cache.c

来自「linux 内核源代码」· C语言 代码 · 共 1,289 行 · 第 1/2 页

/*
 * net/sunrpc/cache.c
 *
 * Generic code for various authentication-related caches
 * used by sunrpc clients and servers.
 *
 * Copyright (C) 2002 Neil Brown <[e-mail]>
 *
 * Released under terms in GPL version 2.  See COPYING.
 *
 */

#include <linux/types.h>
#include <linux/fs.h>
#include <linux/file.h>
#include <linux/slab.h>
#include <linux/signal.h>
#include <linux/sched.h>
#include <linux/kmod.h>
#include <linux/list.h>
#include <linux/module.h>
#include <linux/ctype.h>
#include <asm/uaccess.h>
#include <linux/poll.h>
#include <linux/seq_file.h>
#include <linux/proc_fs.h>
#include <linux/net.h>
#include <linux/workqueue.h>
#include <linux/mutex.h>
#include <asm/ioctls.h>
#include <linux/sunrpc/types.h>
#include <linux/sunrpc/cache.h>
#include <linux/sunrpc/stats.h>

#define	 RPCDBG_FACILITY RPCDBG_CACHE

static int cache_defer_req(struct cache_req *req, struct cache_head *item);
static void cache_revisit_request(struct cache_head *item);

static void cache_init(struct cache_head *h)
{
	time_t now = get_seconds();
	h->next = NULL;
	h->flags = 0;
	kref_init(&h->ref);
	h->expiry_time = now + CACHE_NEW_EXPIRY;
	h->last_refresh = now;
}

struct cache_head *sunrpc_cache_lookup(struct cache_detail *detail,
				       struct cache_head *key, int hash)
{
	struct cache_head **head,  **hp;
	struct cache_head *new = NULL;

	head = &detail->hash_table[hash];

	read_lock(&detail->hash_lock);

	for (hp=head; *hp != NULL ; hp = &(*hp)->next) {
		struct cache_head *tmp = *hp;
		if (detail->match(tmp, key)) {
			cache_get(tmp);
			read_unlock(&detail->hash_lock);
			return tmp;
		}
	}
	read_unlock(&detail->hash_lock);
	/* Didn't find anything, insert an empty entry */

	new = detail->alloc();
	if (!new)
		return NULL;
	/* must fully initialise 'new', else
	 * we might get lose if we need to
	 * cache_put it soon.
	 */
	cache_init(new);
	detail->init(new, key);

	write_lock(&detail->hash_lock);

	/* check if entry appeared while we slept */
	for (hp=head; *hp != NULL ; hp = &(*hp)->next) {
		struct cache_head *tmp = *hp;
		if (detail->match(tmp, key)) {
			cache_get(tmp);
			write_unlock(&detail->hash_lock);
			cache_put(new, detail);
			return tmp;
		}
	}
	new->next = *head;
	*head = new;
	detail->entries++;
	cache_get(new);
	write_unlock(&detail->hash_lock);

	return new;
}
EXPORT_SYMBOL(sunrpc_cache_lookup);


static void queue_loose(struct cache_detail *detail, struct cache_head *ch);

static int cache_fresh_locked(struct cache_head *head, time_t expiry)
{
	head->expiry_time = expiry;
	head->last_refresh = get_seconds();
	return !test_and_set_bit(CACHE_VALID, &head->flags);
}

static void cache_fresh_unlocked(struct cache_head *head,
			struct cache_detail *detail, int new)
{
	if (new)
		cache_revisit_request(head);
	if (test_and_clear_bit(CACHE_PENDING, &head->flags)) {
		cache_revisit_request(head);
		queue_loose(detail, head);
	}
}

struct cache_head *sunrpc_cache_update(struct cache_detail *detail,
				       struct cache_head *new, struct cache_head *old, int hash)
{
	/* The 'old' entry is to be replaced by 'new'.
	 * If 'old' is not VALID, we update it directly,
	 * otherwise we need to replace it
	 */
	struct cache_head **head;
	struct cache_head *tmp;
	int is_new;

