1 /*- 2 * Copyright (c) 2004-2009 University of Zagreb 3 * Copyright (c) 2006-2009 FreeBSD Foundation 4 * All rights reserved. 5 * 6 * This software was developed by the University of Zagreb and the 7 * FreeBSD Foundation under sponsorship by the Stichting NLnet and the 8 * FreeBSD Foundation. 9 * 10 * Copyright (c) 2009 Jeffrey Roberson <jeff@freebsd.org> 11 * Copyright (c) 2009 Robert N. M. Watson 12 * All rights reserved. 13 * 14 * Redistribution and use in source and binary forms, with or without 15 * modification, are permitted provided that the following conditions 16 * are met: 17 * 1. Redistributions of source code must retain the above copyright 18 * notice, this list of conditions and the following disclaimer. 19 * 2. Redistributions in binary form must reproduce the above copyright 20 * notice, this list of conditions and the following disclaimer in the 21 * documentation and/or other materials provided with the distribution. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 26 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 33 * SUCH DAMAGE. 34 */ 35 36 #include <sys/cdefs.h> 37 __FBSDID("$FreeBSD$"); 38 39 #include "opt_ddb.h" 40 #include "opt_kdb.h" 41 #include "opt_kdtrace.h" 42 43 #include <sys/param.h> 44 #include <sys/kdb.h> 45 #include <sys/kernel.h> 46 #include <sys/jail.h> 47 #include <sys/sdt.h> 48 #include <sys/systm.h> 49 #include <sys/sysctl.h> 50 #include <sys/eventhandler.h> 51 #include <sys/linker_set.h> 52 #include <sys/lock.h> 53 #include <sys/malloc.h> 54 #include <sys/proc.h> 55 #include <sys/socket.h> 56 #include <sys/sx.h> 57 #include <sys/sysctl.h> 58 59 #include <machine/stdarg.h> 60 61 #ifdef DDB 62 #include <ddb/ddb.h> 63 #include <ddb/db_sym.h> 64 #endif 65 66 #include <net/if.h> 67 #include <net/if_var.h> 68 #include <net/vnet.h> 69 70 /*- 71 * This file implements core functions for virtual network stacks: 72 * 73 * - Virtual network stack management functions. 74 * 75 * - Virtual network stack memory allocator, which virtualizes global 76 * variables in the network stack 77 * 78 * - Virtualized SYSINIT's/SYSUNINIT's, which allow network stack subsystems 79 * to register startup/shutdown events to be run for each virtual network 80 * stack instance. 81 */ 82 83 MALLOC_DEFINE(M_VNET, "vnet", "network stack control block"); 84 85 /* 86 * The virtual network stack list has two read-write locks, one sleepable and 87 * the other not, so that the list can be stablized and walked in a variety 88 * of network stack contexts. Both must be acquired exclusively to modify 89 * the list, but a read lock of either lock is sufficient to walk the list. 90 */ 91 struct rwlock vnet_rwlock; 92 struct sx vnet_sxlock; 93 94 #define VNET_LIST_WLOCK() do { \ 95 sx_xlock(&vnet_sxlock); \ 96 rw_wlock(&vnet_rwlock); \ 97 } while (0) 98 99 #define VNET_LIST_WUNLOCK() do { \ 100 rw_wunlock(&vnet_rwlock); \ 101 sx_xunlock(&vnet_sxlock); \ 102 } while (0) 103 104 struct vnet_list_head vnet_head; 105 struct vnet *vnet0; 106 107 /* 108 * The virtual network stack allocator provides storage for virtualized 109 * global variables. These variables are defined/declared using the 110 * VNET_DEFINE()/VNET_DECLARE() macros, which place them in the 'set_vnet' 111 * linker set. The details of the implementation are somewhat subtle, but 112 * allow the majority of most network subsystems to maintain 113 * virtualization-agnostic. 