[deliverable/linux.git] / net / sched / cls_tcindex.c

/*
 * net/sched/cls_tcindex.c	Packet classifier for skb->tc_index
 *
 * Written 1998,1999 by Werner Almesberger, EPFL ICA
 */

#include <linux/module.h>
#include <linux/types.h>
#include <linux/kernel.h>
#include <linux/skbuff.h>
#include <linux/errno.h>
#include <linux/slab.h>
#include <net/act_api.h>
#include <net/netlink.h>
#include <net/pkt_cls.h>

/*
 * Passing parameters to the root seems to be done more awkwardly than really
 * necessary. At least, u32 doesn't seem to use such dirty hacks. To be
 * verified. FIXME.
 */

#define PERFECT_HASH_THRESHOLD	64	/* use perfect hash if not bigger */
#define DEFAULT_HASH_SIZE	64	/* optimized for diffserv */


#define	PRIV(tp)	((struct tcindex_data *) (tp)->root)


struct tcindex_filter_result {
	struct tcf_exts		exts;
	struct tcf_result	res;
};

struct tcindex_filter {
	u16 key;
	struct tcindex_filter_result result;
	struct tcindex_filter *next;
};


struct tcindex_data {
	struct tcindex_filter_result *perfect; /* perfect hash; NULL if none */
	struct tcindex_filter **h; /* imperfect hash; only used if !perfect;
				      NULL if unused */
	u16 mask;		/* AND key with mask */
	int shift;		/* shift ANDed key to the right */
	int hash;		/* hash table size; 0 if undefined */
	int alloc_hash;		/* allocated size */
	int fall_through;	/* 0: only classify if explicit match */
};

static const struct tcf_ext_map tcindex_ext_map = {
	.police = TCA_TCINDEX_POLICE,
	.action = TCA_TCINDEX_ACT
};

static inline int
tcindex_filter_is_set(struct tcindex_filter_result *r)
{
	return tcf_exts_is_predicative(&r->exts) || r->res.classid;
}

static struct tcindex_filter_result *
tcindex_lookup(struct tcindex_data *p, u16 key)
{
	struct tcindex_filter *f;

	if (p->perfect)
		return tcindex_filter_is_set(p->perfect + key) ?
			p->perfect + key : NULL;
	else if (p->h) {
		for (f = p->h[key % p->hash]; f; f = f->next)
			if (f->key == key)
				return &f->result;
	}

	return NULL;
}


static int tcindex_classify(struct sk_buff *skb, const struct tcf_proto *tp,
			    struct tcf_result *res)
{
	struct tcindex_data *p = PRIV(tp);
	struct tcindex_filter_result *f;
	int key = (skb->tc_index & p->mask) >> p->shift;

	pr_debug("tcindex_classify(skb %p,tp %p,res %p),p %p\n",
		 skb, tp, res, p);

	f = tcindex_lookup(p, key);
	if (!f) {
		if (!p->fall_through)
			return -1;
		res->classid = TC_H_MAKE(TC_H_MAJ(tp->q->handle), key);
		res->class = 0;
		pr_debug("alg 0x%x\n", res->classid);
		return 0;
	}
	*res = f->res;
	pr_debug("map 0x%x\n", res->classid);

	return tcf_exts_exec(skb, &f->exts, res);
}


static unsigned long tcindex_get(struct tcf_proto *tp, u32 handle)
{
	struct tcindex_data *p = PRIV(tp);
	struct tcindex_filter_result *r;

	pr_debug("tcindex_get(tp %p,handle 0x%08x)\n", tp, handle);
	if (p->perfect && handle >= p->alloc_hash)
		return 0;
	r = tcindex_lookup(p, handle);
	return r && tcindex_filter_is_set(r) ? (unsigned long) r : 0UL;
}


static void tcindex_put(struct tcf_proto *tp, unsigned long f)
{
	pr_debug("tcindex_put(tp %p,f 0x%lx)\n", tp, f);
}


static int tcindex_init(struct tcf_proto *tp)
{
	struct tcindex_data *p;

	pr_debug("tcindex_init(tp %p)\n", tp);
	p = kzalloc(sizeof(struct tcindex_data), GFP_KERNEL);
	if (!p)
		return -ENOMEM;

	p->mask = 0xffff;
	p->hash = DEFAULT_HASH_SIZE;
	p->fall_through = 1;

	tp->root = p;
	return 0;
}


static int
__tcindex_delete(struct tcf_proto *tp, unsigned long arg, int lock)
{
	struct tcindex_data *p = PRIV(tp);
	struct tcindex_filter_result *r = (struct tcindex_filter_result *) arg;
	struct tcindex_filter *f = NULL;

	pr_debug("tcindex_delete(tp %p,arg 0x%lx),p %p,f %p\n", tp, arg, p, f);
	if (p->perfect) {
		if (!r->res.class)
			return -ENOENT;
	} else {
		int i;
		struct tcindex_filter **walk = NULL;

