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  -  FREEBSD11  -  FREEBSD10  -  FREEBSD9  -  FREEBSD92  -  FREEBSD91  -  FREEBSD90  -  FREEBSD8  -  FREEBSD82  -  FREEBSD81  -  FREEBSD80  -  FREEBSD7  -  FREEBSD74  -  FREEBSD73  -  FREEBSD72  -  FREEBSD71  -  FREEBSD70  -  FREEBSD6  -  FREEBSD64  -  FREEBSD63  -  FREEBSD62  -  FREEBSD61  -  FREEBSD60  -  FREEBSD5  -  FREEBSD55  -  FREEBSD54  -  FREEBSD53  -  FREEBSD52  -  FREEBSD51  -  FREEBSD50  -  FREEBSD4  -  FREEBSD3  -  FREEBSD22  -  linux-2.6  -  linux-2.4.22  -  MK83  -  MK84  -  PLAN9  -  DFBSD  -  NETBSD  -  NETBSD5  -  NETBSD4  -  NETBSD3  -  NETBSD20  -  OPENBSD  -  xnu-517  -  xnu-792  -  xnu-792.6.70  -  xnu-1228  -  xnu-1456.1.26  -  xnu-1699.24.8  -  xnu-2050.18.24  -  OPENSOLARIS  -  minix-3-1-1 
SearchContext: -  none  -  3  -  10 

    1 /*-
    2  * Copyright (c) 2004-2009 University of Zagreb
    3  * Copyright (c) 2006-2009 FreeBSD Foundation
    4  * All rights reserved.
    5  *
    6  * This software was developed by the University of Zagreb and the
    7  * FreeBSD Foundation under sponsorship by the Stichting NLnet and the
    8  * FreeBSD Foundation.
    9  *
   10  * Copyright (c) 2009 Jeffrey Roberson <jeff@freebsd.org>
   11  * Copyright (c) 2009 Robert N. M. Watson
   12  * All rights reserved.
   13  *
   14  * Redistribution and use in source and binary forms, with or without
   15  * modification, are permitted provided that the following conditions
   16  * are met:
   17  * 1. Redistributions of source code must retain the above copyright
   18  *    notice, this list of conditions and the following disclaimer.
   19  * 2. Redistributions in binary form must reproduce the above copyright
   20  *    notice, this list of conditions and the following disclaimer in the
   21  *    documentation and/or other materials provided with the distribution.
   22  *
   23  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
   24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
   27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   33  * SUCH DAMAGE.
   34  */
   35 
   36 #include <sys/cdefs.h>
   37 __FBSDID("$FreeBSD: releng/8.1/sys/net/vnet.c 208251 2010-05-18 04:20:35Z bz $");
   38 
   39 #include "opt_ddb.h"
   40 #include "opt_kdb.h"
   41 #include "opt_kdtrace.h"
   42 
   43 #include <sys/param.h>
   44 #include <sys/kdb.h>
   45 #include <sys/kernel.h>
   46 #include <sys/jail.h>
   47 #include <sys/sdt.h>
   48 #include <sys/systm.h>
   49 #include <sys/sysctl.h>
   50 #include <sys/eventhandler.h>
   51 #include <sys/linker_set.h>
   52 #include <sys/lock.h>
   53 #include <sys/malloc.h>
   54 #include <sys/proc.h>
   55 #include <sys/socket.h>
   56 #include <sys/sx.h>
   57 #include <sys/sysctl.h>
   58 
   59 #include <machine/stdarg.h>
   60 
   61 #ifdef DDB
   62 #include <ddb/ddb.h>
   63 #include <ddb/db_sym.h>
   64 #endif
   65 
   66 #include <net/if.h>
   67 #include <net/if_var.h>
   68 #include <net/vnet.h>
   69 
   70 /*-
   71  * This file implements core functions for virtual network stacks:
   72  *
   73  * - Virtual network stack management functions.
   74  *
   75  * - Virtual network stack memory allocator, which virtualizes global
   76  *   variables in the network stack
   77  *
   78  * - Virtualized SYSINIT's/SYSUNINIT's, which allow network stack subsystems
   79  *   to register startup/shutdown events to be run for each virtual network
   80  *   stack instance.
