The Design and Implementation of the FreeBSD Operating System, Second Edition
Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]

FreeBSD/Linux Kernel Cross Reference
sys/net/vnet.c

Version: -  FREEBSD  -  FREEBSD-13-STABLE  -  FREEBSD-13-0  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  l41  -  OPENBSD  -  linux-2.6  -  MK84  -  PLAN9  -  xnu-8792 
SearchContext: -  none  -  3  -  10 

    1 /*-
    2  * SPDX-License-Identifier: BSD-2-Clause-FreeBSD
    3  *
    4  * Copyright (c) 2004-2009 University of Zagreb
    5  * Copyright (c) 2006-2009 FreeBSD Foundation
    6  * All rights reserved.
    7  *
    8  * This software was developed by the University of Zagreb and the
    9  * FreeBSD Foundation under sponsorship by the Stichting NLnet and the
   10  * FreeBSD Foundation.
   11  *
   12  * Copyright (c) 2009 Jeffrey Roberson <jeff@freebsd.org>
   13  * Copyright (c) 2009 Robert N. M. Watson
   14  * All rights reserved.
   15  *
   16  * Redistribution and use in source and binary forms, with or without
   17  * modification, are permitted provided that the following conditions
   18  * are met:
   19  * 1. Redistributions of source code must retain the above copyright
   20  *    notice, this list of conditions and the following disclaimer.
   21  * 2. Redistributions in binary form must reproduce the above copyright
   22  *    notice, this list of conditions and the following disclaimer in the
   23  *    documentation and/or other materials provided with the distribution.
   24  *
   25  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
   26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   28  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
   29  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   30  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   31  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   32  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   34  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   35  * SUCH DAMAGE.
   36  */
   37 
   38 #include <sys/cdefs.h>
   39 __FBSDID("$FreeBSD$");
   40 
   41 #include "opt_ddb.h"
   42 #include "opt_kdb.h"
   43 
   44 #include <sys/param.h>
   45 #include <sys/kdb.h>
   46 #include <sys/kernel.h>
   47 #include <sys/jail.h>
   48 #include <sys/sdt.h>
   49 #include <sys/systm.h>
   50 #include <sys/sysctl.h>
   51 #include <sys/eventhandler.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 FEATURE(vimage, "VIMAGE kernel virtualization");
   84 
   85 static MALLOC_DEFINE(M_VNET, "vnet", "network stack control block");
   86 
   87 /*
   88  * The virtual network stack list has two read-write locks, one sleepable and
   89  * the other not, so that the list can be stablized and walked in a variety
   90  * of network stack contexts.  Both must be acquired exclusively to modify
   91  * the list, but a read lock of either lock is sufficient to walk the list.
   92  */
   93 struct rwlock           vnet_rwlock;
   94 struct sx               vnet_sxlock;
   95 
   96 #define VNET_LIST_WLOCK() do {                                          \
   97         sx_xlock(&vnet_sxlock);                                         \
   98         rw_wlock(&vnet_rwlock);                                         \
   99 } while (0)
  100 
  101 #define VNET_LIST_WUNLOCK() do {                                        \
  102         rw_wunlock(&vnet_rwlock);                                       \
  103         sx_xunlock(&vnet_sxlock);                                       \
  104 } while (0)
  105 
  106 struct vnet_list_head vnet_head;
  107 struct vnet *vnet0;
  108 
  109 /*
  110  * The virtual network stack allocator provides storage for virtualized
  111  * global variables.  These variables are defined/declared using the
  112  * VNET_DEFINE()/VNET_DECLARE() macros, which place them in the 'set_vnet'
  113  * linker set.  The details of the implementation are somewhat subtle, but
  114  * allow the majority of most network subsystems to maintain
  115  * virtualization-agnostic.
