lockd: reject reclaims outside the grace period

[mirror_ubuntu-artful-kernel.git] / fs / lockd / host.c

/*
 * 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/slab.h>
#include <linux/in.h>
#include <linux/in6.h>
#include <linux/sunrpc/clnt.h>
#include <linux/sunrpc/svc.h>
#include <linux/lockd/lockd.h>
#include <linux/lockd/sm_inter.h>
#include <linux/mutex.h>

#include <net/ipv6.h>

#define NLMDBG_FACILITY		NLMDBG_HOSTCACHE
#define NLM_HOST_NRHASH		32
#define NLM_HOST_REBIND		(60 * HZ)
#define NLM_HOST_EXPIRE		(300 * HZ)
#define NLM_HOST_COLLECT	(120 * HZ)

static struct hlist_head	nlm_hosts[NLM_HOST_NRHASH];
static unsigned long		next_gc;
static int			nrhosts;
static DEFINE_MUTEX(nlm_host_mutex);

static void			nlm_gc_hosts(void);
static struct nsm_handle	*nsm_find(const struct sockaddr *sap,
						const size_t salen,
						const char *hostname,
						const size_t hostname_len,
						const int create);

/*
 * Hash function must work well on big- and little-endian platforms
 */
static unsigned int __nlm_hash32(const __be32 n)
{
	unsigned int hash = (__force u32)n ^ ((__force u32)n >> 16);
	return hash ^ (hash >> 8);
}

static unsigned int __nlm_hash_addr4(const struct sockaddr *sap)
{
	const struct sockaddr_in *sin = (struct sockaddr_in *)sap;
	return __nlm_hash32(sin->sin_addr.s_addr);
}

static unsigned int __nlm_hash_addr6(const struct sockaddr *sap)
{
	const struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sap;
	const struct in6_addr addr = sin6->sin6_addr;
	return __nlm_hash32(addr.s6_addr32[0]) ^
	       __nlm_hash32(addr.s6_addr32[1]) ^
	       __nlm_hash32(addr.s6_addr32[2]) ^
	       __nlm_hash32(addr.s6_addr32[3]);
}

static unsigned int nlm_hash_address(const struct sockaddr *sap)
{
	unsigned int hash;

	switch (sap->sa_family) {
	case AF_INET:
		hash = __nlm_hash_addr4(sap);
		break;
	case AF_INET6:
		hash = __nlm_hash_addr6(sap);
		break;
	default:
		hash = 0;
	}
	return hash & (NLM_HOST_NRHASH - 1);
}

static void nlm_clear_port(struct sockaddr *sap)
{
	switch (sap->sa_family) {
	case AF_INET:
		((struct sockaddr_in *)sap)->sin_port = 0;
		break;
	case AF_INET6:
		((struct sockaddr_in6 *)sap)->sin6_port = 0;
		break;
	}
}

static void nlm_display_address(const struct sockaddr *sap,
				char *buf, const size_t len)
{
	const struct sockaddr_in *sin = (struct sockaddr_in *)sap;
	const struct sockaddr_in6 *sin6 = (struct sockaddr_in6 *)sap;

	switch (sap->sa_family) {
	case AF_UNSPEC:
		snprintf(buf, len, "unspecified");
		break;
	case AF_INET:
		snprintf(buf, len, NIPQUAD_FMT, NIPQUAD(sin->sin_addr.s_addr));
		break;
	case AF_INET6:
		if (ipv6_addr_v4mapped(&sin6->sin6_addr))
			snprintf(buf, len, NIPQUAD_FMT,
				 NIPQUAD(sin6->sin6_addr.s6_addr32[3]));
		else
			snprintf(buf, len, NIP6_FMT, NIP6(sin6->sin6_addr));
		break;
	default:
		snprintf(buf, len, "unsupported address family");
		break;
	}
}

