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: stable/11/sys/net/vnet.c 359652 2020-04-06 07:16:31Z hselasky $");
38
39#include "opt_ddb.h"
40#include "opt_kdb.h"
41
42#include <sys/param.h>
43#include <sys/kdb.h>
44#include <sys/kernel.h>
45#include <sys/jail.h>
46#include <sys/sdt.h>
47#include <sys/systm.h>
48#include <sys/sysctl.h>
49#include <sys/eventhandler.h>
50#include <sys/lock.h>
51#include <sys/malloc.h>
52#include <sys/proc.h>
53#include <sys/socket.h>
54#include <sys/sx.h>
55#include <sys/sysctl.h>
56
57#include <machine/stdarg.h>
58
59#ifdef DDB
60#include <ddb/ddb.h>
61#include <ddb/db_sym.h>
62#endif
63
64#include <net/if.h>
65#include <net/if_var.h>
66#include <net/vnet.h>
67
68/*-
69 * This file implements core functions for virtual network stacks:
70 *
71 * - Virtual network stack management functions.
72 *
73 * - Virtual network stack memory allocator, which virtualizes global
74 *   variables in the network stack
75 *
76 * - Virtualized SYSINIT's/SYSUNINIT's, which allow network stack subsystems
77 *   to register startup/shutdown events to be run for each virtual network
78 *   stack instance.
79 */
80
81FEATURE(vimage, "VIMAGE kernel virtualization");
82
83static 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 */
91struct rwlock		vnet_rwlock;
92struct 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
104struct vnet_list_head vnet_head;
105struct 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 * Number of bytes of data in the 'set_vnet' linker set, and hence the total
158 * size of all kernel virtualized global variables, and the malloc(9) type
159 * that will be used to allocate it.
160 */
161#define	VNET_BYTES	(VNET_STOP - VNET_START)
162
163static MALLOC_DEFINE(M_VNET_DATA, "vnet_data", "VNET data");
164
165/*
166 * VNET_MODMIN is the minimum number of bytes we will reserve for the sum of
167 * global variables across all loaded modules.  As this actually sizes an
168 * array declared as a virtualized global variable in the kernel itself, and
169 * we want the virtualized global variable space to be page-sized, we may
170 * have more space than that in practice.
171 */
172#define	VNET_MODMIN	(8 * PAGE_SIZE)
173#define	VNET_SIZE	roundup2(VNET_BYTES, PAGE_SIZE)
174
175/*
176 * Space to store virtualized global variables from loadable kernel modules,
177 * and the free list to manage it.
178 */
179static VNET_DEFINE(char, modspace[VNET_MODMIN]);
180
181/*
182 * Global lists of subsystem constructor and destructors for vnets.  They are
183 * registered via VNET_SYSINIT() and VNET_SYSUNINIT().  Both lists are
184 * protected by the vnet_sysinit_sxlock global lock.
185 */
186static TAILQ_HEAD(vnet_sysinit_head, vnet_sysinit) vnet_constructors =
187	TAILQ_HEAD_INITIALIZER(vnet_constructors);
188static TAILQ_HEAD(vnet_sysuninit_head, vnet_sysinit) vnet_destructors =
189	TAILQ_HEAD_INITIALIZER(vnet_destructors);
190
191struct sx		vnet_sysinit_sxlock;
192
193#define	VNET_SYSINIT_WLOCK()	sx_xlock(&vnet_sysinit_sxlock);
194#define	VNET_SYSINIT_WUNLOCK()	sx_xunlock(&vnet_sysinit_sxlock);
195#define	VNET_SYSINIT_RLOCK()	sx_slock(&vnet_sysinit_sxlock);
196#define	VNET_SYSINIT_RUNLOCK()	sx_sunlock(&vnet_sysinit_sxlock);
197
198struct vnet_data_free {
199	uintptr_t	vnd_start;
200	int		vnd_len;
201	TAILQ_ENTRY(vnet_data_free) vnd_link;
202};
203
204static MALLOC_DEFINE(M_VNET_DATA_FREE, "vnet_data_free",
205    "VNET resource accounting");
206static TAILQ_HEAD(, vnet_data_free) vnet_data_free_head =
207	    TAILQ_HEAD_INITIALIZER(vnet_data_free_head);
208static struct sx vnet_data_free_lock;
209
210SDT_PROVIDER_DEFINE(vnet);
211SDT_PROBE_DEFINE1(vnet, functions, vnet_alloc, entry, "int");
212SDT_PROBE_DEFINE2(vnet, functions, vnet_alloc, alloc, "int",
213    "struct vnet *");
214SDT_PROBE_DEFINE2(vnet, functions, vnet_alloc, return,
215    "int", "struct vnet *");
216SDT_PROBE_DEFINE2(vnet, functions, vnet_destroy, entry,
217    "int", "struct vnet *");
218SDT_PROBE_DEFINE1(vnet, functions, vnet_destroy, return,
219    "int");
220
221#ifdef DDB
222static void db_show_vnet_print_vs(struct vnet_sysinit *, int);
223#endif
224
225/*
226 * Allocate a virtual network stack.