114 * 115 * The virtual network stack allocator handles variables in the base kernel 116 * vs. modules in similar but different ways. In both cases, virtualized 117 * global variables are marked as such by being declared to be part of the 118 * vnet linker set. These "master" copies of global variables serve two 119 * functions: 120 * 121 * (1) They contain static initialization or "default" values for global 122 * variables which will be propagated to each virtual network stack 123 * instance when created. As with normal global variables, they default 124 * to zero-filled. 125 * 126 * (2) They act as unique global names by which the variable can be referred 127 * to, regardless of network stack instance. The single global symbol 128 * will be used to calculate the location of a per-virtual instance 129 * variable at run-time. 130 * 131 * Each virtual network stack instance has a complete copy of each 132 * virtualized global variable, stored in a malloc'd block of memory 133 * referred to by vnet->vnet_data_mem. Critical to the design is that each 134 * per-instance memory block is laid out identically to the master block so 135 * that the offset of each global variable is the same across all blocks. To 136 * optimize run-time access, a precalculated 'base' address, 137 * vnet->vnet_data_base, is stored in each vnet, and is the amount that can 138 * be added to the address of a 'master' instance of a variable to get to the 139 * per-vnet instance. 140 * 141 * Virtualized global variables are handled in a similar manner, but as each 142 * module has its own 'set_vnet' linker set, and we want to keep all 143 * virtualized globals togther, we reserve space in the kernel's linker set 144 * for potential module variables using a per-vnet character array, 145 * 'modspace'. The virtual network stack allocator maintains a free list to 146 * track what space in the array is free (all, initially) and as modules are 147 * linked, allocates portions of the space to specific globals. The kernel 148 * module linker queries the virtual network stack allocator and will 149 * bind references of the global to the location during linking. It also 150 * calls into the virtual network stack allocator, once the memory is 151 * initialized, in order to propagate the new static initializations to all 152 * existing virtual network stack instances so that the soon-to-be executing 153 * module will find every network stack instance with proper default values. 