/*
 * Common host lookup routine for server & client
 */
static struct nlm_host *nlm_lookup_host(int server,
					const struct sockaddr_in *sin,
					int proto, u32 version,
					const char *hostname,
					unsigned int hostname_len,
					const struct sockaddr_in *ssin)
{
	struct hlist_head *chain;
	struct hlist_node *pos;
	struct nlm_host	*host;
	struct nsm_handle *nsm = NULL;

	dprintk("lockd: nlm_lookup_host(proto=%d, vers=%u,"
			" my role is %s, hostname=%.*s)\n",
			proto, version, server ? "server" : "client",
			hostname_len, hostname ? hostname : "<none>");

	mutex_lock(&nlm_host_mutex);

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

	/* 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.
	 */
	chain = &nlm_hosts[nlm_hash_address((struct sockaddr *)sin)];
	hlist_for_each_entry(host, pos, chain, h_hash) {
		if (!nlm_cmp_addr(nlm_addr(host), (struct sockaddr *)sin))
			continue;

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

		if (host->h_proto != proto)
			continue;
		if (host->h_version != version)
			continue;
		if (host->h_server != server)
			continue;
		if (!nlm_cmp_addr(nlm_srcaddr(host), (struct sockaddr *)ssin))
			continue;

		/* Move to head of hash chain. */
		hlist_del(&host->h_hash);
		hlist_add_head(&host->h_hash, chain);

		nlm_get_host(host);
		dprintk("lockd: nlm_lookup_host found host %s (%s)\n",
				host->h_name, host->h_addrbuf);
		goto out;
	}

	/*
	 * The host wasn't in our hash table.  If we don't
	 * have an NSM handle for it yet, create one.
	 */
	if (nsm)
		atomic_inc(&nsm->sm_count);
	else {
		host = NULL;
		nsm = nsm_find((struct sockaddr *)sin, sizeof(*sin),
				hostname, hostname_len, 1);
		if (!nsm) {
			dprintk("lockd: nlm_lookup_host failed; "
				"no nsm handle\n");
			goto out;
		}
	}

	host = kzalloc(sizeof(*host), GFP_KERNEL);
	if (!host) {
		nsm_release(nsm);
		dprintk("lockd: nlm_lookup_host failed; no memory\n");
		goto out;
	}
	host->h_name	   = nsm->sm_name;
	memcpy(nlm_addr(host), sin, sizeof(*sin));
	host->h_addrlen = sizeof(*sin);
	nlm_clear_port(nlm_addr(host));
	memcpy(nlm_srcaddr(host), ssin, sizeof(*ssin));
	host->h_version    = version;
	host->h_proto      = proto;
	host->h_rpcclnt    = NULL;
	mutex_init(&host->h_mutex);
	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);
	init_rwsem(&host->h_rwsem);
	host->h_state      = 0;			/* pseudo NSM state */
	host->h_nsmstate   = 0;			/* real NSM state */
	host->h_nsmhandle  = nsm;
	host->h_server	   = server;
	hlist_add_head(&host->h_hash, chain);
	INIT_LIST_HEAD(&host->h_lockowners);
	spin_lock_init(&host->h_lock);
	INIT_LIST_HEAD(&host->h_granted);
	INIT_LIST_HEAD(&host->h_reclaim);

	nrhosts++;

	nlm_display_address((struct sockaddr *)&host->h_addr,
				host->h_addrbuf, sizeof(host->h_addrbuf));
	nlm_display_address((struct sockaddr *)&host->h_srcaddr,
				host->h_srcaddrbuf, sizeof(host->h_srcaddrbuf));

	dprintk("lockd: nlm_lookup_host created host %s\n",
			host->h_name);

out:
	mutex_unlock(&nlm_host_mutex);
	return host;
}

/*
 * Destroy a host
 */
static void
nlm_destroy_host(struct nlm_host *host)
{
	struct rpc_clnt	*clnt;

	BUG_ON(!list_empty(&host->h_lockowners));
	BUG_ON(atomic_read(&host->h_count));

