host.c 12.5 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
/*
 * linux/fs/lockd/host.c
 *
 * Management for NLM peer hosts. The nlm_host struct is shared
 * between client and server implementation. The only reason to
 * do so is to reduce code bloat.
 *
 * Copyright (C) 1996, Olaf Kirch <okir@monad.swb.de>
 */

#include <linux/types.h>
#include <linux/sched.h>
#include <linux/slab.h>
#include <linux/in.h>
#include <linux/sunrpc/clnt.h>
#include <linux/sunrpc/svc.h>
#include <linux/lockd/lockd.h>
#include <linux/lockd/sm_inter.h>
I
Ingo Molnar 已提交
19
#include <linux/mutex.h>
L
Linus Torvalds 已提交
20 21 22 23 24 25 26 27 28 29 30 31 32


#define NLMDBG_FACILITY		NLMDBG_HOSTCACHE
#define NLM_HOST_MAX		64
#define NLM_HOST_NRHASH		32
#define NLM_ADDRHASH(addr)	(ntohl(addr) & (NLM_HOST_NRHASH-1))
#define NLM_HOST_REBIND		(60 * HZ)
#define NLM_HOST_EXPIRE		((nrhosts > NLM_HOST_MAX)? 300 * HZ : 120 * HZ)
#define NLM_HOST_COLLECT	((nrhosts > NLM_HOST_MAX)? 120 * HZ :  60 * HZ)

static struct nlm_host *	nlm_hosts[NLM_HOST_NRHASH];
static unsigned long		next_gc;
static int			nrhosts;
I
Ingo Molnar 已提交
33
static DEFINE_MUTEX(nlm_host_mutex);
L
Linus Torvalds 已提交
34 35 36


static void			nlm_gc_hosts(void);
37 38
static struct nsm_handle *	__nsm_find(const struct sockaddr_in *,
					const char *, int, int);
L
Linus Torvalds 已提交
39 40 41 42 43

/*
 * Find an NLM server handle in the cache. If there is none, create it.
 */
struct nlm_host *
44 45
nlmclnt_lookup_host(const struct sockaddr_in *sin, int proto, int version,
			const char *hostname, int hostname_len)
L
Linus Torvalds 已提交
46
{
47 48
	return nlm_lookup_host(0, sin, proto, version,
			       hostname, hostname_len);
L
Linus Torvalds 已提交
49 50 51 52 53 54
}

/*
 * Find an NLM client handle in the cache. If there is none, create it.
 */
struct nlm_host *
55 56
nlmsvc_lookup_host(struct svc_rqst *rqstp,
			const char *hostname, int hostname_len)
L
Linus Torvalds 已提交
57 58
{
	return nlm_lookup_host(1, &rqstp->rq_addr,
59 60
			       rqstp->rq_prot, rqstp->rq_vers,
			       hostname, hostname_len);
L
Linus Torvalds 已提交
61 62 63 64 65 66
}

/*
 * Common host lookup routine for server & client
 */
struct nlm_host *
67
nlm_lookup_host(int server, const struct sockaddr_in *sin,
68 69 70
					int proto, int version,
					const char *hostname,
					int hostname_len)
L
Linus Torvalds 已提交
71 72
{
	struct nlm_host	*host, **hp;
73
	struct nsm_handle *nsm = NULL;
L
Linus Torvalds 已提交
74 75
	int		hash;

76 77 78 79 80 81
	dprintk("lockd: nlm_lookup_host(%u.%u.%u.%u, p=%d, v=%d, my role=%s, name=%.*s)\n",
			NIPQUAD(sin->sin_addr.s_addr), proto, version,
			server? "server" : "client",
			hostname_len,
			hostname? hostname : "<none>");

L
Linus Torvalds 已提交
82 83 84 85

	hash = NLM_ADDRHASH(sin->sin_addr.s_addr);

