1195699Srwatson/*- 2196019Srwatson * Copyright (c) 2004-2009 University of Zagreb 3196019Srwatson * Copyright (c) 2006-2009 FreeBSD Foundation 4196019Srwatson * All rights reserved. 5196019Srwatson * 6196019Srwatson * This software was developed by the University of Zagreb and the 7196019Srwatson * FreeBSD Foundation under sponsorship by the Stichting NLnet and the 8196019Srwatson * FreeBSD Foundation. 9196019Srwatson * 10195699Srwatson * Copyright (c) 2009 Jeffrey Roberson <jeff@freebsd.org> 11195699Srwatson * Copyright (c) 2009 Robert N. M. Watson 12195699Srwatson * All rights reserved. 13195699Srwatson * 14195699Srwatson * Redistribution and use in source and binary forms, with or without 15195699Srwatson * modification, are permitted provided that the following conditions 16195699Srwatson * are met: 17195699Srwatson * 1. Redistributions of source code must retain the above copyright 18195699Srwatson * notice, this list of conditions and the following disclaimer. 19195699Srwatson * 2. Redistributions in binary form must reproduce the above copyright 20195699Srwatson * notice, this list of conditions and the following disclaimer in the 21195699Srwatson * documentation and/or other materials provided with the distribution. 22195699Srwatson * 23195699Srwatson * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 24195699Srwatson * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 25195699Srwatson * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 26195699Srwatson * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 27195699Srwatson * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 28195699Srwatson * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 29195699Srwatson * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 30195699Srwatson * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 31195699Srwatson * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 32195699Srwatson * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 33195699Srwatson * SUCH DAMAGE. 34195699Srwatson */ 35195699Srwatson 36195699Srwatson#include <sys/cdefs.h> 37195699Srwatson__FBSDID("$FreeBSD: stable/11/sys/net/vnet.c 359652 2020-04-06 07:16:31Z hselasky $"); 38195699Srwatson 39196019Srwatson#include "opt_ddb.h" 40203483Szec#include "opt_kdb.h" 41196019Srwatson 42195699Srwatson#include <sys/param.h> 43203483Szec#include <sys/kdb.h> 44195699Srwatson#include <sys/kernel.h> 45196019Srwatson#include <sys/jail.h> 46203727Sbz#include <sys/sdt.h> 47195699Srwatson#include <sys/systm.h> 48195699Srwatson#include <sys/sysctl.h> 49205345Sbz#include <sys/eventhandler.h> 50195699Srwatson#include <sys/lock.h> 51195699Srwatson#include <sys/malloc.h> 52195699Srwatson#include <sys/proc.h> 53196019Srwatson#include <sys/socket.h> 54195699Srwatson#include <sys/sx.h> 55195699Srwatson#include <sys/sysctl.h> 56195699Srwatson 57205345Sbz#include <machine/stdarg.h> 58205345Sbz 59196019Srwatson#ifdef DDB 60196019Srwatson#include <ddb/ddb.h> 61203729Sbz#include <ddb/db_sym.h> 62196019Srwatson#endif 63196019Srwatson 64196019Srwatson#include <net/if.h> 65196019Srwatson#include <net/if_var.h> 66195699Srwatson#include <net/vnet.h> 67195699Srwatson 68195699Srwatson/*- 69195972Srwatson * This file implements core functions for virtual network stacks: 70195972Srwatson * 71196025Srwatson * - Virtual network stack management functions. 72196019Srwatson * 73196025Srwatson * - Virtual network stack memory allocator, which virtualizes global 74195972Srwatson * variables in the network stack 75195972Srwatson * 76195972Srwatson * - Virtualized SYSINIT's/SYSUNINIT's, which allow network stack subsystems 77195972Srwatson * to register startup/shutdown events to be run for each virtual network 78195972Srwatson * stack instance. 