bsd/vm/dp_backing_file.c

   1 /*
   2  * Copyright (c) 2000-2010 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 SPARTA, Inc. in 2005 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 <sys/param.h>
  36 #include <sys/systm.h>
  37 #include <sys/lock.h>
  38 #include <sys/proc_internal.h>
  39 #include <sys/kauth.h>
  40 #include <sys/buf.h>
  41 #include <sys/uio.h>
  42 #include <sys/vnode_internal.h>
  43 #include <sys/namei.h>
  44 #include <sys/ubc_internal.h>
  45 #include <sys/malloc.h>
  46 #include <sys/user.h>
  47 #if CONFIG_PROTECT
  48 #include <sys/cprotect.h>
  49 #endif
  50
  51 #include <default_pager/default_pager_types.h>
  52 #include <default_pager/default_pager_object.h>
  53
  54 #include <security/audit/audit.h>
  55 #include <bsm/audit_kevents.h>
  56
  57 #include <mach/mach_types.h>
  58 #include <mach/host_priv.h>
  59 #include <mach/mach_traps.h>
  60 #include <mach/boolean.h>
  61
  62 #include <kern/kern_types.h>
  63 #include <kern/host.h>
  64 #include <kern/task.h>
  65 #include <kern/zalloc.h>
  66 #include <kern/kalloc.h>
  67 #include <kern/assert.h>
  68
  69 #include <libkern/libkern.h>
  70
  71 #include <vm/vm_pageout.h>
  72 #include <vm/vm_map.h>
  73 #include <vm/vm_kern.h>
  74 #include <vm/vnode_pager.h>
  75 #include <vm/vm_protos.h>
  76 #if CONFIG_MACF
  77 #include <security/mac_framework.h>
  78 #endif
  79
  80 /*
  81  * temporary support for delayed instantiation
  82  * of default_pager
  83  */
  84 int default_pager_init_flag = 0;
  85
  86 struct bs_map           bs_port_table[MAX_BACKING_STORE] = {
  87         {0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
  88         {0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
  89         {0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
  90         {0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
  91         {0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
  92         {0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
  93         {0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
  94         {0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
  95         {0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
  96         {0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}};
  97
  98 /* ###################################################### */
  99
 100
 101 /*
 102  *      Routine:        macx_backing_store_recovery
 103  *      Function:
 104  *              Syscall interface to set a tasks privilege
 105  *              level so that it is not subject to
 106  *              macx_backing_store_suspend
 107  */
 108 int
 109 macx_backing_store_recovery(
 110         struct macx_backing_store_recovery_args *args)
 111 {
 112         int             pid = args->pid;
 113         int             error;
 114         struct proc     *p =  current_proc();
 115         boolean_t       funnel_state;
 116
 117         funnel_state = thread_funnel_set(kernel_flock, TRUE);
 118         if ((error = suser(kauth_cred_get(), 0)))
 119                 goto backing_store_recovery_return;
 120
 121         /* for now restrict backing_store_recovery */
 122         /* usage to only present task */
 123         if(pid != proc_selfpid()) {
 124                 error = EINVAL;
 125                 goto backing_store_recovery_return;
 126         }
 127
 128         task_backing_store_privileged(p->task);
 129
 130 backing_store_recovery_return:
 131         (void) thread_funnel_set(kernel_flock, FALSE);
 132         return(error);
 133 }
 134
 135 /*
 136  *      Routine:        macx_backing_store_suspend
 137  *      Function:
 138  *              Syscall interface to stop new demand for
 139  *              backing store when backing store is low
 140  */
 141
 142 int
 143 macx_backing_store_suspend(
 144         struct macx_backing_store_suspend_args *args)
 145 {
 146         boolean_t       suspend = args->suspend;
 147         int             error;
 148         boolean_t       funnel_state;
 149
 150         funnel_state = thread_funnel_set(kernel_flock, TRUE);
 151         if ((error = suser(kauth_cred_get(), 0)))
 152                 goto backing_store_suspend_return;
 153
 154         vm_backing_store_disable(suspend);
 155
 156 backing_store_suspend_return:
 157         (void) thread_funnel_set(kernel_flock, FALSE);
 158         return(error);
 159 }
 160
 161 extern boolean_t backing_store_stop_compaction;
 162
 163 /*
 164  *      Routine:        macx_backing_store_compaction
 165  *      Function:
 166  *              Turn compaction of swap space on or off.  This is
 167  *              used during shutdown/restart so that the kernel
 168  *              doesn't waste time compacting swap files that are
 169  *              about to be deleted anyway.  Compaction is always
 170  *              on by default when the system comes up and is turned
 171  *              off when a shutdown/restart is requested.  It is
 172  *              re-enabled if the shutdown/restart is aborted for any reason.