   81  */
   82 
   83 MALLOC_DEFINE(M_VNET, "vnet", "network stack control block");
   84 
   85 /*
   86  * The virtual network stack list has two read-write locks, one sleepable and
   87  * the other not, so that the list can be stablized and walked in a variety
   88  * of network stack contexts.  Both must be acquired exclusively to modify
   89  * the list, but a read lock of either lock is sufficient to walk the list.
   90  */
   91 struct rwlock           vnet_rwlock;
   92 struct sx               vnet_sxlock;
   93 
   94 #define VNET_LIST_WLOCK() do {                                          \
   95         sx_xlock(&vnet_sxlock);                                         \
   96         rw_wlock(&vnet_rwlock);                                         \
   97 } while (0)
   98 
   99 #define VNET_LIST_WUNLOCK() do {                                        \
  100         rw_wunlock(&vnet_rwlock);                                       \
  101         sx_xunlock(&vnet_sxlock);                                       \
  102 } while (0)
  103 
  104 struct vnet_list_head vnet_head;
  105 struct vnet *vnet0;
  106 
  107 /*
  108  * The virtual network stack allocator provides storage for virtualized
  109  * global variables.  These variables are defined/declared using the
  110  * VNET_DEFINE()/VNET_DECLARE() macros, which place them in the 'set_vnet'
  111  * linker set.  The details of the implementation are somewhat subtle, but
  112  * allow the majority of most network subsystems to maintain
  113  * virtualization-agnostic.
  114  *
  115  * The virtual network stack allocator handles variables in the base kernel
  116  * vs. modules in similar but different ways.  In both cases, virtualized
  117  * global variables are marked as such by being declared to be part of the
  118  * vnet linker set.  These "master" copies of global variables serve two
  119  * functions:
  120  *
  121  * (1) They contain static initialization or "default" values for global
  122  *     variables which will be propagated to each virtual network stack
  123  *     instance when created.  As with normal global variables, they default
  124  *     to zero-filled.
  125  *
  126  * (2) They act as unique global names by which the variable can be referred
  127  *     to, regardless of network stack instance.  The single global symbol
  128  *     will be used to calculate the location of a per-virtual instance
  129  *     variable at run-time.
  130  *
  131  * Each virtual network stack instance has a complete copy of each
  132  * virtualized global variable, stored in a malloc'd block of memory
  133  * referred to by vnet->vnet_data_mem.  Critical to the design is that each
  134  * per-instance memory block is laid out identically to the master block so
  135  * that the offset of each global variable is the same across all blocks.  To
  136  * optimize run-time access, a precalculated 'base' address,
  137  * vnet->vnet_data_base, is stored in each vnet, and is the amount that can
  138  * be added to the address of a 'master' instance of a variable to get to the
  139  * per-vnet instance.
  140  *
  141  * Virtualized global variables are handled in a similar manner, but as each
  142  * module has its own 'set_vnet' linker set, and we want to keep all
  143  * virtualized globals togther, we reserve space in the kernel's linker set
  144  * for potential module variables using a per-vnet character array,
  145  * 'modspace'.  The virtual network stack allocator maintains a free list to
  146  * track what space in the array is free (all, initially) and as modules are
  147  * linked, allocates portions of the space to specific globals.  The kernel
  148  * module linker queries the virtual network stack allocator and will
  149  * bind references of the global to the location during linking.  It also
  150  * calls into the virtual network stack allocator, once the memory is
  151  * initialized, in order to propagate the new static initializations to all
  152  * existing virtual network stack instances so that the soon-to-be executing
  153  * module will find every network stack instance with proper default values.
  154  */
  155 
  156 /*
  157  * Location of the kernel's 'set_vnet' linker set.
