mon.c 14.7 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10
/*
 * linux/fs/lockd/mon.c
 *
 * The kernel statd client.
 *
 * Copyright (C) 1996, Olaf Kirch <okir@monad.swb.de>
 */

#include <linux/types.h>
#include <linux/kernel.h>
11
#include <linux/ktime.h>
12
#include <linux/slab.h>
13

L
Linus Torvalds 已提交
14
#include <linux/sunrpc/clnt.h>
15
#include <linux/sunrpc/addr.h>
16
#include <linux/sunrpc/xprtsock.h>
L
Linus Torvalds 已提交
17 18 19
#include <linux/sunrpc/svc.h>
#include <linux/lockd/lockd.h>

20 21
#include <asm/unaligned.h>

22 23
#include "netns.h"

L
Linus Torvalds 已提交
24
#define NLMDBG_FACILITY		NLMDBG_MONITOR
25 26 27 28 29 30 31 32 33 34 35 36
#define NSM_PROGRAM		100024
#define NSM_VERSION		1

enum {
	NSMPROC_NULL,
	NSMPROC_STAT,
	NSMPROC_MON,
	NSMPROC_UNMON,
	NSMPROC_UNMON_ALL,
	NSMPROC_SIMU_CRASH,
	NSMPROC_NOTIFY,
};
L
Linus Torvalds 已提交
37

38
struct nsm_args {
39
	struct nsm_private	*priv;
40 41 42 43 44
	u32			prog;		/* RPC callback info */
	u32			vers;
	u32			proc;

	char			*mon_name;
45
	char			*nodename;
46 47 48 49 50 51 52
};

struct nsm_res {
	u32			status;
	u32			state;
};

53
static const struct rpc_program	nsm_program;
54 55
static				LIST_HEAD(nsm_handles);
static				DEFINE_SPINLOCK(nsm_lock);
L
Linus Torvalds 已提交
56 57 58 59

/*
 * Local NSM state
 */
60
u32	__read_mostly		nsm_local_state;
61
bool	__read_mostly		nsm_use_hostnames;
L
Linus Torvalds 已提交
62

63 64 65 66 67
static inline struct sockaddr *nsm_addr(const struct nsm_handle *nsm)
{
	return (struct sockaddr *)&nsm->sm_addr;
}

68
static struct rpc_clnt *nsm_create(struct net *net)
C
Chuck Lever 已提交
69 70 71 72 73 74
{
	struct sockaddr_in sin = {
		.sin_family		= AF_INET,
		.sin_addr.s_addr	= htonl(INADDR_LOOPBACK),
	};
	struct rpc_create_args args = {
75
		.net			= net,
76
		.protocol		= XPRT_TRANSPORT_TCP,
C
Chuck Lever 已提交
77 78 79 80 81 82
		.address		= (struct sockaddr *)&sin,
		.addrsize		= sizeof(sin),
		.servername		= "rpc.statd",
		.program		= &nsm_program,
		.version		= NSM_VERSION,
		.authflavor		= RPC_AUTH_NULL,
83
		.flags			= RPC_CLNT_CREATE_NOPING,
C
Chuck Lever 已提交
84 85 86 87 88
	};

	return rpc_create(&args);
}

89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104
static struct rpc_clnt *nsm_client_set(struct lockd_net *ln,
		struct rpc_clnt *clnt)
{
	spin_lock(&ln->nsm_clnt_lock);
	if (ln->nsm_users == 0) {
		if (clnt == NULL)
			goto out;
		ln->nsm_clnt = clnt;
	}
	clnt = ln->nsm_clnt;
	ln->nsm_users++;
out:
	spin_unlock(&ln->nsm_clnt_lock);
	return clnt;
}

105
static struct rpc_clnt *nsm_client_get(struct net *net)
106
{
107
	struct rpc_clnt	*clnt, *new;
108 109
	struct lockd_net *ln = net_generic(net, lockd_net_id);

110 111
	clnt = nsm_client_set(ln, NULL);
	if (clnt != NULL)
112 113
		goto out;