 173  */
 174
 175 int
 176 macx_backing_store_compaction(int flags)
 177 {
 178         int error;
 179
 180         if ((error = suser(kauth_cred_get(), 0)))
 181                 return error;
 182
 183         if (flags & SWAP_COMPACT_DISABLE) {
 184                 backing_store_stop_compaction = TRUE;
 185
 186         } else if (flags & SWAP_COMPACT_ENABLE) {
 187                 backing_store_stop_compaction = FALSE;
 188         }
 189
 190         return 0;
 191 }
 192
 193 /*
 194  *      Routine:        macx_triggers
 195  *      Function:
 196  *              Syscall interface to set the call backs for low and
 197  *              high water marks.
 198  */
 199 int
 200 macx_triggers(
 201         struct macx_triggers_args *args)
 202 {
 203         int     error;
 204
 205         error = suser(kauth_cred_get(), 0);
 206         if (error)
 207                 return error;
 208
 209         return mach_macx_triggers(args);
 210 }
 211
 212
 213 extern boolean_t dp_isssd;
 214
 215 /*
 216  *      Routine:        macx_swapon
 217  *      Function:
 218  *              Syscall interface to add a file to backing store
 219  */
 220 int
 221 macx_swapon(
 222         struct macx_swapon_args *args)
 223 {
 224         int                     size = args->size;
 225         vnode_t                 vp = (vnode_t)NULL;
 226         struct nameidata        nd, *ndp;
 227         register int            error;
 228         kern_return_t           kr;
 229         mach_port_t             backing_store;
 230         memory_object_default_t default_pager;
 231         int                     i;
 232         boolean_t               funnel_state;
 233         off_t                   file_size;
 234         vfs_context_t           ctx = vfs_context_current();
 235         struct proc             *p =  current_proc();
 236         int                     dp_cluster_size;
 237
 238
 239         AUDIT_MACH_SYSCALL_ENTER(AUE_SWAPON);
 240         AUDIT_ARG(value32, args->priority);
 241
 242         funnel_state = thread_funnel_set(kernel_flock, TRUE);
 243         ndp = &nd;
 244
 245         if ((error = suser(kauth_cred_get(), 0)))
 246                 goto swapon_bailout;
 247
 248         /*
 249          * Get a vnode for the paging area.
 250          */
 251         NDINIT(ndp, LOOKUP, OP_LOOKUP, FOLLOW | LOCKLEAF | AUDITVNPATH1,
 252                ((IS_64BIT_PROCESS(p)) ? UIO_USERSPACE64 : UIO_USERSPACE32),
 253                (user_addr_t) args->filename, ctx);
 254
 255         if ((error = namei(ndp)))
 256                 goto swapon_bailout;
 257         nameidone(ndp);
 258         vp = ndp->ni_vp;
 259
 260         if (vp->v_type != VREG) {
 261                 error = EINVAL;
 262                 goto swapon_bailout;
 263         }
 264
 265         /* get file size */
 266         if ((error = vnode_size(vp, &file_size, ctx)) != 0)
 267                 goto swapon_bailout;
 268 #if CONFIG_MACF
 269         vnode_lock(vp);
 270         error = mac_system_check_swapon(vfs_context_ucred(ctx), vp);
 271         vnode_unlock(vp);
 272         if (error)
 273                 goto swapon_bailout;
 274 #endif
 275
 276         /* resize to desired size if it's too small */
 277         if ((file_size < (off_t)size) && ((error = vnode_setsize(vp, (off_t)size, 0, ctx)) != 0))
 278                 goto swapon_bailout;
 279
 280 #if CONFIG_PROTECT
 281         {
 282                 /* initialize content protection keys manually */
 283                 if ((error = cp_handle_vnop(vp, CP_WRITE_ACCESS, 0)) != 0) {
 284                         goto swapon_bailout;
 285                 }
 286         }
 287 #endif
 288
 289
 290         if (default_pager_init_flag == 0) {
 291                 start_def_pager(NULL);
 292                 default_pager_init_flag = 1;
 293         }
 294
 295         /* add new backing store to list */
 296         i = 0;
 297         while(bs_port_table[i].vp != 0) {
 298                 if(i == MAX_BACKING_STORE)
 299                         break;
 300                 i++;
 301         }
 302         if(i == MAX_BACKING_STORE) {
 303                 error = ENOMEM;
 304                 goto swapon_bailout;
 305         }
 306
 307         /* remember the vnode. This vnode has namei() reference */
 308         bs_port_table[i].vp = vp;
 309
 310         /*
 311          * Look to see if we are already paging to this file.