227 */
228struct vnet *
229vnet_alloc(void)
230{
231	struct vnet *vnet;
232
233	SDT_PROBE1(vnet, functions, vnet_alloc, entry, __LINE__);
234	vnet = malloc(sizeof(struct vnet), M_VNET, M_WAITOK | M_ZERO);
235	vnet->vnet_magic_n = VNET_MAGIC_N;
236	vnet->vnet_state = 0;
237	SDT_PROBE2(vnet, functions, vnet_alloc, alloc, __LINE__, vnet);
238
239	/*
240	 * Allocate storage for virtualized global variables and copy in
241	 * initial values form our 'master' copy.
242	 */
243	vnet->vnet_data_mem = malloc(VNET_SIZE, M_VNET_DATA, M_WAITOK);
244	memcpy(vnet->vnet_data_mem, (void *)VNET_START, VNET_BYTES);
245
246	/*
247	 * All use of vnet-specific data will immediately subtract VNET_START
248	 * from the base memory pointer, so pre-calculate that now to avoid
249	 * it on each use.
250	 */
251	vnet->vnet_data_base = (uintptr_t)vnet->vnet_data_mem - VNET_START;
252
253	/* Initialize / attach vnet module instances. */
254	CURVNET_SET_QUIET(vnet);
255	vnet_sysinit();
256	CURVNET_RESTORE();
257
258	VNET_LIST_WLOCK();
259	LIST_INSERT_HEAD(&vnet_head, vnet, vnet_le);
260	VNET_LIST_WUNLOCK();
261
262	SDT_PROBE2(vnet, functions, vnet_alloc, return, __LINE__, vnet);
263	return (vnet);
264}
265
266/*
267 * Destroy a virtual network stack.
268 */
269void
270vnet_destroy(struct vnet *vnet)
271{
272
273	SDT_PROBE2(vnet, functions, vnet_destroy, entry, __LINE__, vnet);
274	KASSERT(vnet->vnet_sockcnt == 0,
275	    ("%s: vnet still has sockets", __func__));
276
277	VNET_LIST_WLOCK();
278	LIST_REMOVE(vnet, vnet_le);
279	VNET_LIST_WUNLOCK();
280
281	CURVNET_SET_QUIET(vnet);
282	vnet_sysuninit();
283	CURVNET_RESTORE();
284
285	/*
286	 * Release storage for the virtual network stack instance.
287	 */
288	free(vnet->vnet_data_mem, M_VNET_DATA);
289	vnet->vnet_data_mem = NULL;
290	vnet->vnet_data_base = 0;
291	vnet->vnet_magic_n = 0xdeadbeef;
292	free(vnet, M_VNET);
293	SDT_PROBE1(vnet, functions, vnet_destroy, return, __LINE__);
294}
295
296/*
297 * Boot time initialization and allocation of virtual network stacks.
298 */
299static void
300vnet_init_prelink(void *arg __unused)
301{
302
303	rw_init(&vnet_rwlock, "vnet_rwlock");
304	sx_init(&vnet_sxlock, "vnet_sxlock");
305	sx_init(&vnet_sysinit_sxlock, "vnet_sysinit_sxlock");
306	LIST_INIT(&vnet_head);
307}
308SYSINIT(vnet_init_prelink, SI_SUB_VNET_PRELINK, SI_ORDER_FIRST,
309    vnet_init_prelink, NULL);
310
311static void
312vnet0_init(void *arg __unused)
313{
314
315	/* Warn people before take off - in case we crash early. */
316	printf("WARNING: VIMAGE (virtualized network stack) is a highly "
317	    "experimental feature.\n");
318
319	/*
320	 * We MUST clear curvnet in vi_init_done() before going SMP,
321	 * otherwise CURVNET_SET() macros would scream about unnecessary
322	 * curvnet recursions.
