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