		for (i = 0; i < p->hash; i++)
			for (walk = p->h+i; *walk; walk = &(*walk)->next)
				if (&(*walk)->result == r)
					goto found;
		return -ENOENT;

found:
		f = *walk;
		if (lock)
			tcf_tree_lock(tp);
		*walk = f->next;
		if (lock)
			tcf_tree_unlock(tp);
	}
	tcf_unbind_filter(tp, &r->res);
	tcf_exts_destroy(tp, &r->exts);
	kfree(f);
	return 0;
}

static int tcindex_delete(struct tcf_proto *tp, unsigned long arg)
{
	return __tcindex_delete(tp, arg, 1);
}

static inline int
valid_perfect_hash(struct tcindex_data *p)
{
	return  p->hash > (p->mask >> p->shift);
}

static const struct nla_policy tcindex_policy[TCA_TCINDEX_MAX + 1] = {
	[TCA_TCINDEX_HASH]		= { .type = NLA_U32 },
	[TCA_TCINDEX_MASK]		= { .type = NLA_U16 },
	[TCA_TCINDEX_SHIFT]		= { .type = NLA_U32 },
	[TCA_TCINDEX_FALL_THROUGH]	= { .type = NLA_U32 },
	[TCA_TCINDEX_CLASSID]		= { .type = NLA_U32 },
};

static int
tcindex_set_parms(struct net *net, struct tcf_proto *tp, unsigned long base,
		  u32 handle, struct tcindex_data *p,
		  struct tcindex_filter_result *r, struct nlattr **tb,
		 struct nlattr *est)
{
	int err, balloc = 0;
	struct tcindex_filter_result new_filter_result, *old_r = r;
	struct tcindex_filter_result cr;
	struct tcindex_data cp;
	struct tcindex_filter *f = NULL; /* make gcc behave */
	struct tcf_exts e;

	err = tcf_exts_validate(net, tp, tb, est, &e, &tcindex_ext_map);
	if (err < 0)
		return err;

	memcpy(&cp, p, sizeof(cp));
	memset(&new_filter_result, 0, sizeof(new_filter_result));

	if (old_r)
		memcpy(&cr, r, sizeof(cr));
	else
		memset(&cr, 0, sizeof(cr));

	if (tb[TCA_TCINDEX_HASH])
		cp.hash = nla_get_u32(tb[TCA_TCINDEX_HASH]);

	if (tb[TCA_TCINDEX_MASK])
		cp.mask = nla_get_u16(tb[TCA_TCINDEX_MASK]);

	if (tb[TCA_TCINDEX_SHIFT])
		cp.shift = nla_get_u32(tb[TCA_TCINDEX_SHIFT]);

	err = -EBUSY;
	/* Hash already allocated, make sure that we still meet the
	 * requirements for the allocated hash.
	 */
	if (cp.perfect) {
		if (!valid_perfect_hash(&cp) ||
		    cp.hash > cp.alloc_hash)
			goto errout;
	} else if (cp.h && cp.hash != cp.alloc_hash)
		goto errout;

	err = -EINVAL;
	if (tb[TCA_TCINDEX_FALL_THROUGH])
		cp.fall_through = nla_get_u32(tb[TCA_TCINDEX_FALL_THROUGH]);

	if (!cp.hash) {
		/* Hash not specified, use perfect hash if the upper limit
		 * of the hashing index is below the threshold.
		 */
		if ((cp.mask >> cp.shift) < PERFECT_HASH_THRESHOLD)
			cp.hash = (cp.mask >> cp.shift) + 1;
		else
			cp.hash = DEFAULT_HASH_SIZE;
	}

	if (!cp.perfect && !cp.h)
		cp.alloc_hash = cp.hash;

	/* Note: this could be as restrictive as if (handle & ~(mask >> shift))
	 * but then, we'd fail handles that may become valid after some future
	 * mask change. While this is extremely unlikely to ever matter,
	 * the check below is safer (and also more backwards-compatible).
	 */
	if (cp.perfect || valid_perfect_hash(&cp))
		if (handle >= cp.alloc_hash)
			goto errout;


	err = -ENOMEM;
	if (!cp.perfect && !cp.h) {
		if (valid_perfect_hash(&cp)) {
			cp.perfect = kcalloc(cp.hash, sizeof(*r), GFP_KERNEL);
			if (!cp.perfect)
				goto errout;
			balloc = 1;
		} else {
			cp.h = kcalloc(cp.hash, sizeof(f), GFP_KERNEL);
			if (!cp.h)
				goto errout;
			balloc = 2;
		}
	}

	if (cp.perfect)
		r = cp.perfect + handle;
	else
		r = tcindex_lookup(&cp, handle) ? : &new_filter_result;

