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         vnet->vnet_state = 0;
  239         SDT_PROBE2(vnet, functions, vnet_alloc, alloc, __LINE__, vnet);
  240 
  241         /*
  242          * Allocate storage for virtualized global variables and copy in
  243          * initial values form our 'master' copy.
  244          */
  245         vnet->vnet_data_mem = malloc(VNET_SIZE, M_VNET_DATA, M_WAITOK);
  246         memcpy(vnet->vnet_data_mem, (void *)VNET_START, VNET_BYTES);
  247 
  248         /*
  249          * All use of vnet-specific data will immediately subtract VNET_START
  250          * from the base memory pointer, so pre-calculate that now to avoid
  251          * it on each use.
  252          */
  253         vnet->vnet_data_base = (uintptr_t)vnet->vnet_data_mem - VNET_START;
  254 
  255         /* Initialize / attach vnet module instances. */
  256         CURVNET_SET_QUIET(vnet);
  257         vnet_sysinit();
  258         CURVNET_RESTORE();
  259 
  260         VNET_LIST_WLOCK();
  261         LIST_INSERT_HEAD(&vnet_head, vnet, vnet_le);
  262         VNET_LIST_WUNLOCK();
  263 
  264         SDT_PROBE2(vnet, functions, vnet_alloc, return, __LINE__, vnet);
  265         return (vnet);
  266 }
  267 
  268 /*
  269  * Destroy a virtual network stack.
  270  */
  271 void
  272 vnet_destroy(struct vnet *vnet)
  273 {
  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         sx_xlock(&ifnet_detach_sxlock);
  285         vnet_sysuninit();
  286         sx_xunlock(&ifnet_detach_sxlock);
  287         CURVNET_RESTORE();
  288 
  289         /*
  290          * Release storage for the virtual network stack instance.
  291          */
  292         free(vnet->vnet_data_mem, M_VNET_DATA);
  293         vnet->vnet_data_mem = NULL;
  294         vnet->vnet_data_base = 0;
  295         vnet->vnet_magic_n = 0xdeadbeef;
  296         free(vnet, M_VNET);
  297         SDT_PROBE1(vnet, functions, vnet_destroy, return, __LINE__);
  298 }
  299 
  300 /*
  301  * Boot time initialization and allocation of virtual network stacks.
  302  */
  303 static void
  304 vnet_init_prelink(void *arg __unused)
  305 {
  306 
  307         rw_init(&vnet_rwlock, "vnet_rwlock");
  308         sx_init(&vnet_sxlock, "vnet_sxlock");
  309         sx_init(&vnet_sysinit_sxlock, "vnet_sysinit_sxlock");
  310         LIST_INIT(&vnet_head);
  311 }
  312 SYSINIT(vnet_init_prelink, SI_SUB_VNET_PRELINK, SI_ORDER_FIRST,
  313     vnet_init_prelink, NULL);
  314 
  315 static void
  316 vnet0_init(void *arg __unused)
  317 {
  318 
  319         if (bootverbose)
  320                 printf("VIMAGE (virtualized network stack) enabled\n");
  321 
  322         /*
  323          * We MUST clear curvnet in vi_init_done() before going SMP,
  324          * otherwise CURVNET_SET() macros would scream about unnecessary
  325          * curvnet recursions.
  326          */
  327         curvnet = prison0.pr_vnet = vnet0 = vnet_alloc();
  328 }
  329 SYSINIT(vnet0_init, SI_SUB_VNET, SI_ORDER_FIRST, vnet0_init, NULL);
  330 
  331 static void
  332 vnet_init_done(void *unused __unused)
  333 {
  334 
  335         curvnet = NULL;
  336 }
  337 SYSINIT(vnet_init_done, SI_SUB_VNET_DONE, SI_ORDER_ANY, vnet_init_done,
  338     NULL);
  339 
  340 /*
  341  * Once on boot, initialize the modspace freelist to entirely cover modspace.