  116  *
  117  * The virtual network stack allocator handles variables in the base kernel
  118  * vs. modules in similar but different ways.  In both cases, virtualized
  119  * global variables are marked as such by being declared to be part of the
  120  * vnet linker set.  These "master" copies of global variables serve two
  121  * functions:
  122  *
  123  * (1) They contain static initialization or "default" values for global
  124  *     variables which will be propagated to each virtual network stack
  125  *     instance when created.  As with normal global variables, they default
  126  *     to zero-filled.
  127  *
  128  * (2) They act as unique global names by which the variable can be referred
  129  *     to, regardless of network stack instance.  The single global symbol
  130  *     will be used to calculate the location of a per-virtual instance
  131  *     variable at run-time.
  132  *
  133  * Each virtual network stack instance has a complete copy of each
  134  * virtualized global variable, stored in a malloc'd block of memory
  135  * referred to by vnet->vnet_data_mem.  Critical to the design is that each
  136  * per-instance memory block is laid out identically to the master block so
  137  * that the offset of each global variable is the same across all blocks.  To
  138  * optimize run-time access, a precalculated 'base' address,
  139  * vnet->vnet_data_base, is stored in each vnet, and is the amount that can
  140  * be added to the address of a 'master' instance of a variable to get to the
  141  * per-vnet instance.
  142  *
  143  * Virtualized global variables are handled in a similar manner, but as each
  144  * module has its own 'set_vnet' linker set, and we want to keep all
  145  * virtualized globals togther, we reserve space in the kernel's linker set
  146  * for potential module variables using a per-vnet character array,
  147  * 'modspace'.  The virtual network stack allocator maintains a free list to
  148  * track what space in the array is free (all, initially) and as modules are
  149  * linked, allocates portions of the space to specific globals.  The kernel
  150  * module linker queries the virtual network stack allocator and will
  151  * bind references of the global to the location during linking.  It also
  152  * calls into the virtual network stack allocator, once the memory is
  153  * initialized, in order to propagate the new static initializations to all
  154  * existing virtual network stack instances so that the soon-to-be executing
  155  * module will find every network stack instance with proper default values.
  156  */
  157 
  158 /*
  159  * Number of bytes of data in the 'set_vnet' linker set, and hence the total
  160  * size of all kernel virtualized global variables, and the malloc(9) type
  161  * that will be used to allocate it.
  162  */
  163 #define VNET_BYTES      (VNET_STOP - VNET_START)
  164 
  165 static MALLOC_DEFINE(M_VNET_DATA, "vnet_data", "VNET data");
  166 
  167 /*
  168  * VNET_MODMIN is the minimum number of bytes we will reserve for the sum of
  169  * global variables across all loaded modules.  As this actually sizes an
  170  * array declared as a virtualized global variable in the kernel itself, and
  171  * we want the virtualized global variable space to be page-sized, we may
  172  * have more space than that in practice.
  173  */
  174 #define VNET_MODMIN     (8 * PAGE_SIZE)
  175 #define VNET_SIZE       roundup2(VNET_BYTES, PAGE_SIZE)
  176 
  177 /*
  178  * Space to store virtualized global variables from loadable kernel modules,
  179  * and the free list to manage it.
  180  */
  181 VNET_DEFINE_STATIC(char, modspace[VNET_MODMIN] __aligned(__alignof(void *)));
  182 
  183 /*
  184  * Global lists of subsystem constructor and destructors for vnets.  They are
  185  * registered via VNET_SYSINIT() and VNET_SYSUNINIT().  Both lists are
  186  * protected by the vnet_sysinit_sxlock global lock.