  158  */
  159 extern uintptr_t        *__start_set_vnet;
  160 extern uintptr_t        *__stop_set_vnet;
  161 
  162 #define VNET_START      (uintptr_t)&__start_set_vnet
  163 #define VNET_STOP       (uintptr_t)&__stop_set_vnet
  164 
  165 /*
  166  * Number of bytes of data in the 'set_vnet' linker set, and hence the total
  167  * size of all kernel virtualized global variables, and the malloc(9) type
  168  * that will be used to allocate it.
  169  */
  170 #define VNET_BYTES      (VNET_STOP - VNET_START)
  171 
  172 MALLOC_DEFINE(M_VNET_DATA, "vnet_data", "VNET data");
  173 
  174 /*
  175  * VNET_MODMIN is the minimum number of bytes we will reserve for the sum of
  176  * global variables across all loaded modules.  As this actually sizes an
  177  * array declared as a virtualized global variable in the kernel itself, and
  178  * we want the virtualized global variable space to be page-sized, we may
  179  * have more space than that in practice.
  180  */
  181 #define VNET_MODMIN     8192
  182 #define VNET_SIZE       roundup2(VNET_BYTES, PAGE_SIZE)
  183 #define VNET_MODSIZE    (VNET_SIZE - (VNET_BYTES - VNET_MODMIN))
  184 
  185 /*
  186  * Space to store virtualized global variables from loadable kernel modules,
  187  * and the free list to manage it.
  188  */
  189 static VNET_DEFINE(char, modspace[VNET_MODMIN]);
  190 
  191 /*
  192  * Global lists of subsystem constructor and destructors for vnets.  They are
  193  * registered via VNET_SYSINIT() and VNET_SYSUNINIT().  Both lists are
  194  * protected by the vnet_sysinit_sxlock global lock.
  195  */
  196 static TAILQ_HEAD(vnet_sysinit_head, vnet_sysinit) vnet_constructors =
  197         TAILQ_HEAD_INITIALIZER(vnet_constructors);
  198 static TAILQ_HEAD(vnet_sysuninit_head, vnet_sysinit) vnet_destructors =
  199         TAILQ_HEAD_INITIALIZER(vnet_destructors);
  200 
  201 struct sx               vnet_sysinit_sxlock;
  202 
  203 #define VNET_SYSINIT_WLOCK()    sx_xlock(&vnet_sysinit_sxlock);
  204 #define VNET_SYSINIT_WUNLOCK()  sx_xunlock(&vnet_sysinit_sxlock);
  205 #define VNET_SYSINIT_RLOCK()    sx_slock(&vnet_sysinit_sxlock);
  206 #define VNET_SYSINIT_RUNLOCK()  sx_sunlock(&vnet_sysinit_sxlock);
  207 
  208 struct vnet_data_free {
  209         uintptr_t       vnd_start;
  210         int             vnd_len;
  211         TAILQ_ENTRY(vnet_data_free) vnd_link;
  212 };
  213 
  214 MALLOC_DEFINE(M_VNET_DATA_FREE, "vnet_data_free", "VNET resource accounting");
  215 static TAILQ_HEAD(, vnet_data_free) vnet_data_free_head =
  216             TAILQ_HEAD_INITIALIZER(vnet_data_free_head);
  217 static struct sx vnet_data_free_lock;
  218 
  219 SDT_PROVIDER_DEFINE(vnet);
  220 SDT_PROBE_DEFINE1(vnet, functions, vnet_alloc, entry, "int");
  221 SDT_PROBE_DEFINE2(vnet, functions, vnet_alloc, alloc, "int", "struct vnet *");
  222 SDT_PROBE_DEFINE2(vnet, functions, vnet_alloc, return, "int", "struct vnet *");
  223 SDT_PROBE_DEFINE2(vnet, functions, vnet_destroy, entry, "int", "struct vnet *");
  224 SDT_PROBE_DEFINE1(vnet, functions, vnet_destroy, return, "int");
  225 
  226 #ifdef DDB
  227 static void db_show_vnet_print_vs(struct vnet_sysinit *, int);
  228 #endif
  229 
  230 /*
  231  * Allocate a virtual network stack.