79196019Srwatson */ 80196019Srwatson 81217203SbzFEATURE(vimage, "VIMAGE kernel virtualization"); 82217203Sbz 83227293Sedstatic MALLOC_DEFINE(M_VNET, "vnet", "network stack control block"); 84196019Srwatson 85196019Srwatson/* 86196019Srwatson * The virtual network stack list has two read-write locks, one sleepable and 87196019Srwatson * the other not, so that the list can be stablized and walked in a variety 88196019Srwatson * of network stack contexts. Both must be acquired exclusively to modify 89196025Srwatson * the list, but a read lock of either lock is sufficient to walk the list. 90196019Srwatson */ 91196019Srwatsonstruct rwlock vnet_rwlock; 92196019Srwatsonstruct sx vnet_sxlock; 93196019Srwatson 94196019Srwatson#define VNET_LIST_WLOCK() do { \ 95196019Srwatson sx_xlock(&vnet_sxlock); \ 96196019Srwatson rw_wlock(&vnet_rwlock); \ 97196019Srwatson} while (0) 98196019Srwatson 99196019Srwatson#define VNET_LIST_WUNLOCK() do { \ 100196019Srwatson rw_wunlock(&vnet_rwlock); \ 101196019Srwatson sx_xunlock(&vnet_sxlock); \ 102196019Srwatson} while (0) 103196019Srwatson 104196019Srwatsonstruct vnet_list_head vnet_head; 105196019Srwatsonstruct vnet *vnet0; 106196019Srwatson 107196019Srwatson/* 108195972Srwatson * The virtual network stack allocator provides storage for virtualized 109195972Srwatson * global variables. These variables are defined/declared using the 110195972Srwatson * VNET_DEFINE()/VNET_DECLARE() macros, which place them in the 'set_vnet' 111195972Srwatson * linker set. The details of the implementation are somewhat subtle, but 112195972Srwatson * allow the majority of most network subsystems to maintain 113195699Srwatson * virtualization-agnostic. 114195699Srwatson * 115195699Srwatson * The virtual network stack allocator handles variables in the base kernel 116195699Srwatson * vs. modules in similar but different ways. In both cases, virtualized 117195699Srwatson * global variables are marked as such by being declared to be part of the 118195699Srwatson * vnet linker set. These "master" copies of global variables serve two 119195699Srwatson * functions: 120195699Srwatson * 121195699Srwatson * (1) They contain static initialization or "default" values for global 122195699Srwatson * variables which will be propagated to each virtual network stack 123195699Srwatson * instance when created. As with normal global variables, they default 124195699Srwatson * to zero-filled. 125195699Srwatson * 126195699Srwatson * (2) They act as unique global names by which the variable can be referred 127195699Srwatson * to, regardless of network stack instance. The single global symbol 128195699Srwatson * will be used to calculate the location of a per-virtual instance 129195699Srwatson * variable at run-time. 130195699Srwatson * 131195699Srwatson * Each virtual network stack instance has a complete copy of each 132195699Srwatson * virtualized global variable, stored in a malloc'd block of memory 133195699Srwatson * referred to by vnet->vnet_data_mem. Critical to the design is that each 134195699Srwatson * per-instance memory block is laid out identically to the master block so 135195699Srwatson * that the offset of each global variable is the same across all blocks. To 136195699Srwatson * optimize run-time access, a precalculated 'base' address, 137195699Srwatson * vnet->vnet_data_base, is stored in each vnet, and is the amount that can 138195699Srwatson * be added to the address of a 'master' instance of a variable to get to the 139195699Srwatson * per-vnet instance. 140195699Srwatson * 141195699Srwatson * Virtualized global variables are handled in a similar manner, but as each 142195699Srwatson * module has its own 'set_vnet' linker set, and we want to keep all 143195699Srwatson * virtualized globals togther, we reserve space in the kernel's linker set 144195699Srwatson * for potential module variables using a per-vnet character array, 145195699Srwatson * 'modspace'. The virtual network stack allocator maintains a free list to 146195699Srwatson * track what space in the array is free (all, initially) and as modules are 147195699Srwatson * linked, allocates portions of the space to specific globals. The kernel 148195699Srwatson * module linker queries the virtual network stack allocator and will 149195699Srwatson * bind references of the global to the location during linking. It also 150195699Srwatson * calls into the virtual network stack allocator, once the memory is 151195699Srwatson * initialized, in order to propagate the new static initializations to all 152195699Srwatson * existing virtual network stack instances so that the soon-to-be executing 153195699Srwatson * module will find every network stack instance with proper default values. 