	/* Lock hash table */
I
Ingo Molnar 已提交
86
	mutex_lock(&nlm_host_mutex);
L
Linus Torvalds 已提交
87 88 89 90

	if (time_after_eq(jiffies, next_gc))
		nlm_gc_hosts();

91 92 93 94 95 96 97
	/* We may keep several nlm_host objects for a peer, because each
	 * nlm_host is identified by
	 * (address, protocol, version, server/client)
	 * We could probably simplify this a little by putting all those
	 * different NLM rpc_clients into one single nlm_host object.
	 * This would allow us to have one nlm_host per address.
	 */
L
Linus Torvalds 已提交
98
	for (hp = &nlm_hosts[hash]; (host = *hp) != 0; hp = &host->h_next) {
99 100 101 102 103 104 105
		if (!nlm_cmp_addr(&host->h_addr, sin))
			continue;

		/* See if we have an NSM handle for this client */
		if (!nsm && (nsm = host->h_nsmhandle) != 0)
			atomic_inc(&nsm->sm_count);

L
Linus Torvalds 已提交
106 107 108 109 110 111 112
		if (host->h_proto != proto)
			continue;
		if (host->h_version != version)
			continue;
		if (host->h_server != server)
			continue;

113 114 115 116
		if (hp != nlm_hosts + hash) {
			*hp = host->h_next;
			host->h_next = nlm_hosts[hash];
			nlm_hosts[hash] = host;
L
Linus Torvalds 已提交
117
		}
118 119
		nlm_get_host(host);
		goto out;
L
Linus Torvalds 已提交
120 121
	}

122 123 124 125 126
	/* Sadly, the host isn't in our hash table yet. See if
	 * we have an NSM handle for it. If not, create one.
	 */
	if (!nsm && !(nsm = nsm_find(sin, hostname, hostname_len)))
		goto out;
L
Linus Torvalds 已提交
127

128
	host = kzalloc(sizeof(*host), GFP_KERNEL);
129 130 131 132 133
	if (!host) {
		nsm_release(nsm);
		goto out;
	}
	host->h_name	   = nsm->sm_name;
L
Linus Torvalds 已提交
134 135 136 137 138
	host->h_addr       = *sin;
	host->h_addr.sin_port = 0;	/* ouch! */
	host->h_version    = version;
	host->h_proto      = proto;
	host->h_rpcclnt    = NULL;
T
Trond Myklebust 已提交
139
	mutex_init(&host->h_mutex);
L
Linus Torvalds 已提交
140 141 142 143
	host->h_nextrebind = jiffies + NLM_HOST_REBIND;
	host->h_expires    = jiffies + NLM_HOST_EXPIRE;
	atomic_set(&host->h_count, 1);
	init_waitqueue_head(&host->h_gracewait);
T
Trond Myklebust 已提交
144
	init_rwsem(&host->h_rwsem);
L
Linus Torvalds 已提交
145 146
	host->h_state      = 0;			/* pseudo NSM state */
	host->h_nsmstate   = 0;			/* real NSM state */
147
	host->h_nsmhandle  = nsm;
L
Linus Torvalds 已提交
148 149 150 151 152
	host->h_server	   = server;
	host->h_next       = nlm_hosts[hash];
	nlm_hosts[hash]    = host;
	INIT_LIST_HEAD(&host->h_lockowners);
	spin_lock_init(&host->h_lock);
153 154
	INIT_LIST_HEAD(&host->h_granted);
	INIT_LIST_HEAD(&host->h_reclaim);
L
Linus Torvalds 已提交
155 156 157 158

	if (++nrhosts > NLM_HOST_MAX)
		next_gc = 0;