  232  */
  233 struct vnet *
  234 vnet_alloc(void)
  235 {
  236         struct vnet *vnet;
  237 
  238         SDT_PROBE1(vnet, functions, vnet_alloc, entry, __LINE__);
  239         vnet = malloc(sizeof(struct vnet), M_VNET, M_WAITOK | M_ZERO);
  240         vnet->vnet_magic_n = VNET_MAGIC_N;
  241         SDT_PROBE2(vnet, functions, vnet_alloc, alloc, __LINE__, vnet);
  242 
  243         /*
  244          * Allocate storage for virtualized global variables and copy in
  245          * initial values form our 'master' copy.
  246          */
  247         vnet->vnet_data_mem = malloc(VNET_SIZE, M_VNET_DATA, M_WAITOK);
  248         memcpy(vnet->vnet_data_mem, (void *)VNET_START, VNET_BYTES);
  249 
  250         /*
  251          * All use of vnet-specific data will immediately subtract VNET_START
  252          * from the base memory pointer, so pre-calculate that now to avoid
  253          * it on each use.
  254          */
  255         vnet->vnet_data_base = (uintptr_t)vnet->vnet_data_mem - VNET_START;
  256 
  257         /* Initialize / attach vnet module instances. */
  258         CURVNET_SET_QUIET(vnet);
  259         vnet_sysinit();
  260         CURVNET_RESTORE();
  261 
  262         VNET_LIST_WLOCK();
  263         LIST_INSERT_HEAD(&vnet_head, vnet, vnet_le);
  264         VNET_LIST_WUNLOCK();
  265 
  266         SDT_PROBE2(vnet, functions, vnet_alloc, return, __LINE__, vnet);
  267         return (vnet);
  268 }
  269 
  270 /*
  271  * Destroy a virtual network stack.
  272  */
  273 void
  274 vnet_destroy(struct vnet *vnet)
  275 {
  276         struct ifnet *ifp, *nifp;
  277 
  278         SDT_PROBE2(vnet, functions, vnet_destroy, entry, __LINE__, vnet);
  279         KASSERT(vnet->vnet_sockcnt == 0,
  280             ("%s: vnet still has sockets", __func__));
  281 
  282         VNET_LIST_WLOCK();
  283         LIST_REMOVE(vnet, vnet_le);
  284         VNET_LIST_WUNLOCK();
  285 
  286         CURVNET_SET_QUIET(vnet);
  287 
  288         /* Return all inherited interfaces to their parent vnets. */
  289         TAILQ_FOREACH_SAFE(ifp, &V_ifnet, if_link, nifp) {
  290                 if (ifp->if_home_vnet != ifp->if_vnet)
  291                         if_vmove(ifp, ifp->if_home_vnet);
  292         }
  293 
  294         vnet_sysuninit();
  295         CURVNET_RESTORE();
  296 
  297         /*
  298          * Release storage for the virtual network stack instance.
  299          */
  300         free(vnet->vnet_data_mem, M_VNET_DATA);
  301         vnet->vnet_data_mem = NULL;
  302         vnet->vnet_data_base = 0;
  303         vnet->vnet_magic_n = 0xdeadbeef;
  304         free(vnet, M_VNET);
  305         SDT_PROBE1(vnet, functions, vnet_destroy, return, __LINE__);
  306 }
  307 
  308 /*
  309  * Boot time initialization and allocation of virtual network stacks.
  310  */
  311 static void
  312 vnet_init_prelink(void *arg)
  313 {
  314 
  315         rw_init(&vnet_rwlock, "vnet_rwlock");
  316         sx_init(&vnet_sxlock, "vnet_sxlock");
  317         sx_init(&vnet_sysinit_sxlock, "vnet_sysinit_sxlock");
  318         LIST_INIT(&vnet_head);
  319 }
  320 SYSINIT(vnet_init_prelink, SI_SUB_VNET_PRELINK, SI_ORDER_FIRST,
  321     vnet_init_prelink, NULL);
  322 
  323 static void
  324 vnet0_init(void *arg)
  325 {
  326 
  327         /* Warn people before take off - in case we crash early. */
  328         printf("WARNING: VIMAGE (virtualized network stack) is a highly "
  329             "experimental feature.\n");
  330 
  331         /*
  332          * We MUST clear curvnet in vi_init_done() before going SMP,
  333          * otherwise CURVNET_SET() macros would scream about unnecessary
  334          * curvnet recursions.