	/*
	 * Release NSM handle and unmonitor host.
	 */
	nsm_unmonitor(host);

	clnt = host->h_rpcclnt;
	if (clnt != NULL)
		rpc_shutdown_client(clnt);
	kfree(host);
}

/*
 * Find an NLM server handle in the cache. If there is none, create it.
 */
struct nlm_host *nlmclnt_lookup_host(const struct sockaddr_in *sin,
				     int proto, u32 version,
				     const char *hostname,
				     unsigned int hostname_len)
{
	const struct sockaddr_in source = {
		.sin_family	= AF_UNSPEC,
	};

	return nlm_lookup_host(0, sin, proto, version,
			       hostname, hostname_len, &source);
}

/*
 * Find an NLM client handle in the cache. If there is none, create it.
 */
struct nlm_host *
nlmsvc_lookup_host(struct svc_rqst *rqstp,
			const char *hostname, unsigned int hostname_len)
{
	const struct sockaddr_in source = {
		.sin_family	= AF_INET,
		.sin_addr	= rqstp->rq_daddr.addr,
	};

	return nlm_lookup_host(1, svc_addr_in(rqstp),
			       rqstp->rq_prot, rqstp->rq_vers,
			       hostname, hostname_len, &source);
}

/*
 * 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 %s (%s), my addr=%s\n",
			host->h_name, host->h_addrbuf, host->h_srcaddrbuf);

	/* Lock host handle */
	mutex_lock(&host->h_mutex);

	/* If we've already created an RPC client, check whether
	 * RPC rebind is required
	 */
	if ((clnt = host->h_rpcclnt) != NULL) {
		if (time_after_eq(jiffies, host->h_nextrebind)) {
			rpc_force_rebind(clnt);
			host->h_nextrebind = jiffies + NLM_HOST_REBIND;
			dprintk("lockd: next rebind in %lu jiffies\n",
					host->h_nextrebind - jiffies);
		}
	} else {
		unsigned long increment = nlmsvc_timeout;
		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	= nlm_addr(host),
			.addrsize	= host->h_addrlen,
			.saddress	= nlm_srcaddr(host),
			.timeout	= &timeparms,
			.servername	= host->h_name,
			.program	= &nlm_program,
			.version	= host->h_version,
			.authflavor	= RPC_AUTH_UNIX,
			.flags		= (RPC_CLNT_CREATE_NOPING |
					   RPC_CLNT_CREATE_AUTOBIND),
		};

		/*
		 * lockd retries server side blocks automatically so we want
		 * those to be soft RPC calls. Client side calls need to be
		 * hard RPC tasks.
		 */
		if (!host->h_server)
			args.flags |= RPC_CLNT_CREATE_HARDRTRY;

		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;
		}
	}

	mutex_unlock(&host->h_mutex);
	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)) {
		rpc_force_rebind(host->h_rpcclnt);
		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);
		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));
		}
	}
}

/*
 * We were notified that the host indicated by address &sin
 * has rebooted.
 * Release all resources held by that peer.
 */
void nlm_host_rebooted(const struct sockaddr_in *sin,
				const char *hostname,
				unsigned int hostname_len,
				u32 new_state)
{
	struct hlist_head *chain;
	struct hlist_node *pos;
	struct nsm_handle *nsm;
	struct nlm_host	*host;

	nsm = nsm_find((struct sockaddr *)sin, sizeof(*sin),
			hostname, hostname_len, 0);
	if (nsm == NULL) {
		dprintk("lockd: never saw rebooted peer '%.*s' before\n",
				hostname_len, hostname);
		return;
	}

	dprintk("lockd: nlm_host_rebooted(%.*s, %s)\n",
			hostname_len, hostname, nsm->sm_addrbuf);

	/* 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 (chain = nlm_hosts; chain < nlm_hosts + NLM_HOST_NRHASH; ++chain) {
		hlist_for_each_entry(host, pos, chain, h_hash) {
			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;
			}
		}
	}

	mutex_unlock(&nlm_host_mutex);
}

/*
 * Shut down the hosts module.
 * Note that this routine is called only at server shutdown time.
 */
void
nlm_shutdown_hosts(void)
{
	struct hlist_head *chain;
	struct hlist_node *pos;
	struct nlm_host	*host;

	dprintk("lockd: shutting down host module\n");
	mutex_lock(&nlm_host_mutex);