	if (r == &new_filter_result) {
		f = kzalloc(sizeof(*f), GFP_KERNEL);
		if (!f)
			goto errout_alloc;
	}

	if (tb[TCA_TCINDEX_CLASSID]) {
		cr.res.classid = nla_get_u32(tb[TCA_TCINDEX_CLASSID]);
		tcf_bind_filter(tp, &cr.res, base);
	}

	tcf_exts_change(tp, &cr.exts, &e);

	tcf_tree_lock(tp);
	if (old_r && old_r != r)
		memset(old_r, 0, sizeof(*old_r));

	memcpy(p, &cp, sizeof(cp));
	memcpy(r, &cr, sizeof(cr));

	if (r == &new_filter_result) {
		struct tcindex_filter **fp;

		f->key = handle;
		f->result = new_filter_result;
		f->next = NULL;
		for (fp = p->h+(handle % p->hash); *fp; fp = &(*fp)->next)
			/* nothing */;
		*fp = f;
	}
	tcf_tree_unlock(tp);

	return 0;

errout_alloc:
	if (balloc == 1)
		kfree(cp.perfect);
	else if (balloc == 2)
		kfree(cp.h);
errout:
	tcf_exts_destroy(tp, &e);
	return err;
}

static int
tcindex_change(struct net *net, struct sk_buff *in_skb,
	       struct tcf_proto *tp, unsigned long base, u32 handle,
	       struct nlattr **tca, unsigned long *arg)
{
	struct nlattr *opt = tca[TCA_OPTIONS];
	struct nlattr *tb[TCA_TCINDEX_MAX + 1];
	struct tcindex_data *p = PRIV(tp);
	struct tcindex_filter_result *r = (struct tcindex_filter_result *) *arg;
	int err;

	pr_debug("tcindex_change(tp %p,handle 0x%08x,tca %p,arg %p),opt %p,"
	    "p %p,r %p,*arg 0x%lx\n",
	    tp, handle, tca, arg, opt, p, r, arg ? *arg : 0L);

	if (!opt)
		return 0;

	err = nla_parse_nested(tb, TCA_TCINDEX_MAX, opt, tcindex_policy);
	if (err < 0)
		return err;

	return tcindex_set_parms(net, tp, base, handle, p, r, tb,
				 tca[TCA_RATE]);
}


static void tcindex_walk(struct tcf_proto *tp, struct tcf_walker *walker)
{
	struct tcindex_data *p = PRIV(tp);
	struct tcindex_filter *f, *next;
	int i;

	pr_debug("tcindex_walk(tp %p,walker %p),p %p\n", tp, walker, p);
	if (p->perfect) {
		for (i = 0; i < p->hash; i++) {
			if (!p->perfect[i].res.class)
				continue;
			if (walker->count >= walker->skip) {
				if (walker->fn(tp,
				    (unsigned long) (p->perfect+i), walker)
				     < 0) {
					walker->stop = 1;
					return;
				}
			}
			walker->count++;
		}
	}
	if (!p->h)
		return;
	for (i = 0; i < p->hash; i++) {
		for (f = p->h[i]; f; f = next) {
			next = f->next;
			if (walker->count >= walker->skip) {
				if (walker->fn(tp, (unsigned long) &f->result,
				    walker) < 0) {
					walker->stop = 1;
					return;
				}
			}
			walker->count++;
		}
	}
}


static int tcindex_destroy_element(struct tcf_proto *tp,
    unsigned long arg, struct tcf_walker *walker)
{
	return __tcindex_delete(tp, arg, 0);
}


static void tcindex_destroy(struct tcf_proto *tp)
{
	struct tcindex_data *p = PRIV(tp);
	struct tcf_walker walker;

	pr_debug("tcindex_destroy(tp %p),p %p\n", tp, p);
	walker.count = 0;
	walker.skip = 0;
	walker.fn = &tcindex_destroy_element;
	tcindex_walk(tp, &walker);
	kfree(p->perfect);
	kfree(p->h);
	kfree(p);
	tp->root = NULL;
}


static int tcindex_dump(struct tcf_proto *tp, unsigned long fh,
    struct sk_buff *skb, struct tcmsg *t)
{
	struct tcindex_data *p = PRIV(tp);
	struct tcindex_filter_result *r = (struct tcindex_filter_result *) fh;
	unsigned char *b = skb_tail_pointer(skb);
	struct nlattr *nest;

	pr_debug("tcindex_dump(tp %p,fh 0x%lx,skb %p,t %p),p %p,r %p,b %p\n",
		 tp, fh, skb, t, p, r, b);
	pr_debug("p->perfect %p p->h %p\n", p->perfect, p->h);

	nest = nla_nest_start(skb, TCA_OPTIONS);
	if (nest == NULL)
		goto nla_put_failure;