  335          */
  336         curvnet = prison0.pr_vnet = vnet0 = vnet_alloc();
  337 }
  338 SYSINIT(vnet0_init, SI_SUB_VNET, SI_ORDER_FIRST, vnet0_init, NULL);
  339 
  340 static void
  341 vnet_init_done(void *unused)
  342 {
  343 
  344         curvnet = NULL;
  345 }
  346 
  347 SYSINIT(vnet_init_done, SI_SUB_VNET_DONE, SI_ORDER_FIRST, vnet_init_done,
  348     NULL);
  349 
  350 /*
  351  * Once on boot, initialize the modspace freelist to entirely cover modspace.
  352  */
  353 static void
  354 vnet_data_startup(void *dummy __unused)
  355 {
  356         struct vnet_data_free *df;
  357 
  358         df = malloc(sizeof(*df), M_VNET_DATA_FREE, M_WAITOK | M_ZERO);
  359         df->vnd_start = (uintptr_t)&VNET_NAME(modspace);
  360         df->vnd_len = VNET_MODMIN;
  361         TAILQ_INSERT_HEAD(&vnet_data_free_head, df, vnd_link);
  362         sx_init(&vnet_data_free_lock, "vnet_data alloc lock");
  363 }
  364 SYSINIT(vnet_data, SI_SUB_KLD, SI_ORDER_FIRST, vnet_data_startup, 0);
  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  * Variants on sysctl_handle_foo that know how to handle virtualized global
  472  * variables: if 'arg1' is a pointer, then we transform it to the local vnet
  473  * offset.
  474  */
  475 int
  476 vnet_sysctl_handle_int(SYSCTL_HANDLER_ARGS)
  477 {
  478 
  479         if (arg1 != NULL)
  480                 arg1 = (void *)(curvnet->vnet_data_base + (uintptr_t)arg1);
  481         return (sysctl_handle_int(oidp, arg1, arg2, req));
  482 }
  483 
  484 int
  485 vnet_sysctl_handle_opaque(SYSCTL_HANDLER_ARGS)
  486 {
  487 
  488         if (arg1 != NULL)
  489                 arg1 = (void *)(curvnet->vnet_data_base + (uintptr_t)arg1);
  490         return (sysctl_handle_opaque(oidp, arg1, arg2, req));
  491 }
  492 
  493 int
  494 vnet_sysctl_handle_string(SYSCTL_HANDLER_ARGS)
  495 {
  496 
  497         if (arg1 != NULL)
  498                 arg1 = (void *)(curvnet->vnet_data_base + (uintptr_t)arg1);
  499         return (sysctl_handle_string(oidp, arg1, arg2, req));
  500 }
  501 
  502 int
  503 vnet_sysctl_handle_uint(SYSCTL_HANDLER_ARGS)
  504 {
  505 
  506         if (arg1 != NULL)
  507                 arg1 = (void *)(curvnet->vnet_data_base + (uintptr_t)arg1);
  508         return (sysctl_handle_int(oidp, arg1, arg2, req));
  509 }
  510 
  511 /*
  512  * Support for special SYSINIT handlers registered via VNET_SYSINIT()
  513  * and VNET_SYSUNINIT().
  514  */
  515 void
  516 vnet_register_sysinit(void *arg)
  517 {
  518         struct vnet_sysinit *vs, *vs2;  
  519         struct vnet *vnet;
  520 
  521         vs = arg;
  522         KASSERT(vs->subsystem > SI_SUB_VNET, ("vnet sysinit too early"));
  523 
  524         /* Add the constructor to the global list of vnet constructors. */
  525         VNET_SYSINIT_WLOCK();
  526         TAILQ_FOREACH(vs2, &vnet_constructors, link) {
  527                 if (vs2->subsystem > vs->subsystem)
  528                         break;
  529                 if (vs2->subsystem == vs->subsystem && vs2->order > vs->order)
  530                         break;
  531         }
  532         if (vs2 != NULL)
  533                 TAILQ_INSERT_BEFORE(vs2, vs, link);
  534         else
  535                 TAILQ_INSERT_TAIL(&vnet_constructors, vs, link);
  536 
  537         /*
  538          * Invoke the constructor on all the existing vnets when it is
  539          * registered.