	if (!test_bit(CACHE_VALID, &old->flags)) {
		write_lock(&detail->hash_lock);
		if (!test_bit(CACHE_VALID, &old->flags)) {
			if (test_bit(CACHE_NEGATIVE, &new->flags))
				set_bit(CACHE_NEGATIVE, &old->flags);
			else
				detail->update(old, new);
			is_new = cache_fresh_locked(old, new->expiry_time);
			write_unlock(&detail->hash_lock);
			cache_fresh_unlocked(old, detail, is_new);
			return old;
		}
		write_unlock(&detail->hash_lock);
	}
	/* We need to insert a new entry */
	tmp = detail->alloc();
	if (!tmp) {
		cache_put(old, detail);
		return NULL;
	}
	cache_init(tmp);
	detail->init(tmp, old);
	head = &detail->hash_table[hash];

	write_lock(&detail->hash_lock);
	if (test_bit(CACHE_NEGATIVE, &new->flags))
		set_bit(CACHE_NEGATIVE, &tmp->flags);
	else
		detail->update(tmp, new);
	tmp->next = *head;
	*head = tmp;
	detail->entries++;
	cache_get(tmp);
	is_new = cache_fresh_locked(tmp, new->expiry_time);
	cache_fresh_locked(old, 0);
	write_unlock(&detail->hash_lock);
	cache_fresh_unlocked(tmp, detail, is_new);
	cache_fresh_unlocked(old, detail, 0);
	cache_put(old, detail);
	return tmp;
}
EXPORT_SYMBOL(sunrpc_cache_update);

static int cache_make_upcall(struct cache_detail *detail, struct cache_head *h);
/*
 * This is the generic cache management routine for all
 * the authentication caches.
 * It checks the currency of a cache item and will (later)
 * initiate an upcall to fill it if needed.
 *
 *
 * Returns 0 if the cache_head can be used, or cache_puts it and returns
 * -EAGAIN if upcall is pending,
 * -ETIMEDOUT if upcall failed and should be retried,
 * -ENOENT if cache entry was negative
 */
int cache_check(struct cache_detail *detail,
		    struct cache_head *h, struct cache_req *rqstp)
{
	int rv;
	long refresh_age, age;

	/* First decide return status as best we can */
	if (!test_bit(CACHE_VALID, &h->flags) ||
	    h->expiry_time < get_seconds())
		rv = -EAGAIN;
	else if (detail->flush_time > h->last_refresh)
		rv = -EAGAIN;
	else {
		/* entry is valid */
		if (test_bit(CACHE_NEGATIVE, &h->flags))
			rv = -ENOENT;
		else rv = 0;
	}

	/* now see if we want to start an upcall */
	refresh_age = (h->expiry_time - h->last_refresh);
	age = get_seconds() - h->last_refresh;

	if (rqstp == NULL) {
		if (rv == -EAGAIN)
			rv = -ENOENT;
	} else if (rv == -EAGAIN || age > refresh_age/2) {
		dprintk("RPC:       Want update, refage=%ld, age=%ld\n",
				refresh_age, age);
		if (!test_and_set_bit(CACHE_PENDING, &h->flags)) {
			switch (cache_make_upcall(detail, h)) {
			case -EINVAL:
				clear_bit(CACHE_PENDING, &h->flags);
				if (rv == -EAGAIN) {
					set_bit(CACHE_NEGATIVE, &h->flags);
					cache_fresh_unlocked(h, detail,
					     cache_fresh_locked(h, get_seconds()+CACHE_NEW_EXPIRY));
					rv = -ENOENT;
				}
				break;

			case -EAGAIN:
				clear_bit(CACHE_PENDING, &h->flags);
				cache_revisit_request(h);
				break;
			}
		}
	}

	if (rv == -EAGAIN)
		if (cache_defer_req(rqstp, h) != 0)
			rv = -ETIMEDOUT;