154195699Srwatson */ 155195699Srwatson 156195699Srwatson/* 157195699Srwatson * Number of bytes of data in the 'set_vnet' linker set, and hence the total 158195699Srwatson * size of all kernel virtualized global variables, and the malloc(9) type 159195699Srwatson * that will be used to allocate it. 160195699Srwatson */ 161195699Srwatson#define VNET_BYTES (VNET_STOP - VNET_START) 162195699Srwatson 163227293Sedstatic MALLOC_DEFINE(M_VNET_DATA, "vnet_data", "VNET data"); 164195699Srwatson 165195699Srwatson/* 166195699Srwatson * VNET_MODMIN is the minimum number of bytes we will reserve for the sum of 167195699Srwatson * global variables across all loaded modules. As this actually sizes an 168195699Srwatson * array declared as a virtualized global variable in the kernel itself, and 169195699Srwatson * we want the virtualized global variable space to be page-sized, we may 170195699Srwatson * have more space than that in practice. 171195699Srwatson */ 172340051Sbz#define VNET_MODMIN (8 * PAGE_SIZE) 173195699Srwatson#define VNET_SIZE roundup2(VNET_BYTES, PAGE_SIZE) 174195699Srwatson 175195699Srwatson/* 176195699Srwatson * Space to store virtualized global variables from loadable kernel modules, 177195699Srwatson * and the free list to manage it. 178195699Srwatson */ 179215701Sdimstatic VNET_DEFINE(char, modspace[VNET_MODMIN]); 180195699Srwatson 181195837Srwatson/* 182196025Srwatson * Global lists of subsystem constructor and destructors for vnets. They are 183196633Szec * registered via VNET_SYSINIT() and VNET_SYSUNINIT(). Both lists are 184196633Szec * protected by the vnet_sysinit_sxlock global lock. 185195837Srwatson */ 186195837Srwatsonstatic TAILQ_HEAD(vnet_sysinit_head, vnet_sysinit) vnet_constructors = 187195837Srwatson TAILQ_HEAD_INITIALIZER(vnet_constructors); 188195837Srwatsonstatic TAILQ_HEAD(vnet_sysuninit_head, vnet_sysinit) vnet_destructors = 189195837Srwatson TAILQ_HEAD_INITIALIZER(vnet_destructors); 190195837Srwatson 191196633Szecstruct sx vnet_sysinit_sxlock; 192196633Szec 193196633Szec#define VNET_SYSINIT_WLOCK() sx_xlock(&vnet_sysinit_sxlock); 194196633Szec#define VNET_SYSINIT_WUNLOCK() sx_xunlock(&vnet_sysinit_sxlock); 195196633Szec#define VNET_SYSINIT_RLOCK() sx_slock(&vnet_sysinit_sxlock); 196196633Szec#define VNET_SYSINIT_RUNLOCK() sx_sunlock(&vnet_sysinit_sxlock); 197196633Szec 198195699Srwatsonstruct vnet_data_free { 199195699Srwatson uintptr_t vnd_start; 200195699Srwatson int vnd_len; 201195699Srwatson TAILQ_ENTRY(vnet_data_free) vnd_link; 202195699Srwatson}; 203195699Srwatson 204227293Sedstatic MALLOC_DEFINE(M_VNET_DATA_FREE, "vnet_data_free", 205227293Sed "VNET resource accounting"); 206195699Srwatsonstatic TAILQ_HEAD(, vnet_data_free) vnet_data_free_head = 207195699Srwatson TAILQ_HEAD_INITIALIZER(vnet_data_free_head); 208195699Srwatsonstatic struct sx vnet_data_free_lock; 209195699Srwatson 210203727SbzSDT_PROVIDER_DEFINE(vnet); 211258622SavgSDT_PROBE_DEFINE1(vnet, functions, vnet_alloc, entry, "int"); 212258622SavgSDT_PROBE_DEFINE2(vnet, functions, vnet_alloc, alloc, "int", 213211616Srpaulo "struct vnet *"); 214258622SavgSDT_PROBE_DEFINE2(vnet, functions, vnet_alloc, return, 215211616Srpaulo "int", "struct vnet *"); 216258622SavgSDT_PROBE_DEFINE2(vnet, functions, vnet_destroy, entry, 217211616Srpaulo "int", "struct vnet *"); 218258675SglebiusSDT_PROBE_DEFINE1(vnet, functions, vnet_destroy, return, 219211616Srpaulo "int"); 220203727Sbz 221203729Sbz#ifdef DDB 222203729Sbzstatic void db_show_vnet_print_vs(struct vnet_sysinit *, int); 223203729Sbz#endif 224203729Sbz 225195699Srwatson/* 226196019Srwatson * Allocate a virtual network stack. 227196019Srwatson */ 228196019Srwatsonstruct vnet * 229196019Srwatsonvnet_alloc(void) 230196019Srwatson{ 231196019Srwatson struct vnet *vnet; 232196019Srwatson 233203727Sbz SDT_PROBE1(vnet, functions, vnet_alloc, entry, __LINE__); 234196019Srwatson vnet = malloc(sizeof(struct vnet), M_VNET, M_WAITOK | M_ZERO); 235196019Srwatson vnet->vnet_magic_n = VNET_MAGIC_N; 236300155Sbz vnet->vnet_state = 0; 237203727Sbz SDT_PROBE2(vnet, functions, vnet_alloc, alloc, __LINE__, vnet); 238196019Srwatson 239196024Srwatson /* 240196024Srwatson * Allocate storage for virtualized global variables and copy in 241196024Srwatson * initial values form our 'master' copy. 242196024Srwatson */ 243196024Srwatson vnet->vnet_data_mem = malloc(VNET_SIZE, M_VNET_DATA, M_WAITOK); 244196024Srwatson memcpy(vnet->vnet_data_mem, (void *)VNET_START, VNET_BYTES); 245196024Srwatson 