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 void kprintf(const char *fmt, ...);
  81
  82 /*
  83  * temporary support for delayed instantiation
  84  * of default_pager
  85  */
  86 int default_pager_init_flag = 0;
  87
  88 struct bs_map           bs_port_table[MAX_BACKING_STORE] = {
  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         {0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},
  98         {0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0},{0,0}};
  99
 100 /* ###################################################### */
 101
 102
 103 /*
 104  *      Routine:        macx_backing_store_recovery
 105  *      Function:
 106  *              Syscall interface to set a tasks privilege
 107  *              level so that it is not subject to
 108  *              macx_backing_store_suspend
 109  */
 110 int
 111 macx_backing_store_recovery(
 112         struct macx_backing_store_recovery_args *args)
 113 {
 114         int             pid = args->pid;
 115         int             error;
 116         struct proc     *p =  current_proc();
 117         boolean_t       funnel_state;
 118
 119         funnel_state = thread_funnel_set(kernel_flock, TRUE);
 120         if ((error = suser(kauth_cred_get(), 0)))
 121                 goto backing_store_recovery_return;
 122
 123         /* for now restrict backing_store_recovery */
 124         /* usage to only present task */
 125         if(pid != proc_selfpid()) {
 126                 error = EINVAL;
 127                 goto backing_store_recovery_return;
 128         }
 129
 130         task_backing_store_privileged(p->task);
 131
 132 backing_store_recovery_return:
 133         (void) thread_funnel_set(kernel_flock, FALSE);
 134         return(error);
 135 }
 136
 137 /*
 138  *      Routine:        macx_backing_store_suspend
 139  *      Function:
 140  *              Syscall interface to stop new demand for
 141  *              backing store when backing store is low
 142  */
 143
 144 int
 145 macx_backing_store_suspend(
 146         struct macx_backing_store_suspend_args *args)
 147 {
 148         boolean_t       suspend = args->suspend;
 149         int             error;
 150         boolean_t       funnel_state;
 151
 152         funnel_state = thread_funnel_set(kernel_flock, TRUE);
 153         if ((error = suser(kauth_cred_get(), 0)))
 154                 goto backing_store_suspend_return;
 155
 156         vm_backing_store_disable(suspend);
 157
 158 backing_store_suspend_return:
 159         (void) thread_funnel_set(kernel_flock, FALSE);
 160         return(error);
 161 }
 162
 163 extern boolean_t backing_store_stop_compaction;
 164 extern boolean_t compressor_store_stop_compaction;
 165
 166 /*
 167  *      Routine:        macx_backing_store_compaction
 168  *      Function:
 169  *              Turn compaction of swap space on or off.  This is
 170  *              used during shutdown/restart so that the kernel
 171  *              doesn't waste time compacting swap files that are
 172  *              about to be deleted anyway.  Compaction is always
 173  *              on by default when the system comes up and is turned
 174  *              off when a shutdown/restart is requested.  It is
 175  *              re-enabled if the shutdown/restart is aborted for any reason.
 176  */
 177
 178 int
 179 macx_backing_store_compaction(int flags)
 180 {
 181         int error;
 182
 183         if ((error = suser(kauth_cred_get(), 0)))
 184                 return error;
 185
 186         if (flags & SWAP_COMPACT_DISABLE) {
 187                 backing_store_stop_compaction = TRUE;
 188                 compressor_store_stop_compaction = TRUE;
 189
 190                 kprintf("backing_store_stop_compaction = TRUE\n");
 191
 192         } else if (flags & SWAP_COMPACT_ENABLE) {
 193                 backing_store_stop_compaction = FALSE;
 194                 compressor_store_stop_compaction = FALSE;
 195
 196                 kprintf("backing_store_stop_compaction = FALSE\n");
 197         }
 198
 199         return 0;
 200 }
 201
 202 /*
 203  *      Routine:        macx_triggers
 204  *      Function:
 205  *              Syscall interface to set the call backs for low and
 206  *              high water marks.
