bsd/dev/ppc/systemcalls.c

   1 /*
   2  * Copyright (c) 2000-2008 Apple Inc. All rights reserved.
   3  *
   4  * @APPLE_OSREFERENCE_LICENSE_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 License
  10  * may not be used to create, or enable the creation or redistribution of,
  11  * unlawful or unlicensed copies of an Apple operating system, or to
  12  * circumvent, violate, or enable the circumvention or violation of, any
  13  * terms of an Apple operating system software license agreement.
  14  *
  15  * Please obtain a copy of the License at
  16  * http://www.opensource.apple.com/apsl/ and read it before using this file.
  17  *
  18  * The Original Code and all software distributed under the License are
  19  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  20  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  21  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  22  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  23  * Please see the License for the specific language governing rights and
  24  * limitations under the License.
  25  *
  26  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  27  */
  28 /*
  29  * NOTICE: This file was modified by McAfee Research in 2004 to introduce
  30  * support for mandatory and extensible security protections.  This notice
  31  * is included in support of clause 2.2 (b) of the Apple Public License,
  32  * Version 2.0.
  33  */
  34
  35 #include <mach/mach_traps.h>
  36
  37 #include <kern/task.h>
  38 #include <kern/thread.h>
  39 #include <kern/assert.h>
  40 #include <kern/clock.h>
  41 #include <kern/locks.h>
  42 #include <kern/sched_prim.h>
  43 #include <mach/machine/thread_status.h>
  44 #include <mach/thread_act.h>
  45 #include <ppc/savearea.h>
  46
  47 #include <sys/kernel.h>
  48 #include <sys/vm.h>
  49 #include <sys/proc_internal.h>
  50 #include <sys/syscall.h>
  51 #include <sys/systm.h>
  52 #include <sys/user.h>
  53 #include <sys/errno.h>
  54 #include <sys/kdebug.h>
  55 #include <sys/sysent.h>
  56 #include <sys/sysproto.h>
  57 #include <sys/kauth.h>
  58
  59 #include <security/audit/audit.h>
  60
  61 #if CONFIG_DTRACE
  62 extern int32_t dtrace_systrace_syscall(struct proc *, void *, int *);
  63 extern void dtrace_systrace_syscall_return(unsigned short, int, int *);
  64 #endif
  65
  66 extern void
  67 unix_syscall(struct savearea *regs);
  68
  69 extern struct savearea *
  70 find_user_regs(
  71         thread_t act);
  72
  73 extern lck_spin_t * tz_slock;
  74
  75 /*
  76  * Function:    unix_syscall
  77  *
  78  * Inputs:      regs    - pointer to Process Control Block
  79  *
  80  * Outputs:     none
  81  */
  82 void
  83 unix_syscall(struct savearea    *regs)
  84 {
  85         thread_t                        thread_act;
  86         struct uthread          *uthread;
  87         struct proc                     *proc;
  88         struct sysent           *callp;
  89         int                                     error;
  90         unsigned int            code;
  91         boolean_t                       flavor;
  92
  93         flavor = (((unsigned int)regs->save_r0) == 0)? 1: 0;
  94
  95         if (flavor)
  96                 code = regs->save_r3;
  97         else
  98                 code = regs->save_r0;
  99
 100         if (kdebug_enable && (code != 180)) {
 101                 if (flavor)
 102                         KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_START,
 103                                 regs->save_r4, regs->save_r5, regs->save_r6, regs->save_r7, 0);
 104                 else
 105                         KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_START,
 106                                 regs->save_r3, regs->save_r4, regs->save_r5, regs->save_r6, 0);
 107         }
 108         thread_act = current_thread();
 109         uthread = get_bsdthread_info(thread_act);
 110
 111         if (!(uthread->uu_flag & UT_VFORK))
 112                 proc = (struct proc *)get_bsdtask_info(current_task());
 113         else
 114                 proc = current_proc();
 115
 116         /* Make sure there is a process associated with this task */
 117         if (proc == NULL) {
 118                 regs->save_r3 = (long long)EPERM;
 119                 /* set the "pc" to execute cerror routine */
 120                 regs->save_srr0 -= 4;
 121                 task_terminate_internal(current_task());
 122                 thread_exception_return();
 123                 /* NOTREACHED */
 124         }
 125
 126         /*
 127          * Delayed binding of thread credential to process credential, if we
 128          * are not running with an explicitly set thread credential.