246196024Srwatson /* 247196024Srwatson * All use of vnet-specific data will immediately subtract VNET_START 248196024Srwatson * from the base memory pointer, so pre-calculate that now to avoid 249196024Srwatson * it on each use. 250196024Srwatson */ 251196024Srwatson vnet->vnet_data_base = (uintptr_t)vnet->vnet_data_mem - VNET_START; 252196024Srwatson 253196019Srwatson /* Initialize / attach vnet module instances. */ 254196019Srwatson CURVNET_SET_QUIET(vnet); 255196019Srwatson vnet_sysinit(); 256196019Srwatson CURVNET_RESTORE(); 257196019Srwatson 258196633Szec VNET_LIST_WLOCK(); 259196019Srwatson LIST_INSERT_HEAD(&vnet_head, vnet, vnet_le); 260196019Srwatson VNET_LIST_WUNLOCK(); 261196019Srwatson 262203727Sbz SDT_PROBE2(vnet, functions, vnet_alloc, return, __LINE__, vnet); 263196019Srwatson return (vnet); 264196019Srwatson} 265196019Srwatson 266196019Srwatson/* 267196019Srwatson * Destroy a virtual network stack. 268196019Srwatson */ 269196019Srwatsonvoid 270196019Srwatsonvnet_destroy(struct vnet *vnet) 271196019Srwatson{ 272196019Srwatson 273203727Sbz SDT_PROBE2(vnet, functions, vnet_destroy, entry, __LINE__, vnet); 274196019Srwatson KASSERT(vnet->vnet_sockcnt == 0, 275196019Srwatson ("%s: vnet still has sockets", __func__)); 276196019Srwatson 277196019Srwatson VNET_LIST_WLOCK(); 278196019Srwatson LIST_REMOVE(vnet, vnet_le); 279196633Szec VNET_LIST_WUNLOCK(); 280196019Srwatson 281196019Srwatson CURVNET_SET_QUIET(vnet); 282196019Srwatson vnet_sysuninit(); 283196019Srwatson CURVNET_RESTORE(); 284196019Srwatson 285196024Srwatson /* 286196024Srwatson * Release storage for the virtual network stack instance. 287196024Srwatson */ 288196024Srwatson free(vnet->vnet_data_mem, M_VNET_DATA); 289196024Srwatson vnet->vnet_data_mem = NULL; 290196024Srwatson vnet->vnet_data_base = 0; 291196019Srwatson vnet->vnet_magic_n = 0xdeadbeef; 292196019Srwatson free(vnet, M_VNET); 293203727Sbz SDT_PROBE1(vnet, functions, vnet_destroy, return, __LINE__); 294196019Srwatson} 295196019Srwatson 296196019Srwatson/* 297196019Srwatson * Boot time initialization and allocation of virtual network stacks. 298196019Srwatson */ 299196019Srwatsonstatic void 300300001Sbzvnet_init_prelink(void *arg __unused) 301196019Srwatson{ 302196019Srwatson 303196019Srwatson rw_init(&vnet_rwlock, "vnet_rwlock"); 304196019Srwatson sx_init(&vnet_sxlock, "vnet_sxlock"); 305196633Szec sx_init(&vnet_sysinit_sxlock, "vnet_sysinit_sxlock"); 306196019Srwatson LIST_INIT(&vnet_head); 307196019Srwatson} 308196019SrwatsonSYSINIT(vnet_init_prelink, SI_SUB_VNET_PRELINK, SI_ORDER_FIRST, 309196019Srwatson vnet_init_prelink, NULL); 310196019Srwatson 311196019Srwatsonstatic void 312300001Sbzvnet0_init(void *arg __unused) 313196019Srwatson{ 314196019Srwatson 315196026Srwatson /* Warn people before take off - in case we crash early. */ 316196026Srwatson printf("WARNING: VIMAGE (virtualized network stack) is a highly " 317196026Srwatson "experimental feature.\n"); 318196026Srwatson 319196019Srwatson /* 320196019Srwatson * We MUST clear curvnet in vi_init_done() before going SMP, 321196019Srwatson * otherwise CURVNET_SET() macros would scream about unnecessary 322196019Srwatson * curvnet recursions. 323196019Srwatson */ 324196019Srwatson curvnet = prison0.pr_vnet = vnet0 = vnet_alloc(); 325196019Srwatson} 326196019SrwatsonSYSINIT(vnet0_init, SI_SUB_VNET, SI_ORDER_FIRST, vnet0_init, NULL); 327196019Srwatson 328196019Srwatsonstatic void 329300001Sbzvnet_init_done(void *unused __unused) 330196019Srwatson{ 331196019Srwatson 332196019Srwatson curvnet = NULL; 333196019Srwatson} 334302054SbzSYSINIT(vnet_init_done, SI_SUB_VNET_DONE, SI_ORDER_ANY, vnet_init_done, 335196019Srwatson NULL); 336196019Srwatson 337196019Srwatson/* 338195699Srwatson * Once on boot, initialize the modspace freelist to entirely cover modspace. 