  540          */
  541         VNET_FOREACH(vnet) {
  542                 CURVNET_SET_QUIET(vnet);
  543                 vs->func(vs->arg);
  544                 CURVNET_RESTORE();
  545         }
  546         VNET_SYSINIT_WUNLOCK();
  547 }
  548 
  549 void
  550 vnet_deregister_sysinit(void *arg)
  551 {
  552         struct vnet_sysinit *vs;
  553 
  554         vs = arg;
  555 
  556         /* Remove the constructor from the global list of vnet constructors. */
  557         VNET_SYSINIT_WLOCK();
  558         TAILQ_REMOVE(&vnet_constructors, vs, link);
  559         VNET_SYSINIT_WUNLOCK();
  560 }
  561 
  562 void
  563 vnet_register_sysuninit(void *arg)
  564 {
  565         struct vnet_sysinit *vs, *vs2;
  566 
  567         vs = arg;
  568 
  569         /* Add the destructor to the global list of vnet destructors. */
  570         VNET_SYSINIT_WLOCK();
  571         TAILQ_FOREACH(vs2, &vnet_destructors, link) {
  572                 if (vs2->subsystem > vs->subsystem)
  573                         break;
  574                 if (vs2->subsystem == vs->subsystem && vs2->order > vs->order)
  575                         break;
  576         }
  577         if (vs2 != NULL)
  578                 TAILQ_INSERT_BEFORE(vs2, vs, link);
  579         else
  580                 TAILQ_INSERT_TAIL(&vnet_destructors, vs, link);
  581         VNET_SYSINIT_WUNLOCK();
  582 }
  583 
  584 void
  585 vnet_deregister_sysuninit(void *arg)
  586 {
  587         struct vnet_sysinit *vs;
  588         struct vnet *vnet;
  589 
  590         vs = arg;
  591 
  592         /*
  593          * Invoke the destructor on all the existing vnets when it is
  594          * deregistered.
  595          */
  596         VNET_SYSINIT_WLOCK();
  597         VNET_FOREACH(vnet) {
  598                 CURVNET_SET_QUIET(vnet);
  599                 vs->func(vs->arg);
  600                 CURVNET_RESTORE();
  601         }
  602 
  603         /* Remove the destructor from the global list of vnet destructors. */
  604         TAILQ_REMOVE(&vnet_destructors, vs, link);
  605         VNET_SYSINIT_WUNLOCK();
  606 }
  607 
  608 /*
  609  * Invoke all registered vnet constructors on the current vnet.  Used during
  610  * vnet construction.  The caller is responsible for ensuring the new vnet is
  611  * the current vnet and that the vnet_sysinit_sxlock lock is locked.
  612  */
  613 void
  614 vnet_sysinit(void)
  615 {
  616         struct vnet_sysinit *vs;
  617 
  618         VNET_SYSINIT_RLOCK();
  619         TAILQ_FOREACH(vs, &vnet_constructors, link) {
  620                 vs->func(vs->arg);
  621         }
  622         VNET_SYSINIT_RUNLOCK();
  623 }
  624 
  625 /*
  626  * Invoke all registered vnet destructors on the current vnet.  Used during
  627  * vnet destruction.  The caller is responsible for ensuring the dying vnet
  628  * the current vnet and that the vnet_sysinit_sxlock lock is locked.
  629  */
  630 void
  631 vnet_sysuninit(void)
  632 {
  633         struct vnet_sysinit *vs;
  634 
  635         VNET_SYSINIT_RLOCK();
  636         TAILQ_FOREACH_REVERSE(vs, &vnet_destructors, vnet_sysuninit_head,
  637             link) {
  638                 vs->func(vs->arg);
  639         }
  640         VNET_SYSINIT_RUNLOCK();
  641 }
  642 
  643 /*
  644  * EVENTHANDLER(9) extensions.