323	 */
324	curvnet = prison0.pr_vnet = vnet0 = vnet_alloc();
325}
326SYSINIT(vnet0_init, SI_SUB_VNET, SI_ORDER_FIRST, vnet0_init, NULL);
327
328static void
329vnet_init_done(void *unused __unused)
330{
331
332	curvnet = NULL;
333}
334SYSINIT(vnet_init_done, SI_SUB_VNET_DONE, SI_ORDER_ANY, vnet_init_done,
335    NULL);
336
337/*
338 * Once on boot, initialize the modspace freelist to entirely cover modspace.
339 */
340static void
341vnet_data_startup(void *dummy __unused)
342{
343	struct vnet_data_free *df;
344
345	df = malloc(sizeof(*df), M_VNET_DATA_FREE, M_WAITOK | M_ZERO);
346	df->vnd_start = (uintptr_t)&VNET_NAME(modspace);
347	df->vnd_len = VNET_MODMIN;
348	TAILQ_INSERT_HEAD(&vnet_data_free_head, df, vnd_link);
349	sx_init(&vnet_data_free_lock, "vnet_data alloc lock");
350}
351SYSINIT(vnet_data, SI_SUB_KLD, SI_ORDER_FIRST, vnet_data_startup, NULL);
352
353/* Dummy VNET_SYSINIT to make sure we always reach the final end state. */
354static void
355vnet_sysinit_done(void *unused __unused)
356{
357
358	return;
359}
360VNET_SYSINIT(vnet_sysinit_done, SI_SUB_VNET_DONE, SI_ORDER_ANY,
361    vnet_sysinit_done, NULL);
362
363/*
364 * When a module is loaded and requires storage for a virtualized global
365 * variable, allocate space from the modspace free list.  This interface
366 * should be used only by the kernel linker.
367 */
368void *
369vnet_data_alloc(int size)
370{
371	struct vnet_data_free *df;
372	void *s;
373
374	s = NULL;
375	size = roundup2(size, sizeof(void *));
376	sx_xlock(&vnet_data_free_lock);
377	TAILQ_FOREACH(df, &vnet_data_free_head, vnd_link) {
378		if (df->vnd_len < size)
379			continue;
380		if (df->vnd_len == size) {
381			s = (void *)df->vnd_start;
382			TAILQ_REMOVE(&vnet_data_free_head, df, vnd_link);
383			free(df, M_VNET_DATA_FREE);
384			break;
385		}
386		s = (void *)df->vnd_start;
387		df->vnd_len -= size;
388		df->vnd_start = df->vnd_start + size;
389		break;
390	}
391	sx_xunlock(&vnet_data_free_lock);
392
393	return (s);
394}
395
396/*
397 * Free space for a virtualized global variable on module unload.
398 */
399void
400vnet_data_free(void *start_arg, int size)
401{
402	struct vnet_data_free *df;
403	struct vnet_data_free *dn;
404	uintptr_t start;
405	uintptr_t end;
406
407	size = roundup2(size, sizeof(void *));
408	start = (uintptr_t)start_arg;
409	end = start + size;
410	/*
411	 * Free a region of space and merge it with as many neighbors as
412	 * possible.  Keeping the list sorted simplifies this operation.
413	 */
414	sx_xlock(&vnet_data_free_lock);
415	TAILQ_FOREACH(df, &vnet_data_free_head, vnd_link) {
416		if (df->vnd_start > end)
417			break;
418		/*
419		 * If we expand at the end of an entry we may have to merge
420		 * it with the one following it as well.
421		 */
422		if (df->vnd_start + df->vnd_len == start) {
423			df->vnd_len += size;
424			dn = TAILQ_NEXT(df, vnd_link);
425			if (df->vnd_start + df->vnd_len == dn->vnd_start) {
426				df->vnd_len += dn->vnd_len;
427				TAILQ_REMOVE(&vnet_data_free_head, dn,
428				    vnd_link);
429				free(dn, M_VNET_DATA_FREE);
430			}
431			sx_xunlock(&vnet_data_free_lock);
432			return;
433		}
434		if (df->vnd_start == end) {
435			df->vnd_start = start;
436			df->vnd_len += size;
437			sx_xunlock(&vnet_data_free_lock);
438			return;
439		}
440	}
441	dn = malloc(sizeof(*df), M_VNET_DATA_FREE, M_WAITOK | M_ZERO);
442	dn->vnd_start = start;
443	dn->vnd_len = size;
444	if (df)
445		TAILQ_INSERT_BEFORE(df, dn, vnd_link);
446	else
447		TAILQ_INSERT_TAIL(&vnet_data_free_head, dn, vnd_link);
448	sx_xunlock(&vnet_data_free_lock);
449}
450
451/*
452 * When a new virtualized global variable has been allocated, propagate its
453 * initial value to each already-allocated virtual network stack instance.