339195699Srwatson */ 340195699Srwatsonstatic void 341195699Srwatsonvnet_data_startup(void *dummy __unused) 342195699Srwatson{ 343195699Srwatson struct vnet_data_free *df; 344195699Srwatson 345195699Srwatson df = malloc(sizeof(*df), M_VNET_DATA_FREE, M_WAITOK | M_ZERO); 346195699Srwatson df->vnd_start = (uintptr_t)&VNET_NAME(modspace); 347208100Sbz df->vnd_len = VNET_MODMIN; 348195699Srwatson TAILQ_INSERT_HEAD(&vnet_data_free_head, df, vnd_link); 349195699Srwatson sx_init(&vnet_data_free_lock, "vnet_data alloc lock"); 350195699Srwatson} 351359652ShselaskySYSINIT(vnet_data, SI_SUB_KLD, SI_ORDER_FIRST, vnet_data_startup, NULL); 352195699Srwatson 353300151Sbz/* Dummy VNET_SYSINIT to make sure we always reach the final end state. */ 354300151Sbzstatic void 355300151Sbzvnet_sysinit_done(void *unused __unused) 356300151Sbz{ 357300151Sbz 358300151Sbz return; 359300151Sbz} 360300151SbzVNET_SYSINIT(vnet_sysinit_done, SI_SUB_VNET_DONE, SI_ORDER_ANY, 361300151Sbz vnet_sysinit_done, NULL); 362300151Sbz 363195699Srwatson/* 364195699Srwatson * When a module is loaded and requires storage for a virtualized global 365195699Srwatson * variable, allocate space from the modspace free list. This interface 366195699Srwatson * should be used only by the kernel linker. 367195699Srwatson */ 368195699Srwatsonvoid * 369195699Srwatsonvnet_data_alloc(int size) 370195699Srwatson{ 371195699Srwatson struct vnet_data_free *df; 372195699Srwatson void *s; 373195699Srwatson 374195699Srwatson s = NULL; 375195699Srwatson size = roundup2(size, sizeof(void *)); 376195699Srwatson sx_xlock(&vnet_data_free_lock); 377195699Srwatson TAILQ_FOREACH(df, &vnet_data_free_head, vnd_link) { 378195699Srwatson if (df->vnd_len < size) 379195699Srwatson continue; 380195699Srwatson if (df->vnd_len == size) { 381195699Srwatson s = (void *)df->vnd_start; 382195699Srwatson TAILQ_REMOVE(&vnet_data_free_head, df, vnd_link); 383195699Srwatson free(df, M_VNET_DATA_FREE); 384195699Srwatson break; 385195699Srwatson } 386195699Srwatson s = (void *)df->vnd_start; 387195699Srwatson df->vnd_len -= size; 388195699Srwatson df->vnd_start = df->vnd_start + size; 389195699Srwatson break; 390195699Srwatson } 391195699Srwatson sx_xunlock(&vnet_data_free_lock); 392195699Srwatson 393195699Srwatson return (s); 394195699Srwatson} 395195699Srwatson 396195699Srwatson/* 397195699Srwatson * Free space for a virtualized global variable on module unload. 398195699Srwatson */ 399195699Srwatsonvoid 400195699Srwatsonvnet_data_free(void *start_arg, int size) 401195699Srwatson{ 402195699Srwatson struct vnet_data_free *df; 403195699Srwatson struct vnet_data_free *dn; 404195699Srwatson uintptr_t start; 405195699Srwatson uintptr_t end; 406195699Srwatson 407195699Srwatson size = roundup2(size, sizeof(void *)); 408195699Srwatson start = (uintptr_t)start_arg; 409195699Srwatson end = start + size; 410195699Srwatson /* 411195699Srwatson * Free a region of space and merge it with as many neighbors as 412195699Srwatson * possible. Keeping the list sorted simplifies this operation. 413195699Srwatson */ 414195699Srwatson sx_xlock(&vnet_data_free_lock); 415195699Srwatson TAILQ_FOREACH(df, &vnet_data_free_head, vnd_link) { 416195699Srwatson if (df->vnd_start > end) 417195699Srwatson break; 418195699Srwatson /* 419196025Srwatson * If we expand at the end of an entry we may have to merge 420196025Srwatson * it with the one following it as well. 421195699Srwatson */ 422195699Srwatson if (df->vnd_start + df->vnd_len == start) { 423195699Srwatson df->vnd_len += size; 424195699Srwatson dn = TAILQ_NEXT(df, vnd_link); 425195699Srwatson if (df->vnd_start + df->vnd_len == dn->vnd_start) { 426195699Srwatson df->vnd_len += dn->vnd_len; 427196025Srwatson TAILQ_REMOVE(&vnet_data_free_head, dn, 428196025Srwatson vnd_link); 429195699Srwatson free(dn, M_VNET_DATA_FREE); 430195699Srwatson } 431195699Srwatson sx_xunlock(&vnet_data_free_lock); 432195699Srwatson return; 433195699Srwatson } 434195699Srwatson if (df->vnd_start == end) { 435195699Srwatson df->vnd_start = start; 436195699Srwatson df->vnd_len += size; 437195699Srwatson sx_xunlock(&vnet_data_free_lock); 438195699Srwatson return; 439195699Srwatson } 440195699Srwatson } 441195699Srwatson dn = malloc(sizeof(*df), M_VNET_DATA_FREE, M_WAITOK | M_ZERO); 442195699Srwatson dn->vnd_start = start; 443195699Srwatson dn->vnd_len = size; 444195699Srwatson if (df) 445195699Srwatson TAILQ_INSERT_BEFORE(df, dn, vnd_link); 446195699Srwatson else 447195699Srwatson TAILQ_INSERT_TAIL(&vnet_data_free_head, dn, vnd_link); 448195699Srwatson sx_xunlock(&vnet_data_free_lock); 449195699Srwatson} 450195699Srwatson 451195699Srwatson/* 452195699Srwatson * When a new virtualized global variable has been allocated, propagate its 453195699Srwatson * initial value to each already-allocated virtual network stack instance. 