	if (!fh) {
		t->tcm_handle = ~0; /* whatever ... */
		if (nla_put_u32(skb, TCA_TCINDEX_HASH, p->hash) ||
		    nla_put_u16(skb, TCA_TCINDEX_MASK, p->mask) ||
		    nla_put_u32(skb, TCA_TCINDEX_SHIFT, p->shift) ||
		    nla_put_u32(skb, TCA_TCINDEX_FALL_THROUGH, p->fall_through))
			goto nla_put_failure;
		nla_nest_end(skb, nest);
	} else {
		if (p->perfect) {
			t->tcm_handle = r-p->perfect;
		} else {
			struct tcindex_filter *f;
			int i;

			t->tcm_handle = 0;
			for (i = 0; !t->tcm_handle && i < p->hash; i++) {
				for (f = p->h[i]; !t->tcm_handle && f;
				     f = f->next) {
					if (&f->result == r)
						t->tcm_handle = f->key;
				}
			}
		}
		pr_debug("handle = %d\n", t->tcm_handle);
		if (r->res.class &&
		    nla_put_u32(skb, TCA_TCINDEX_CLASSID, r->res.classid))
			goto nla_put_failure;

		if (tcf_exts_dump(skb, &r->exts, &tcindex_ext_map) < 0)
			goto nla_put_failure;
		nla_nest_end(skb, nest);

		if (tcf_exts_dump_stats(skb, &r->exts, &tcindex_ext_map) < 0)
			goto nla_put_failure;
	}

	return skb->len;

nla_put_failure:
	nlmsg_trim(skb, b);
	return -1;
}

static struct tcf_proto_ops cls_tcindex_ops __read_mostly = {
	.kind		=	"tcindex",
	.classify	=	tcindex_classify,
	.init		=	tcindex_init,
	.destroy	=	tcindex_destroy,
	.get		=	tcindex_get,
	.put		=	tcindex_put,
	.change		=	tcindex_change,
	.delete		=	tcindex_delete,
	.walk		=	tcindex_walk,
	.dump		=	tcindex_dump,
	.owner		=	THIS_MODULE,
};

static int __init init_tcindex(void)
{
	return register_tcf_proto_ops(&cls_tcindex_ops);
}

static void __exit exit_tcindex(void)
{
	unregister_tcf_proto_ops(&cls_tcindex_ops);
}