 207  */
 208 int
 209 macx_triggers(
 210         struct macx_triggers_args *args)
 211 {
 212         int     error;
 213
 214         error = suser(kauth_cred_get(), 0);
 215         if (error)
 216                 return error;
 217
 218         return mach_macx_triggers(args);
 219 }
 220
 221
 222 extern boolean_t dp_isssd;
 223 extern void vm_swap_init(void);
 224 extern int vm_compressor_mode;
 225
 226 /*
 227  * In the compressed pager world, the swapfiles are created by the kernel.
 228  * Well, all except the first one. That swapfile is absorbed by the kernel at
 229  * the end of the macx_swapon function (if swap is enabled). That's why
 230  * we allow the first invocation of macx_swapon to succeed.
 231  *
 232  * If the compressor pool is running low, the kernel messages the dynamic pager
 233  * on the port it has registered with the kernel. That port can transport 1 of 2
 234  * pieces of information to dynamic pager: create a swapfile or delete a swapfile.
 235  *
 236  * We choose to transmit the former. So, that message tells dynamic pager
 237  * to create a swapfile and activate it by calling macx_swapon.
 238  *
 239  * We deny this new macx_swapon request. That leads dynamic pager to interpret the
 240  * failure as a serious error and notify all it's clients that swap is running low.
 241  * That's how we get the loginwindow "Resume / Force Quit Applications" dialog to appear.
 242  *
 243  * NOTE:
 244  * If the kernel has already created multiple swapfiles by the time the compressor
 245  * pool is running low (and it has to play this trick), dynamic pager won't be able to
 246  * create a file in user-space and, that too will lead to a similar notification blast
 247  * to all of it's clients. So, that behaves as desired too.
 248  */
 249 boolean_t       macx_swapon_allowed = TRUE;
 250
 251 /*
 252  *      Routine:        macx_swapon
 253  *      Function:
 254  *              Syscall interface to add a file to backing store
 255  */
 256 int
 257 macx_swapon(
 258         struct macx_swapon_args *args)
 259 {
 260         int                     size = args->size;
 261         vnode_t                 vp = (vnode_t)NULL;
 262         struct nameidata        nd, *ndp;
 263         register int            error;
 264         kern_return_t           kr;
 265         mach_port_t             backing_store;
 266         memory_object_default_t default_pager;
 267         int                     i;
 268         boolean_t               funnel_state;
 269         off_t                   file_size;
 270         vfs_context_t           ctx = vfs_context_current();
 271         struct proc             *p =  current_proc();
 272         int                     dp_cluster_size;
 273
 274         if (COMPRESSED_PAGER_IS_ACTIVE) {
 275                 if (macx_swapon_allowed == FALSE) {
 276                         return EINVAL;
 277                 } else {
 278                         if ((vm_compressor_mode == VM_PAGER_COMPRESSOR_WITH_SWAP) ||
 279                             (vm_compressor_mode == VM_PAGER_FREEZER_COMPRESSOR_WITH_SWAP)) {
 280                                 vm_swap_init();
 281                         }
 282
 283                         macx_swapon_allowed = FALSE;
 284                         return 0;
 285                 }
 286         }
 287
 288         AUDIT_MACH_SYSCALL_ENTER(AUE_SWAPON);
 289         AUDIT_ARG(value32, args->priority);
 290
 291         funnel_state = thread_funnel_set(kernel_flock, TRUE);
 292         ndp = &nd;
 293
 294         if ((error = suser(kauth_cred_get(), 0)))
 295                 goto swapon_bailout;
 296
 297         /*
 298          * Get a vnode for the paging area.