114 115 116 117 118 119 120
	clnt = new = nsm_create(net);
	if (IS_ERR(clnt))
		goto out;

	clnt = nsm_client_set(ln, new);
	if (clnt != new)
		rpc_shutdown_client(new);
121 122 123 124
out:
	return clnt;
}

125
static void nsm_client_put(struct net *net)
126 127
{
	struct lockd_net *ln = net_generic(net, lockd_net_id);
128
	struct rpc_clnt	*clnt = NULL;
129 130

	spin_lock(&ln->nsm_clnt_lock);
131 132 133 134
	ln->nsm_users--;
	if (ln->nsm_users == 0) {
		clnt = ln->nsm_clnt;
		ln->nsm_clnt = NULL;
135 136
	}
	spin_unlock(&ln->nsm_clnt_lock);
137
	if (clnt != NULL)
138 139 140
		rpc_shutdown_client(clnt);
}

141
static int nsm_mon_unmon(struct nsm_handle *nsm, u32 proc, struct nsm_res *res,
142
			 struct rpc_clnt *clnt)
L
Linus Torvalds 已提交
143 144
{
	int		status;
145
	struct nsm_args args = {
146
		.priv		= &nsm->sm_priv,
147 148 149
		.prog		= NLM_PROGRAM,
		.vers		= 3,
		.proc		= NLMPROC_NSM_NOTIFY,
150
		.mon_name	= nsm->sm_mon_name,
151
		.nodename	= clnt->cl_nodename,
152
	};
C
Chuck Lever 已提交
153 154 155 156
	struct rpc_message msg = {
		.rpc_argp	= &args,
		.rpc_resp	= res,
	};
L
Linus Torvalds 已提交
157 158 159

	memset(res, 0, sizeof(*res));

C
Chuck Lever 已提交
160
	msg.rpc_proc = &clnt->cl_procinfo[proc];
161
	status = rpc_call_sync(clnt, &msg, RPC_TASK_SOFTCONN);
L
Linus Torvalds 已提交
162
	if (status < 0)
163 164
		dprintk("lockd: NSM upcall RPC failed, status=%d\n",
				status);
L
Linus Torvalds 已提交
165 166 167 168 169
	else
		status = 0;
	return status;
}

170 171 172 173 174 175 176 177 178 179
/**
 * nsm_monitor - Notify a peer in case we reboot
 * @host: pointer to nlm_host of peer to notify
 *
 * If this peer is not already monitored, this function sends an
 * upcall to the local rpc.statd to record the name/address of
 * the peer to notify in case we reboot.
 *
 * Returns zero if the peer is monitored by the local rpc.statd;
 * otherwise a negative errno value is returned.
L
Linus Torvalds 已提交
180
 */
181
int nsm_monitor(const struct nlm_host *host)
L
Linus Torvalds 已提交
182
{
183
	struct nsm_handle *nsm = host->h_nsmhandle;
L
Linus Torvalds 已提交
184 185
	struct nsm_res	res;
	int		status;
186
	struct rpc_clnt *clnt;
L
Linus Torvalds 已提交
187

188
	dprintk("lockd: nsm_monitor(%s)\n", nsm->sm_name);
189 190

	if (nsm->sm_monitored)
191
		return 0;
L
Linus Torvalds 已提交
192

193 194 195 196 197 198
	/*
	 * Choose whether to record the caller_name or IP address of
	 * this peer in the local rpc.statd's database.
	 */
	nsm->sm_mon_name = nsm_use_hostnames ? nsm->sm_name : nsm->sm_addrbuf;

199 200 201 202 203 204 205 206 207
	clnt = nsm_client_get(host->net);
	if (IS_ERR(clnt)) {
		status = PTR_ERR(clnt);
		dprintk("lockd: failed to create NSM upcall transport, "
				"status=%d, net=%p\n", status, host->net);
		return status;
	}

	status = nsm_mon_unmon(nsm, NSMPROC_MON, &res, clnt);
208
	if (unlikely(res.status != 0))
209
		status = -EIO;
210
	if (unlikely(status < 0)) {
211
		printk(KERN_NOTICE "lockd: cannot monitor %s\n", nsm->sm_name);
212 213 214 215 216 217 218 219 220
		return status;
	}

	nsm->sm_monitored = 1;
	if (unlikely(nsm_local_state != res.state)) {
		nsm_local_state = res.state;
		dprintk("lockd: NSM state changed to %d\n", nsm_local_state);
	}
	return 0;
L
Linus Torvalds 已提交
221 222
}