  187  */
  188 static TAILQ_HEAD(vnet_sysinit_head, vnet_sysinit) vnet_constructors =
  189         TAILQ_HEAD_INITIALIZER(vnet_constructors);
  190 static TAILQ_HEAD(vnet_sysuninit_head, vnet_sysinit) vnet_destructors =
  191         TAILQ_HEAD_INITIALIZER(vnet_destructors);
  192 
  193 struct sx               vnet_sysinit_sxlock;
  194 
  195 #define VNET_SYSINIT_WLOCK()    sx_xlock(&vnet_sysinit_sxlock);
  196 #define VNET_SYSINIT_WUNLOCK()  sx_xunlock(&vnet_sysinit_sxlock);
  197 #define VNET_SYSINIT_RLOCK()    sx_slock(&vnet_sysinit_sxlock);
  198 #define VNET_SYSINIT_RUNLOCK()  sx_sunlock(&vnet_sysinit_sxlock);
  199 
  200 struct vnet_data_free {
  201         uintptr_t       vnd_start;
  202         int             vnd_len;
  203         TAILQ_ENTRY(vnet_data_free) vnd_link;
  204 };
  205 
  206 static MALLOC_DEFINE(M_VNET_DATA_FREE, "vnet_data_free",
  207     "VNET resource accounting");
  208 static TAILQ_HEAD(, vnet_data_free) vnet_data_free_head =
  209             TAILQ_HEAD_INITIALIZER(vnet_data_free_head);
  210 static struct sx vnet_data_free_lock;
  211 
  212 SDT_PROVIDER_DEFINE(vnet);
  213 SDT_PROBE_DEFINE1(vnet, functions, vnet_alloc, entry, "int");
  214 SDT_PROBE_DEFINE2(vnet, functions, vnet_alloc, alloc, "int",
  215     "struct vnet *");
  216 SDT_PROBE_DEFINE2(vnet, functions, vnet_alloc, return,
  217     "int", "struct vnet *");
  218 SDT_PROBE_DEFINE2(vnet, functions, vnet_destroy, entry,
  219     "int", "struct vnet *");
  220 SDT_PROBE_DEFINE1(vnet, functions, vnet_destroy, return,
  221     "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 
  274         SDT_PROBE2(vnet, functions, vnet_destroy, entry, __LINE__, vnet);
  275         KASSERT(vnet->vnet_sockcnt == 0,
  276             ("%s: vnet still has sockets", __func__));
  277 
  278         VNET_LIST_WLOCK();
  279         LIST_REMOVE(vnet, vnet_le);
  280         VNET_LIST_WUNLOCK();
  281 
  282         /* Signal that VNET is being shutdown. */
  283         vnet->vnet_shutdown = true;
  284 
  285         CURVNET_SET_QUIET(vnet);
  286         sx_xlock(&ifnet_detach_sxlock);
  287         vnet_sysuninit();
  288         sx_xunlock(&ifnet_detach_sxlock);
  289         CURVNET_RESTORE();
  290 
  291         /*
  292          * Release storage for the virtual network stack instance.
  293          */
  294         free(vnet->vnet_data_mem, M_VNET_DATA);
  295         vnet->vnet_data_mem = NULL;
  296         vnet->vnet_data_base = 0;
  297         vnet->vnet_magic_n = 0xdeadbeef;
  298         free(vnet, M_VNET);
  299         SDT_PROBE1(vnet, functions, vnet_destroy, return, __LINE__);
  300 }
  301 
  302 /*
  303  * Boot time initialization and allocation of virtual network stacks.
  304  */
  305 static void
  306 vnet_init_prelink(void *arg __unused)
  307 {
  308 
  309         rw_init(&vnet_rwlock, "vnet_rwlock");
  310         sx_init(&vnet_sxlock, "vnet_sxlock");
  311         sx_init(&vnet_sysinit_sxlock, "vnet_sysinit_sxlock");
  312         LIST_INIT(&vnet_head);
  313 }
  314 SYSINIT(vnet_init_prelink, SI_SUB_VNET_PRELINK, SI_ORDER_FIRST,
  315     vnet_init_prelink, NULL);
  316 
  317 static void
  318 vnet0_init(void *arg __unused)
  319 {
  320 
  321         if (bootverbose)
  322                 printf("VIMAGE (virtualized network stack) enabled\n");
  323 
  324         /*
  325          * We MUST clear curvnet in vi_init_done() before going SMP,
  326          * otherwise CURVNET_SET() macros would scream about unnecessary
  327          * curvnet recursions.