  645  */
  646 /*
  647  * Invoke the eventhandler function originally registered with the possibly
  648  * registered argument for all virtual network stack instances.
  649  *
  650  * This iterator can only be used for eventhandlers that do not take any
  651  * additional arguments, as we do ignore the variadic arguments from the
  652  * EVENTHANDLER_INVOKE() call.
  653  */
  654 void
  655 vnet_global_eventhandler_iterator_func(void *arg, ...)
  656 {
  657         VNET_ITERATOR_DECL(vnet_iter);
  658         struct eventhandler_entry_vimage *v_ee;
  659 
  660         /*
  661          * There is a bug here in that we should actually cast things to
  662          * (struct eventhandler_entry_ ## name *)  but that's not easily
  663          * possible in here so just re-using the variadic version we
  664          * defined for the generic vimage case.
  665          */
  666         v_ee = arg;
  667         VNET_LIST_RLOCK();
  668         VNET_FOREACH(vnet_iter) {
  669                 CURVNET_SET(vnet_iter);
  670                 ((vimage_iterator_func_t)v_ee->func)(v_ee->ee_arg);
  671                 CURVNET_RESTORE();
  672         }
  673         VNET_LIST_RUNLOCK();
  674 }
  675 
  676 #ifdef VNET_DEBUG
  677 struct vnet_recursion {
  678         SLIST_ENTRY(vnet_recursion)      vnr_le;
  679         const char                      *prev_fn;
  680         const char                      *where_fn;
  681         int                              where_line;
  682         struct vnet                     *old_vnet;
  683         struct vnet                     *new_vnet;
  684 };
  685 
  686 static SLIST_HEAD(, vnet_recursion) vnet_recursions =
  687     SLIST_HEAD_INITIALIZER(vnet_recursions);
  688 
  689 static void
  690 vnet_print_recursion(struct vnet_recursion *vnr, int brief)
  691 {
  692 
  693         if (!brief)
  694                 printf("CURVNET_SET() recursion in ");
  695         printf("%s() line %d, prev in %s()", vnr->where_fn, vnr->where_line,
  696             vnr->prev_fn);
  697         if (brief)
  698                 printf(", ");
  699         else
  700                 printf("\n    ");
  701         printf("%p -> %p\n", vnr->old_vnet, vnr->new_vnet);
  702 }
  703 
  704 void
  705 vnet_log_recursion(struct vnet *old_vnet, const char *old_fn, int line)
  706 {
  707         struct vnet_recursion *vnr;
  708 
  709         /* Skip already logged recursion events. */
  710         SLIST_FOREACH(vnr, &vnet_recursions, vnr_le)
  711                 if (vnr->prev_fn == old_fn &&
  712                     vnr->where_fn == curthread->td_vnet_lpush &&
  713                     vnr->where_line == line &&
  714                     (vnr->old_vnet == vnr->new_vnet) == (curvnet == old_vnet))
  715                         return;
  716 
  717         vnr = malloc(sizeof(*vnr), M_VNET, M_NOWAIT | M_ZERO);
  718         if (vnr == NULL)
  719                 panic("%s: malloc failed", __func__);
  720         vnr->prev_fn = old_fn;
  721         vnr->where_fn = curthread->td_vnet_lpush;
  722         vnr->where_line = line;
  723         vnr->old_vnet = old_vnet;
  724         vnr->new_vnet = curvnet;
  725 
  726         SLIST_INSERT_HEAD(&vnet_recursions, vnr, vnr_le);
  727 
  728         vnet_print_recursion(vnr, 0);
  729 #ifdef KDB
  730         kdb_backtrace();
  731 #endif
  732 }
  733 #endif /* VNET_DEBUG */
  734 
  735 /*
  736  * DDB(4).