module_init(init_tcindex)
module_exit(exit_tcindex)
MODULE_LICENSE("GPL");
Commit	Line	Data
1da177e4 LT	1	/*
	2	* net/sched/cls_tcindex.c Packet classifier for skb->tc_index
	3	*
	4	* Written 1998,1999 by Werner Almesberger, EPFL ICA
	5	*/
	6
1da177e4 LT	7	#include <linux/module.h>
	8	#include <linux/types.h>
	9	#include <linux/kernel.h>
	10	#include <linux/skbuff.h>
	11	#include <linux/errno.h>
5a0e3ad6	12	#include <linux/slab.h>
1da177e4	13	#include <net/act_api.h>
dc5fc579	14	#include <net/netlink.h>
1da177e4	15	#include <net/pkt_cls.h>
1da177e4	16
1da177e4 LT	17	/*
	18	* Passing parameters to the root seems to be done more awkwardly than really
	19	* necessary. At least, u32 doesn't seem to use such dirty hacks. To be
	20	* verified. FIXME.
	21	*/
	22
	23	#define PERFECT_HASH_THRESHOLD 64 /* use perfect hash if not bigger */
	24	#define DEFAULT_HASH_SIZE 64 /* optimized for diffserv */
	25
	26
1da177e4 LT	27	#define PRIV(tp) ((struct tcindex_data *) (tp)->root)
	28
	29
	30	struct tcindex_filter_result {
	31	struct tcf_exts exts;
	32	struct tcf_result res;
	33	};
	34
	35	struct tcindex_filter {
	36	u16 key;
	37	struct tcindex_filter_result result;
	38	struct tcindex_filter *next;
	39	};
	40
	41
	42	struct tcindex_data {
	43	struct tcindex_filter_result perfect; / perfect hash; NULL if none */
	44	struct tcindex_filter *h; / imperfect hash; only used if !perfect;
	45	NULL if unused */
	46	u16 mask; /* AND key with mask */
	47	int shift; /* shift ANDed key to the right */
	48	int hash; /* hash table size; 0 if undefined */
	49	int alloc_hash; /* allocated size */
	50	int fall_through; /* 0: only classify if explicit match */
	51	};
	52
5239008b	53	static const struct tcf_ext_map tcindex_ext_map = {
1da177e4 LT	54	.police = TCA_TCINDEX_POLICE,
	55	.action = TCA_TCINDEX_ACT
	56	};
	57
	58	static inline int
	59	tcindex_filter_is_set(struct tcindex_filter_result *r)
	60	{
	61	return tcf_exts_is_predicative(&r->exts) \|\| r->res.classid;
	62	}
	63
	64	static struct tcindex_filter_result *
	65	tcindex_lookup(struct tcindex_data *p, u16 key)
	66	{
	67	struct tcindex_filter *f;
	68
	69	if (p->perfect)
	70	return tcindex_filter_is_set(p->perfect + key) ?
	71	p->perfect + key : NULL;
	72	else if (p->h) {
	73	for (f = p->h[key % p->hash]; f; f = f->next)
	74	if (f->key == key)
	75	return &f->result;
	76	}
	77
	78	return NULL;
	79	}
	80
	81
dc7f9f6e	82	static int tcindex_classify(struct sk_buff skb, const struct tcf_proto tp,
1da177e4 LT	83	struct tcf_result *res)
	84	{
	85	struct tcindex_data *p = PRIV(tp);
	86	struct tcindex_filter_result *f;
	87	int key = (skb->tc_index & p->mask) >> p->shift;
	88
aa767bfe SH	89	pr_debug("tcindex_classify(skb %p,tp %p,res %p),p %p\n",
aa767bfe SH	90	skb, tp, res, p);
1da177e4 LT	91
	92	f = tcindex_lookup(p, key);
	93	if (!f) {
	94	if (!p->fall_through)
	95	return -1;
	96	res->classid = TC_H_MAKE(TC_H_MAJ(tp->q->handle), key);
	97	res->class = 0;
aa767bfe	98	pr_debug("alg 0x%x\n", res->classid);
1da177e4 LT	99	return 0;
	100	}
	101	*res = f->res;
aa767bfe	102	pr_debug("map 0x%x\n", res->classid);
1da177e4 LT	103
	104	return tcf_exts_exec(skb, &f->exts, res);
	105	}
	106
	107
	108	static unsigned long tcindex_get(struct tcf_proto *tp, u32 handle)
	109	{
	110	struct tcindex_data *p = PRIV(tp);
	111	struct tcindex_filter_result *r;
	112
aa767bfe	113	pr_debug("tcindex_get(tp %p,handle 0x%08x)\n", tp, handle);
1da177e4 LT	114	if (p->perfect && handle >= p->alloc_hash)
	115	return 0;
	116	r = tcindex_lookup(p, handle);
	117	return r && tcindex_filter_is_set(r) ? (unsigned long) r : 0UL;
	118	}
	119
	120
	121	static void tcindex_put(struct tcf_proto *tp, unsigned long f)
	122	{
aa767bfe	123	pr_debug("tcindex_put(tp %p,f 0x%lx)\n", tp, f);
1da177e4 LT	124	}
	125
	126
	127	static int tcindex_init(struct tcf_proto *tp)