  342  */
  343 static void
  344 vnet_data_startup(void *dummy __unused)
  345 {
  346         struct vnet_data_free *df;
  347 
  348         df = malloc(sizeof(*df), M_VNET_DATA_FREE, M_WAITOK | M_ZERO);
  349         df->vnd_start = (uintptr_t)&VNET_NAME(modspace);
  350         df->vnd_len = VNET_MODMIN;
  351         TAILQ_INSERT_HEAD(&vnet_data_free_head, df, vnd_link);
  352         sx_init(&vnet_data_free_lock, "vnet_data alloc lock");
  353 }
  354 SYSINIT(vnet_data, SI_SUB_KLD, SI_ORDER_FIRST, vnet_data_startup, NULL);
  355 
  356 /* Dummy VNET_SYSINIT to make sure we always reach the final end state. */
  357 static void
  358 vnet_sysinit_done(void *unused __unused)
  359 {
  360 
  361         return;
  362 }
  363 VNET_SYSINIT(vnet_sysinit_done, SI_SUB_VNET_DONE, SI_ORDER_ANY,
  364     vnet_sysinit_done, NULL);
  365 
  366 /*
  367  * When a module is loaded and requires storage for a virtualized global
  368  * variable, allocate space from the modspace free list.  This interface
  369  * should be used only by the kernel linker.
  370  */
  371 void *
  372 vnet_data_alloc(int size)
  373 {
  374         struct vnet_data_free *df;
  375         void *s;
  376 
  377         s = NULL;
  378         size = roundup2(size, sizeof(void *));
  379         sx_xlock(&vnet_data_free_lock);
  380         TAILQ_FOREACH(df, &vnet_data_free_head, vnd_link) {
  381                 if (df->vnd_len < size)
  382                         continue;
  383                 if (df->vnd_len == size) {
  384                         s = (void *)df->vnd_start;
  385                         TAILQ_REMOVE(&vnet_data_free_head, df, vnd_link);
  386                         free(df, M_VNET_DATA_FREE);
  387                         break;
  388                 }
  389                 s = (void *)df->vnd_start;
  390                 df->vnd_len -= size;
  391                 df->vnd_start = df->vnd_start + size;
  392                 break;
  393         }
  394         sx_xunlock(&vnet_data_free_lock);
  395 
  396         return (s);
  397 }
  398 
  399 /*
  400  * Free space for a virtualized global variable on module unload.
  401  */
  402 void
  403 vnet_data_free(void *start_arg, int size)
  404 {
  405         struct vnet_data_free *df;
  406         struct vnet_data_free *dn;
  407         uintptr_t start;
  408         uintptr_t end;
  409 
  410         size = roundup2(size, sizeof(void *));
  411         start = (uintptr_t)start_arg;
  412         end = start + size;
  413         /*
  414          * Free a region of space and merge it with as many neighbors as
  415          * possible.  Keeping the list sorted simplifies this operation.
  416          */
  417         sx_xlock(&vnet_data_free_lock);
  418         TAILQ_FOREACH(df, &vnet_data_free_head, vnd_link) {
  419                 if (df->vnd_start > end)
  420                         break;
  421                 /*
  422                  * If we expand at the end of an entry we may have to merge
  423                  * it with the one following it as well.
  424                  */
  425                 if (df->vnd_start + df->vnd_len == start) {
  426                         df->vnd_len += size;
  427                         dn = TAILQ_NEXT(df, vnd_link);
  428                         if (df->vnd_start + df->vnd_len == dn->vnd_start) {
  429                                 df->vnd_len += dn->vnd_len;
  430                                 TAILQ_REMOVE(&vnet_data_free_head, dn,
  431                                     vnd_link);
  432                                 free(dn, M_VNET_DATA_FREE);
  433                         }
  434                         sx_xunlock(&vnet_data_free_lock);
  435                         return;
  436                 }
  437                 if (df->vnd_start == end) {
  438                         df->vnd_start = start;
  439                         df->vnd_len += size;
  440                         sx_xunlock(&vnet_data_free_lock);
  441                         return;
  442                 }
  443         }
  444         dn = malloc(sizeof(*df), M_VNET_DATA_FREE, M_WAITOK | M_ZERO);
  445         dn->vnd_start = start;
  446         dn->vnd_len = size;
  447         if (df)
  448                 TAILQ_INSERT_BEFORE(df, dn, vnd_link);
  449         else
  450                 TAILQ_INSERT_TAIL(&vnet_data_free_head, dn, vnd_link);
  451         sx_xunlock(&vnet_data_free_lock);
  452 }
  453 
  454 /*
  455  * When a new virtualized global variable has been allocated, propagate its
  456  * initial value to each already-allocated virtual network stack instance.
  457  */
  458 void
  459 vnet_data_copy(void *start, int size)
  460 {
  461         struct vnet *vnet;
  462 
  463         VNET_LIST_RLOCK();
  464         LIST_FOREACH(vnet, &vnet_head, vnet_le)
  465                 memcpy((void *)((uintptr_t)vnet->vnet_data_base +
  466                     (uintptr_t)start), start, size);
  467         VNET_LIST_RUNLOCK();
  468 }
  469 
  470 /*
  471  * Support for special SYSINIT handlers registered via VNET_SYSINIT()
  472  * and VNET_SYSUNINIT().