	/* First, make all hosts eligible for gc */
	dprintk("lockd: nuking all hosts...\n");
	for (chain = nlm_hosts; chain < nlm_hosts + NLM_HOST_NRHASH; ++chain) {
		hlist_for_each_entry(host, pos, chain, h_hash) {
			host->h_expires = jiffies - 1;
			if (host->h_rpcclnt) {
				rpc_shutdown_client(host->h_rpcclnt);
				host->h_rpcclnt = NULL;
			}
		}
	}

	/* Then, perform a garbage collection pass */
	nlm_gc_hosts();
	mutex_unlock(&nlm_host_mutex);

	/* 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 (chain = nlm_hosts; chain < nlm_hosts + NLM_HOST_NRHASH; ++chain) {
			hlist_for_each_entry(host, pos, chain, h_hash) {
				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 hlist_head *chain;
	struct hlist_node *pos, *next;
	struct nlm_host	*host;

	dprintk("lockd: host garbage collection\n");
	for (chain = nlm_hosts; chain < nlm_hosts + NLM_HOST_NRHASH; ++chain) {
		hlist_for_each_entry(host, pos, chain, h_hash)
			host->h_inuse = 0;
	}

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

	for (chain = nlm_hosts; chain < nlm_hosts + NLM_HOST_NRHASH; ++chain) {
		hlist_for_each_entry_safe(host, pos, next, chain, h_hash) {
			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);
				continue;
			}
			dprintk("lockd: delete host %s\n", host->h_name);
			hlist_del_init(&host->h_hash);

			nlm_destroy_host(host);
			nrhosts--;
		}
	}

	next_gc = jiffies + NLM_HOST_COLLECT;
}


/*
 * Manage NSM handles
 */
static LIST_HEAD(nsm_handles);
static DEFINE_SPINLOCK(nsm_lock);

static struct nsm_handle *nsm_find(const struct sockaddr *sap,
				   const size_t salen,
				   const char *hostname,
				   const size_t hostname_len,
				   const int create)
{
	struct nsm_handle *nsm = NULL;
	struct nsm_handle *pos;

	if (!sap)
		return NULL;

	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);
	list_for_each_entry(pos, &nsm_handles, sm_link) {

		if (hostname && nsm_use_hostnames) {
			if (strlen(pos->sm_name) != hostname_len
			 || memcmp(pos->sm_name, hostname, hostname_len))
				continue;
		} else if (!nlm_cmp_addr(nsm_addr(pos), sap))
			continue;
		atomic_inc(&pos->sm_count);
		kfree(nsm);
		nsm = pos;
		goto found;
	}
	if (nsm) {
		list_add(&nsm->sm_link, &nsm_handles);
		goto found;
	}
	spin_unlock(&nsm_lock);

	if (!create)
		return NULL;

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

	memcpy(nsm_addr(nsm), sap, salen);
	nsm->sm_addrlen = salen;
	nsm->sm_name = (char *) (nsm + 1);
	memcpy(nsm->sm_name, hostname, hostname_len);
	nsm->sm_name[hostname_len] = '\0';
	nlm_display_address((struct sockaddr *)&nsm->sm_addr,
				nsm->sm_addrbuf, sizeof(nsm->sm_addrbuf));
	atomic_set(&nsm->sm_count, 1);
	goto retry;

found:
	spin_unlock(&nsm_lock);
	return nsm;
}

/*
 * Release an NSM handle
 */
void
nsm_release(struct nsm_handle *nsm)
{
	if (!nsm)
		return;
	if (atomic_dec_and_lock(&nsm->sm_count, &nsm_lock)) {
		list_del(&nsm->sm_link);
		spin_unlock(&nsm_lock);
		kfree(nsm);
	}
}