 299          */
 300         NDINIT(ndp, LOOKUP, OP_LOOKUP, FOLLOW | LOCKLEAF | AUDITVNPATH1,
 301                ((IS_64BIT_PROCESS(p)) ? UIO_USERSPACE64 : UIO_USERSPACE32),
 302                (user_addr_t) args->filename, ctx);
 303
 304         if ((error = namei(ndp)))
 305                 goto swapon_bailout;
 306         nameidone(ndp);
 307         vp = ndp->ni_vp;
 308
 309         if (vp->v_type != VREG) {
 310                 error = EINVAL;
 311                 goto swapon_bailout;
 312         }
 313
 314         /* get file size */
 315         if ((error = vnode_size(vp, &file_size, ctx)) != 0)
 316                 goto swapon_bailout;
 317 #if CONFIG_MACF
 318         vnode_lock(vp);
 319         error = mac_system_check_swapon(vfs_context_ucred(ctx), vp);
 320         vnode_unlock(vp);
 321         if (error)
 322                 goto swapon_bailout;
 323 #endif
 324
 325         /* resize to desired size if it's too small */
 326         if ((file_size < (off_t)size) && ((error = vnode_setsize(vp, (off_t)size, 0, ctx)) != 0))
 327                 goto swapon_bailout;
 328
 329 #if CONFIG_PROTECT
 330         {
 331                 /* initialize content protection keys manually */
 332                 if ((error = cp_handle_vnop(vp, CP_WRITE_ACCESS, 0)) != 0) {
 333                         goto swapon_bailout;
 334                 }
 335         }
 336 #endif
 337
 338
 339         if (default_pager_init_flag == 0) {
 340                 start_def_pager(NULL);
 341                 default_pager_init_flag = 1;
 342         }
 343
 344         /* add new backing store to list */
 345         i = 0;
 346         while(bs_port_table[i].vp != 0) {
 347                 if(i == MAX_BACKING_STORE)
 348                         break;
 349                 i++;
 350         }
 351         if(i == MAX_BACKING_STORE) {
 352                 error = ENOMEM;
 353                 goto swapon_bailout;
 354         }
 355
 356         /* remember the vnode. This vnode has namei() reference */
 357         bs_port_table[i].vp = vp;
 358
 359         /*
 360          * Look to see if we are already paging to this file.
 361          */
 362         /* make certain the copy send of kernel call will work */
 363         default_pager = MEMORY_OBJECT_DEFAULT_NULL;
 364         kr = host_default_memory_manager(host_priv_self(), &default_pager, 0);
 365         if(kr != KERN_SUCCESS) {
 366            error = EAGAIN;
 367            bs_port_table[i].vp = 0;
 368            goto swapon_bailout;
 369         }
 370
 371         if ((dp_isssd = vnode_pager_isSSD(vp)) == TRUE) {
 372                 /*
 373                  * keep the cluster size small since the
 374                  * seek cost is effectively 0 which means
 375                  * we don't care much about fragmentation
 376                  */
 377                 dp_cluster_size = 2 * PAGE_SIZE;
 378         } else {
 379                 /*
 380                  * use the default cluster size
 381                  */
 382                 dp_cluster_size = 0;
 383         }
 384         kr = default_pager_backing_store_create(default_pager,
 385                                         -1, /* default priority */
 386                                         dp_cluster_size,
 387                                         &backing_store);
 388         memory_object_default_deallocate(default_pager);
 389
 390         if(kr != KERN_SUCCESS) {
 391            error = ENOMEM;
 392            bs_port_table[i].vp = 0;
 393            goto swapon_bailout;
 394         }
 395
 396         /* Mark this vnode as being used for swapfile */
 397         vnode_lock_spin(vp);
 398         SET(vp->v_flag, VSWAP);
 399         vnode_unlock(vp);
 400
 401         /*
 402          * NOTE: we are able to supply PAGE_SIZE here instead of
 403          *      an actual record size or block number because:
 404          *      a: we do not support offsets from the beginning of the
 405          *              file (allowing for non page size/record modulo offsets.
 406          *      b: because allow paging will be done modulo page size
 407          */
 408
 409         kr = default_pager_add_file(backing_store, (vnode_ptr_t) vp,
 410                                 PAGE_SIZE, (int)(file_size/PAGE_SIZE));
 411         if(kr != KERN_SUCCESS) {
 412            bs_port_table[i].vp = 0;
 413            if(kr == KERN_INVALID_ARGUMENT)
 414                 error = EINVAL;
 415            else
 416                 error = ENOMEM;
 417
 418            /* This vnode is not to be used for swapfile */
 419            vnode_lock_spin(vp);
 420            CLR(vp->v_flag, VSWAP);
 421            vnode_unlock(vp);
 422
 423            goto swapon_bailout;
 424         }
 425         bs_port_table[i].bs = (void *)backing_store;
 426         error = 0;
 427
 428         ubc_setthreadcred(vp, p, current_thread());
 429
 430         /*
 431          * take a long term reference on the vnode to keep
 432          * vnreclaim() away from this vnode.