  328          */
  329         curvnet = prison0.pr_vnet = vnet0 = vnet_alloc();
  330 }
  331 SYSINIT(vnet0_init, SI_SUB_VNET, SI_ORDER_FIRST, vnet0_init, NULL);
  332 
  333 static void
  334 vnet_init_done(void *unused __unused)
  335 {
  336 
  337         curvnet = NULL;
  338 }
  339 SYSINIT(vnet_init_done, SI_SUB_VNET_DONE, SI_ORDER_ANY, vnet_init_done,
  340     NULL);
  341 
  342 /*
  343  * Once on boot, initialize the modspace freelist to entirely cover modspace.
  344  */
  345 static void
  346 vnet_data_startup(void *dummy __unused)
  347 {
  348         struct vnet_data_free *df;
  349 
  350         df = malloc(sizeof(*df), M_VNET_DATA_FREE, M_WAITOK | M_ZERO);
  351         df->vnd_start = (uintptr_t)&VNET_NAME(modspace);
  352         df->vnd_len = VNET_MODMIN;
  353         TAILQ_INSERT_HEAD(&vnet_data_free_head, df, vnd_link);
  354         sx_init(&vnet_data_free_lock, "vnet_data alloc lock");
  355 }
  356 SYSINIT(vnet_data, SI_SUB_KLD, SI_ORDER_FIRST, vnet_data_startup, NULL);
  357 
  358 /* Dummy VNET_SYSINIT to make sure we always reach the final end state. */
  359 static void
  360 vnet_sysinit_done(void *unused __unused)
  361 {
  362 
  363         return;
  364 }
  365 VNET_SYSINIT(vnet_sysinit_done, SI_SUB_VNET_DONE, SI_ORDER_ANY,
  366     vnet_sysinit_done, NULL);
  367 
  368 /*
  369  * When a module is loaded and requires storage for a virtualized global
  370  * variable, allocate space from the modspace free list.  This interface
  371  * should be used only by the kernel linker.
  372  */
  373 void *
  374 vnet_data_alloc(int size)
  375 {
  376         struct vnet_data_free *df;
  377         void *s;
  378 
  379         s = NULL;
  380         size = roundup2(size, sizeof(void *));
  381         sx_xlock(&vnet_data_free_lock);
  382         TAILQ_FOREACH(df, &vnet_data_free_head, vnd_link) {
  383                 if (df->vnd_len < size)
  384                         continue;
  385                 if (df->vnd_len == size) {
  386                         s = (void *)df->vnd_start;
  387                         TAILQ_REMOVE(&vnet_data_free_head, df, vnd_link);
  388                         free(df, M_VNET_DATA_FREE);
  389                         break;
  390                 }
  391                 s = (void *)df->vnd_start;
  392                 df->vnd_len -= size;
  393                 df->vnd_start = df->vnd_start + size;
  394                 break;
  395         }
  396         sx_xunlock(&vnet_data_free_lock);
  397 
  398         return (s);
  399 }
  400 
  401 /*
  402  * Free space for a virtualized global variable on module unload.
  403  */
  404 void
  405 vnet_data_free(void *start_arg, int size)
  406 {
  407         struct vnet_data_free *df;
  408         struct vnet_data_free *dn;
  409         uintptr_t start;
  410         uintptr_t end;
  411 
  412         size = roundup2(size, sizeof(void *));
  413         start = (uintptr_t)start_arg;
  414         end = start + size;
  415         /*
  416          * Free a region of space and merge it with as many neighbors as
  417          * possible.  Keeping the list sorted simplifies this operation.
  418          */
  419         sx_xlock(&vnet_data_free_lock);
  420         TAILQ_FOREACH(df, &vnet_data_free_head, vnd_link) {
  421                 if (df->vnd_start > end)
  422                         break;
  423                 /*
  424                  * If we expand at the end of an entry we may have to merge
  425                  * it with the one following it as well.
