 400                          * until we can revise the ktrace API safely.
 401                          */
 402                         ktrsysret(proc, code, error, uthread->uu_rval[1]);
 403                         break;
 404                 default:
 405                         ktrsysret(proc, code, error, uthread->uu_rval[0]);
 406                         break;
 407                 }
 408         }
 409
 410         cancel_enable = callp->sy_cancel;
 411
 412         if (cancel_enable == _SYSCALL_CANCEL_NONE)
 413                 uthread->uu_flag &= ~UT_NOTCANCELPT;
 414
 415         exit_funnel_section();
 416
 417         if (uthread->uu_lowpri_delay) {
 418                 /*
 419                  * task is marked as a low priority I/O type
 420                  * and the I/O we issued while in this system call
 421                  * collided with normal I/O operations... we'll
 422                  * delay in order to mitigate the impact of this
 423                  * task on the normal operation of the system
 424                  */
 425                 IOSleep(uthread->uu_lowpri_delay);
 426                 uthread->uu_lowpri_delay = 0;
 427         }
 428         if (kdebug_enable && (code != 180)) {
 429                 if (callp->sy_return_type == _SYSCALL_RET_SSIZE_T)
 430                         KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_END,
 431                                               error, uthread->uu_rval[1], 0, 0, 0);
 432                 else
 433                         KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_END,
 434                                               error, uthread->uu_rval[0], uthread->uu_rval[1], 0, 0);
 435         }
 436
 437         thread_exception_return();
 438         /* NOTREACHED */
 439 }
 440
 441 #ifdef JOE_DEBUG
 442 joe_debug(char *p) {
 443
 444         printf("%s\n", p);
 445 }
 446 #endif