	if (rv)
		cache_put(h, detail);
	return rv;
}

/*
 * caches need to be periodically cleaned.
 * For this we maintain a list of cache_detail and
 * a current pointer into that list and into the table
 * for that entry.
 *
 * Each time clean_cache is called it finds the next non-empty entry
 * in the current table and walks the list in that entry
 * looking for entries that can be removed.
 *
 * An entry gets removed if:
 * - The expiry is before current time
 * - The last_refresh time is before the flush_time for that cache
 *
 * later we might drop old entries with non-NEVER expiry if that table
 * is getting 'full' for some definition of 'full'
 *
 * The question of "how often to scan a table" is an interesting one
 * and is answered in part by the use of the "nextcheck" field in the
 * cache_detail.
 * When a scan of a table begins, the nextcheck field is set to a time
 * that is well into the future.
 * While scanning, if an expiry time is found that is earlier than the
 * current nextcheck time, nextcheck is set to that expiry time.
 * If the flush_time is ever set to a time earlier than the nextcheck
 * time, the nextcheck time is then set to that flush_time.
 *
 * A table is then only scanned if the current time is at least
 * the nextcheck time.
 *
 */

static LIST_HEAD(cache_list);
static DEFINE_SPINLOCK(cache_list_lock);
static struct cache_detail *current_detail;
static int current_index;

static const struct file_operations cache_file_operations;
static const struct file_operations content_file_operations;
static const struct file_operations cache_flush_operations;

static void do_cache_clean(struct work_struct *work);
static DECLARE_DELAYED_WORK(cache_cleaner, do_cache_clean);

void cache_register(struct cache_detail *cd)
{
	cd->proc_ent = proc_mkdir(cd->name, proc_net_rpc);
	if (cd->proc_ent) {
		struct proc_dir_entry *p;
		cd->proc_ent->owner = cd->owner;
		cd->channel_ent = cd->content_ent = NULL;

		p = create_proc_entry("flush", S_IFREG|S_IRUSR|S_IWUSR,
				      cd->proc_ent);
		cd->flush_ent =  p;
		if (p) {
			p->proc_fops = &cache_flush_operations;
			p->owner = cd->owner;
			p->data = cd;
		}

		if (cd->cache_request || cd->cache_parse) {
			p = create_proc_entry("channel", S_IFREG|S_IRUSR|S_IWUSR,
					      cd->proc_ent);
			cd->channel_ent = p;
			if (p) {
				p->proc_fops = &cache_file_operations;
				p->owner = cd->owner;
				p->data = cd;
			}
		}
		if (cd->cache_show) {
			p = create_proc_entry("content", S_IFREG|S_IRUSR|S_IWUSR,
					      cd->proc_ent);
			cd->content_ent = p;
			if (p) {
				p->proc_fops = &content_file_operations;
				p->owner = cd->owner;
				p->data = cd;
			}
		}
	}
	rwlock_init(&cd->hash_lock);
	INIT_LIST_HEAD(&cd->queue);
	spin_lock(&cache_list_lock);
	cd->nextcheck = 0;
	cd->entries = 0;
	atomic_set(&cd->readers, 0);
	cd->last_close = 0;
	cd->last_warn = -1;
	list_add(&cd->others, &cache_list);
	spin_unlock(&cache_list_lock);

	/* start the cleaning process */
	schedule_delayed_work(&cache_cleaner, 0);
}

int cache_unregister(struct cache_detail *cd)
{
	cache_purge(cd);
	spin_lock(&cache_list_lock);
	write_lock(&cd->hash_lock);
	if (cd->entries || atomic_read(&cd->inuse)) {
		write_unlock(&cd->hash_lock);
		spin_unlock(&cache_list_lock);
		return -EBUSY;
	}
	if (current_detail == cd)
		current_detail = NULL;
	list_del_init(&cd->others);
	write_unlock(&cd->hash_lock);
	spin_unlock(&cache_list_lock);
	if (cd->proc_ent) {
		if (cd->flush_ent)
			remove_proc_entry("flush", cd->proc_ent);
		if (cd->channel_ent)
			remove_proc_entry("channel", cd->proc_ent);
		if (cd->content_ent)
			remove_proc_entry("content", cd->proc_ent);

		cd->proc_ent = NULL;
		remove_proc_entry(cd->name, proc_net_rpc);
	}
	if (list_empty(&cache_list)) {
		/* module must be being unloaded so its safe to kill the worker */
		cancel_delayed_work_sync(&cache_cleaner);
	}
	return 0;
}