	128	{
	129	struct tcindex_data *p;
	130
aa767bfe SH	131	pr_debug("tcindex_init(tp %p)\n", tp);
aa767bfe SH	132	p = kzalloc(sizeof(struct tcindex_data), GFP_KERNEL);
1da177e4 LT	133	if (!p)
	134	return -ENOMEM;
	135
1da177e4 LT	136	p->mask = 0xffff;
	137	p->hash = DEFAULT_HASH_SIZE;
	138	p->fall_through = 1;
	139
	140	tp->root = p;
	141	return 0;
	142	}
	143
	144
	145	static int
	146	__tcindex_delete(struct tcf_proto *tp, unsigned long arg, int lock)
	147	{
	148	struct tcindex_data *p = PRIV(tp);
	149	struct tcindex_filter_result r = (struct tcindex_filter_result ) arg;
	150	struct tcindex_filter *f = NULL;
	151
aa767bfe	152	pr_debug("tcindex_delete(tp %p,arg 0x%lx),p %p,f %p\n", tp, arg, p, f);
1da177e4 LT	153	if (p->perfect) {
	154	if (!r->res.class)
	155	return -ENOENT;
	156	} else {
	157	int i;
	158	struct tcindex_filter **walk = NULL;
	159
	160	for (i = 0; i < p->hash; i++)
	161	for (walk = p->h+i; walk; walk = &(walk)->next)
	162	if (&(*walk)->result == r)
	163	goto found;
	164	return -ENOENT;
	165
	166	found:
	167	f = *walk;
	168	if (lock)
	169	tcf_tree_lock(tp);
	170	*walk = f->next;
	171	if (lock)
	172	tcf_tree_unlock(tp);
	173	}
	174	tcf_unbind_filter(tp, &r->res);
	175	tcf_exts_destroy(tp, &r->exts);
a51482bd	176	kfree(f);
1da177e4 LT	177	return 0;
	178	}
	179
	180	static int tcindex_delete(struct tcf_proto *tp, unsigned long arg)
	181	{
	182	return __tcindex_delete(tp, arg, 1);
	183	}
	184
	185	static inline int
	186	valid_perfect_hash(struct tcindex_data *p)
	187	{
	188	return p->hash > (p->mask >> p->shift);
	189	}
	190
6fa8c014 PM	191	static const struct nla_policy tcindex_policy[TCA_TCINDEX_MAX + 1] = {
	192	[TCA_TCINDEX_HASH] = { .type = NLA_U32 },
	193	[TCA_TCINDEX_MASK] = { .type = NLA_U16 },
	194	[TCA_TCINDEX_SHIFT] = { .type = NLA_U32 },
	195	[TCA_TCINDEX_FALL_THROUGH] = { .type = NLA_U32 },
	196	[TCA_TCINDEX_CLASSID] = { .type = NLA_U32 },
	197	};
	198
1da177e4	199	static int
c1b52739 BL	200	tcindex_set_parms(struct net net, struct tcf_proto tp, unsigned long base,
	201	u32 handle, struct tcindex_data *p,
	202	struct tcindex_filter_result r, struct nlattr *tb,
	203	struct nlattr *est)
1da177e4 LT	204	{
	205	int err, balloc = 0;
	206	struct tcindex_filter_result new_filter_result, *old_r = r;
	207	struct tcindex_filter_result cr;
	208	struct tcindex_data cp;
	209	struct tcindex_filter f = NULL; / make gcc behave */
	210	struct tcf_exts e;
	211
c1b52739	212	err = tcf_exts_validate(net, tp, tb, est, &e, &tcindex_ext_map);
1da177e4 LT	213	if (err < 0)
1da177e4 LT	214	return err;
10297b99	215
1da177e4 LT	216	memcpy(&cp, p, sizeof(cp));
	217	memset(&new_filter_result, 0, sizeof(new_filter_result));
	218
	219	if (old_r)
	220	memcpy(&cr, r, sizeof(cr));
	221	else
	222	memset(&cr, 0, sizeof(cr));
	223
6fa8c014	224	if (tb[TCA_TCINDEX_HASH])
1587bac4	225	cp.hash = nla_get_u32(tb[TCA_TCINDEX_HASH]);
1da177e4	226
6fa8c014	227	if (tb[TCA_TCINDEX_MASK])
1587bac4	228	cp.mask = nla_get_u16(tb[TCA_TCINDEX_MASK]);
1da177e4	229
6fa8c014	230	if (tb[TCA_TCINDEX_SHIFT])
1587bac4	231	cp.shift = nla_get_u32(tb[TCA_TCINDEX_SHIFT]);
1da177e4 LT	232
	233	err = -EBUSY;
	234	/* Hash already allocated, make sure that we still meet the
	235	* requirements for the allocated hash.
	236	*/
	237	if (cp.perfect) {
	238	if (!valid_perfect_hash(&cp) \|\|
	239	cp.hash > cp.alloc_hash)
	240	goto errout;
	241	} else if (cp.h && cp.hash != cp.alloc_hash)
	242	goto errout;
	243
	244	err = -EINVAL;
6fa8c014	245	if (tb[TCA_TCINDEX_FALL_THROUGH])
1587bac4	246	cp.fall_through = nla_get_u32(tb[TCA_TCINDEX_FALL_THROUGH]);
1da177e4 LT	247
	248	if (!cp.hash) {
	249	/* Hash not specified, use perfect hash if the upper limit
	250	* of the hashing index is below the threshold.
	251	*/
	252	if ((cp.mask >> cp.shift) < PERFECT_HASH_THRESHOLD)
cc7ec456	253	cp.hash = (cp.mask >> cp.shift) + 1;