  473  */
  474 void
  475 vnet_register_sysinit(void *arg)
  476 {
  477         struct vnet_sysinit *vs, *vs2;  
  478         struct vnet *vnet;
  479 
  480         vs = arg;
  481         KASSERT(vs->subsystem > SI_SUB_VNET, ("vnet sysinit too early"));
  482 
  483         /* Add the constructor to the global list of vnet constructors. */
  484         VNET_SYSINIT_WLOCK();
  485         TAILQ_FOREACH(vs2, &vnet_constructors, link) {
  486                 if (vs2->subsystem > vs->subsystem)
  487                         break;
  488                 if (vs2->subsystem == vs->subsystem && vs2->order > vs->order)
  489                         break;
  490         }
  491         if (vs2 != NULL)
  492                 TAILQ_INSERT_BEFORE(vs2, vs, link);
  493         else
  494                 TAILQ_INSERT_TAIL(&vnet_constructors, vs, link);
  495 
  496         /*
  497          * Invoke the constructor on all the existing vnets when it is
  498          * registered.
  499          */
  500         VNET_FOREACH(vnet) {
  501                 CURVNET_SET_QUIET(vnet);
  502                 vs->func(vs->arg);
  503                 CURVNET_RESTORE();
  504         }
  505         VNET_SYSINIT_WUNLOCK();
  506 }
  507 
  508 void
  509 vnet_deregister_sysinit(void *arg)
  510 {
  511         struct vnet_sysinit *vs;
  512 
  513         vs = arg;
  514 
  515         /* Remove the constructor from the global list of vnet constructors. */
  516         VNET_SYSINIT_WLOCK();
  517         TAILQ_REMOVE(&vnet_constructors, vs, link);
  518         VNET_SYSINIT_WUNLOCK();
  519 }
  520 
  521 void
  522 vnet_register_sysuninit(void *arg)
  523 {
  524         struct vnet_sysinit *vs, *vs2;
  525 
  526         vs = arg;
  527 
  528         /* Add the destructor to the global list of vnet destructors. */
  529         VNET_SYSINIT_WLOCK();
  530         TAILQ_FOREACH(vs2, &vnet_destructors, link) {
  531                 if (vs2->subsystem > vs->subsystem)
  532                         break;
  533                 if (vs2->subsystem == vs->subsystem && vs2->order > vs->order)
  534                         break;
  535         }
  536         if (vs2 != NULL)
  537                 TAILQ_INSERT_BEFORE(vs2, vs, link);
  538         else
  539                 TAILQ_INSERT_TAIL(&vnet_destructors, vs, link);
  540         VNET_SYSINIT_WUNLOCK();
  541 }
  542 
  543 void
  544 vnet_deregister_sysuninit(void *arg)
  545 {
  546         struct vnet_sysinit *vs;
  547         struct vnet *vnet;
  548 
  549         vs = arg;
  550 
  551         /*
  552          * Invoke the destructor on all the existing vnets when it is
  553          * deregistered.
  554          */
  555         VNET_SYSINIT_WLOCK();
  556         VNET_FOREACH(vnet) {
  557                 CURVNET_SET_QUIET(vnet);
  558                 vs->func(vs->arg);
  559                 CURVNET_RESTORE();
  560         }
  561 
  562         /* Remove the destructor from the global list of vnet destructors. */
  563         TAILQ_REMOVE(&vnet_destructors, vs, link);
  564         VNET_SYSINIT_WUNLOCK();
  565 }
  566 
  567 /*
  568  * Invoke all registered vnet constructors on the current vnet.  Used during
  569  * vnet construction.  The caller is responsible for ensuring the new vnet is
  570  * the current vnet and that the vnet_sysinit_sxlock lock is locked.
  571  */
  572 void
  573 vnet_sysinit(void)
  574 {
  575         struct vnet_sysinit *vs;
  576 
  577         VNET_SYSINIT_RLOCK();
  578         TAILQ_FOREACH(vs, &vnet_constructors, link) {
  579                 curvnet->vnet_state = vs->subsystem;
  580                 vs->func(vs->arg);
  581         }
  582         VNET_SYSINIT_RUNLOCK();
  583 }
  584 
  585 /*
  586  * Invoke all registered vnet destructors on the current vnet.  Used during
  587  * vnet destruction.  The caller is responsible for ensuring the dying vnet
  588  * the current vnet and that the vnet_sysinit_sxlock lock is locked.