223 224 225 226 227 228 229
/**
 * nsm_unmonitor - Unregister peer notification
 * @host: pointer to nlm_host of peer to stop monitoring
 *
 * If this peer is monitored, this function sends an upcall to
 * tell the local rpc.statd not to send this peer a notification
 * when we reboot.
L
Linus Torvalds 已提交
230
 */
231
void nsm_unmonitor(const struct nlm_host *host)
L
Linus Torvalds 已提交
232
{
233
	struct nsm_handle *nsm = host->h_nsmhandle;
L
Linus Torvalds 已提交
234
	struct nsm_res	res;
235
	int status;
L
Linus Torvalds 已提交
236

237 238
	if (atomic_read(&nsm->sm_count) == 1
	 && nsm->sm_monitored && !nsm->sm_sticky) {
239 240
		struct lockd_net *ln = net_generic(host->net, lockd_net_id);

241
		dprintk("lockd: nsm_unmonitor(%s)\n", nsm->sm_name);
242

243
		status = nsm_mon_unmon(nsm, NSMPROC_UNMON, &res, ln->nsm_clnt);
244 245
		if (res.status != 0)
			status = -EIO;
246
		if (status < 0)
247
			printk(KERN_NOTICE "lockd: cannot unmonitor %s\n",
248
					nsm->sm_name);
249 250
		else
			nsm->sm_monitored = 0;
251 252

		nsm_client_put(host->net);
253
	}
L
Linus Torvalds 已提交
254 255
}

C
Chuck Lever 已提交
256 257 258 259 260 261 262 263 264 265 266 267
static struct nsm_handle *nsm_lookup_hostname(const char *hostname,
					      const size_t len)
{
	struct nsm_handle *nsm;

	list_for_each_entry(nsm, &nsm_handles, sm_link)
		if (strlen(nsm->sm_name) == len &&
		    memcmp(nsm->sm_name, hostname, len) == 0)
			return nsm;
	return NULL;
}

268 269 270 271 272
static struct nsm_handle *nsm_lookup_addr(const struct sockaddr *sap)
{
	struct nsm_handle *nsm;

	list_for_each_entry(nsm, &nsm_handles, sm_link)
273
		if (rpc_cmp_addr(nsm_addr(nsm), sap))
274 275 276 277
			return nsm;
	return NULL;
}

C
Chuck Lever 已提交
278 279 280 281 282 283 284 285 286 287 288
static struct nsm_handle *nsm_lookup_priv(const struct nsm_private *priv)
{
	struct nsm_handle *nsm;

	list_for_each_entry(nsm, &nsm_handles, sm_link)
		if (memcmp(nsm->sm_priv.data, priv->data,
					sizeof(priv->data)) == 0)
			return nsm;
	return NULL;
}

289 290 291 292 293 294
/*
 * Construct a unique cookie to match this nsm_handle to this monitored
 * host.  It is passed to the local rpc.statd via NSMPROC_MON, and
 * returned via NLMPROC_SM_NOTIFY, in the "priv" field of these
 * requests.
 *
295 296 297 298 299 300 301 302 303 304
 * The NSM protocol requires that these cookies be unique while the
 * system is running.  We prefer a stronger requirement of making them
 * unique across reboots.  If user space bugs cause a stale cookie to
 * be sent to the kernel, it could cause the wrong host to lose its
 * lock state if cookies were not unique across reboots.
 *
 * The cookies are exposed only to local user space via loopback.  They
 * do not appear on the physical network.  If we want greater security
 * for some reason, nsm_init_private() could perform a one-way hash to
 * obscure the contents of the cookie.
305 306 307
 */
static void nsm_init_private(struct nsm_handle *nsm)
{
308 309
	u64 *p = (u64 *)&nsm->sm_priv.data;
	struct timespec ts;
310
	s64 ns;
311 312

	ktime_get_ts(&ts);
313 314 315
	ns = timespec_to_ns(&ts);
	put_unaligned(ns, p);
	put_unaligned((unsigned long)nsm, p + 1);
316 317
}