159
out:
I
Ingo Molnar 已提交
160
	mutex_unlock(&nlm_host_mutex);
L
Linus Torvalds 已提交
161 162 163 164 165 166 167 168 169 170
	return host;
}

struct nlm_host *
nlm_find_client(void)
{
	/* find a nlm_host for a client for which h_killed == 0.
	 * and return it
	 */
	int hash;
I
Ingo Molnar 已提交
171
	mutex_lock(&nlm_host_mutex);
L
Linus Torvalds 已提交
172 173 174 175 176 177
	for (hash = 0 ; hash < NLM_HOST_NRHASH; hash++) {
		struct nlm_host *host, **hp;
		for (hp = &nlm_hosts[hash]; (host = *hp) != 0; hp = &host->h_next) {
			if (host->h_server &&
			    host->h_killed == 0) {
				nlm_get_host(host);
I
Ingo Molnar 已提交
178
				mutex_unlock(&nlm_host_mutex);
L
Linus Torvalds 已提交
179 180 181 182
				return host;
			}
		}
	}
I
Ingo Molnar 已提交
183
	mutex_unlock(&nlm_host_mutex);
L
Linus Torvalds 已提交
184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199
	return NULL;
}

				
/*
 * Create the NLM RPC client for an NLM peer
 */
struct rpc_clnt *
nlm_bind_host(struct nlm_host *host)
{
	struct rpc_clnt	*clnt;

	dprintk("lockd: nlm_bind_host(%08x)\n",
			(unsigned)ntohl(host->h_addr.sin_addr.s_addr));

	/* Lock host handle */
T
Trond Myklebust 已提交
200
	mutex_lock(&host->h_mutex);
L
Linus Torvalds 已提交
201 202 203 204 205

	/* If we've already created an RPC client, check whether
	 * RPC rebind is required
	 */
	if ((clnt = host->h_rpcclnt) != NULL) {
206
		if (time_after_eq(jiffies, host->h_nextrebind)) {
207
			rpc_force_rebind(clnt);
L
Linus Torvalds 已提交
208 209 210 211 212
			host->h_nextrebind = jiffies + NLM_HOST_REBIND;
			dprintk("lockd: next rebind in %ld jiffies\n",
					host->h_nextrebind - jiffies);
		}
	} else {
213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239
		unsigned long increment = nlmsvc_timeout * HZ;
		struct rpc_timeout timeparms = {
			.to_initval	= increment,
			.to_increment	= increment,
			.to_maxval	= increment * 6UL,
			.to_retries	= 5U,
		};
		struct rpc_create_args args = {
			.protocol	= host->h_proto,
			.address	= (struct sockaddr *)&host->h_addr,
			.addrsize	= sizeof(host->h_addr),
			.timeout	= &timeparms,
			.servername	= host->h_name,
			.program	= &nlm_program,
			.version	= host->h_version,
			.authflavor	= RPC_AUTH_UNIX,
			.flags		= (RPC_CLNT_CREATE_HARDRTRY |
					   RPC_CLNT_CREATE_AUTOBIND),
		};

		clnt = rpc_create(&args);
		if (!IS_ERR(clnt))
			host->h_rpcclnt = clnt;
		else {
			printk("lockd: couldn't create RPC handle for %s\n", host->h_name);
			clnt = NULL;
		}
L
Linus Torvalds 已提交
240 241
	}

T
Trond Myklebust 已提交
242
	mutex_unlock(&host->h_mutex);
L
Linus Torvalds 已提交
243 244 245 246 247 248 249 250 251 252 253
	return clnt;
}

/*
 * Force a portmap lookup of the remote lockd port
 */
void
nlm_rebind_host(struct nlm_host *host)
{
	dprintk("lockd: rebind host %s\n", host->h_name);
	if (host->h_rpcclnt && time_after_eq(jiffies, host->h_nextrebind)) {
254
		rpc_force_rebind(host->h_rpcclnt);
L
Linus Torvalds 已提交
255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279
		host->h_nextrebind = jiffies + NLM_HOST_REBIND;
	}
}

/*
 * Increment NLM host count
 */
struct nlm_host * nlm_get_host(struct nlm_host *host)
{
	if (host) {
		dprintk("lockd: get host %s\n", host->h_name);
		atomic_inc(&host->h_count);
		host->h_expires = jiffies + NLM_HOST_EXPIRE;
	}
	return host;
}