 129          */
 130         kauth_cred_uthread_update(uthread, proc);
 131
 132         callp = (code >= NUM_SYSENT) ? &sysent[63] : &sysent[code];
 133
 134         if (callp->sy_narg != 0) {
 135                 void            *regsp;
 136                 sy_munge_t      *mungerp;
 137
 138                 if (IS_64BIT_PROCESS(proc)) {
 139                         /* XXX Turn 64 bit unsafe calls into nosys() */
 140                         if (callp->sy_flags & UNSAFE_64BIT) {
 141                                 callp = &sysent[63];
 142                                 goto unsafe;
 143                         }
 144                         mungerp = callp->sy_arg_munge64;
 145                 }
 146                 else {
 147                         mungerp = callp->sy_arg_munge32;
 148                 }
 149                 if ( !flavor) {
 150                         regsp = (void *) &regs->save_r3;
 151                 } else {
 152                         /* indirect system call consumes an argument so only 7 are supported */
 153                         if (callp->sy_narg > 7) {
 154                                 callp = &sysent[63];
 155                                 goto unsafe;
 156                         }
 157                         regsp = (void *) &regs->save_r4;
 158                 }
 159                 /* call syscall argument munger to copy in arguments (see xnu/bsd/dev/ppc/munge.s) */
 160                 (*mungerp)(regsp, (void *) &uthread->uu_arg[0]);
 161         }
 162
 163 unsafe:
 164
 165         uthread->uu_flag |= UT_NOTCANCELPT;
 166
 167         uthread->uu_rval[0] = 0;
 168
 169         /*
 170          * r4 is volatile, if we set it to regs->save_r4 here the child
 171          * will have parents r4 after execve
 172          */
 173         uthread->uu_rval[1] = 0;
 174
 175         error = 0;
 176
 177         /*
 178          * PPC runtime calls cerror after every unix system call, so
 179          * assume no error and adjust the "pc" to skip this call.
 180          * It will be set back to the cerror call if an error is detected.
 181          */
 182         regs->save_srr0 += 4;
 183
 184 #ifdef JOE_DEBUG
 185         uthread->uu_iocount = 0;
 186         uthread->uu_vpindex = 0;
 187 #endif
 188         AUDIT_SYSCALL_ENTER(code, proc, uthread);
 189         error = (*(callp->sy_call))(proc, (void *)uthread->uu_arg, &(uthread->uu_rval[0]));
 190         AUDIT_SYSCALL_EXIT(code, proc, uthread, error);
 191 #if CONFIG_MACF
 192         mac_thread_userret(code, error, thread_act);
 193 #endif
 194
 195
 196 #ifdef JOE_DEBUG
 197         if (uthread->uu_iocount)
 198                 printf("system call returned with uu_iocount != 0\n");
 199 #endif
 200 #if CONFIG_DTRACE
 201         uthread->t_dtrace_errno = error;
 202 #endif /* CONFIG_DTRACE */
 203
 204         regs = find_user_regs(thread_act);
 205
 206         if (error == ERESTART) {
 207                 regs->save_srr0 -= 8;
 208         } else if (error != EJUSTRETURN) {
 209                 if (error) {
 210                         regs->save_r3 = (long long)error;
 211                         /* set the "pc" to execute cerror routine */
 212                         regs->save_srr0 -= 4;
 213                 } else { /* (not error) */
 214                         switch (callp->sy_return_type) {
 215                         case _SYSCALL_RET_INT_T:
 216                                 regs->save_r3 = uthread->uu_rval[0];
 217                                 regs->save_r4 = uthread->uu_rval[1];
 218                                 break;
 219                         case _SYSCALL_RET_UINT_T:
 220                                 regs->save_r3 = ((u_int)uthread->uu_rval[0]);
 221                                 regs->save_r4 = ((u_int)uthread->uu_rval[1]);
 222                                 break;
 223                         case _SYSCALL_RET_OFF_T:
 224                                 /* off_t returns 64 bits split across two registers for 32 bit */
 225                                 /* process and in one register for 64 bit process */
 226                                 if (IS_64BIT_PROCESS(proc)) {
 227                                         u_int64_t       *retp = (u_int64_t *)&uthread->uu_rval[0];
 228                                         regs->save_r3 = *retp;
 229                                         regs->save_r4 = 0;
 230                                 }
 231                                 else {
 232                                         regs->save_r3 = uthread->uu_rval[0];
 233                                         regs->save_r4 = uthread->uu_rval[1];
 234                                 }
 235                                 break;
 236                         case _SYSCALL_RET_ADDR_T:
 237                         case _SYSCALL_RET_SIZE_T:
 238                         case _SYSCALL_RET_SSIZE_T:
 239                                 /* the variable length return types (user_addr_t, user_ssize_t,
 240                                  * and user_size_t) are always the largest possible size in the
 241                                  * kernel (we use uu_rval[0] and [1] as one 64 bit value).