318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333
static struct nsm_handle *nsm_create_handle(const struct sockaddr *sap,
					    const size_t salen,
					    const char *hostname,
					    const size_t hostname_len)
{
	struct nsm_handle *new;

	new = kzalloc(sizeof(*new) + hostname_len + 1, GFP_KERNEL);
	if (unlikely(new == NULL))
		return NULL;

	atomic_set(&new->sm_count, 1);
	new->sm_name = (char *)(new + 1);
	memcpy(nsm_addr(new), sap, salen);
	new->sm_addrlen = salen;
	nsm_init_private(new);
334 335 336 337 338

	if (rpc_ntop(nsm_addr(new), new->sm_addrbuf,
					sizeof(new->sm_addrbuf)) == 0)
		(void)snprintf(new->sm_addrbuf, sizeof(new->sm_addrbuf),
				"unsupported address family");
339 340 341 342 343 344
	memcpy(new->sm_name, hostname, hostname_len);
	new->sm_name[hostname_len] = '\0';

	return new;
}

345
/**
346
 * nsm_get_handle - Find or create a cached nsm_handle
347 348 349 350 351
 * @sap: pointer to socket address of handle to find
 * @salen: length of socket address
 * @hostname: pointer to C string containing hostname to find
 * @hostname_len: length of C string
 *
352
 * Behavior is modulated by the global nsm_use_hostnames variable.
353
 *
354 355 356 357
 * Returns a cached nsm_handle after bumping its ref count, or
 * returns a fresh nsm_handle if a handle that matches @sap and/or
 * @hostname cannot be found in the handle cache.  Returns NULL if
 * an error occurs.
358
 */
359 360 361
struct nsm_handle *nsm_get_handle(const struct sockaddr *sap,
				  const size_t salen, const char *hostname,
				  const size_t hostname_len)
362
{
363
	struct nsm_handle *cached, *new = NULL;
364 365 366 367 368 369 370 371 372 373 374 375

	if (hostname && memchr(hostname, '/', hostname_len) != NULL) {
		if (printk_ratelimit()) {
			printk(KERN_WARNING "Invalid hostname \"%.*s\" "
					    "in NFS lock request\n",
				(int)hostname_len, hostname);
		}
		return NULL;
	}

retry:
	spin_lock(&nsm_lock);
376 377 378 379 380 381 382 383 384 385 386 387 388 389 390

	if (nsm_use_hostnames && hostname != NULL)
		cached = nsm_lookup_hostname(hostname, hostname_len);
	else
		cached = nsm_lookup_addr(sap);

	if (cached != NULL) {
		atomic_inc(&cached->sm_count);
		spin_unlock(&nsm_lock);
		kfree(new);
		dprintk("lockd: found nsm_handle for %s (%s), "
				"cnt %d\n", cached->sm_name,
				cached->sm_addrbuf,
				atomic_read(&cached->sm_count));
		return cached;
391
	}
392 393 394 395

	if (new != NULL) {
		list_add(&new->sm_link, &nsm_handles);
		spin_unlock(&nsm_lock);
396
		dprintk("lockd: created nsm_handle for %s (%s)\n",
397 398
				new->sm_name, new->sm_addrbuf);
		return new;
399
	}
400

401 402
	spin_unlock(&nsm_lock);

403 404
	new = nsm_create_handle(sap, salen, hostname, hostname_len);
	if (unlikely(new == NULL))
405 406 407 408
		return NULL;
	goto retry;
}

C
Chuck Lever 已提交
409 410 411 412
/**
 * nsm_reboot_lookup - match NLMPROC_SM_NOTIFY arguments to an nsm_handle
 * @info: pointer to NLMPROC_SM_NOTIFY arguments
 *
413 414 415
 * Returns a matching nsm_handle if found in the nsm cache. The returned
 * nsm_handle's reference count is bumped. Otherwise returns NULL if some
 * error occurred.
C
Chuck Lever 已提交
416 417 418 419 420 421 422
 */
struct nsm_handle *nsm_reboot_lookup(const struct nlm_reboot *info)
{
	struct nsm_handle *cached;

	spin_lock(&nsm_lock);