1da177e4 LT	254	else
	255	cp.hash = DEFAULT_HASH_SIZE;
	256	}
	257
	258	if (!cp.perfect && !cp.h)
	259	cp.alloc_hash = cp.hash;
	260
	261	/* Note: this could be as restrictive as if (handle & ~(mask >> shift))
	262	* but then, we'd fail handles that may become valid after some future
	263	* mask change. While this is extremely unlikely to ever matter,
	264	* the check below is safer (and also more backwards-compatible).
	265	*/
	266	if (cp.perfect \|\| valid_perfect_hash(&cp))
	267	if (handle >= cp.alloc_hash)
	268	goto errout;
	269
	270
	271	err = -ENOMEM;
	272	if (!cp.perfect && !cp.h) {
	273	if (valid_perfect_hash(&cp)) {
0da974f4	274	cp.perfect = kcalloc(cp.hash, sizeof(*r), GFP_KERNEL);
1da177e4 LT	275	if (!cp.perfect)
1da177e4 LT	276	goto errout;
1da177e4 LT	277	balloc = 1;
1da177e4 LT	278	} else {
0da974f4	279	cp.h = kcalloc(cp.hash, sizeof(f), GFP_KERNEL);
1da177e4 LT	280	if (!cp.h)
1da177e4 LT	281	goto errout;
1da177e4 LT	282	balloc = 2;
	283	}
	284	}
	285
	286	if (cp.perfect)
	287	r = cp.perfect + handle;
	288	else
	289	r = tcindex_lookup(&cp, handle) ? : &new_filter_result;
	290
	291	if (r == &new_filter_result) {
0da974f4	292	f = kzalloc(sizeof(*f), GFP_KERNEL);
1da177e4 LT	293	if (!f)
1da177e4 LT	294	goto errout_alloc;
10297b99	295	}
1da177e4	296
add93b61	297	if (tb[TCA_TCINDEX_CLASSID]) {
1587bac4	298	cr.res.classid = nla_get_u32(tb[TCA_TCINDEX_CLASSID]);
1da177e4	299	tcf_bind_filter(tp, &cr.res, base);
10297b99	300	}
1da177e4 LT	301
	302	tcf_exts_change(tp, &cr.exts, &e);
	303
	304	tcf_tree_lock(tp);
	305	if (old_r && old_r != r)
	306	memset(old_r, 0, sizeof(*old_r));
	307
	308	memcpy(p, &cp, sizeof(cp));
	309	memcpy(r, &cr, sizeof(cr));
	310
	311	if (r == &new_filter_result) {
	312	struct tcindex_filter **fp;
	313
	314	f->key = handle;
	315	f->result = new_filter_result;
	316	f->next = NULL;
	317	for (fp = p->h+(handle % p->hash); fp; fp = &(fp)->next)
	318	/* nothing */;
	319	*fp = f;
10297b99	320	}
1da177e4 LT	321	tcf_tree_unlock(tp);
	322
	323	return 0;
	324
	325	errout_alloc:
	326	if (balloc == 1)
	327	kfree(cp.perfect);
	328	else if (balloc == 2)
	329	kfree(cp.h);
	330	errout:
	331	tcf_exts_destroy(tp, &e);
	332	return err;
	333	}
	334
	335	static int
c1b52739	336	tcindex_change(struct net net, struct sk_buff in_skb,
af4c6641	337	struct tcf_proto *tp, unsigned long base, u32 handle,
add93b61	338	struct nlattr *tca, unsigned long arg)
1da177e4	339	{
add93b61 PM	340	struct nlattr *opt = tca[TCA_OPTIONS];
add93b61 PM	341	struct nlattr *tb[TCA_TCINDEX_MAX + 1];
1da177e4 LT	342	struct tcindex_data *p = PRIV(tp);
1da177e4 LT	343	struct tcindex_filter_result r = (struct tcindex_filter_result ) *arg;
cee63723	344	int err;
1da177e4	345
aa767bfe	346	pr_debug("tcindex_change(tp %p,handle 0x%08x,tca %p,arg %p),opt %p,"
1da177e4 LT	347	"p %p,r %p,*arg 0x%lx\n",
	348	tp, handle, tca, arg, opt, p, r, arg ? *arg : 0L);
	349
	350	if (!opt)
	351	return 0;
	352
6fa8c014	353	err = nla_parse_nested(tb, TCA_TCINDEX_MAX, opt, tcindex_policy);
cee63723 PM	354	if (err < 0)
cee63723 PM	355	return err;
1da177e4	356
c1b52739 BL	357	return tcindex_set_parms(net, tp, base, handle, p, r, tb,
c1b52739 BL	358	tca[TCA_RATE]);
1da177e4 LT	359	}
	360
	361
	362	static void tcindex_walk(struct tcf_proto tp, struct tcf_walker walker)
	363	{
	364	struct tcindex_data *p = PRIV(tp);
aa767bfe	365	struct tcindex_filter f, next;
1da177e4 LT	366	int i;
1da177e4 LT	367
aa767bfe	368	pr_debug("tcindex_walk(tp %p,walker %p),p %p\n", tp, walker, p);
1da177e4 LT	369	if (p->perfect) {
	370	for (i = 0; i < p->hash; i++) {
	371	if (!p->perfect[i].res.class)
	372	continue;
	373	if (walker->count >= walker->skip) {
	374	if (walker->fn(tp,
	375	(unsigned long) (p->perfect+i), walker)
	376	< 0) {
	377	walker->stop = 1;
	378	return;
	379	}
	380	}
	381	walker->count++;