 433          */
 434         vnode_ref(vp);
 435
 436 swapon_bailout:
 437         if (vp) {
 438                 vnode_put(vp);
 439         }
 440         (void) thread_funnel_set(kernel_flock, FALSE);
 441         AUDIT_MACH_SYSCALL_EXIT(error);
 442
 443         if (error)
 444                 printf("macx_swapon FAILED - %d\n", error);
 445         else
 446                 printf("macx_swapon SUCCESS\n");
 447
 448         return(error);
 449 }
 450
 451 /*
 452  *      Routine:        macx_swapoff
 453  *      Function:
 454  *              Syscall interface to remove a file from backing store
 455  */
 456 int
 457 macx_swapoff(
 458         struct macx_swapoff_args *args)
 459 {
 460         __unused int    flags = args->flags;
 461         kern_return_t   kr;
 462         mach_port_t     backing_store;
 463
 464         struct vnode            *vp = 0;
 465         struct nameidata        nd, *ndp;
 466         struct proc             *p =  current_proc();
 467         int                     i;
 468         int                     error;
 469         boolean_t               funnel_state;
 470         vfs_context_t ctx = vfs_context_current();
 471         int                     orig_iopol_disk;
 472
 473         AUDIT_MACH_SYSCALL_ENTER(AUE_SWAPOFF);
 474
 475         funnel_state = thread_funnel_set(kernel_flock, TRUE);
 476         backing_store = NULL;
 477         ndp = &nd;
 478
 479         if ((error = suser(kauth_cred_get(), 0)))
 480                 goto swapoff_bailout;
 481
 482         /*
 483          * Get the vnode for the paging area.
 484          */
 485         NDINIT(ndp, LOOKUP, OP_LOOKUP, FOLLOW | LOCKLEAF | AUDITVNPATH1,
 486                ((IS_64BIT_PROCESS(p)) ? UIO_USERSPACE64 : UIO_USERSPACE32),
 487                (user_addr_t) args->filename, ctx);
 488
 489         if ((error = namei(ndp)))
 490                 goto swapoff_bailout;
 491         nameidone(ndp);
 492         vp = ndp->ni_vp;
 493
 494         if (vp->v_type != VREG) {
 495                 error = EINVAL;
 496                 goto swapoff_bailout;
 497         }
 498 #if CONFIG_MACF
 499         vnode_lock(vp);
 500         error = mac_system_check_swapoff(vfs_context_ucred(ctx), vp);
 501         vnode_unlock(vp);
 502         if (error)
 503                 goto swapoff_bailout;
 504 #endif
 505
 506         for(i = 0; i < MAX_BACKING_STORE; i++) {
 507                 if(bs_port_table[i].vp == vp) {
 508                         break;
 509                 }
 510         }
 511         if (i == MAX_BACKING_STORE) {
 512                 error = EINVAL;
 513                 goto swapoff_bailout;
 514         }
 515         backing_store = (mach_port_t)bs_port_table[i].bs;
 516
 517         orig_iopol_disk = proc_get_task_policy(current_task(), current_thread(),
 518                                                TASK_POLICY_INTERNAL, TASK_POLICY_IOPOL);
 519
 520         proc_set_task_policy(current_task(), current_thread(), TASK_POLICY_INTERNAL,
 521                              TASK_POLICY_IOPOL, IOPOL_THROTTLE);
 522
 523         kr = default_pager_backing_store_delete(backing_store);
 524
 525         proc_set_task_policy(current_task(), current_thread(), TASK_POLICY_INTERNAL,
 526                              TASK_POLICY_IOPOL, orig_iopol_disk);
 527
 528         switch (kr) {
 529                 case KERN_SUCCESS:
 530                         error = 0;
 531                         bs_port_table[i].vp = 0;
 532                         /* This vnode is no longer used for swapfile */
 533                         vnode_lock_spin(vp);
 534                         CLR(vp->v_flag, VSWAP);
 535                         vnode_unlock(vp);
 536
 537                         /* get rid of macx_swapon() "long term" reference */
 538                         vnode_rele(vp);
 539
 540                         break;
 541                 case KERN_FAILURE:
 542                         error = EAGAIN;
 543                         break;
 544                 default:
 545                         error = EAGAIN;