 242                                  */
 243                                 {
 244                                         user_addr_t *retp = (user_addr_t *)&uthread->uu_rval[0];
 245                                         regs->save_r3 = *retp;
 246                                         regs->save_r4 = 0;
 247                                 }
 248                                 break;
 249                         case _SYSCALL_RET_NONE:
 250                                 break;
 251                         default:
 252                                 panic("unix_syscall: unknown return type");
 253                                 break;
 254                         }
 255                 }
 256         }
 257         /* else  (error == EJUSTRETURN) { nothing } */
 258
 259
 260         uthread->uu_flag &= ~UT_NOTCANCELPT;
 261
 262         /* panic if funnel is held */
 263         syscall_exit_funnelcheck();
 264
 265         if (uthread->uu_lowpri_window) {
 266                 /*
 267                  * task is marked as a low priority I/O type
 268                  * and the I/O we issued while in this system call
 269                  * collided with normal I/O operations... we'll
 270                  * delay in order to mitigate the impact of this
 271                  * task on the normal operation of the system
 272                  */
 273                 throttle_lowpri_io(TRUE);
 274         }
 275         if (kdebug_enable && (code != 180)) {
 276
 277                 if (callp->sy_return_type == _SYSCALL_RET_SSIZE_T)
 278                         KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_END,
 279                                               error, uthread->uu_rval[1], 0, proc->p_pid, 0);
 280                 else
 281                         KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_END,
 282                                               error, uthread->uu_rval[0], uthread->uu_rval[1], proc->p_pid, 0);
 283         }
 284
 285         thread_exception_return();
 286         /* NOTREACHED */
 287 }
 288
 289 void
 290 unix_syscall_return(int error)
 291 {
 292         thread_t                                        thread_act;
 293         struct uthread                          *uthread;
 294         struct proc                                     *proc;
 295         struct savearea                         *regs;
 296         unsigned int                            code;
 297         struct sysent                           *callp;
 298
 299         thread_act = current_thread();
 300         proc = current_proc();
 301         uthread = get_bsdthread_info(thread_act);
 302
 303         regs = find_user_regs(thread_act);
 304
 305         if (regs->save_r0 != 0)
 306                 code = regs->save_r0;
 307         else
 308                 code = regs->save_r3;
 309
 310         callp = (code >= NUM_SYSENT) ? &sysent[63] : &sysent[code];
 311
 312 #if CONFIG_DTRACE
 313         if (callp->sy_call == dtrace_systrace_syscall)
 314                 dtrace_systrace_syscall_return( code, error, uthread->uu_rval );
 315 #endif /* CONFIG_DTRACE */
 316         AUDIT_SYSCALL_EXIT(code, proc, uthread, error);
 317
 318         /*
 319          * Get index into sysent table
 320          */
 321         if (error == ERESTART) {
 322                 regs->save_srr0 -= 8;
 323         } else if (error != EJUSTRETURN) {
 324                 if (error) {
 325                         regs->save_r3 = (long long)error;
 326                         /* set the "pc" to execute cerror routine */
 327                         regs->save_srr0 -= 4;
 328                 } else { /* (not error) */
 329                         switch (callp->sy_return_type) {
 330                         case _SYSCALL_RET_INT_T:
 331                                 regs->save_r3 = uthread->uu_rval[0];
 332                                 regs->save_r4 = uthread->uu_rval[1];
 333                                 break;
 334                         case _SYSCALL_RET_UINT_T:
 335                                 regs->save_r3 = ((u_int)uthread->uu_rval[0]);
 336                                 regs->save_r4 = ((u_int)uthread->uu_rval[1]);
 337                                 break;
 338                         case _SYSCALL_RET_OFF_T:
 339                                 /* off_t returns 64 bits split across two registers for 32 bit */
 340                                 /* process and in one register for 64 bit process */
 341                                 if (IS_64BIT_PROCESS(proc)) {
 342                                         u_int64_t       *retp = (u_int64_t *)&uthread->uu_rval[0];
 343                                         regs->save_r3 = *retp;
 344                                 }
 345                                 else {
 346                                         regs->save_r3 = uthread->uu_rval[0];
 347                                         regs->save_r4 = uthread->uu_rval[1];
 348                                 }
 349                                 break;
 350                         case _SYSCALL_RET_ADDR_T:
 351                         case _SYSCALL_RET_SIZE_T:
 352                         case _SYSCALL_RET_SSIZE_T:
 353                                 /* the variable length return types (user_addr_t, user_ssize_t,
 354                                  * and user_size_t) are always the largest possible size in the
 355                                  * kernel (we use uu_rval[0] and [1] as one 64 bit value).
 356                                  */
 357                                 {
 358                                         u_int64_t       *retp = (u_int64_t *)&uthread->uu_rval[0];
 359                                         regs->save_r3 = *retp;
 360                                 }
 361                                 break;
 362                         case _SYSCALL_RET_NONE:
 363                                 break;
 364                         default:
 365                                 panic("unix_syscall: unknown return type");
 366                                 break;
 367                         }
 368                 }
 369         }
 370         /* else  (error == EJUSTRETURN) { nothing } */
 371
 372
 373         uthread->uu_flag &= ~UT_NOTCANCELPT;
 374
 375         /* panic if funnel is held */
 376         syscall_exit_funnelcheck();
 377
 378         if (uthread->uu_lowpri_window) {
 379                 /*
 380                  * task is marked as a low priority I/O type
 381                  * and the I/O we issued while in this system call
 382                  * collided with normal I/O operations... we'll
 383                  * delay in order to mitigate the impact of this
 384                  * task on the normal operation of the system
 385                  */
 386                 throttle_lowpri_io(TRUE);
 387         }
 388         if (kdebug_enable && (code != 180)) {
 389                 if (callp->sy_return_type == _SYSCALL_RET_SSIZE_T)
 390                         KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_END,
 391                                               error, uthread->uu_rval[1], 0, proc->p_pid, 0);
 392                 else
 393                         KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_END,
 394                                               error, uthread->uu_rval[0], uthread->uu_rval[1], proc->p_pid, 0);
 395         }
 396
 397         thread_exception_return();
 398         /* NOTREACHED */
 399 }
 400
 401 void
 402 munge_lwww(
 403         const void      *in32,
 404         void            *out64)
 405 {
 406         const uint32_t  *arg32;
 407         uint64_t        *arg64;
 408
 409         arg32 = (const uint32_t *) in32;
 410         arg64 = (uint64_t *) out64;
 411
 412         arg64[3] = arg32[9];    /* lwwW */
 413         arg64[2] = arg32[7];    /* lwWw */
 414         arg64[1] = arg32[5];    /* lWww */
 415         arg64[0] = ((uint64_t) arg32[1]) << 32; /* Lwww (hi) */
 416         arg64[0] |= (uint64_t) arg32[3];        /* Lwww (lo) */
 417 }
 418
 419 void
 420 munge_lw(
 421         const void      *in32,
 422         void            *out64)
 423 {
 424         const uint32_t  *arg32;
 425         uint64_t        *arg64;
 426
 427         arg32 = (const uint32_t *) in32;
 428         arg64 = (uint64_t *) out64;
 429
 430         arg64[1] = arg32[5];    /* lW */
 431         arg64[0] = ((uint64_t) arg32[1]) << 32; /* Lw (hi) */
 432         arg64[0] |= (uint64_t) arg32[3];        /* Lw (lo) */
 433 }