/* clean cache tries to find something to clean
 * and cleans it.
 * It returns 1 if it cleaned something,
 *            0 if it didn't find anything this time
 *           -1 if it fell off the end of the list.
 */
static int cache_clean(void)
{
	int rv = 0;
	struct list_head *next;

	spin_lock(&cache_list_lock);

	/* find a suitable table if we don't already have one */
	while (current_detail == NULL ||
	    current_index >= current_detail->hash_size) {
		if (current_detail)
			next = current_detail->others.next;
		else
			next = cache_list.next;
		if (next == &cache_list) {
			current_detail = NULL;
			spin_unlock(&cache_list_lock);
			return -1;
		}
		current_detail = list_entry(next, struct cache_detail, others);
		if (current_detail->nextcheck > get_seconds())
			current_index = current_detail->hash_size;
		else {
			current_index = 0;
			current_detail->nextcheck = get_seconds()+30*60;
		}
	}

	/* find a non-empty bucket in the table */
	while (current_detail &&
	       current_index < current_detail->hash_size &&
	       current_detail->hash_table[current_index] == NULL)
		current_index++;

	/* find a cleanable entry in the bucket and clean it, or set to next bucket */

	if (current_detail && current_index < current_detail->hash_size) {
		struct cache_head *ch, **cp;
		struct cache_detail *d;

		write_lock(&current_detail->hash_lock);

		/* Ok, now to clean this strand */

		cp = & current_detail->hash_table[current_index];
		ch = *cp;
		for (; ch; cp= & ch->next, ch= *cp) {
			if (current_detail->nextcheck > ch->expiry_time)
				current_detail->nextcheck = ch->expiry_time+1;
			if (ch->expiry_time >= get_seconds()
			    && ch->last_refresh >= current_detail->flush_time
				)
				continue;
			if (test_and_clear_bit(CACHE_PENDING, &ch->flags))
				queue_loose(current_detail, ch);

			if (atomic_read(&ch->ref.refcount) == 1)
				break;
		}
		if (ch) {
			*cp = ch->next;
			ch->next = NULL;
			current_detail->entries--;
			rv = 1;
		}
		write_unlock(&current_detail->hash_lock);
		d = current_detail;
		if (!ch)
			current_index ++;
		spin_unlock(&cache_list_lock);
		if (ch)
			cache_put(ch, d);
	} else
		spin_unlock(&cache_list_lock);

	return rv;
}

/*
 * We want to regularly clean the cache, so we need to schedule some work ...
 */
static void do_cache_clean(struct work_struct *work)
{
	int delay = 5;
	if (cache_clean() == -1)
		delay = 30*HZ;

	if (list_empty(&cache_list))
		delay = 0;

	if (delay)
		schedule_delayed_work(&cache_cleaner, delay);
}


/*
 * Clean all caches promptly.  This just calls cache_clean
 * repeatedly until we are sure that every cache has had a chance to
 * be fully cleaned
 */
void cache_flush(void)
{
	while (cache_clean() != -1)
		cond_resched();
	while (cache_clean() != -1)
		cond_resched();
}

void cache_purge(struct cache_detail *detail)
{
	detail->flush_time = LONG_MAX;
	detail->nextcheck = get_seconds();
	cache_flush();
	detail->flush_time = 1;
}



/*
 * Deferral and Revisiting of Requests.
 *
 * If a cache lookup finds a pending entry, we
 * need to defer the request and revisit it later.
 * All deferred requests are stored in a hash table,
 * indexed by "struct cache_head *".
 * As it may be wasteful to store a whole request
 * structure, we allow the request to provide a
 * deferred form, which must contain a
 * 'struct cache_deferred_req'
 * This cache_deferred_req contains a method to allow
 * it to be revisited when cache info is available
 */