 312          */
 313         /* make certain the copy send of kernel call will work */
 314         default_pager = MEMORY_OBJECT_DEFAULT_NULL;
 315         kr = host_default_memory_manager(host_priv_self(), &default_pager, 0);
 316         if(kr != KERN_SUCCESS) {
 317            error = EAGAIN;
 318            bs_port_table[i].vp = 0;
 319            goto swapon_bailout;
 320         }
 321
 322 #if CONFIG_EMBEDDED
 323         dp_cluster_size = 1 * PAGE_SIZE;
 324 #else
 325         if ((dp_isssd = vnode_pager_isSSD(vp)) == TRUE) {
 326                 /*
 327                  * keep the cluster size small since the
 328                  * seek cost is effectively 0 which means
 329                  * we don't care much about fragmentation
 330                  */
 331                 dp_cluster_size = 2 * PAGE_SIZE;
 332         } else {
 333                 /*
 334                  * use the default cluster size
 335                  */
 336                 dp_cluster_size = 0;
 337         }
 338 #endif
 339         kr = default_pager_backing_store_create(default_pager,
 340                                         -1, /* default priority */
 341                                         dp_cluster_size,
 342                                         &backing_store);
 343         memory_object_default_deallocate(default_pager);
 344
 345         if(kr != KERN_SUCCESS) {
 346            error = ENOMEM;
 347            bs_port_table[i].vp = 0;
 348            goto swapon_bailout;
 349         }
 350
 351         /* Mark this vnode as being used for swapfile */
 352         vnode_lock_spin(vp);
 353         SET(vp->v_flag, VSWAP);
 354         vnode_unlock(vp);
 355
 356         /*
 357          * NOTE: we are able to supply PAGE_SIZE here instead of
 358          *      an actual record size or block number because:
 359          *      a: we do not support offsets from the beginning of the
 360          *              file (allowing for non page size/record modulo offsets.
 361          *      b: because allow paging will be done modulo page size
 362          */
 363
 364         kr = default_pager_add_file(backing_store, (vnode_ptr_t) vp,
 365                                 PAGE_SIZE, (int)(file_size/PAGE_SIZE));
 366         if(kr != KERN_SUCCESS) {
 367            bs_port_table[i].vp = 0;
 368            if(kr == KERN_INVALID_ARGUMENT)
 369                 error = EINVAL;
 370            else
 371                 error = ENOMEM;
 372
 373            /* This vnode is not to be used for swapfile */
 374            vnode_lock_spin(vp);
 375            CLR(vp->v_flag, VSWAP);
 376            vnode_unlock(vp);
 377
 378            goto swapon_bailout;
 379         }
 380         bs_port_table[i].bs = (void *)backing_store;
 381         error = 0;
 382
 383         ubc_setthreadcred(vp, p, current_thread());
 384
 385         /*
 386          * take a long term reference on the vnode to keep
 387          * vnreclaim() away from this vnode.
 388          */
 389         vnode_ref(vp);
 390
 391 swapon_bailout:
 392         if (vp) {
 393                 vnode_put(vp);
 394         }
 395         (void) thread_funnel_set(kernel_flock, FALSE);
 396         AUDIT_MACH_SYSCALL_EXIT(error);
 397
 398         if (error)
 399                 printf("macx_swapon FAILED - %d\n", error);
 400         else
 401                 printf("macx_swapon SUCCESS\n");
 402
 403         return(error);
 404 }
 405
 406 /*
 407  *      Routine:        macx_swapoff
 408  *      Function:
 409  *              Syscall interface to remove a file from backing store
 410  */
 411 int
 412 macx_swapoff(
 413         struct macx_swapoff_args *args)
 414 {
 415         __unused int    flags = args->flags;
 416         kern_return_t   kr;
 417         mach_port_t     backing_store;
 418
 419         struct vnode            *vp = 0;
 420         struct nameidata        nd, *ndp;
 421         struct proc             *p =  current_proc();
 422         int                     i;
 423         int                     error;
 424         boolean_t               funnel_state;
 425         vfs_context_t ctx = vfs_context_current();
 426         struct uthread  *ut;
 427         int                     orig_iopol_disk;
 428
 429         AUDIT_MACH_SYSCALL_ENTER(AUE_SWAPOFF);
 430
 431         funnel_state = thread_funnel_set(kernel_flock, TRUE);
 432         backing_store = NULL;
 433         ndp = &nd;
 434
 435         if ((error = suser(kauth_cred_get(), 0)))
 436                 goto swapoff_bailout;
 437
 438         /*
 439          * Get the vnode for the paging area.
