bsd/dev/ppc/systemcalls.c

   1 /*
   2  * Copyright (c) 2000-2005 Apple Computer, Inc. All rights reserved.
   3  *
   4  * @APPLE_LICENSE_OSREFERENCE_HEADER_START@
   5  *
   6  * This file contains Original Code and/or Modifications of Original Code
   7  * as defined in and that are subject to the Apple Public Source License
   8  * Version 2.0 (the 'License'). You may not use this file except in
   9  * compliance with the License.  The rights granted to you under the
  10  * License may not be used to create, or enable the creation or
  11  * redistribution of, unlawful or unlicensed copies of an Apple operating
  12  * system, or to circumvent, violate, or enable the circumvention or
  13  * violation of, any terms of an Apple operating system software license
  14  * agreement.
  15  *
  16  * Please obtain a copy of the License at
  17  * http://www.opensource.apple.com/apsl/ and read it before using this
  18  * file.
  19  *
  20  * The Original Code and all software distributed under the License are
  21  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  22  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  23  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  24  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  25  * Please see the License for the specific language governing rights and
  26  * limitations under the License.
  27  *
  28  * @APPLE_LICENSE_OSREFERENCE_HEADER_END@
  29  */
  30
  31 #include <kern/task.h>
  32 #include <kern/thread.h>
  33 #include <kern/assert.h>
  34 #include <kern/clock.h>
  35 #include <kern/locks.h>
  36 #include <kern/sched_prim.h>
  37 #include <mach/machine/thread_status.h>
  38 #include <ppc/savearea.h>
  39
  40 #include <sys/kernel.h>
  41 #include <sys/vm.h>
  42 #include <sys/proc_internal.h>
  43 #include <sys/syscall.h>
  44 #include <sys/systm.h>
  45 #include <sys/user.h>
  46 #include <sys/errno.h>
  47 #include <sys/ktrace.h>
  48 #include <sys/kdebug.h>
  49 #include <sys/sysent.h>
  50 #include <sys/sysproto.h>
  51 #include <sys/kauth.h>
  52
  53 #include <bsm/audit_kernel.h>
  54
  55 extern void
  56 unix_syscall(struct savearea *regs);
  57 void
  58 unix_syscall_return(int error);
  59
  60 extern struct savearea *
  61 find_user_regs(
  62         thread_t act);
  63
  64 extern void enter_funnel_section(funnel_t *funnel_lock);
  65 extern void exit_funnel_section(void);
  66
  67 /*
  68  * Function:    unix_syscall
  69  *
  70  * Inputs:      regs    - pointer to Process Control Block
  71  *
  72  * Outputs:     none
  73  */
  74 void
  75 unix_syscall(struct savearea    *regs)
  76 {
  77         thread_t                        thread_act;
  78         struct uthread          *uthread;
  79         struct proc                     *proc;
  80         struct sysent           *callp;
  81         int                                     error;
  82         unsigned short          code;
  83         boolean_t                       flavor;
  84         int funnel_type;
  85         unsigned int cancel_enable;
  86
  87         flavor = (((unsigned int)regs->save_r0) == 0)? 1: 0;
  88
  89         if (flavor)
  90                 code = regs->save_r3;
  91         else
  92                 code = regs->save_r0;
  93
  94         if (kdebug_enable && (code != 180)) {
  95                 if (flavor)
  96                         KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_START,
  97                                 regs->save_r4, regs->save_r5, regs->save_r6, regs->save_r7, 0);
  98                 else
  99                         KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_START,
 100                                 regs->save_r3, regs->save_r4, regs->save_r5, regs->save_r6, 0);
 101         }
 102         thread_act = current_thread();
 103         uthread = get_bsdthread_info(thread_act);
 104
 105         if (!(uthread->uu_flag & UT_VFORK))
 106                 proc = (struct proc *)get_bsdtask_info(current_task());
 107         else
 108                 proc = current_proc();
 109
 110         /* Make sure there is a process associated with this task */
 111         if (proc == NULL) {
 112                 regs->save_r3 = (long long)EPERM;
 113                 /* set the "pc" to execute cerror routine */
 114                 regs->save_srr0 -= 4;
 115                 task_terminate_internal(current_task());
 116                 thread_exception_return();
 117                 /* NOTREACHED */
 118         }
 119
 120         /*
 121          * Delayed binding of thread credential to process credential, if we
 122          * are not running with an explicitly set thread credential.