  426                  */
  427                 if (df->vnd_start + df->vnd_len == start) {
  428                         df->vnd_len += size;
  429                         dn = TAILQ_NEXT(df, vnd_link);
  430                         if (df->vnd_start + df->vnd_len == dn->vnd_start) {
  431                                 df->vnd_len += dn->vnd_len;
  432                                 TAILQ_REMOVE(&vnet_data_free_head, dn,
  433                                     vnd_link);
  434                                 free(dn, M_VNET_DATA_FREE);
  435                         }
  436                         sx_xunlock(&vnet_data_free_lock);
  437                         return;
  438                 }
  439                 if (df->vnd_start == end) {
  440                         df->vnd_start = start;
  441                         df->vnd_len += size;
  442                         sx_xunlock(&vnet_data_free_lock);
  443                         return;
  444                 }
  445         }
  446         dn = malloc(sizeof(*df), M_VNET_DATA_FREE, M_WAITOK | M_ZERO);
  447         dn->vnd_start = start;
  448         dn->vnd_len = size;
  449         if (df)
  450                 TAILQ_INSERT_BEFORE(df, dn, vnd_link);
  451         else
  452                 TAILQ_INSERT_TAIL(&vnet_data_free_head, dn, vnd_link);
  453         sx_xunlock(&vnet_data_free_lock);
  454 }
  455 
  456 /*
  457  * When a new virtualized global variable has been allocated, propagate its
  458  * initial value to each already-allocated virtual network stack instance.
  459  */
  460 void
  461 vnet_data_copy(void *start, int size)
  462 {
  463         struct vnet *vnet;
  464 
  465         VNET_LIST_RLOCK();
  466         LIST_FOREACH(vnet, &vnet_head, vnet_le)
  467                 memcpy((void *)((uintptr_t)vnet->vnet_data_base +
  468                     (uintptr_t)start), start, size);
  469         VNET_LIST_RUNLOCK();
  470 }
  471 
  472 /*
  473  * Support for special SYSINIT handlers registered via VNET_SYSINIT()
  474  * and VNET_SYSUNINIT().
  475  */
  476 void
  477 vnet_register_sysinit(void *arg)
  478 {
  479         struct vnet_sysinit *vs, *vs2;  
  480         struct vnet *vnet;
  481 
  482         vs = arg;
  483         KASSERT(vs->subsystem > SI_SUB_VNET, ("vnet sysinit too early"));
  484 
  485         /* Add the constructor to the global list of vnet constructors. */
  486         VNET_SYSINIT_WLOCK();
  487         TAILQ_FOREACH(vs2, &vnet_constructors, link) {
  488                 if (vs2->subsystem > vs->subsystem)
  489                         break;
  490                 if (vs2->subsystem == vs->subsystem && vs2->order > vs->order)
  491                         break;
  492         }
  493         if (vs2 != NULL)
  494                 TAILQ_INSERT_BEFORE(vs2, vs, link);
  495         else
  496                 TAILQ_INSERT_TAIL(&vnet_constructors, vs, link);
  497 
  498         /*
  499          * Invoke the constructor on all the existing vnets when it is
  500          * registered.