154 */ 155 156 /* 157 * Number of bytes of data in the 'set_vnet' linker set, and hence the total 158 * size of all kernel virtualized global variables, and the malloc(9) type 159 * that will be used to allocate it. 160 */ 161 #define VNET_BYTES (VNET_STOP - VNET_START) 162 163 MALLOC_DEFINE(M_VNET_DATA, "vnet_data", "VNET data"); 164 165 /* 166 * VNET_MODMIN is the minimum number of bytes we will reserve for the sum of 167 * global variables across all loaded modules. As this actually sizes an 168 * array declared as a virtualized global variable in the kernel itself, and 169 * we want the virtualized global variable space to be page-sized, we may 170 * have more space than that in practice. 171 */ 172 #define VNET_MODMIN 8192 173 #define VNET_SIZE roundup2(VNET_BYTES, PAGE_SIZE) 174 #define VNET_MODSIZE (VNET_SIZE - (VNET_BYTES - VNET_MODMIN)) 175 176 /* 177 * Space to store virtualized global variables from loadable kernel modules, 178 * and the free list to manage it. 179 */ 180 static VNET_DEFINE(char, modspace[VNET_MODMIN]); 181 182 /* 183 * Global lists of subsystem constructor and destructors for vnets. They are 184 * registered via VNET_SYSINIT() and VNET_SYSUNINIT(). Both lists are 185 * protected by the vnet_sysinit_sxlock global lock. 186 */ 187 static TAILQ_HEAD(vnet_sysinit_head, vnet_sysinit) vnet_constructors = 188 TAILQ_HEAD_INITIALIZER(vnet_constructors); 189 static TAILQ_HEAD(vnet_sysuninit_head, vnet_sysinit) vnet_destructors = 190 TAILQ_HEAD_INITIALIZER(vnet_destructors); 191 192 struct sx vnet_sysinit_sxlock; 193 194 #define VNET_SYSINIT_WLOCK() sx_xlock(&vnet_sysinit_sxlock); 195 #define VNET_SYSINIT_WUNLOCK() sx_xunlock(&vnet_sysinit_sxlock); 196 #define VNET_SYSINIT_RLOCK() sx_slock(&vnet_sysinit_sxlock); 197 #define VNET_SYSINIT_RUNLOCK() sx_sunlock(&vnet_sysinit_sxlock); 198 199 struct vnet_data_free { 200 uintptr_t vnd_start; 201 int vnd_len; 202 TAILQ_ENTRY(vnet_data_free) vnd_link; 203 }; 204 205 MALLOC_DEFINE(M_VNET_DATA_FREE, "vnet_data_free", "VNET resource accounting"); 206 static TAILQ_HEAD(, vnet_data_free) vnet_data_free_head = 207 TAILQ_HEAD_INITIALIZER(vnet_data_free_head); 208 static struct sx vnet_data_free_lock; 209 210 SDT_PROVIDER_DEFINE(vnet); 211 SDT_PROBE_DEFINE1(vnet, functions, vnet_alloc, entry, entry, "int"); 212 SDT_PROBE_DEFINE2(vnet, functions, vnet_alloc, alloc, alloc, "int", 213 "struct vnet *"); 214 SDT_PROBE_DEFINE2(vnet, functions, vnet_alloc, return, return, 215 "int", "struct vnet *"); 216 SDT_PROBE_DEFINE2(vnet, functions, vnet_destroy, entry, entry, 217 "int", "struct vnet *"); 218 SDT_PROBE_DEFINE1(vnet, functions, vnet_destroy, return, entry, 219 "int"); 220 221 #ifdef DDB 222 static void db_show_vnet_print_vs(struct vnet_sysinit *, int); 223 #endif 224 225 /* 226 * Allocate a virtual network stack. 227 */ 228 struct vnet * 229 vnet_alloc(void) 230 { 231 struct vnet *vnet; 232 233 SDT_PROBE1(vnet, functions, vnet_alloc, entry, __LINE__); 234 vnet = malloc(sizeof(struct vnet), M_VNET, M_WAITOK | M_ZERO); 235 vnet->vnet_magic_n = VNET_MAGIC_N; 236 SDT_PROBE2(vnet, functions, vnet_alloc, alloc, __LINE__, vnet); 237 238 /* 239 * Allocate storage for virtualized global variables and copy in 240 * initial values form our 'master' copy. 241 */ 242 vnet->vnet_data_mem = malloc(VNET_SIZE, M_VNET_DATA, M_WAITOK); 243 memcpy(vnet->vnet_data_mem, (void *)VNET_START, VNET_BYTES); 244 245 /* 246 * All use of vnet-specific data will