423
	cached = nsm_lookup_priv(&info->priv);
C
Chuck Lever 已提交
424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439
	if (unlikely(cached == NULL)) {
		spin_unlock(&nsm_lock);
		dprintk("lockd: never saw rebooted peer '%.*s' before\n",
				info->len, info->mon);
		return cached;
	}

	atomic_inc(&cached->sm_count);
	spin_unlock(&nsm_lock);

	dprintk("lockd: host %s (%s) rebooted, cnt %d\n",
			cached->sm_name, cached->sm_addrbuf,
			atomic_read(&cached->sm_count));
	return cached;
}

440 441 442 443 444 445 446 447 448 449
/**
 * nsm_release - Release an NSM handle
 * @nsm: pointer to handle to be released
 *
 */
void nsm_release(struct nsm_handle *nsm)
{
	if (atomic_dec_and_lock(&nsm->sm_count, &nsm_lock)) {
		list_del(&nsm->sm_link);
		spin_unlock(&nsm_lock);
450 451
		dprintk("lockd: destroyed nsm_handle for %s (%s)\n",
				nsm->sm_name, nsm->sm_addrbuf);
452 453 454 455
		kfree(nsm);
	}
}

L
Linus Torvalds 已提交
456 457
/*
 * XDR functions for NSM.
458 459 460
 *
 * See http://www.opengroup.org/ for details on the Network
 * Status Monitor wire protocol.
L
Linus Torvalds 已提交
461 462
 */

463
static void encode_nsm_string(struct xdr_stream *xdr, const char *string)
464
{
465 466 467
	const u32 len = strlen(string);
	__be32 *p;

468
	p = xdr_reserve_space(xdr, 4 + len);
469
	xdr_encode_opaque(p, string, len);
470 471
}

472 473 474
/*
 * "mon_name" specifies the host to be monitored.
 */
475
static void encode_mon_name(struct xdr_stream *xdr, const struct nsm_args *argp)
476
{
477
	encode_nsm_string(xdr, argp->mon_name);
478 479
}

480 481 482
/*
 * The "my_id" argument specifies the hostname and RPC procedure
 * to be called when the status manager receives notification
483
 * (via the NLMPROC_SM_NOTIFY call) that the state of host "mon_name"
484 485
 * has changed.
 */
486
static void encode_my_id(struct xdr_stream *xdr, const struct nsm_args *argp)
487
{
488 489
	__be32 *p;

490
	encode_nsm_string(xdr, argp->nodename);
491 492 493 494
	p = xdr_reserve_space(xdr, 4 + 4 + 4);
	*p++ = cpu_to_be32(argp->prog);
	*p++ = cpu_to_be32(argp->vers);
	*p = cpu_to_be32(argp->proc);
495 496
}

497 498
/*
 * The "mon_id" argument specifies the non-private arguments
499
 * of an NSMPROC_MON or NSMPROC_UNMON call.
500
 */
501
static void encode_mon_id(struct xdr_stream *xdr, const struct nsm_args *argp)
502
{
503 504
	encode_mon_name(xdr, argp);
	encode_my_id(xdr, argp);
505 506
}

507 508
/*
 * The "priv" argument may contain private information required
509 510
 * by the NSMPROC_MON call. This information will be supplied in the
 * NLMPROC_SM_NOTIFY call.
511
 */
512
static void encode_priv(struct xdr_stream *xdr, const struct nsm_args *argp)
513
{
514 515 516
	__be32 *p;

	p = xdr_reserve_space(xdr, SM_PRIV_SIZE);
517
	xdr_encode_opaque_fixed(p, argp->priv->data, SM_PRIV_SIZE);
518 519
}

520 521
static void nsm_xdr_enc_mon(struct rpc_rqst *req, struct xdr_stream *xdr,
			    const struct nsm_args *argp)
L
Linus Torvalds 已提交
522
{
523 524
	encode_mon_id(xdr, argp);
	encode_priv(xdr, argp);
L
Linus Torvalds 已提交
525 526
}