454 */
455void
456vnet_data_copy(void *start, int size)
457{
458	struct vnet *vnet;
459
460	VNET_LIST_RLOCK();
461	LIST_FOREACH(vnet, &vnet_head, vnet_le)
462		memcpy((void *)((uintptr_t)vnet->vnet_data_base +
463		    (uintptr_t)start), start, size);
464	VNET_LIST_RUNLOCK();
465}
466
467/*
468 * Support for special SYSINIT handlers registered via VNET_SYSINIT()
469 * and VNET_SYSUNINIT().
470 */
471void
472vnet_register_sysinit(void *arg)
473{
474	struct vnet_sysinit *vs, *vs2;
475	struct vnet *vnet;
476
477	vs = arg;
478	KASSERT(vs->subsystem > SI_SUB_VNET, ("vnet sysinit too early"));
479
480	/* Add the constructor to the global list of vnet constructors. */
481	VNET_SYSINIT_WLOCK();
482	TAILQ_FOREACH(vs2, &vnet_constructors, link) {
483		if (vs2->subsystem > vs->subsystem)
484			break;
485		if (vs2->subsystem == vs->subsystem && vs2->order > vs->order)
486			break;
487	}
488	if (vs2 != NULL)
489		TAILQ_INSERT_BEFORE(vs2, vs, link);
490	else
491		TAILQ_INSERT_TAIL(&vnet_constructors, vs, link);
492
493	/*
494	 * Invoke the constructor on all the existing vnets when it is
495	 * registered.
496	 */
497	VNET_FOREACH(vnet) {
498		CURVNET_SET_QUIET(vnet);
499		vs->func(vs->arg);
500		CURVNET_RESTORE();
501	}
502	VNET_SYSINIT_WUNLOCK();
503}
504
505void
506vnet_deregister_sysinit(void *arg)
507{
508	struct vnet_sysinit *vs;
509
510	vs = arg;
511
512	/* Remove the constructor from the global list of vnet constructors. */
513	VNET_SYSINIT_WLOCK();
514	TAILQ_REMOVE(&vnet_constructors, vs, link);
515	VNET_SYSINIT_WUNLOCK();
516}
517
518void
519vnet_register_sysuninit(void *arg)
520{
521	struct vnet_sysinit *vs, *vs2;
522
523	vs = arg;
524
525	/* Add the destructor to the global list of vnet destructors. */
526	VNET_SYSINIT_WLOCK();
527	TAILQ_FOREACH(vs2, &vnet_destructors, link) {
528		if (vs2->subsystem > vs->subsystem)
529			break;
530		if (vs2->subsystem == vs->subsystem && vs2->order > vs->order)
531			break;
532	}
533	if (vs2 != NULL)
534		TAILQ_INSERT_BEFORE(vs2, vs, link);
535	else
536		TAILQ_INSERT_TAIL(&vnet_destructors, vs, link);
537	VNET_SYSINIT_WUNLOCK();
538}
539
540void
541vnet_deregister_sysuninit(void *arg)
542{
543	struct vnet_sysinit *vs;
544	struct vnet *vnet;
545
546	vs = arg;
547
548	/*
549	 * Invoke the destructor on all the existing vnets when it is
550	 * deregistered.
551	 */
552	VNET_SYSINIT_WLOCK();
553	VNET_FOREACH(vnet) {
554		CURVNET_SET_QUIET(vnet);
555		vs->func(vs->arg);
556		CURVNET_RESTORE();
557	}
558
559	/* Remove the destructor from the global list of vnet destructors. */
560	TAILQ_REMOVE(&vnet_destructors, vs, link);
561	VNET_SYSINIT_WUNLOCK();
562}
563
564/*
565 * Invoke all registered vnet constructors on the current vnet.  Used during
566 * vnet construction.  The caller is responsible for ensuring the new vnet is
567 * the current vnet and that the vnet_sysinit_sxlock lock is locked.