  589  */
  590 void
  591 vnet_sysuninit(void)
  592 {
  593         struct vnet_sysinit *vs;
  594 
  595         VNET_SYSINIT_RLOCK();
  596         TAILQ_FOREACH_REVERSE(vs, &vnet_destructors, vnet_sysuninit_head,
  597             link) {
  598                 curvnet->vnet_state = vs->subsystem;
  599                 vs->func(vs->arg);
  600         }
  601         VNET_SYSINIT_RUNLOCK();
  602 }
  603 
  604 /*
  605  * EVENTHANDLER(9) extensions.
  606  */
  607 /*
  608  * Invoke the eventhandler function originally registered with the possibly
  609  * registered argument for all virtual network stack instances.
  610  *
  611  * This iterator can only be used for eventhandlers that do not take any
  612  * additional arguments, as we do ignore the variadic arguments from the
  613  * EVENTHANDLER_INVOKE() call.
  614  */
  615 void
  616 vnet_global_eventhandler_iterator_func(void *arg, ...)
  617 {
  618         VNET_ITERATOR_DECL(vnet_iter);
  619         struct eventhandler_entry_vimage *v_ee;
  620 
  621         /*
  622          * There is a bug here in that we should actually cast things to
  623          * (struct eventhandler_entry_ ## name *)  but that's not easily
  624          * possible in here so just re-using the variadic version we
  625          * defined for the generic vimage case.
  626          */
  627         v_ee = arg;
  628         VNET_LIST_RLOCK();
  629         VNET_FOREACH(vnet_iter) {
  630                 CURVNET_SET(vnet_iter);
  631                 ((vimage_iterator_func_t)v_ee->func)(v_ee->ee_arg);
  632                 CURVNET_RESTORE();
  633         }
  634         VNET_LIST_RUNLOCK();
  635 }
  636 
  637 #ifdef VNET_DEBUG
  638 struct vnet_recursion {
  639         SLIST_ENTRY(vnet_recursion)      vnr_le;
  640         const char                      *prev_fn;
  641         const char                      *where_fn;
  642         int                              where_line;
  643         struct vnet                     *old_vnet;
  644         struct vnet                     *new_vnet;
  645 };
  646 
  647 static SLIST_HEAD(, vnet_recursion) vnet_recursions =
  648     SLIST_HEAD_INITIALIZER(vnet_recursions);
  649 
  650 static void
  651 vnet_print_recursion(struct vnet_recursion *vnr, int brief)
  652 {
  653 
  654         if (!brief)
  655                 printf("CURVNET_SET() recursion in ");
  656         printf("%s() line %d, prev in %s()", vnr->where_fn, vnr->where_line,
  657             vnr->prev_fn);
  658         if (brief)
  659                 printf(", ");
  660         else
  661                 printf("\n    ");
  662         printf("%p -> %p\n", vnr->old_vnet, vnr->new_vnet);
  663 }
  664 
  665 void
  666 vnet_log_recursion(struct vnet *old_vnet, const char *old_fn, int line)
  667 {
  668         struct vnet_recursion *vnr;
  669 
  670         /* Skip already logged recursion events. */
  671         SLIST_FOREACH(vnr, &vnet_recursions, vnr_le)
  672                 if (vnr->prev_fn == old_fn &&
  673                     vnr->where_fn == curthread->td_vnet_lpush &&
  674                     vnr->where_line == line &&
  675                     (vnr->old_vnet == vnr->new_vnet) == (curvnet == old_vnet))
  676                         return;
  677 
  678         vnr = malloc(sizeof(*vnr), M_VNET, M_NOWAIT | M_ZERO);
  679         if (vnr == NULL)
  680                 panic("%s: malloc failed", __func__);
  681         vnr->prev_fn = old_fn;
  682         vnr->where_fn = curthread->td_vnet_lpush;
  683         vnr->where_line = line;
  684         vnr->old_vnet = old_vnet;
  685         vnr->new_vnet = curvnet;
  686 
  687         SLIST_INSERT_HEAD(&vnet_recursions, vnr, vnr_le);
  688 
  689         vnet_print_recursion(vnr, 0);
  690 #ifdef KDB
  691         kdb_backtrace();
  692 #endif
  693 }
  694 #endif /* VNET_DEBUG */
  695 
  696 /*
  697  * DDB(4).
  698  */
  699 #ifdef DDB
  700 static void
  701 db_vnet_print(struct vnet *vnet)
  702 {
  703 
  704         db_printf("vnet            = %p\n", vnet);
  705         db_printf(" vnet_magic_n   = %#08x (%s, orig %#08x)\n",
  706             vnet->vnet_magic_n,
  707             (vnet->vnet_magic_n == VNET_MAGIC_N) ?