/*
 * Release NLM host after use
 */
void nlm_release_host(struct nlm_host *host)
{
	if (host != NULL) {
		dprintk("lockd: release host %s\n", host->h_name);
		BUG_ON(atomic_read(&host->h_count) < 0);
280 281 282 283 284
		if (atomic_dec_and_test(&host->h_count)) {
			BUG_ON(!list_empty(&host->h_lockowners));
			BUG_ON(!list_empty(&host->h_granted));
			BUG_ON(!list_empty(&host->h_reclaim));
		}
L
Linus Torvalds 已提交
285 286 287
	}
}

288 289 290 291 292
/*
 * We were notified that the host indicated by address &sin
 * has rebooted.
 * Release all resources held by that peer.
 */
293 294 295
void nlm_host_rebooted(const struct sockaddr_in *sin,
				const char *hostname, int hostname_len,
				u32 new_state)
296
{
297 298 299
	struct nsm_handle *nsm;
	struct nlm_host	*host, **hp;
	int		hash;
300

301 302 303 304 305
	dprintk("lockd: nlm_host_rebooted(%s, %u.%u.%u.%u)\n",
			hostname, NIPQUAD(sin->sin_addr));

	/* Find the NSM handle for this peer */
	if (!(nsm = __nsm_find(sin, hostname, hostname_len, 0)))
306 307
		return;

308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340
	/* When reclaiming locks on this peer, make sure that
	 * we set up a new notification */
	nsm->sm_monitored = 0;

	/* Mark all hosts tied to this NSM state as having rebooted.
	 * We run the loop repeatedly, because we drop the host table
	 * lock for this.
	 * To avoid processing a host several times, we match the nsmstate.
	 */
again:	mutex_lock(&nlm_host_mutex);
	for (hash = 0; hash < NLM_HOST_NRHASH; hash++) {
		for (hp = &nlm_hosts[hash]; (host = *hp); hp = &host->h_next) {
			if (host->h_nsmhandle == nsm
			 && host->h_nsmstate != new_state) {
				host->h_nsmstate = new_state;
				host->h_state++;

				nlm_get_host(host);
				mutex_unlock(&nlm_host_mutex);

				if (host->h_server) {
					/* We're server for this guy, just ditch
					 * all the locks he held. */
					nlmsvc_free_host_resources(host);
				} else {
					/* He's the server, initiate lock recovery. */
					nlmclnt_recovery(host);
				}

				nlm_release_host(host);
				goto again;
			}
		}
341
	}
342 343

	mutex_unlock(&nlm_host_mutex);
344 345
}

L
Linus Torvalds 已提交
346 347 348 349 350 351 352 353 354 355 356
/*
 * Shut down the hosts module.
 * Note that this routine is called only at server shutdown time.
 */
void
nlm_shutdown_hosts(void)
{
	struct nlm_host	*host;
	int		i;

	dprintk("lockd: shutting down host module\n");
I
Ingo Molnar 已提交
357
	mutex_lock(&nlm_host_mutex);
L
Linus Torvalds 已提交
358 359 360 361 362 363 364 365 366 367

	/* First, make all hosts eligible for gc */
	dprintk("lockd: nuking all hosts...\n");
	for (i = 0; i < NLM_HOST_NRHASH; i++) {
		for (host = nlm_hosts[i]; host; host = host->h_next)
			host->h_expires = jiffies - 1;
	}

	/* Then, perform a garbage collection pass */
	nlm_gc_hosts();
I
Ingo Molnar 已提交
368
	mutex_unlock(&nlm_host_mutex);
L
Linus Torvalds 已提交
369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417

	/* complain if any hosts are left */
	if (nrhosts) {
		printk(KERN_WARNING "lockd: couldn't shutdown host module!\n");
		dprintk("lockd: %d hosts left:\n", nrhosts);
		for (i = 0; i < NLM_HOST_NRHASH; i++) {
			for (host = nlm_hosts[i]; host; host = host->h_next) {
				dprintk("       %s (cnt %d use %d exp %ld)\n",
					host->h_name, atomic_read(&host->h_count),
					host->h_inuse, host->h_expires);
			}
		}
	}
}