454195699Srwatson */ 455195699Srwatsonvoid 456195699Srwatsonvnet_data_copy(void *start, int size) 457195699Srwatson{ 458196020Srwatson struct vnet *vnet; 459195699Srwatson 460196020Srwatson VNET_LIST_RLOCK(); 461196020Srwatson LIST_FOREACH(vnet, &vnet_head, vnet_le) 462196020Srwatson memcpy((void *)((uintptr_t)vnet->vnet_data_base + 463196020Srwatson (uintptr_t)start), start, size); 464196020Srwatson VNET_LIST_RUNLOCK(); 465195699Srwatson} 466195699Srwatson 467195699Srwatson/* 468195837Srwatson * Support for special SYSINIT handlers registered via VNET_SYSINIT() 469195837Srwatson * and VNET_SYSUNINIT(). 470195837Srwatson */ 471195837Srwatsonvoid 472195837Srwatsonvnet_register_sysinit(void *arg) 473195837Srwatson{ 474195837Srwatson struct vnet_sysinit *vs, *vs2; 475195837Srwatson struct vnet *vnet; 476195837Srwatson 477195837Srwatson vs = arg; 478195837Srwatson KASSERT(vs->subsystem > SI_SUB_VNET, ("vnet sysinit too early")); 479195837Srwatson 480195837Srwatson /* Add the constructor to the global list of vnet constructors. */ 481196633Szec VNET_SYSINIT_WLOCK(); 482195837Srwatson TAILQ_FOREACH(vs2, &vnet_constructors, link) { 483195837Srwatson if (vs2->subsystem > vs->subsystem) 484195837Srwatson break; 485195837Srwatson if (vs2->subsystem == vs->subsystem && vs2->order > vs->order) 486195837Srwatson break; 487195837Srwatson } 488195837Srwatson if (vs2 != NULL) 489195837Srwatson TAILQ_INSERT_BEFORE(vs2, vs, link); 490195837Srwatson else 491195837Srwatson TAILQ_INSERT_TAIL(&vnet_constructors, vs, link); 492195837Srwatson 493195837Srwatson /* 494195837Srwatson * Invoke the constructor on all the existing vnets when it is 495195837Srwatson * registered. 496195837Srwatson */ 497195837Srwatson VNET_FOREACH(vnet) { 498195837Srwatson CURVNET_SET_QUIET(vnet); 499195837Srwatson vs->func(vs->arg); 500195837Srwatson CURVNET_RESTORE(); 501195837Srwatson } 502196633Szec VNET_SYSINIT_WUNLOCK(); 503195837Srwatson} 504195837Srwatson 505195837Srwatsonvoid 506195837Srwatsonvnet_deregister_sysinit(void *arg) 507195837Srwatson{ 508195837Srwatson struct vnet_sysinit *vs; 509195837Srwatson 510195837Srwatson vs = arg; 511195837Srwatson 512195837Srwatson /* Remove the constructor from the global list of vnet constructors. */ 513196633Szec VNET_SYSINIT_WLOCK(); 514195837Srwatson TAILQ_REMOVE(&vnet_constructors, vs, link); 515196633Szec VNET_SYSINIT_WUNLOCK(); 516195837Srwatson} 517195837Srwatson 518195837Srwatsonvoid 519195837Srwatsonvnet_register_sysuninit(void *arg) 520195837Srwatson{ 521195837Srwatson struct vnet_sysinit *vs, *vs2; 522195837Srwatson 523195837Srwatson vs = arg; 524195837Srwatson 525195837Srwatson /* Add the destructor to the global list of vnet destructors. */ 526196633Szec VNET_SYSINIT_WLOCK(); 527195837Srwatson TAILQ_FOREACH(vs2, &vnet_destructors, link) { 528195837Srwatson if (vs2->subsystem > vs->subsystem) 529195837Srwatson break; 530195837Srwatson if (vs2->subsystem == vs->subsystem && vs2->order > vs->order) 531195837Srwatson break; 532195837Srwatson } 533195837Srwatson if (vs2 != NULL) 534195837Srwatson TAILQ_INSERT_BEFORE(vs2, vs, link); 535195837Srwatson else 536195837Srwatson TAILQ_INSERT_TAIL(&vnet_destructors, vs, link); 537196633Szec VNET_SYSINIT_WUNLOCK(); 538195837Srwatson} 539195837Srwatson 540195837Srwatsonvoid 541195837Srwatsonvnet_deregister_sysuninit(void *arg) 542195837Srwatson{ 543195837Srwatson struct vnet_sysinit *vs; 544195837Srwatson struct vnet *vnet; 545195837Srwatson 546195837Srwatson vs = arg; 547195837Srwatson 548195837Srwatson /* 549195837Srwatson * Invoke the destructor on all the existing vnets when it is 550195837Srwatson * deregistered. 551195837Srwatson */ 552196633Szec VNET_SYSINIT_WLOCK(); 553195837Srwatson VNET_FOREACH(vnet) { 554195837Srwatson CURVNET_SET_QUIET(vnet); 555195837Srwatson vs->func(vs->arg); 556195837Srwatson CURVNET_RESTORE(); 557195837Srwatson } 558195837Srwatson 559195837Srwatson /* Remove the destructor from the global list of vnet destructors. */ 560195837Srwatson TAILQ_REMOVE(&vnet_destructors, vs, link); 561196633Szec VNET_SYSINIT_WUNLOCK(); 562195837Srwatson} 563195837Srwatson 564195837Srwatson/* 565196025Srwatson * Invoke all registered vnet constructors on the current vnet. Used during 566196025Srwatson * vnet construction. The caller is responsible for ensuring the new vnet is 567196633Szec * the current vnet and that the vnet_sysinit_sxlock lock is locked. 