immediately subtract VNET_START 247 * from the base memory pointer, so pre-calculate that now to avoid 248 * it on each use. 249 */ 250 vnet->vnet_data_base = (uintptr_t)vnet->vnet_data_mem - VNET_START; 251 252 /* Initialize / attach vnet module instances. */ 253 CURVNET_SET_QUIET(vnet); 254 vnet_sysinit(); 255 CURVNET_RESTORE(); 256 257 VNET_LIST_WLOCK(); 258 LIST_INSERT_HEAD(&vnet_head, vnet, vnet_le); 259 VNET_LIST_WUNLOCK(); 260 261 SDT_PROBE2(vnet, functions, vnet_alloc, return, __LINE__, vnet); 262 return (vnet); 263 } 264 265 /* 266 * Destroy a virtual network stack. 267 */ 268 void 269 vnet_destroy(struct vnet *vnet) 270 { 271 struct ifnet *ifp, *nifp; 272 273 SDT_PROBE2(vnet, functions, vnet_destroy, entry, __LINE__, vnet); 274 KASSERT(vnet->vnet_sockcnt == 0, 275 ("%s: vnet still has sockets", __func__)); 276 277 VNET_LIST_WLOCK(); 278 LIST_REMOVE(vnet, vnet_le); 279 VNET_LIST_WUNLOCK(); 280 281 CURVNET_SET_QUIET(vnet); 282 283 /* Return all inherited interfaces to their parent vnets. */ 284 TAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) { 285 if (ifp->if_home_vnet != ifp->if_vnet) 286 if_vmove(ifp, ifp->if_home_vnet); 287 } 288 289 vnet_sysuninit(); 290 CURVNET_RESTORE(); 291 292 /* 293 * Release storage for the virtual network stack instance. 294 */ 295 free(vnet->vnet_data_mem, M_VNET_DATA); 296 vnet->vnet_data_mem = NULL; 297 vnet->vnet_data_base = 0; 298 vnet->vnet_magic_n = 0xdeadbeef; 299 free(vnet, M_VNET); 300 SDT_PROBE1(vnet, functions, vnet_destroy, return, __LINE__); 301 } 302 303 /* 304 * Boot time initialization and allocation of virtual network stacks. 305 */ 306 static void 307 vnet_init_prelink(void *arg) 308 { 309 310 rw_init(&vnet_rwlock, "vnet_rwlock"); 311 sx_init(&vnet_sxlock, "vnet_sxlock"); 312 sx_init(&vnet_sysinit_sxlock, "vnet_sysinit_sxlock"); 313 LIST_INIT(&vnet_head); 314 } 315 SYSINIT(vnet_init_prelink, SI_SUB_VNET_PRELINK, SI_ORDER_FIRST, 316 vnet_init_prelink, NULL); 317 318 static void 319 vnet0_init(void *arg) 320 { 321 322 /* Warn people before take off - in case we crash early. */ 323 printf("WARNING: VIMAGE (virtualized network stack) is a highly " 324 "experimental feature.\n"); 325 326 /* 327 * We MUST clear curvnet in vi_init_done() before going SMP, 328 * otherwise CURVNET_SET() macros would scream about unnecessary 329 * curvnet recursions. 330 */ 331 curvnet = prison0.pr_vnet = vnet0 = vnet_alloc(); 332 } 333 SYSINIT(vnet0_init, SI_SUB_VNET, SI_ORDER_FIRST, vnet0_init, NULL); 334 335 static void 336 vnet_init_done(void *unused) 337 { 338 339 curvnet = NULL; 340 } 341 342 SYSINIT(vnet_init_done, SI_SUB_VNET_DONE, SI_ORDER_FIRST, vnet_init_done, 343 NULL); 344 345 /* 346 * Once on boot, initialize the modspace freelist to entirely cover modspace. 347 */ 348 static void 349 vnet_data_startup(void *dummy __unused) 350 { 351 struct vnet_data_free *df; 352 353 df = malloc(sizeof(*df), M_VNET_DATA_FREE, M_WAITOK | M_ZERO); 354 df->vnd_start = (uintptr_t)&VNET_NAME(modspace); 355 df->vnd_len = VNET_MODMIN; 356 TAILQ_INSERT_HEAD(&vnet_data_free_head, df, vnd_link); 357 sx_init(&vnet_data_free_lock, "vnet_data alloc lock"); 358 } 359 SYSINIT(vnet_data, SI_SUB_KLD, SI_ORDER_FIRST, vnet_data_startup, 0); 360 361 /* 362 * When a module is loaded and requires storage for a virtualized global 363 * variable, allocate space from the modspace free list. This interface 364 * should be used only by the kernel linker. 