/*
 * Garbage collect any unused NLM hosts.
 * This GC combines reference counting for async operations with
 * mark & sweep for resources held by remote clients.
 */
static void
nlm_gc_hosts(void)
{
	struct nlm_host	**q, *host;
	struct rpc_clnt	*clnt;
	int		i;

	dprintk("lockd: host garbage collection\n");
	for (i = 0; i < NLM_HOST_NRHASH; i++) {
		for (host = nlm_hosts[i]; host; host = host->h_next)
			host->h_inuse = 0;
	}

	/* Mark all hosts that hold locks, blocks or shares */
	nlmsvc_mark_resources();

	for (i = 0; i < NLM_HOST_NRHASH; i++) {
		q = &nlm_hosts[i];
		while ((host = *q) != NULL) {
			if (atomic_read(&host->h_count) || host->h_inuse
			 || time_before(jiffies, host->h_expires)) {
				dprintk("nlm_gc_hosts skipping %s (cnt %d use %d exp %ld)\n",
					host->h_name, atomic_read(&host->h_count),
					host->h_inuse, host->h_expires);
				q = &host->h_next;
				continue;
			}
			dprintk("lockd: delete host %s\n", host->h_name);
			*q = host->h_next;
418 419 420 421 422 423

			/*
			 * Unmonitor unless host was invalidated (i.e. lockd restarted)
			 */
			nsm_unmonitor(host);

L
Linus Torvalds 已提交
424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440
			if ((clnt = host->h_rpcclnt) != NULL) {
				if (atomic_read(&clnt->cl_users)) {
					printk(KERN_WARNING
						"lockd: active RPC handle\n");
					clnt->cl_dead = 1;
				} else {
					rpc_destroy_client(host->h_rpcclnt);
				}
			}
			kfree(host);
			nrhosts--;
		}
	}

	next_gc = jiffies + NLM_HOST_COLLECT;
}

441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520

/*
 * Manage NSM handles
 */
static LIST_HEAD(nsm_handles);
static DECLARE_MUTEX(nsm_sema);

static struct nsm_handle *
__nsm_find(const struct sockaddr_in *sin,
		const char *hostname, int hostname_len,
		int create)
{
	struct nsm_handle *nsm = NULL;
	struct list_head *pos;

	if (!sin)
		return NULL;

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

	down(&nsm_sema);
	list_for_each(pos, &nsm_handles) {
		nsm = list_entry(pos, struct nsm_handle, sm_link);

		if (!nlm_cmp_addr(&nsm->sm_addr, sin))
			continue;
		atomic_inc(&nsm->sm_count);
		goto out;
	}

	if (!create) {
		nsm = NULL;
		goto out;
	}

	nsm = kzalloc(sizeof(*nsm) + hostname_len + 1, GFP_KERNEL);
	if (nsm != NULL) {
		nsm->sm_addr = *sin;
		nsm->sm_name = (char *) (nsm + 1);
		memcpy(nsm->sm_name, hostname, hostname_len);
		nsm->sm_name[hostname_len] = '\0';
		atomic_set(&nsm->sm_count, 1);

		list_add(&nsm->sm_link, &nsm_handles);
	}

out:	up(&nsm_sema);
	return nsm;
}

struct nsm_handle *
nsm_find(const struct sockaddr_in *sin, const char *hostname, int hostname_len)
{
	return __nsm_find(sin, hostname, hostname_len, 1);
}

/*
 * Release an NSM handle
 */
void
nsm_release(struct nsm_handle *nsm)
{
	if (!nsm)
		return;
	if (atomic_dec_and_test(&nsm->sm_count)) {
		down(&nsm_sema);
		if (atomic_read(&nsm->sm_count) == 0) {
			list_del(&nsm->sm_link);
			kfree(nsm);
		}
		up(&nsm_sema);
	}
}