568 */
569void
570vnet_sysinit(void)
571{
572	struct vnet_sysinit *vs;
573
574	VNET_SYSINIT_RLOCK();
575	TAILQ_FOREACH(vs, &vnet_constructors, link) {
576		curvnet->vnet_state = vs->subsystem;
577		vs->func(vs->arg);
578	}
579	VNET_SYSINIT_RUNLOCK();
580}
581
582/*
583 * Invoke all registered vnet destructors on the current vnet.  Used during
584 * vnet destruction.  The caller is responsible for ensuring the dying vnet
585 * the current vnet and that the vnet_sysinit_sxlock lock is locked.
586 */
587void
588vnet_sysuninit(void)
589{
590	struct vnet_sysinit *vs;
591
592	VNET_SYSINIT_RLOCK();
593	TAILQ_FOREACH_REVERSE(vs, &vnet_destructors, vnet_sysuninit_head,
594	    link) {
595		curvnet->vnet_state = vs->subsystem;
596		vs->func(vs->arg);
597	}
598	VNET_SYSINIT_RUNLOCK();
599}
600
601/*
602 * EVENTHANDLER(9) extensions.
603 */
604/*
605 * Invoke the eventhandler function originally registered with the possibly
606 * registered argument for all virtual network stack instances.
607 *
608 * This iterator can only be used for eventhandlers that do not take any
609 * additional arguments, as we do ignore the variadic arguments from the
610 * EVENTHANDLER_INVOKE() call.
611 */
612void
613vnet_global_eventhandler_iterator_func(void *arg, ...)
614{
615	VNET_ITERATOR_DECL(vnet_iter);
616	struct eventhandler_entry_vimage *v_ee;
617
618	/*
619	 * There is a bug here in that we should actually cast things to
620	 * (struct eventhandler_entry_ ## name *)  but that's not easily
621	 * possible in here so just re-using the variadic version we
622	 * defined for the generic vimage case.
623	 */
624	v_ee = arg;
625	VNET_LIST_RLOCK();
626	VNET_FOREACH(vnet_iter) {
627		CURVNET_SET(vnet_iter);
628		((vimage_iterator_func_t)v_ee->func)(v_ee->ee_arg);
629		CURVNET_RESTORE();
630	}
631	VNET_LIST_RUNLOCK();
632}
633
634#ifdef VNET_DEBUG
635struct vnet_recursion {
636	SLIST_ENTRY(vnet_recursion)	 vnr_le;
637	const char			*prev_fn;
638	const char			*where_fn;
639	int				 where_line;
640	struct vnet			*old_vnet;
641	struct vnet			*new_vnet;
642};
643
644static SLIST_HEAD(, vnet_recursion) vnet_recursions =
645    SLIST_HEAD_INITIALIZER(vnet_recursions);
646
647static void
648vnet_print_recursion(struct vnet_recursion *vnr, int brief)
649{
650
651	if (!brief)
652		printf("CURVNET_SET() recursion in ");
653	printf("%s() line %d, prev in %s()", vnr->where_fn, vnr->where_line,
654	    vnr->prev_fn);
655	if (brief)
656		printf(", ");
657	else
658		printf("\n    ");
659	printf("%p -> %p\n", vnr->old_vnet, vnr->new_vnet);
660}
661
662void
663vnet_log_recursion(struct vnet *old_vnet, const char *old_fn, int line)
664{
665	struct vnet_recursion *vnr;
666
667	/* Skip already logged recursion events. */
668	SLIST_FOREACH(vnr, &vnet_recursions, vnr_le)
669		if (vnr->prev_fn == old_fn &&
670		    vnr->where_fn == curthread->td_vnet_lpush &&
671		    vnr->where_line == line &&
672		    (vnr->old_vnet == vnr->new_vnet) == (curvnet == old_vnet))
673			return;
674
675	vnr = malloc(sizeof(*vnr), M_VNET, M_NOWAIT | M_ZERO);
676	if (vnr == NULL)
677		panic("%s: malloc failed", __func__);
678	vnr->prev_fn = old_fn;
679	vnr->where_fn = curthread->td_vnet_lpush;
680	vnr->where_line = line;
681	vnr->old_vnet = old_vnet;
682	vnr->new_vnet = curvnet;
683
684	SLIST_INSERT_HEAD(&vnet_recursions, vnr, vnr_le);
685
686	vnet_print_recursion(vnr, 0);
687#ifdef KDB
688	kdb_backtrace();
689#endif
690}
691#endif /* VNET_DEBUG */
692
693/*
694 * DDB(4).