  501          */
  502         VNET_FOREACH(vnet) {
  503                 CURVNET_SET_QUIET(vnet);
  504                 vs->func(vs->arg);
  505                 CURVNET_RESTORE();
  506         }
  507         VNET_SYSINIT_WUNLOCK();
  508 }
  509 
  510 void
  511 vnet_deregister_sysinit(void *arg)
  512 {
  513         struct vnet_sysinit *vs;
  514 
  515         vs = arg;
  516 
  517         /* Remove the constructor from the global list of vnet constructors. */
  518         VNET_SYSINIT_WLOCK();
  519         TAILQ_REMOVE(&vnet_constructors, vs, link);
  520         VNET_SYSINIT_WUNLOCK();
  521 }
  522 
  523 void
  524 vnet_register_sysuninit(void *arg)
  525 {
  526         struct vnet_sysinit *vs, *vs2;
  527 
  528         vs = arg;
  529 
  530         /* Add the destructor to the global list of vnet destructors. */
  531         VNET_SYSINIT_WLOCK();
  532         TAILQ_FOREACH(vs2, &vnet_destructors, link) {
  533                 if (vs2->subsystem > vs->subsystem)
  534                         break;
  535                 if (vs2->subsystem == vs->subsystem && vs2->order > vs->order)
  536                         break;
  537         }
  538         if (vs2 != NULL)
  539                 TAILQ_INSERT_BEFORE(vs2, vs, link);
  540         else
  541                 TAILQ_INSERT_TAIL(&vnet_destructors, vs, link);
  542         VNET_SYSINIT_WUNLOCK();
  543 }
  544 
  545 void
  546 vnet_deregister_sysuninit(void *arg)
  547 {
  548         struct vnet_sysinit *vs;
  549         struct vnet *vnet;
  550 
  551         vs = arg;
  552 
  553         /*
  554          * Invoke the destructor on all the existing vnets when it is
  555          * deregistered.
  556          */
  557         VNET_SYSINIT_WLOCK();
  558         VNET_FOREACH(vnet) {
  559                 CURVNET_SET_QUIET(vnet);
  560                 vs->func(vs->arg);
  561                 CURVNET_RESTORE();
  562         }
  563 
  564         /* Remove the destructor from the global list of vnet destructors. */
  565         TAILQ_REMOVE(&vnet_destructors, vs, link);
  566         VNET_SYSINIT_WUNLOCK();
  567 }
  568 
  569 /*
  570  * Invoke all registered vnet constructors on the current vnet.  Used during
  571  * vnet construction.  The caller is responsible for ensuring the new vnet is
  572  * the current vnet and that the vnet_sysinit_sxlock lock is locked.
  573  */
  574 void
  575 vnet_sysinit(void)
  576 {
  577         struct vnet_sysinit *vs;
  578 
  579         VNET_SYSINIT_RLOCK();
  580         TAILQ_FOREACH(vs, &vnet_constructors, link) {
  581                 curvnet->vnet_state = vs->subsystem;
  582                 vs->func(vs->arg);
  583         }
  584         VNET_SYSINIT_RUNLOCK();
  585 }
  586 
  587 /*
  588  * Invoke all registered vnet destructors on the current vnet.  Used during
  589  * vnet destruction.  The caller is responsible for ensuring the dying vnet
  590  * the current vnet and that the vnet_sysinit_sxlock lock is locked.
  591  */
  592 void
  593 vnet_sysuninit(void)
  594 {
  595         struct vnet_sysinit *vs;
  596 
  597         VNET_SYSINIT_RLOCK();
  598         TAILQ_FOREACH_REVERSE(vs, &vnet_destructors, vnet_sysuninit_head,
  599             link) {
  600                 curvnet->vnet_state = vs->subsystem;
  601                 vs->func(vs->arg);
  602         }
  603         VNET_SYSINIT_RUNLOCK();
  604 }
  605 
  606 /*
  607  * EVENTHANDLER(9) extensions.
  608  */
  609 /*
  610  * Invoke the eventhandler function originally registered with the possibly
  611  * registered argument for all virtual network stack instances.
  612  *
  613  * This iterator can only be used for eventhandlers that do not take any
  614  * additional arguments, as we do ignore the variadic arguments from the
  615  * EVENTHANDLER_INVOKE() call.
  616  */
  617 void
  618 vnet_global_eventhandler_iterator_func(void *arg, ...)
  619 {
  620         VNET_ITERATOR_DECL(vnet_iter);
  621         struct eventhandler_entry_vimage *v_ee;
  622 
  623         /*
  624          * There is a bug here in that we should actually cast things to
  625          * (struct eventhandler_entry_ ## name *)  but that's not easily
  626          * possible in here so just re-using the variadic version we
  627          * defined for the generic vimage case.