527 528
static void nsm_xdr_enc_unmon(struct rpc_rqst *req, struct xdr_stream *xdr,
			      const struct nsm_args *argp)
L
Linus Torvalds 已提交
529
{
530
	encode_mon_id(xdr, argp);
L
Linus Torvalds 已提交
531 532
}

533 534 535
static int nsm_xdr_dec_stat_res(struct rpc_rqst *rqstp,
				struct xdr_stream *xdr,
				struct nsm_res *resp)
L
Linus Torvalds 已提交
536
{
537
	__be32 *p;
538

539
	p = xdr_inline_decode(xdr, 4 + 4);
540 541
	if (unlikely(p == NULL))
		return -EIO;
542 543
	resp->status = be32_to_cpup(p++);
	resp->state = be32_to_cpup(p);
544

545 546
	dprintk("lockd: %s status %d state %d\n",
		__func__, resp->status, resp->state);
L
Linus Torvalds 已提交
547 548 549
	return 0;
}

550 551 552
static int nsm_xdr_dec_stat(struct rpc_rqst *rqstp,
			    struct xdr_stream *xdr,
			    struct nsm_res *resp)
L
Linus Torvalds 已提交
553
{
554
	__be32 *p;
555

556
	p = xdr_inline_decode(xdr, 4);
557 558
	if (unlikely(p == NULL))
		return -EIO;
559
	resp->state = be32_to_cpup(p);
560

561
	dprintk("lockd: %s state %d\n", __func__, resp->state);
L
Linus Torvalds 已提交
562 563 564 565
	return 0;
}

#define SM_my_name_sz	(1+XDR_QUADLEN(SM_MAXSTRLEN))
566 567 568
#define SM_my_id_sz	(SM_my_name_sz+3)
#define SM_mon_name_sz	(1+XDR_QUADLEN(SM_MAXSTRLEN))
#define SM_mon_id_sz	(SM_mon_name_sz+SM_my_id_sz)
569 570
#define SM_priv_sz	(XDR_QUADLEN(SM_PRIV_SIZE))
#define SM_mon_sz	(SM_mon_id_sz+SM_priv_sz)
L
Linus Torvalds 已提交
571 572 573 574
#define SM_monres_sz	2
#define SM_unmonres_sz	1

static struct rpc_procinfo	nsm_procedures[] = {
575 576
[NSMPROC_MON] = {
		.p_proc		= NSMPROC_MON,
577
		.p_encode	= (kxdreproc_t)nsm_xdr_enc_mon,
578
		.p_decode	= (kxdrdproc_t)nsm_xdr_dec_stat_res,
579 580
		.p_arglen	= SM_mon_sz,
		.p_replen	= SM_monres_sz,
581
		.p_statidx	= NSMPROC_MON,
582
		.p_name		= "MONITOR",
L
Linus Torvalds 已提交
583
	},
584 585
[NSMPROC_UNMON] = {
		.p_proc		= NSMPROC_UNMON,
586
		.p_encode	= (kxdreproc_t)nsm_xdr_enc_unmon,
587
		.p_decode	= (kxdrdproc_t)nsm_xdr_dec_stat,
588 589
		.p_arglen	= SM_mon_id_sz,
		.p_replen	= SM_unmonres_sz,
590
		.p_statidx	= NSMPROC_UNMON,
591
		.p_name		= "UNMONITOR",
L
Linus Torvalds 已提交
592 593 594
	},
};

595
static const struct rpc_version nsm_version1 = {
596 597
		.number		= 1,
		.nrprocs	= ARRAY_SIZE(nsm_procedures),
L
Linus Torvalds 已提交
598 599 600
		.procs		= nsm_procedures
};

601
static const struct rpc_version *nsm_version[] = {
L
Linus Torvalds 已提交
602 603 604 605 606
	[1] = &nsm_version1,
};

static struct rpc_stat		nsm_stats;

607
static const struct rpc_program nsm_program = {
L
Linus Torvalds 已提交
608
		.name		= "statd",
609
		.number		= NSM_PROGRAM,
610
		.nrvers		= ARRAY_SIZE(nsm_version),
L
Linus Torvalds 已提交
611 612 613
		.version	= nsm_version,
		.stats		= &nsm_stats
};