 123          */
 124         if (uthread->uu_ucred != proc->p_ucred &&
 125             (uthread->uu_flag & UT_SETUID) == 0) {
 126                 kauth_cred_t old = uthread->uu_ucred;
 127                 proc_lock(proc);
 128                 uthread->uu_ucred = proc->p_ucred;
 129                 kauth_cred_ref(uthread->uu_ucred);
 130                 proc_unlock(proc);
 131                 if (old != NOCRED)
 132                         kauth_cred_rele(old);
 133         }
 134
 135         callp = (code >= nsysent) ? &sysent[63] : &sysent[code];
 136
 137         if (callp->sy_narg != 0) {
 138                 void            *regsp;
 139                 sy_munge_t      *mungerp;
 140
 141                 if (IS_64BIT_PROCESS(proc)) {
 142                         /* XXX Turn 64 bit unsafe calls into nosys() */
 143                         if (callp->sy_funnel & UNSAFE_64BIT) {
 144                                 callp = &sysent[63];
 145                                 goto unsafe;
 146                         }
 147                         mungerp = callp->sy_arg_munge64;
 148                 }
 149                 else {
 150                         mungerp = callp->sy_arg_munge32;
 151                 }
 152                 if ( !flavor) {
 153                         regsp = (void *) &regs->save_r3;
 154                 } else {
 155                         /* indirect system call consumes an argument so only 7 are supported */
 156                         if (callp->sy_narg > 7) {
 157                                 callp = &sysent[63];
 158                                 goto unsafe;
 159                         }
 160                         regsp = (void *) &regs->save_r4;
 161                 }
 162                 /* call syscall argument munger to copy in arguments (see xnu/bsd/dev/ppc/munge.s) */
 163                 (*mungerp)(regsp, (void *) &uthread->uu_arg[0]);
 164         }
 165
 166 unsafe:
 167         cancel_enable = callp->sy_cancel;
 168
 169         if (cancel_enable == _SYSCALL_CANCEL_NONE) {
 170                         uthread->uu_flag |= UT_NOTCANCELPT;
 171         } else {
 172                 if((uthread->uu_flag & (UT_CANCELDISABLE | UT_CANCEL | UT_CANCELED)) == UT_CANCEL) {
 173                         if (cancel_enable == _SYSCALL_CANCEL_PRE) {
 174                                         /* system call cancelled; return to handle cancellation */
 175                                         regs->save_r3 = (long long)EINTR;
 176                                         thread_exception_return();
 177                                         /* NOTREACHED */
 178                         } else {
 179                         thread_abort_safely(thread_act);
 180                         }
 181                 }
 182         }
 183
 184         funnel_type = (int)(callp->sy_funnel & FUNNEL_MASK);
 185         if (funnel_type == KERNEL_FUNNEL)
 186                  enter_funnel_section(kernel_flock);
 187
 188         uthread->uu_rval[0] = 0;
 189
 190         /*
 191          * r4 is volatile, if we set it to regs->save_r4 here the child
 192          * will have parents r4 after execve
 193          */
 194         uthread->uu_rval[1] = 0;
 195
 196         error = 0;
 197
 198         /*
 199          * PPC runtime calls cerror after every unix system call, so
 200          * assume no error and adjust the "pc" to skip this call.
 201          * It will be set back to the cerror call if an error is detected.
 202          */
 203         regs->save_srr0 += 4;
 204
 205         if (KTRPOINT(proc, KTR_SYSCALL))
 206                 ktrsyscall(proc, code, callp->sy_narg, uthread->uu_arg);
 207
 208 #ifdef JOE_DEBUG
 209         uthread->uu_iocount = 0;
 210         uthread->uu_vpindex = 0;
 211 #endif
 212         AUDIT_SYSCALL_ENTER(code, proc, uthread);
 213         error = (*(callp->sy_call))(proc, (void *)uthread->uu_arg, &(uthread->uu_rval[0]));
 214         AUDIT_SYSCALL_EXIT(error, proc, uthread);
 215
 216 #ifdef JOE_DEBUG
 217         if (uthread->uu_iocount)
 218                 joe_debug("system call returned with uu_iocount != 0");
 219 #endif
 220         regs = find_user_regs(thread_act);
 221
 222         if (error == ERESTART) {
 223                 regs->save_srr0 -= 8;
 224         } else if (error != EJUSTRETURN) {
 225                 if (error) {
 226                         regs->save_r3 = (long long)error;
 227                         /* set the "pc" to execute cerror routine */
 228                         regs->save_srr0 -= 4;
 229                 } else { /* (not error) */
 230                         switch (callp->sy_return_type) {
 231                         case _SYSCALL_RET_INT_T:
 232                                 regs->save_r3 = uthread->uu_rval[0];
 233                                 regs->save_r4 = uthread->uu_rval[1];
 234                                 break;
 235                         case _SYSCALL_RET_UINT_T:
 236                                 regs->save_r3 = ((u_int)uthread->uu_rval[0]);
 237                                 regs->save_r4 = ((u_int)uthread->uu_rval[1]);
 238                                 break;
 239                         case _SYSCALL_RET_OFF_T:
 240                                 /* off_t returns 64 bits split across two registers for 32 bit */
 241                                 /* process and in one register for 64 bit process */
 242                                 if (IS_64BIT_PROCESS(proc)) {
 243                                         u_int64_t       *retp = (u_int64_t *)&uthread->uu_rval[0];
 244                                         regs->save_r3 = *retp;
 245                                         regs->save_r4 = 0;
 246                                 }
 247                                 else {
 248                                         regs->save_r3 = uthread->uu_rval[0];
 249                                         regs->save_r4 = uthread->uu_rval[1];
 250                                 }
 251                                 break;
 252                         case _SYSCALL_RET_ADDR_T:
 253                         case _SYSCALL_RET_SIZE_T:
 254                         case _SYSCALL_RET_SSIZE_T:
 255                                 /* the variable length return types (user_addr_t, user_ssize_t,
 256                                  * and user_size_t) are always the largest possible size in the
 257                                  * kernel (we use uu_rval[0] and [1] as one 64 bit value).
 258                                  */
 259                                 {
 260                                         user_addr_t *retp = (user_addr_t *)&uthread->uu_rval[0];
 261                                         regs->save_r3 = *retp;
 262                                         regs->save_r4 = 0;
 263                                 }
 264                                 break;
 265                         case _SYSCALL_RET_NONE:
 266                                 break;
 267                         default:
 268                                 panic("unix_syscall: unknown return type");
 269                                 break;
 270                         }
 271                 }
 272         }
 273         /* else  (error == EJUSTRETURN) { nothing } */
 274
 275
 276         if (KTRPOINT(proc, KTR_SYSRET)) {
 277                 switch(callp->sy_return_type) {
 278                 case _SYSCALL_RET_ADDR_T:
 279                 case _SYSCALL_RET_SIZE_T:
 280                 case _SYSCALL_RET_SSIZE_T:
 281                         /*
 282                          * Trace the value of the least significant bits,
 283                          * until we can revise the ktrace API safely.
 284                          */
 285                         ktrsysret(proc, code, error, uthread->uu_rval[1]);
 286                         break;
 287                 default:
 288                         ktrsysret(proc, code, error, uthread->uu_rval[0]);
 289                         break;
 290                 }
 291         }
 292
 293         if (cancel_enable == _SYSCALL_CANCEL_NONE)
 294                 uthread->uu_flag &= ~UT_NOTCANCELPT;
 295
 296         exit_funnel_section();
 297
 298         if (uthread->uu_lowpri_delay) {
 299                 /*
 300                  * task is marked as a low priority I/O type
 301                  * and the I/O we issued while in this system call
 302                  * collided with normal I/O operations... we'll
 303                  * delay in order to mitigate the impact of this
 304                  * task on the normal operation of the system
 305                  */
 306                 IOSleep(uthread->uu_lowpri_delay);
 307                 uthread->uu_lowpri_delay = 0;
 308         }
 309         if (kdebug_enable && (code != 180)) {
 310
 311                 if (callp->sy_return_type == _SYSCALL_RET_SSIZE_T)
 312                         KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_END,
 313                                               error, uthread->uu_rval[1], 0, 0, 0);
 314                 else
 315                         KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_END,
 316                                               error, uthread->uu_rval[0], uthread->uu_rval[1], 0, 0);
 317         }
 318
 319         thread_exception_return();
 320         /* NOTREACHED */
 321 }
 322
 323 void
 324 unix_syscall_return(int error)
 325 {
 326         thread_t                                        thread_act;
 327         struct uthread                          *uthread;
 328         struct proc                                     *proc;
 329         struct savearea                         *regs;
 330         unsigned short                          code;
 331         struct sysent                           *callp;
 332         unsigned int cancel_enable;
 333
 334         thread_act = current_thread();
 335         proc = current_proc();
 336         uthread = get_bsdthread_info(thread_act);
 337
 338         regs = find_user_regs(thread_act);
 339
 340         if (regs->save_r0 != 0)
 341                 code = regs->save_r0;
 342         else
 343                 code = regs->save_r3;
 344
 345         callp = (code >= nsysent) ? &sysent[63] : &sysent[code];
 346
 347         /*
 348          * Get index into sysent table
 349          */
 350         if (error == ERESTART) {