568195837Srwatson */ 569195837Srwatsonvoid 570195837Srwatsonvnet_sysinit(void) 571195837Srwatson{ 572195837Srwatson struct vnet_sysinit *vs; 573195837Srwatson 574196633Szec VNET_SYSINIT_RLOCK(); 575195837Srwatson TAILQ_FOREACH(vs, &vnet_constructors, link) { 576300155Sbz curvnet->vnet_state = vs->subsystem; 577195837Srwatson vs->func(vs->arg); 578195837Srwatson } 579196633Szec VNET_SYSINIT_RUNLOCK(); 580195837Srwatson} 581195837Srwatson 582195837Srwatson/* 583196025Srwatson * Invoke all registered vnet destructors on the current vnet. Used during 584196025Srwatson * vnet destruction. The caller is responsible for ensuring the dying vnet 585196633Szec * the current vnet and that the vnet_sysinit_sxlock lock is locked. 586195837Srwatson */ 587195837Srwatsonvoid 588195837Srwatsonvnet_sysuninit(void) 589195837Srwatson{ 590195837Srwatson struct vnet_sysinit *vs; 591195837Srwatson 592196633Szec VNET_SYSINIT_RLOCK(); 593195837Srwatson TAILQ_FOREACH_REVERSE(vs, &vnet_destructors, vnet_sysuninit_head, 594195837Srwatson link) { 595300155Sbz curvnet->vnet_state = vs->subsystem; 596195837Srwatson vs->func(vs->arg); 597195837Srwatson } 598196633Szec VNET_SYSINIT_RUNLOCK(); 599195837Srwatson} 600196019Srwatson 601205345Sbz/* 602205345Sbz * EVENTHANDLER(9) extensions. 603205345Sbz */ 604205345Sbz/* 605205345Sbz * Invoke the eventhandler function originally registered with the possibly 606205345Sbz * registered argument for all virtual network stack instances. 607205345Sbz * 608205345Sbz * This iterator can only be used for eventhandlers that do not take any 609205345Sbz * additional arguments, as we do ignore the variadic arguments from the 610205345Sbz * EVENTHANDLER_INVOKE() call. 611205345Sbz */ 612205345Sbzvoid 613205345Sbzvnet_global_eventhandler_iterator_func(void *arg, ...) 614205345Sbz{ 615205345Sbz VNET_ITERATOR_DECL(vnet_iter); 616205345Sbz struct eventhandler_entry_vimage *v_ee; 617205345Sbz 618205345Sbz /* 619205345Sbz * There is a bug here in that we should actually cast things to 620205345Sbz * (struct eventhandler_entry_ ## name *) but that's not easily 621205345Sbz * possible in here so just re-using the variadic version we 622205345Sbz * defined for the generic vimage case. 623205345Sbz */ 624205345Sbz v_ee = arg; 625205345Sbz VNET_LIST_RLOCK(); 626205345Sbz VNET_FOREACH(vnet_iter) { 627205345Sbz CURVNET_SET(vnet_iter); 628205345Sbz ((vimage_iterator_func_t)v_ee->func)(v_ee->ee_arg); 629205345Sbz CURVNET_RESTORE(); 630205345Sbz } 631205345Sbz VNET_LIST_RUNLOCK(); 632205345Sbz} 633205345Sbz 634203483Szec#ifdef VNET_DEBUG 635203483Szecstruct vnet_recursion { 636203483Szec SLIST_ENTRY(vnet_recursion) vnr_le; 637203483Szec const char *prev_fn; 638203483Szec const char *where_fn; 639203483Szec int where_line; 640203483Szec struct vnet *old_vnet; 641203483Szec struct vnet *new_vnet; 642203483Szec}; 643203483Szec 644203483Szecstatic SLIST_HEAD(, vnet_recursion) vnet_recursions = 645203483Szec SLIST_HEAD_INITIALIZER(vnet_recursions); 646203483Szec 647203483Szecstatic void 648203483Szecvnet_print_recursion(struct vnet_recursion *vnr, int brief) 649203483Szec{ 650203483Szec 651203483Szec if (!brief) 652203483Szec printf("CURVNET_SET() recursion in "); 653203483Szec printf("%s() line %d, prev in %s()", vnr->where_fn, vnr->where_line, 654203483Szec vnr->prev_fn); 655203483Szec if (brief) 656203483Szec printf(", "); 657203483Szec else 658203483Szec printf("\n "); 659203483Szec printf("%p -> %p\n", vnr->old_vnet, vnr->new_vnet); 660203483Szec} 661203483Szec 662203483Szecvoid 663203483Szecvnet_log_recursion(struct vnet *old_vnet, const char *old_fn, int line) 664203483Szec{ 665203483Szec struct vnet_recursion *vnr; 666203483Szec 667203483Szec /* Skip already logged recursion events. */ 668203483Szec SLIST_FOREACH(vnr, &vnet_recursions, vnr_le) 669203483Szec if (vnr->prev_fn == old_fn && 670203483Szec vnr->where_fn == curthread->td_vnet_lpush && 671203483Szec vnr->where_line == line && 672203483Szec (vnr->old_vnet == vnr->new_vnet) == (curvnet == old_vnet)) 673203483Szec return; 674203483Szec 675203483Szec vnr = malloc(sizeof(*vnr), M_VNET, M_NOWAIT | M_ZERO); 676203483Szec if (vnr == NULL) 677203483Szec panic("%s: malloc failed", __func__); 678203483Szec vnr->prev_fn = old_fn; 679203483Szec vnr->where_fn = curthread->td_vnet_lpush; 680203483Szec vnr->where_line = line; 681203483Szec vnr->old_vnet = old_vnet; 682203483Szec vnr->new_vnet = curvnet; 683203483Szec 684203483Szec SLIST_INSERT_HEAD(&vnet_recursions, vnr, vnr_le); 685203483Szec 686203483Szec vnet_print_recursion(vnr, 0); 687203483Szec#ifdef KDB 688203483Szec kdb_backtrace(); 689203483Szec#endif 690203483Szec} 691203483Szec#endif /* VNET_DEBUG */ 692203483Szec 693205345Sbz/* 694205345Sbz * DDB(4). 