365 */ 366 void * 367 vnet_data_alloc(int size) 368 { 369 struct vnet_data_free *df; 370 void *s; 371 372 s = NULL; 373 size = roundup2(size, sizeof(void *)); 374 sx_xlock(&vnet_data_free_lock); 375 TAILQ_FOREACH(df, &vnet_data_free_head, vnd_link) { 376 if (df->vnd_len < size) 377 continue; 378 if (df->vnd_len == size) { 379 s = (void *)df->vnd_start; 380 TAILQ_REMOVE(&vnet_data_free_head, df, vnd_link); 381 free(df, M_VNET_DATA_FREE); 382 break; 383 } 384 s = (void *)df->vnd_start; 385 df->vnd_len -= size; 386 df->vnd_start = df->vnd_start + size; 387 break; 388 } 389 sx_xunlock(&vnet_data_free_lock); 390 391 return (s); 392 } 393 394 /* 395 * Free space for a virtualized global variable on module unload. 396 */ 397 void 398 vnet_data_free(void *start_arg, int size) 399 { 400 struct vnet_data_free *df; 401 struct vnet_data_free *dn; 402 uintptr_t start; 403 uintptr_t end; 404 405 size = roundup2(size, sizeof(void *)); 406 start = (uintptr_t)start_arg; 407 end = start + size; 408 /* 409 * Free a region of space and merge it with as many neighbors as 410 * possible. Keeping the list sorted simplifies this operation. 411 */ 412 sx_xlock(&vnet_data_free_lock); 413 TAILQ_FOREACH(df, &vnet_data_free_head, vnd_link) { 414 if (df->vnd_start > end) 415 break; 416 /* 417 * If we expand at the end of an entry we may have to merge 418 * it with the one following it as well. 419 */ 420 if (df->vnd_start + df->vnd_len == start) { 421 df->vnd_len += size; 422 dn = TAILQ_NEXT(df, vnd_link); 423 if (df->vnd_start + df->vnd_len == dn->vnd_start) { 424 df->vnd_len += dn->vnd_len; 425 TAILQ_REMOVE(&vnet_data_free_head, dn, 426 vnd_link); 427 free(dn, M_VNET_DATA_FREE); 428 } 429 sx_xunlock(&vnet_data_free_lock); 430 return; 431 } 432 if (df->vnd_start == end) { 433 df->vnd_start = start; 434 df->vnd_len += size; 435 sx_xunlock(&vnet_data_free_lock); 436 return; 437 } 438 } 439 dn = malloc(sizeof(*df), M_VNET_DATA_FREE, M_WAITOK | M_ZERO); 440 dn->vnd_start = start; 441 dn->vnd_len = size; 442 if (df) 443 TAILQ_INSERT_BEFORE(df, dn, vnd_link); 444 else 445 TAILQ_INSERT_TAIL(&vnet_data_free_head, dn, vnd_link); 446 sx_xunlock(&vnet_data_free_lock); 447 } 448 449 /* 450 * When a new virtualized global variable has been allocated, propagate its 451 * initial value to each already-allocated virtual network stack instance. 452 */ 453 void 454 vnet_data_copy(void *start, int size) 455 { 456 struct vnet *vnet; 457 458 VNET_LIST_RLOCK(); 459 LIST_FOREACH(vnet, &vnet_head, vnet_le) 460 memcpy((void *)((uintptr_t)vnet->vnet_data_base + 461 (uintptr_t)start), start, size); 462 VNET_LIST_RUNLOCK(); 463 } 464 465 /* 466 * Variants on sysctl_handle_foo that know how to handle virtualized global 467 * variables: if 'arg1' is a pointer, then we transform it to the local vnet 468 * offset. 