 546                         break;
 547         }
 548
 549 swapoff_bailout:
 550         /* get rid of macx_swapoff() namei() reference */
 551         if (vp)
 552                 vnode_put(vp);
 553
 554         (void) thread_funnel_set(kernel_flock, FALSE);
 555         AUDIT_MACH_SYSCALL_EXIT(error);
 556
 557         if (error)
 558                 printf("macx_swapoff FAILED - %d\n", error);
 559         else
 560                 printf("macx_swapoff SUCCESS\n");
 561
 562         return(error);
 563 }
 564
 565 /*
 566  *      Routine:        macx_swapinfo
 567  *      Function:
 568  *              Syscall interface to get general swap statistics
 569  */
 570 extern uint64_t vm_swap_get_total_space(void);
 571 extern uint64_t vm_swap_get_used_space(void);
 572 extern uint64_t vm_swap_get_free_space(void);
 573 extern boolean_t vm_swap_up;
 574
 575 int
 576 macx_swapinfo(
 577         memory_object_size_t    *total_p,
 578         memory_object_size_t    *avail_p,
 579         vm_size_t               *pagesize_p,
 580         boolean_t               *encrypted_p)
 581 {
 582         int                     error;
 583         memory_object_default_t default_pager;
 584         default_pager_info_64_t dpi64;
 585         kern_return_t           kr;
 586
 587         error = 0;
 588         if (COMPRESSED_PAGER_IS_ACTIVE) {
 589
 590                 if (vm_swap_up == TRUE) {
 591
 592                         *total_p = vm_swap_get_total_space();
 593                         *avail_p = vm_swap_get_free_space();
 594                         *pagesize_p = PAGE_SIZE_64;
 595                         *encrypted_p = TRUE;
 596
 597                 } else {
 598
 599                         *total_p = 0;
 600                         *avail_p = 0;
 601                         *pagesize_p = 0;
 602                         *encrypted_p = FALSE;
 603                 }
 604         } else {
 605
 606                 /*
 607                  * Get a handle on the default pager.
 608                  */
 609                 default_pager = MEMORY_OBJECT_DEFAULT_NULL;
 610                 kr = host_default_memory_manager(host_priv_self(), &default_pager, 0);
 611                 if (kr != KERN_SUCCESS) {
 612                         error = EAGAIN; /* XXX why EAGAIN ? */
 613                         goto done;
 614                 }
 615                 if (default_pager == MEMORY_OBJECT_DEFAULT_NULL) {
 616                         /*
 617                          * The default pager has not initialized yet,
 618                          * so it can't be using any swap space at all.
 619                          */
 620                         *total_p = 0;
 621                         *avail_p = 0;
 622                         *pagesize_p = 0;
 623                         *encrypted_p = FALSE;
 624                         goto done;
 625                 }
 626
 627                 /*
 628                  * Get swap usage data from default pager.
 629                  */
 630                 kr = default_pager_info_64(default_pager, &dpi64);
 631                 if (kr != KERN_SUCCESS) {
 632                         error = ENOTSUP;
 633                         goto done;
 634                 }
 635
 636                 /*
 637                  * Provide default pager info to caller.
 638                  */
 639                 *total_p = dpi64.dpi_total_space;
 640                 *avail_p = dpi64.dpi_free_space;
 641                 *pagesize_p = dpi64.dpi_page_size;
 642                 if (dpi64.dpi_flags & DPI_ENCRYPTED) {
 643                         *encrypted_p = TRUE;
 644                 } else {
 645                         *encrypted_p = FALSE;
 646                 }
 647
 648 done:
 649                 if (default_pager != MEMORY_OBJECT_DEFAULT_NULL) {
 650                         /* release our handle on default pager */
 651                         memory_object_default_deallocate(default_pager);
 652                 }
 653         }
 654         return error;
 655 }