#define	DFR_HASHSIZE	(PAGE_SIZE/sizeof(struct list_head))
#define	DFR_HASH(item)	((((long)item)>>4 ^ (((long)item)>>13)) % DFR_HASHSIZE)

#define	DFR_MAX	300	/* ??? */

static DEFINE_SPINLOCK(cache_defer_lock);
static LIST_HEAD(cache_defer_list);
static struct list_head cache_defer_hash[DFR_HASHSIZE];
static int cache_defer_cnt;

static int cache_defer_req(struct cache_req *req, struct cache_head *item)
{
	struct cache_deferred_req *dreq;
	int hash = DFR_HASH(item);

	if (cache_defer_cnt >= DFR_MAX) {
		/* too much in the cache, randomly drop this one,
		 * or continue and drop the oldest below
		 */
		if (net_random()&1)
			return -ETIMEDOUT;
	}
	dreq = req->defer(req);
	if (dreq == NULL)
		return -ETIMEDOUT;

	dreq->item = item;
	dreq->recv_time = get_seconds();

	spin_lock(&cache_defer_lock);

	list_add(&dreq->recent, &cache_defer_list);

	if (cache_defer_hash[hash].next == NULL)
		INIT_LIST_HEAD(&cache_defer_hash[hash]);
	list_add(&dreq->hash, &cache_defer_hash[hash]);

	/* it is in, now maybe clean up */
	dreq = NULL;
	if (++cache_defer_cnt > DFR_MAX) {
		dreq = list_entry(cache_defer_list.prev,
				  struct cache_deferred_req, recent);
		list_del(&dreq->recent);
		list_del(&dreq->hash);
		cache_defer_cnt--;
	}
	spin_unlock(&cache_defer_lock);

	if (dreq) {
		/* there was one too many */
		dreq->revisit(dreq, 1);
	}
	if (!test_bit(CACHE_PENDING, &item->flags)) {
		/* must have just been validated... */
		cache_revisit_request(item);
	}
	return 0;
}

static void cache_revisit_request(struct cache_head *item)
{
	struct cache_deferred_req *dreq;
	struct list_head pending;

	struct list_head *lp;
	int hash = DFR_HASH(item);

	INIT_LIST_HEAD(&pending);
	spin_lock(&cache_defer_lock);

	lp = cache_defer_hash[hash].next;
	if (lp) {
		while (lp != &cache_defer_hash[hash]) {
			dreq = list_entry(lp, struct cache_deferred_req, hash);
			lp = lp->next;
			if (dreq->item == item) {
				list_del(&dreq->hash);
				list_move(&dreq->recent, &pending);
				cache_defer_cnt--;
			}
		}
	}
	spin_unlock(&cache_defer_lock);

	while (!list_empty(&pending)) {
		dreq = list_entry(pending.next, struct cache_deferred_req, recent);
		list_del_init(&dreq->recent);
		dreq->revisit(dreq, 0);
	}
}

void cache_clean_deferred(void *owner)
{
	struct cache_deferred_req *dreq, *tmp;
	struct list_head pending;


	INIT_LIST_HEAD(&pending);
	spin_lock(&cache_defer_lock);

	list_for_each_entry_safe(dreq, tmp, &cache_defer_list, recent) {
		if (dreq->owner == owner) {
			list_del(&dreq->hash);
			list_move(&dreq->recent, &pending);
			cache_defer_cnt--;
		}
	}
	spin_unlock(&cache_defer_lock);

	while (!list_empty(&pending)) {
		dreq = list_entry(pending.next, struct cache_deferred_req, recent);
		list_del_init(&dreq->recent);
		dreq->revisit(dreq, 1);
	}
}

/*
 * communicate with user-space
 *
 * We have a magic /proc file - /proc/sunrpc/cache
 * On read, you get a full request, or block
 * On write, an update request is processed
 * Poll works if anything to read, and always allows write
 *
 * Implemented by linked list of requests.  Each open file has
 * a ->private that also exists in this list.  New request are added
 * to the end and may wakeup and preceding readers.
 * New readers are added to the head.  If, on read, an item is found with