 351                 regs->save_srr0 -= 8;
 352         } else if (error != EJUSTRETURN) {
 353                 if (error) {
 354                         regs->save_r3 = (long long)error;
 355                         /* set the "pc" to execute cerror routine */
 356                         regs->save_srr0 -= 4;
 357                 } else { /* (not error) */
 358                         switch (callp->sy_return_type) {
 359                         case _SYSCALL_RET_INT_T:
 360                                 regs->save_r3 = uthread->uu_rval[0];
 361                                 regs->save_r4 = uthread->uu_rval[1];
 362                                 break;
 363                         case _SYSCALL_RET_UINT_T:
 364                                 regs->save_r3 = ((u_int)uthread->uu_rval[0]);
 365                                 regs->save_r4 = ((u_int)uthread->uu_rval[1]);
 366                                 break;
 367                         case _SYSCALL_RET_OFF_T:
 368                                 /* off_t returns 64 bits split across two registers for 32 bit */
 369                                 /* process and in one register for 64 bit process */
 370                                 if (IS_64BIT_PROCESS(proc)) {
 371                                         u_int64_t       *retp = (u_int64_t *)&uthread->uu_rval[0];
 372                                         regs->save_r3 = *retp;
 373                                 }
 374                                 else {
 375                                         regs->save_r3 = uthread->uu_rval[0];
 376                                         regs->save_r4 = uthread->uu_rval[1];
 377                                 }
 378                                 break;
 379                         case _SYSCALL_RET_ADDR_T:
 380                         case _SYSCALL_RET_SIZE_T:
 381                         case _SYSCALL_RET_SSIZE_T:
 382                                 /* the variable length return types (user_addr_t, user_ssize_t,
 383                                  * and user_size_t) are always the largest possible size in the
 384                                  * kernel (we use uu_rval[0] and [1] as one 64 bit value).
 385                                  */
 386                                 {
 387                                         u_int64_t       *retp = (u_int64_t *)&uthread->uu_rval[0];
 388                                         regs->save_r3 = *retp;
 389                                 }
 390                                 break;
 391                         case _SYSCALL_RET_NONE:
 392                                 break;
 393                         default:
 394                                 panic("unix_syscall: unknown return type");
 395                                 break;
 396                         }
 397                 }
 398         }
 399         /* else  (error == EJUSTRETURN) { nothing } */
 400
 401         if (KTRPOINT(proc, KTR_SYSRET)) {
 402                 switch(callp->sy_return_type) {
 403                 case _SYSCALL_RET_ADDR_T:
 404                 case _SYSCALL_RET_SIZE_T:
 405                 case _SYSCALL_RET_SSIZE_T:
 406                         /*
 407                          * Trace the value of the least significant bits,
 408                          * until we can revise the ktrace API safely.
 409                          */
 410                         ktrsysret(proc, code, error, uthread->uu_rval[1]);
 411                         break;
 412                 default:
 413                         ktrsysret(proc, code, error, uthread->uu_rval[0]);
 414                         break;
 415                 }
 416         }
 417
 418         cancel_enable = callp->sy_cancel;
 419
 420         if (cancel_enable == _SYSCALL_CANCEL_NONE)
 421                 uthread->uu_flag &= ~UT_NOTCANCELPT;
 422
 423         exit_funnel_section();
 424
 425         if (uthread->uu_lowpri_delay) {
 426                 /*
 427                  * task is marked as a low priority I/O type
 428                  * and the I/O we issued while in this system call
 429                  * collided with normal I/O operations... we'll
 430                  * delay in order to mitigate the impact of this
 431                  * task on the normal operation of the system
 432                  */
 433                 IOSleep(uthread->uu_lowpri_delay);
 434                 uthread->uu_lowpri_delay = 0;
 435         }
 436         if (kdebug_enable && (code != 180)) {
 437                 if (callp->sy_return_type == _SYSCALL_RET_SSIZE_T)
 438                         KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_END,
 439                                               error, uthread->uu_rval[1], 0, 0, 0);
 440                 else
 441                         KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_END,
 442                                               error, uthread->uu_rval[0], uthread->uu_rval[1], 0, 0);
 443         }
 444
 445         thread_exception_return();
 446         /* NOTREACHED */
 447 }
 448
 449 #ifdef JOE_DEBUG
 450 joe_debug(char *p) {
 451
 452         printf("%s\n", p);
 453 }
 454 #endif