695205345Sbz */ 696196019Srwatson#ifdef DDB 697300148Sbzstatic void 698300148Sbzdb_vnet_print(struct vnet *vnet) 699196019Srwatson{ 700300148Sbz 701300148Sbz db_printf("vnet = %p\n", vnet); 702300148Sbz db_printf(" vnet_magic_n = %#08x (%s, orig %#08x)\n", 703300148Sbz vnet->vnet_magic_n, 704300148Sbz (vnet->vnet_magic_n == VNET_MAGIC_N) ? 705300148Sbz "ok" : "mismatch", VNET_MAGIC_N); 706300148Sbz db_printf(" vnet_ifcnt = %u\n", vnet->vnet_ifcnt); 707300148Sbz db_printf(" vnet_sockcnt = %u\n", vnet->vnet_sockcnt); 708300148Sbz db_printf(" vnet_data_mem = %p\n", vnet->vnet_data_mem); 709300148Sbz db_printf(" vnet_data_base = %#jx\n", 710300148Sbz (uintmax_t)vnet->vnet_data_base); 711300155Sbz db_printf(" vnet_state = %#08x\n", vnet->vnet_state); 712300148Sbz db_printf("\n"); 713300148Sbz} 714300148Sbz 715300148SbzDB_SHOW_ALL_COMMAND(vnets, db_show_all_vnets) 716300148Sbz{ 717196019Srwatson VNET_ITERATOR_DECL(vnet_iter); 718196019Srwatson 719196019Srwatson VNET_FOREACH(vnet_iter) { 720300148Sbz db_vnet_print(vnet_iter); 721196129Sbz if (db_pager_quit) 722196129Sbz break; 723196019Srwatson } 724196019Srwatson} 725203483Szec 726300148SbzDB_SHOW_COMMAND(vnet, db_show_vnet) 727300148Sbz{ 728300148Sbz 729300148Sbz if (!have_addr) { 730300148Sbz db_printf("usage: show vnet <struct vnet *>\n"); 731300148Sbz return; 732300148Sbz } 733300148Sbz 734300148Sbz db_vnet_print((struct vnet *)addr); 735300148Sbz} 736300148Sbz 737203729Sbzstatic void 738203729Sbzdb_show_vnet_print_vs(struct vnet_sysinit *vs, int ddb) 739203729Sbz{ 740203729Sbz const char *vsname, *funcname; 741203729Sbz c_db_sym_t sym; 742203729Sbz db_expr_t offset; 743203729Sbz 744203729Sbz#define xprint(...) \ 745203729Sbz if (ddb) \ 746203729Sbz db_printf(__VA_ARGS__); \ 747203729Sbz else \ 748203729Sbz printf(__VA_ARGS__) 749203729Sbz 750203729Sbz if (vs == NULL) { 751203729Sbz xprint("%s: no vnet_sysinit * given\n", __func__); 752203729Sbz return; 753203729Sbz } 754203729Sbz 755203729Sbz sym = db_search_symbol((vm_offset_t)vs, DB_STGY_ANY, &offset); 756203729Sbz db_symbol_values(sym, &vsname, NULL); 757203729Sbz sym = db_search_symbol((vm_offset_t)vs->func, DB_STGY_PROC, &offset); 758203729Sbz db_symbol_values(sym, &funcname, NULL); 759203729Sbz xprint("%s(%p)\n", (vsname != NULL) ? vsname : "", vs); 760300148Sbz xprint(" %#08x %#08x\n", vs->subsystem, vs->order); 761203729Sbz xprint(" %p(%s)(%p)\n", 762203729Sbz vs->func, (funcname != NULL) ? funcname : "", vs->arg); 763203729Sbz#undef xprint 764203729Sbz} 765203729Sbz 766203729SbzDB_SHOW_COMMAND(vnet_sysinit, db_show_vnet_sysinit) 767203729Sbz{ 768203729Sbz struct vnet_sysinit *vs; 769203729Sbz 770203729Sbz db_printf("VNET_SYSINIT vs Name(Ptr)\n"); 771203729Sbz db_printf(" Subsystem Order\n"); 772203729Sbz db_printf(" Function(Name)(Arg)\n"); 773203729Sbz TAILQ_FOREACH(vs, &vnet_constructors, link) { 774203729Sbz db_show_vnet_print_vs(vs, 1); 775203729Sbz if (db_pager_quit) 776203729Sbz break; 777203729Sbz } 778203729Sbz} 779203729Sbz 780203729SbzDB_SHOW_COMMAND(vnet_sysuninit, db_show_vnet_sysuninit) 781203729Sbz{ 782203729Sbz struct vnet_sysinit *vs; 783203729Sbz 784203729Sbz db_printf("VNET_SYSUNINIT vs Name(Ptr)\n"); 785203729Sbz db_printf(" Subsystem Order\n"); 786203729Sbz db_printf(" Function(Name)(Arg)\n"); 787203729Sbz TAILQ_FOREACH_REVERSE(vs, &vnet_destructors, vnet_sysuninit_head, 788203729Sbz link) { 789203729Sbz db_show_vnet_print_vs(vs, 1); 790203729Sbz if (db_pager_quit) 791203729Sbz break; 792203729Sbz } 793203729Sbz} 794203729Sbz 795203483Szec#ifdef VNET_DEBUG 796203483SzecDB_SHOW_COMMAND(vnetrcrs, db_show_vnetrcrs) 797203483Szec{ 798203483Szec struct vnet_recursion *vnr; 799203483Szec 800203483Szec SLIST_FOREACH(vnr, &vnet_recursions, vnr_le) 801203483Szec vnet_print_recursion(vnr, 1); 802203483Szec} 803196019Srwatson#endif 804203483Szec#endif /* DDB */ 805