  737  */
  738 #ifdef DDB
  739 DB_SHOW_COMMAND(vnets, db_show_vnets)
  740 {
  741         VNET_ITERATOR_DECL(vnet_iter);
  742 
  743         VNET_FOREACH(vnet_iter) {
  744                 db_printf("vnet            = %p\n", vnet_iter);
  745                 db_printf(" vnet_magic_n   = 0x%x (%s, orig 0x%x)\n",
  746                     vnet_iter->vnet_magic_n,
  747                     (vnet_iter->vnet_magic_n == VNET_MAGIC_N) ?
  748                         "ok" : "mismatch", VNET_MAGIC_N);
  749                 db_printf(" vnet_ifcnt     = %u\n", vnet_iter->vnet_ifcnt);
  750                 db_printf(" vnet_sockcnt   = %u\n", vnet_iter->vnet_sockcnt);
  751                 db_printf(" vnet_data_mem  = %p\n", vnet_iter->vnet_data_mem);
  752                 db_printf(" vnet_data_base = 0x%jx\n",
  753                     (uintmax_t)vnet_iter->vnet_data_base);
  754                 db_printf("\n");
  755                 if (db_pager_quit)
  756                         break;
  757         }
  758 }
  759 
  760 static void
  761 db_show_vnet_print_vs(struct vnet_sysinit *vs, int ddb)
  762 {
  763         const char *vsname, *funcname;
  764         c_db_sym_t sym;
  765         db_expr_t  offset;
  766 
  767 #define xprint(...)                                                     \
  768         if (ddb)                                                        \
  769                 db_printf(__VA_ARGS__);                                 \
  770         else                                                            \
  771                 printf(__VA_ARGS__)
  772 
  773         if (vs == NULL) {
  774                 xprint("%s: no vnet_sysinit * given\n", __func__);
  775                 return;
  776         }
  777 
  778         sym = db_search_symbol((vm_offset_t)vs, DB_STGY_ANY, &offset);
  779         db_symbol_values(sym, &vsname, NULL);
  780         sym = db_search_symbol((vm_offset_t)vs->func, DB_STGY_PROC, &offset);
  781         db_symbol_values(sym, &funcname, NULL);
  782         xprint("%s(%p)\n", (vsname != NULL) ? vsname : "", vs);
  783         xprint("  0x%08x 0x%08x\n", vs->subsystem, vs->order);
  784         xprint("  %p(%s)(%p)\n",
  785             vs->func, (funcname != NULL) ? funcname : "", vs->arg);
  786 #undef xprint
  787 }
  788 
  789 DB_SHOW_COMMAND(vnet_sysinit, db_show_vnet_sysinit)
  790 {
  791         struct vnet_sysinit *vs;
  792 
  793         db_printf("VNET_SYSINIT vs Name(Ptr)\n");
  794         db_printf("  Subsystem  Order\n");
  795         db_printf("  Function(Name)(Arg)\n");
  796         TAILQ_FOREACH(vs, &vnet_constructors, link) {
  797                 db_show_vnet_print_vs(vs, 1);
  798                 if (db_pager_quit)
  799                         break;
  800         }
  801 }
  802 
  803 DB_SHOW_COMMAND(vnet_sysuninit, db_show_vnet_sysuninit)
  804 {
  805         struct vnet_sysinit *vs;
  806 
  807         db_printf("VNET_SYSUNINIT vs Name(Ptr)\n");
  808         db_printf("  Subsystem  Order\n");
  809         db_printf("  Function(Name)(Arg)\n");
  810         TAILQ_FOREACH_REVERSE(vs, &vnet_destructors, vnet_sysuninit_head,
  811             link) {
  812                 db_show_vnet_print_vs(vs, 1);
  813                 if (db_pager_quit)
  814                         break;
  815         }
  816 }
  817 
  818 #ifdef VNET_DEBUG
  819 DB_SHOW_COMMAND(vnetrcrs, db_show_vnetrcrs)
  820 {
  821         struct vnet_recursion *vnr;
  822 
  823         SLIST_FOREACH(vnr, &vnet_recursions, vnr_le)
  824                 vnet_print_recursion(vnr, 1);
  825 }
  826 #endif
  827 #endif /* DDB */

Cache object: 8ca6039b180cbf81f7ebff2b0994a581


[ 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.