  628          */
  629         v_ee = arg;
  630         VNET_LIST_RLOCK();
  631         VNET_FOREACH(vnet_iter) {
  632                 CURVNET_SET(vnet_iter);
  633                 ((vimage_iterator_func_t)v_ee->func)(v_ee->ee_arg);
  634                 CURVNET_RESTORE();
  635         }
  636         VNET_LIST_RUNLOCK();
  637 }
  638 
  639 #ifdef VNET_DEBUG
  640 struct vnet_recursion {
  641         SLIST_ENTRY(vnet_recursion)      vnr_le;
  642         const char                      *prev_fn;
  643         const char                      *where_fn;
  644         int                              where_line;
  645         struct vnet                     *old_vnet;
  646         struct vnet                     *new_vnet;
  647 };
  648 
  649 static SLIST_HEAD(, vnet_recursion) vnet_recursions =
  650     SLIST_HEAD_INITIALIZER(vnet_recursions);
  651 
  652 static void
  653 vnet_print_recursion(struct vnet_recursion *vnr, int brief)
  654 {
  655 
  656         if (!brief)
  657                 printf("CURVNET_SET() recursion in ");
  658         printf("%s() line %d, prev in %s()", vnr->where_fn, vnr->where_line,
  659             vnr->prev_fn);
  660         if (brief)
  661                 printf(", ");
  662         else
  663                 printf("\n    ");
  664         printf("%p -> %p\n", vnr->old_vnet, vnr->new_vnet);
  665 }
  666 
  667 void
  668 vnet_log_recursion(struct vnet *old_vnet, const char *old_fn, int line)
  669 {
  670         struct vnet_recursion *vnr;
  671 
  672         /* Skip already logged recursion events. */
  673         SLIST_FOREACH(vnr, &vnet_recursions, vnr_le)
  674                 if (vnr->prev_fn == old_fn &&
  675                     vnr->where_fn == curthread->td_vnet_lpush &&
  676                     vnr->where_line == line &&
  677                     (vnr->old_vnet == vnr->new_vnet) == (curvnet == old_vnet))
  678                         return;
  679 
  680         vnr = malloc(sizeof(*vnr), M_VNET, M_NOWAIT | M_ZERO);
  681         if (vnr == NULL)
  682                 panic("%s: malloc failed", __func__);
  683         vnr->prev_fn = old_fn;
  684         vnr->where_fn = curthread->td_vnet_lpush;
  685         vnr->where_line = line;
  686         vnr->old_vnet = old_vnet;
  687         vnr->new_vnet = curvnet;
  688 
  689         SLIST_INSERT_HEAD(&vnet_recursions, vnr, vnr_le);
  690 
  691         vnet_print_recursion(vnr, 0);
  692 #ifdef KDB
  693         kdb_backtrace();
  694 #endif
  695 }
  696 #endif /* VNET_DEBUG */
  697 
  698 /*
  699  * DDB(4).
  700  */
  701 #ifdef DDB
  702 static void
  703 db_vnet_print(struct vnet *vnet)
  704 {
  705 
  706         db_printf("vnet            = %p\n", vnet);
  707         db_printf(" vnet_magic_n   = %#08x (%s, orig %#08x)\n",
  708             vnet->vnet_magic_n,
  709             (vnet->vnet_magic_n == VNET_MAGIC_N) ?