695 */
696#ifdef DDB
697static void
698db_vnet_print(struct vnet *vnet)
699{
700
701	db_printf("vnet            = %p\n", vnet);
702	db_printf(" vnet_magic_n   = %#08x (%s, orig %#08x)\n",
703	    vnet->vnet_magic_n,
704	    (vnet->vnet_magic_n == VNET_MAGIC_N) ?
705		"ok" : "mismatch", VNET_MAGIC_N);
706	db_printf(" vnet_ifcnt     = %u\n", vnet->vnet_ifcnt);
707	db_printf(" vnet_sockcnt   = %u\n", vnet->vnet_sockcnt);
708	db_printf(" vnet_data_mem  = %p\n", vnet->vnet_data_mem);
709	db_printf(" vnet_data_base = %#jx\n",
710	    (uintmax_t)vnet->vnet_data_base);
711	db_printf(" vnet_state     = %#08x\n", vnet->vnet_state);
712	db_printf("\n");
713}
714
715DB_SHOW_ALL_COMMAND(vnets, db_show_all_vnets)
716{
717	VNET_ITERATOR_DECL(vnet_iter);
718
719	VNET_FOREACH(vnet_iter) {
720		db_vnet_print(vnet_iter);
721		if (db_pager_quit)
722			break;
723	}
724}
725
726DB_SHOW_COMMAND(vnet, db_show_vnet)
727{
728
729	if (!have_addr) {
730		db_printf("usage: show vnet <struct vnet *>\n");
731		return;
732	}
733
734	db_vnet_print((struct vnet *)addr);
735}
736
737static void
738db_show_vnet_print_vs(struct vnet_sysinit *vs, int ddb)
739{
740	const char *vsname, *funcname;
741	c_db_sym_t sym;
742	db_expr_t  offset;
743
744#define xprint(...)							\
745	if (ddb)							\
746		db_printf(__VA_ARGS__);					\
747	else								\
748		printf(__VA_ARGS__)
749
750	if (vs == NULL) {
751		xprint("%s: no vnet_sysinit * given\n", __func__);
752		return;
753	}
754
755	sym = db_search_symbol((vm_offset_t)vs, DB_STGY_ANY, &offset);
756	db_symbol_values(sym, &vsname, NULL);
757	sym = db_search_symbol((vm_offset_t)vs->func, DB_STGY_PROC, &offset);
758	db_symbol_values(sym, &funcname, NULL);
759	xprint("%s(%p)\n", (vsname != NULL) ? vsname : "", vs);
760	xprint("  %#08x %#08x\n", vs->subsystem, vs->order);
761	xprint("  %p(%s)(%p)\n",
762	    vs->func, (funcname != NULL) ? funcname : "", vs->arg);
763#undef xprint
764}
765
766DB_SHOW_COMMAND(vnet_sysinit, db_show_vnet_sysinit)
767{
768	struct vnet_sysinit *vs;
769
770	db_printf("VNET_SYSINIT vs Name(Ptr)\n");
771	db_printf("  Subsystem  Order\n");
772	db_printf("  Function(Name)(Arg)\n");
773	TAILQ_FOREACH(vs, &vnet_constructors, link) {
774		db_show_vnet_print_vs(vs, 1);
775		if (db_pager_quit)
776			break;
777	}
778}
779
780DB_SHOW_COMMAND(vnet_sysuninit, db_show_vnet_sysuninit)
781{
782	struct vnet_sysinit *vs;
783
784	db_printf("VNET_SYSUNINIT vs Name(Ptr)\n");
785	db_printf("  Subsystem  Order\n");
786	db_printf("  Function(Name)(Arg)\n");
787	TAILQ_FOREACH_REVERSE(vs, &vnet_destructors, vnet_sysuninit_head,
788	    link) {
789		db_show_vnet_print_vs(vs, 1);
790		if (db_pager_quit)
791			break;
792	}
793}
794
795#ifdef VNET_DEBUG
796DB_SHOW_COMMAND(vnetrcrs, db_show_vnetrcrs)
797{
798	struct vnet_recursion *vnr;
799
800	SLIST_FOREACH(vnr, &vnet_recursions, vnr_le)
801		vnet_print_recursion(vnr, 1);
802}
803#endif
804#endif /* DDB */
805