 440          */
 441         NDINIT(ndp, LOOKUP, OP_LOOKUP, FOLLOW | LOCKLEAF | AUDITVNPATH1,
 442                ((IS_64BIT_PROCESS(p)) ? UIO_USERSPACE64 : UIO_USERSPACE32),
 443                (user_addr_t) args->filename, ctx);
 444
 445         if ((error = namei(ndp)))
 446                 goto swapoff_bailout;
 447         nameidone(ndp);
 448         vp = ndp->ni_vp;
 449
 450         if (vp->v_type != VREG) {
 451                 error = EINVAL;
 452                 goto swapoff_bailout;
 453         }
 454 #if CONFIG_MACF
 455         vnode_lock(vp);
 456         error = mac_system_check_swapoff(vfs_context_ucred(ctx), vp);
 457         vnode_unlock(vp);
 458         if (error)
 459                 goto swapoff_bailout;
 460 #endif
 461
 462         for(i = 0; i < MAX_BACKING_STORE; i++) {
 463                 if(bs_port_table[i].vp == vp) {
 464                         break;
 465                 }
 466         }
 467         if (i == MAX_BACKING_STORE) {
 468                 error = EINVAL;
 469                 goto swapoff_bailout;
 470         }
 471         backing_store = (mach_port_t)bs_port_table[i].bs;
 472
 473         ut = get_bsdthread_info(current_thread());
 474
 475         orig_iopol_disk = proc_get_thread_selfdiskacc();
 476         proc_apply_thread_selfdiskacc(IOPOL_THROTTLE);
 477
 478         kr = default_pager_backing_store_delete(backing_store);
 479
 480         proc_apply_thread_selfdiskacc(orig_iopol_disk);
 481
 482         switch (kr) {
 483                 case KERN_SUCCESS:
 484                         error = 0;
 485                         bs_port_table[i].vp = 0;
 486                         /* This vnode is no longer used for swapfile */
 487                         vnode_lock_spin(vp);
 488                         CLR(vp->v_flag, VSWAP);
 489                         vnode_unlock(vp);
 490
 491                         /* get rid of macx_swapon() "long term" reference */
 492                         vnode_rele(vp);
 493
 494                         break;
 495                 case KERN_FAILURE:
 496                         error = EAGAIN;
 497                         break;
 498                 default:
 499                         error = EAGAIN;
 500                         break;
 501         }
 502
 503 swapoff_bailout:
 504         /* get rid of macx_swapoff() namei() reference */
 505         if (vp)
 506                 vnode_put(vp);
 507
 508         (void) thread_funnel_set(kernel_flock, FALSE);
 509         AUDIT_MACH_SYSCALL_EXIT(error);
 510
 511         if (error)
 512                 printf("macx_swapoff FAILED - %d\n", error);
 513         else
 514                 printf("macx_swapoff SUCCESS\n");
 515
 516         return(error);
 517 }
 518
 519 /*
 520  *      Routine:        macx_swapinfo
 521  *      Function:
 522  *              Syscall interface to get general swap statistics
 523  */
 524 int
 525 macx_swapinfo(
 526         memory_object_size_t    *total_p,
 527         memory_object_size_t    *avail_p,
 528         vm_size_t               *pagesize_p,
 529         boolean_t               *encrypted_p)
 530 {
 531         int                     error;
 532         memory_object_default_t default_pager;
 533         default_pager_info_64_t dpi64;
 534         kern_return_t           kr;
 535
 536         error = 0;
 537
 538         /*
 539          * Get a handle on the default pager.
 540          */
 541         default_pager = MEMORY_OBJECT_DEFAULT_NULL;
 542         kr = host_default_memory_manager(host_priv_self(), &default_pager, 0);
 543         if (kr != KERN_SUCCESS) {
 544                 error = EAGAIN; /* XXX why EAGAIN ? */
 545                 goto done;
 546         }
 547         if (default_pager == MEMORY_OBJECT_DEFAULT_NULL) {
 548                 /*
 549                  * The default pager has not initialized yet,
 550                  * so it can't be using any swap space at all.
 551                  */
 552                 *total_p = 0;
 553                 *avail_p = 0;
 554                 *pagesize_p = 0;
 555                 *encrypted_p = FALSE;
 556                 goto done;
 557         }
 558
 559         /*
 560          * Get swap usage data from default pager.
 561          */
 562         kr = default_pager_info_64(default_pager, &dpi64);
 563         if (kr != KERN_SUCCESS) {
 564                 error = ENOTSUP;
 565                 goto done;
 566         }
 567
 568         /*
 569          * Provide default pager info to caller.
 570          */
 571         *total_p = dpi64.dpi_total_space;
 572         *avail_p = dpi64.dpi_free_space;
 573         *pagesize_p = dpi64.dpi_page_size;
 574         if (dpi64.dpi_flags & DPI_ENCRYPTED) {
 575                 *encrypted_p = TRUE;
 576         } else {
 577                 *encrypted_p = FALSE;
 578         }
 579
 580 done:
 581         if (default_pager != MEMORY_OBJECT_DEFAULT_NULL) {
 582                 /* release our handle on default pager */
 583                 memory_object_default_deallocate(default_pager);
 584         }
 585         return error;
 586 }