  710                 "ok" : "mismatch", VNET_MAGIC_N);
  711         db_printf(" vnet_ifcnt     = %u\n", vnet->vnet_ifcnt);
  712         db_printf(" vnet_sockcnt   = %u\n", vnet->vnet_sockcnt);
  713         db_printf(" vnet_data_mem  = %p\n", vnet->vnet_data_mem);
  714         db_printf(" vnet_data_base = %#jx\n",
  715             (uintmax_t)vnet->vnet_data_base);
  716         db_printf(" vnet_state     = %#08x\n", vnet->vnet_state);
  717         db_printf(" vnet_shutdown  = %#03x\n", vnet->vnet_shutdown);
  718         db_printf("\n");
  719 }
  720 
  721 DB_SHOW_ALL_COMMAND(vnets, db_show_all_vnets)
  722 {
  723         VNET_ITERATOR_DECL(vnet_iter);
  724 
  725         VNET_FOREACH(vnet_iter) {
  726                 db_vnet_print(vnet_iter);
  727                 if (db_pager_quit)
  728                         break;
  729         }
  730 }
  731 
  732 DB_SHOW_COMMAND(vnet, db_show_vnet)
  733 {
  734 
  735         if (!have_addr) {
  736                 db_printf("usage: show vnet <struct vnet *>\n");
  737                 return;
  738         }
  739 
  740         db_vnet_print((struct vnet *)addr);
  741 }
  742 
  743 static void
  744 db_show_vnet_print_vs(struct vnet_sysinit *vs, int ddb)
  745 {
  746         const char *vsname, *funcname;
  747         c_db_sym_t sym;
  748         db_expr_t  offset;
  749 
  750 #define xprint(...)                                                     \
  751         if (ddb)                                                        \
  752                 db_printf(__VA_ARGS__);                                 \
  753         else                                                            \
  754                 printf(__VA_ARGS__)
  755 
  756         if (vs == NULL) {
  757                 xprint("%s: no vnet_sysinit * given\n", __func__);
  758                 return;
  759         }
  760 
  761         sym = db_search_symbol((vm_offset_t)vs, DB_STGY_ANY, &offset);
  762         db_symbol_values(sym, &vsname, NULL);
  763         sym = db_search_symbol((vm_offset_t)vs->func, DB_STGY_PROC, &offset);
  764         db_symbol_values(sym, &funcname, NULL);
  765         xprint("%s(%p)\n", (vsname != NULL) ? vsname : "", vs);
  766         xprint("  %#08x %#08x\n", vs->subsystem, vs->order);
  767         xprint("  %p(%s)(%p)\n",
  768             vs->func, (funcname != NULL) ? funcname : "", vs->arg);
  769 #undef xprint
  770 }
  771 
  772 DB_SHOW_COMMAND_FLAGS(vnet_sysinit, db_show_vnet_sysinit, DB_CMD_MEMSAFE)
  773 {
  774         struct vnet_sysinit *vs;
  775 
  776         db_printf("VNET_SYSINIT vs Name(Ptr)\n");
  777         db_printf("  Subsystem  Order\n");
  778         db_printf("  Function(Name)(Arg)\n");
  779         TAILQ_FOREACH(vs, &vnet_constructors, link) {
  780                 db_show_vnet_print_vs(vs, 1);
  781                 if (db_pager_quit)
  782                         break;
  783         }
  784 }
  785 
  786 DB_SHOW_COMMAND_FLAGS(vnet_sysuninit, db_show_vnet_sysuninit, DB_CMD_MEMSAFE)
  787 {
  788         struct vnet_sysinit *vs;
  789 
  790         db_printf("VNET_SYSUNINIT vs Name(Ptr)\n");
  791         db_printf("  Subsystem  Order\n");
  792         db_printf("  Function(Name)(Arg)\n");
  793         TAILQ_FOREACH_REVERSE(vs, &vnet_destructors, vnet_sysuninit_head,
  794             link) {
  795                 db_show_vnet_print_vs(vs, 1);
  796                 if (db_pager_quit)
  797                         break;
  798         }
  799 }
  800 
  801 #ifdef VNET_DEBUG
  802 DB_SHOW_COMMAND_FLAGS(vnetrcrs, db_show_vnetrcrs, DB_CMD_MEMSAFE)
  803 {
  804         struct vnet_recursion *vnr;
  805 
  806         SLIST_FOREACH(vnr, &vnet_recursions, vnr_le)
  807                 vnet_print_recursion(vnr, 1);
  808 }
  809 #endif
  810 #endif /* DDB */

Cache object: 55b697416569e5cb967995673a00ce84


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]


This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.