  708                 "ok" : "mismatch", VNET_MAGIC_N);
  709         db_printf(" vnet_ifcnt     = %u\n", vnet->vnet_ifcnt);
  710         db_printf(" vnet_sockcnt   = %u\n", vnet->vnet_sockcnt);
  711         db_printf(" vnet_data_mem  = %p\n", vnet->vnet_data_mem);
  712         db_printf(" vnet_data_base = %#jx\n",
  713             (uintmax_t)vnet->vnet_data_base);
  714         db_printf(" vnet_state     = %#08x\n", vnet->vnet_state);
  715         db_printf("\n");
  716 }
  717 
  718 DB_SHOW_ALL_COMMAND(vnets, db_show_all_vnets)
  719 {
  720         VNET_ITERATOR_DECL(vnet_iter);
  721 
  722         VNET_FOREACH(vnet_iter) {
  723                 db_vnet_print(vnet_iter);
  724                 if (db_pager_quit)
  725                         break;
  726         }
  727 }
  728 
  729 DB_SHOW_COMMAND(vnet, db_show_vnet)
  730 {
  731 
  732         if (!have_addr) {
  733                 db_printf("usage: show vnet <struct vnet *>\n");
  734                 return;
  735         }
  736 
  737         db_vnet_print((struct vnet *)addr);
  738 }
  739 
  740 static void
  741 db_show_vnet_print_vs(struct vnet_sysinit *vs, int ddb)
  742 {
  743         const char *vsname, *funcname;
  744         c_db_sym_t sym;
  745         db_expr_t  offset;
  746 
  747 #define xprint(...)                                                     \
  748         if (ddb)                                                        \
  749                 db_printf(__VA_ARGS__);                                 \
  750         else                                                            \
  751                 printf(__VA_ARGS__)
  752 
  753         if (vs == NULL) {
  754                 xprint("%s: no vnet_sysinit * given\n", __func__);
  755                 return;
  756         }
  757 
  758         sym = db_search_symbol((vm_offset_t)vs, DB_STGY_ANY, &offset);
  759         db_symbol_values(sym, &vsname, NULL);
  760         sym = db_search_symbol((vm_offset_t)vs->func, DB_STGY_PROC, &offset);
  761         db_symbol_values(sym, &funcname, NULL);
  762         xprint("%s(%p)\n", (vsname != NULL) ? vsname : "", vs);
  763         xprint("  %#08x %#08x\n", vs->subsystem, vs->order);
  764         xprint("  %p(%s)(%p)\n",
  765             vs->func, (funcname != NULL) ? funcname : "", vs->arg);
  766 #undef xprint
  767 }
  768 
  769 DB_SHOW_COMMAND(vnet_sysinit, db_show_vnet_sysinit)
  770 {
  771         struct vnet_sysinit *vs;
  772 
  773         db_printf("VNET_SYSINIT vs Name(Ptr)\n");
  774         db_printf("  Subsystem  Order\n");
  775         db_printf("  Function(Name)(Arg)\n");
  776         TAILQ_FOREACH(vs, &vnet_constructors, link) {
  777                 db_show_vnet_print_vs(vs, 1);
  778                 if (db_pager_quit)
  779                         break;
  780         }
  781 }
  782 
  783 DB_SHOW_COMMAND(vnet_sysuninit, db_show_vnet_sysuninit)
  784 {
  785         struct vnet_sysinit *vs;
  786 
  787         db_printf("VNET_SYSUNINIT vs Name(Ptr)\n");
  788         db_printf("  Subsystem  Order\n");
  789         db_printf("  Function(Name)(Arg)\n");
  790         TAILQ_FOREACH_REVERSE(vs, &vnet_destructors, vnet_sysuninit_head,
  791             link) {
  792                 db_show_vnet_print_vs(vs, 1);
  793                 if (db_pager_quit)
  794                         break;
  795         }
  796 }
  797 
  798 #ifdef VNET_DEBUG
  799 DB_SHOW_COMMAND(vnetrcrs, db_show_vnetrcrs)
  800 {
  801         struct vnet_recursion *vnr;
  802 
  803         SLIST_FOREACH(vnr, &vnet_recursions, vnr_le)
  804                 vnet_print_recursion(vnr, 1);
  805 }
  806 #endif
  807 #endif /* DDB */

Cache object: c3e7462a5ec997a79d4f0c193ca421e1


[ 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.