469 */ 470 int 471 vnet_sysctl_handle_int(SYSCTL_HANDLER_ARGS) 472 { 473 474 if (arg1 != NULL) 475 arg1 = (void *)(curvnet->vnet_data_base + (uintptr_t)arg1); 476 return (sysctl_handle_int(oidp, arg1, arg2, req)); 477 } 478 479 int 480 vnet_sysctl_handle_opaque(SYSCTL_HANDLER_ARGS) 481 { 482 483 if (arg1 != NULL) 484 arg1 = (void *)(curvnet->vnet_data_base + (uintptr_t)arg1); 485 return (sysctl_handle_opaque(oidp, arg1, arg2, req)); 486 } 487 488 int 489 vnet_sysctl_handle_string(SYSCTL_HANDLER_ARGS) 490 { 491 492 if (arg1 != NULL) 493 arg1 = (void *)(curvnet->vnet_data_base + (uintptr_t)arg1); 494 return (sysctl_handle_string(oidp, arg1, arg2, req)); 495 } 496 497 int 498 vnet_sysctl_handle_uint(SYSCTL_HANDLER_ARGS) 499 { 500 501 if (arg1 != NULL) 502 arg1 = (void *)(curvnet->vnet_data_base + (uintptr_t)arg1); 503 return (sysctl_handle_int(oidp, arg1, arg2, req)); 504 } 505 506 /* 507 * Support for special SYSINIT handlers registered via VNET_SYSINIT() 508 * and VNET_SYSUNINIT(). 509 */ 510 void 511 vnet_register_sysinit(void *arg) 512 { 513 struct vnet_sysinit *vs, *vs2; 514 struct vnet *vnet; 515 516 vs = arg; 517 KASSERT(vs->subsystem > SI_SUB_VNET, ("vnet sysinit too early")); 518 519 /* Add the constructor to the global list of vnet constructors. */ 520 VNET_SYSINIT_WLOCK(); 521 TAILQ_FOREACH(vs2, &vnet_constructors, link) { 522 if (vs2->subsystem > vs->subsystem) 523 break; 524 if (vs2->subsystem == vs->subsystem && vs2->order > vs->order) 525 break; 526 } 527 if (vs2 != NULL) 528 TAILQ_INSERT_BEFORE(vs2, vs, link); 529 else 530 TAILQ_INSERT_TAIL(&vnet_constructors, vs, link); 531 532 /* 533 * Invoke the constructor on all the existing vnets when it is 534 * registered. 535 */ 536 VNET_FOREACH(vnet) { 537 CURVNET_SET_QUIET(vnet); 538 vs->func(vs->arg); 539 CURVNET_RESTORE(); 540 } 541 VNET_SYSINIT_WUNLOCK(); 542 } 543 544 void 545 vnet_deregister_sysinit(void *arg) 546 { 547 struct vnet_sysinit *vs; 548 549 vs = arg; 550 551 /* Remove the constructor from the global list of vnet constructors. */ 552 VNET_SYSINIT_WLOCK(); 553 TAILQ_REMOVE(&vnet_constructors, vs, link); 554 VNET_SYSINIT_WUNLOCK(); 555 } 556 557 void 558 vnet_register_sysuninit(void *arg) 559 { 560 struct vnet_sysinit *vs, *vs2; 561 562 vs = arg; 563 564 /* Add the destructor to the global list of vnet destructors. */ 565 VNET_SYSINIT_WLOCK(); 566 TAILQ_FOREACH(vs2, &vnet_destructors, link) { 567 if (vs2->subsystem > vs->subsystem) 568 break; 569 if (vs2->subsystem == vs->subsystem && vs2->order > vs->order) 570 break; 571 } 572 if (vs2 != NULL) 573 TAILQ_INSERT_BEFORE(vs2, vs, link); 574 else 575 TAILQ_INSERT_TAIL(&vnet_destructors, vs, link); 576 VNET_SYSINIT_WUNLOCK(); 577 } 578 579 void 580 vnet_deregister_sysuninit(void *arg) 581 { 582 struct vnet_sysinit *vs; 583 struct vnet *vnet; 584 585 vs = arg; 586 587 /* 588 * Invoke the destructor on all the existing vnets when it is 589 * deregistered. 