	382	}
	383	}
	384	if (!p->h)
	385	return;
	386	for (i = 0; i < p->hash; i++) {
	387	for (f = p->h[i]; f; f = next) {
	388	next = f->next;
	389	if (walker->count >= walker->skip) {
aa767bfe	390	if (walker->fn(tp, (unsigned long) &f->result,
1da177e4 LT	391	walker) < 0) {
	392	walker->stop = 1;
	393	return;
	394	}
	395	}
	396	walker->count++;
	397	}
	398	}
	399	}
	400
	401
	402	static int tcindex_destroy_element(struct tcf_proto *tp,
	403	unsigned long arg, struct tcf_walker *walker)
	404	{
	405	return __tcindex_delete(tp, arg, 0);
	406	}
	407
	408
	409	static void tcindex_destroy(struct tcf_proto *tp)
	410	{
	411	struct tcindex_data *p = PRIV(tp);
	412	struct tcf_walker walker;
	413
aa767bfe	414	pr_debug("tcindex_destroy(tp %p),p %p\n", tp, p);
1da177e4 LT	415	walker.count = 0;
	416	walker.skip = 0;
	417	walker.fn = &tcindex_destroy_element;
aa767bfe	418	tcindex_walk(tp, &walker);
a51482bd JJ	419	kfree(p->perfect);
a51482bd JJ	420	kfree(p->h);
1da177e4 LT	421	kfree(p);
	422	tp->root = NULL;
	423	}
	424
	425
	426	static int tcindex_dump(struct tcf_proto *tp, unsigned long fh,
	427	struct sk_buff skb, struct tcmsg t)
	428	{
	429	struct tcindex_data *p = PRIV(tp);
	430	struct tcindex_filter_result r = (struct tcindex_filter_result ) fh;
27a884dc	431	unsigned char *b = skb_tail_pointer(skb);
4b3550ef	432	struct nlattr *nest;
1da177e4	433
aa767bfe SH	434	pr_debug("tcindex_dump(tp %p,fh 0x%lx,skb %p,t %p),p %p,r %p,b %p\n",
	435	tp, fh, skb, t, p, r, b);
	436	pr_debug("p->perfect %p p->h %p\n", p->perfect, p->h);
4b3550ef PM	437
	438	nest = nla_nest_start(skb, TCA_OPTIONS);
	439	if (nest == NULL)
	440	goto nla_put_failure;
	441
1da177e4 LT	442	if (!fh) {
1da177e4 LT	443	t->tcm_handle = ~0; /* whatever ... */
1b34ec43 DM	444	if (nla_put_u32(skb, TCA_TCINDEX_HASH, p->hash) \|\|
	445	nla_put_u16(skb, TCA_TCINDEX_MASK, p->mask) \|\|
	446	nla_put_u32(skb, TCA_TCINDEX_SHIFT, p->shift) \|\|
	447	nla_put_u32(skb, TCA_TCINDEX_FALL_THROUGH, p->fall_through))
	448	goto nla_put_failure;
4b3550ef	449	nla_nest_end(skb, nest);
1da177e4 LT	450	} else {
	451	if (p->perfect) {
	452	t->tcm_handle = r-p->perfect;
	453	} else {
	454	struct tcindex_filter *f;
	455	int i;
	456
	457	t->tcm_handle = 0;
	458	for (i = 0; !t->tcm_handle && i < p->hash; i++) {
	459	for (f = p->h[i]; !t->tcm_handle && f;
	460	f = f->next) {
	461	if (&f->result == r)
	462	t->tcm_handle = f->key;
	463	}
	464	}
	465	}
aa767bfe	466	pr_debug("handle = %d\n", t->tcm_handle);
1b34ec43 DM	467	if (r->res.class &&
	468	nla_put_u32(skb, TCA_TCINDEX_CLASSID, r->res.classid))
	469	goto nla_put_failure;
1da177e4 LT	470
1da177e4 LT	471	if (tcf_exts_dump(skb, &r->exts, &tcindex_ext_map) < 0)
add93b61	472	goto nla_put_failure;
4b3550ef	473	nla_nest_end(skb, nest);
1da177e4 LT	474
1da177e4 LT	475	if (tcf_exts_dump_stats(skb, &r->exts, &tcindex_ext_map) < 0)
add93b61	476	goto nla_put_failure;
1da177e4	477	}
10297b99	478
1da177e4 LT	479	return skb->len;
1da177e4 LT	480
add93b61	481	nla_put_failure:
dc5fc579	482	nlmsg_trim(skb, b);
1da177e4 LT	483	return -1;
	484	}
	485
2eb9d75c	486	static struct tcf_proto_ops cls_tcindex_ops __read_mostly = {
1da177e4 LT	487	.kind = "tcindex",
	488	.classify = tcindex_classify,
	489	.init = tcindex_init,
	490	.destroy = tcindex_destroy,
	491	.get = tcindex_get,
	492	.put = tcindex_put,
	493	.change = tcindex_change,
	494	.delete = tcindex_delete,
	495	.walk = tcindex_walk,
	496	.dump = tcindex_dump,
	497	.owner = THIS_MODULE,
	498	};
	499
	500	static int __init init_tcindex(void)
	501	{
	502	return register_tcf_proto_ops(&cls_tcindex_ops);
	503	}
	504
10297b99	505	static void __exit exit_tcindex(void)
1da177e4 LT	506	{
	507	unregister_tcf_proto_ops(&cls_tcindex_ops);
	508	}
	509
	510	module_init(init_tcindex)
	511	module_exit(exit_tcindex)
	512	MODULE_LICENSE("GPL");