590 */ 591 VNET_SYSINIT_WLOCK(); 592 VNET_FOREACH(vnet) { 593 CURVNET_SET_QUIET(vnet); 594 vs->func(vs->arg); 595 CURVNET_RESTORE(); 596 } 597 598 /* Remove the destructor from the global list of vnet destructors. */ 599 TAILQ_REMOVE(&vnet_destructors, vs, link); 600 VNET_SYSINIT_WUNLOCK(); 601 } 602 603 /* 604 * Invoke all registered vnet constructors on the current vnet. Used during 605 * vnet construction. The caller is responsible for ensuring the new vnet is 606 * the current vnet and that the vnet_sysinit_sxlock lock is locked. 607 */ 608 void 609 vnet_sysinit(void) 610 { 611 struct vnet_sysinit *vs; 612 613 VNET_SYSINIT_RLOCK(); 614 TAILQ_FOREACH(vs, &vnet_constructors, link) { 615 vs->func(vs->arg); 616 } 617 VNET_SYSINIT_RUNLOCK(); 618 } 619 620 /* 621 * Invoke all registered vnet destructors on the current vnet. Used during 622 * vnet destruction. The caller is responsible for ensuring the dying vnet 623 * the current vnet and that the vnet_sysinit_sxlock lock is locked. 624 */ 625 void 626 vnet_sysuninit(void) 627 { 628 struct vnet_sysinit *vs; 629 630 VNET_SYSINIT_RLOCK(); 631 TAILQ_FOREACH_REVERSE(vs, &vnet_destructors, vnet_sysuninit_head, 632 link) { 633 vs->func(vs->arg); 634 } 635 VNET_SYSINIT_RUNLOCK(); 636 } 637 638 /* 639 * EVENTHANDLER(9) extensions. 640 */ 641 /* 642 * Invoke the eventhandler function originally registered with the possibly 643 * registered argument for all virtual network stack instances. 644 * 645 * This iterator can only be used for eventhandlers that do not take any 646 * additional arguments, as we do ignore the variadic arguments from the 647 * EVENTHANDLER_INVOKE() call. 648 */ 649 void 650 vnet_global_eventhandler_iterator_func(void *arg, ...) 651 { 652 VNET_ITERATOR_DECL(vnet_iter); 653 struct eventhandler_entry_vimage *v_ee; 654 655 /* 656 * There is a bug here in that we should actually cast things to 657 * (struct eventhandler_entry_ ## name *) but that's not easily 658 * possible in here so just re-using the variadic version we 659 * defined for the generic vimage case. 660 */ 661 v_ee = arg; 662 VNET_LIST_RLOCK(); 663 VNET_FOREACH(vnet_iter) { 664 CURVNET_SET(vnet_iter); 665 ((vimage_iterator_func_t)v_ee->func)(v_ee->ee_arg); 666 CURVNET_RESTORE(); 667 } 668 VNET_LIST_RUNLOCK(); 669 } 670 671 #ifdef VNET_DEBUG 672 struct vnet_recursion { 673 SLIST_ENTRY(vnet_recursion) vnr_le; 674 const char *prev_fn; 675 const char *where_fn; 676 int where_line; 677 struct vnet *old_vnet; 678 struct vnet *new_vnet; 679 }; 680 681 static SLIST_HEAD(, vnet_recursion) vnet_recursions = 682 SLIST_HEAD_INITIALIZER(vnet_recursions); 683 684 static void 685 vnet_print_recursion(struct vnet_recursion *vnr, int brief) 686 { 687 688 if (!brief) 689 printf("CURVNET_SET() recursion in "); 690 printf("%s() line %d, prev in %s()", vnr->where_fn, vnr->where_line, 691 vnr->prev_fn); 692 if (brief) 693 printf(", "); 694 else 695 printf("\n "); 696 printf("%p -> %p\n", vnr->old_vnet, vnr->new_vnet); 697 } 698 699 void 700 vnet_log_recursion(struct vnet *old_vnet, const char *old_fn, int line) 701 { 702 struct vnet_recursion *vnr; 703 704 /* Skip already logged recursion events. */ 705 SLIST_FOREACH(vnr, &vnet_recursions, vnr_le) 706 if (vnr->prev_fn == old_fn && 707 vnr->where_fn == curthread->td_vnet_lpush && 708 vnr->where_line == line && 709 (vnr->old_vnet == vnr->new_vnet) == (curvnet == old_vnet)) 710 return; 711 712 vnr = malloc(sizeof(*vnr), M_VNET, M_NOWAIT | M_ZERO); 713 if (vnr == NULL) 714 panic("%s: malloc failed", __func__); 715 vnr->prev_fn = old_fn; 716 vnr->where_fn = curthread->td_vnet_lpush; 717 vnr->where_line = line; 718 vnr->old_vnet = old_vnet; 719 vnr->new_vnet = curvnet; 720 721 SLIST_INSERT_HEAD(&vnet_recursions, vnr, vnr_le); 722 723 vnet_print_recursion(vnr, 0); 724 #ifdef KDB 725 kdb_backtrace(); 726 #endif 727 } 728 #endif /* VNET_DEBUG */ 729 730 /* 731 * DDB(4). 732 */ 733 #ifdef DDB 734 DB_SHOW_COMMAND(vnets, db_show_vnets) 735 { 736 VNET_ITERATOR_DECL(vnet_iter); 737 738 VNET_FOREACH(vnet_iter) { 739 db_printf("vnet = %p\n", vnet_iter); 740 db_printf(" vnet_magic_n = 0x%x (%s, orig 0x%x)\n", 741 vnet_iter->vnet_magic_n, 742 (vnet_iter->vnet_magic_n == VNET_MAGIC_N) ? 743 "ok" : "mismatch", VNET_MAGIC_N); 744 db_printf(" vnet_ifcnt = %u\n", vnet_iter->vnet_ifcnt); 745 db_printf(" vnet_sockcnt = %u\n", vnet_iter->vnet_sockcnt); 746 db_printf(" vnet_data_mem = %p\n", vnet_iter->vnet_data_mem); 747 db_printf(" vnet_data_base = 0x%jx\n", 748 (uintmax_t)vnet_iter->vnet_data_base); 749 db_printf("\n"); 750 if (db_pager_quit) 751 break; 752 } 753 } 754 755 static void 756 db_show_vnet_print_vs(struct vnet_sysinit *vs, int ddb) 757 { 758 const char *vsname, *funcname; 759 c_db_sym_t sym; 760 db_expr_t offset; 761 762 #define xprint(...) \ 763 if (ddb) \ 764 db_printf(__VA_ARGS__); \ 765 else \ 766 printf(__VA_ARGS__) 767 768 if (vs == NULL) { 769 xprint("%s: no vnet_sysinit * given\n", __func__); 770 return; 771 } 772 773 sym = db_search_symbol((vm_offset_t)vs, DB_STGY_ANY, &offset); 774 db_symbol_values(sym, &vsname, NULL); 775 sym = db_search_symbol((vm_offset_t)vs->func, DB_STGY_PROC, &offset); 776 db_symbol_values(sym, &funcname, NULL); 777 xprint("%s(%p)\n", (vsname != NULL) ? vsname : "", vs); 778 xprint(" 0x%08x 0x%08x\n", vs->subsystem, vs->order); 779 xprint(" %p(%s)(%p)\n", 780 vs->func, (funcname != NULL) ? funcname : "", vs->arg); 781 #undef xprint 782 } 783 784 DB_SHOW_COMMAND(vnet_sysinit, db_show_vnet_sysinit) 785 { 786 struct vnet_sysinit *vs; 787 788 db_printf("VNET_SYSINIT vs Name(Ptr)\n"); 789 db_printf(" Subsystem Order\n"); 790 db_printf(" Function(Name)(Arg)\n"); 791 TAILQ_FOREACH(vs, &vnet_constructors, link) { 792 db_show_vnet_print_vs(vs, 1); 793 if (db_pager_quit) 794 break; 795 } 796 } 797 798 DB_SHOW_COMMAND(vnet_sysuninit, db_show_vnet_sysuninit) 799 { 800 struct vnet_sysinit *vs; 801 802 db_printf("VNET_SYSUNINIT vs Name(Ptr)\n"); 803 db_printf(" Subsystem Order\n"); 804 db_printf(" Function(Name)(Arg)\n"); 805 TAILQ_FOREACH_REVERSE(vs, &vnet_destructors, vnet_sysuninit_head, 806 link) { 807 db_show_vnet_print_vs(vs, 1); 808 if (db_pager_quit) 809 break; 810 } 811 } 812 813 #ifdef VNET_DEBUG 814 DB_SHOW_COMMAND(vnetrcrs, db_show_vnetrcrs) 815 { 816 struct vnet_recursion *vnr; 817 818 SLIST_FOREACH(vnr, &vnet_recursions, vnr_le) 819 vnet_print_recursion(vnr, 1); 820 } 821 #endif 822 #endif /* DDB */ 823