cregit-Linux how code gets into the kernel

Release 4.11 net/sched/em_meta.c

Directory: net/sched
/*
 * net/sched/em_meta.c  Metadata ematch
 *
 *              This program is free software; you can redistribute it and/or
 *              modify it under the terms of the GNU General Public License
 *              as published by the Free Software Foundation; either version
 *              2 of the License, or (at your option) any later version.
 *
 * Authors:     Thomas Graf <tgraf@suug.ch>
 *
 * ==========================================================================
 *
 *      The metadata ematch compares two meta objects where each object
 *      represents either a meta value stored in the kernel or a static
 *      value provided by userspace. The objects are not provided by
 *      userspace itself but rather a definition providing the information
 *      to build them. Every object is of a certain type which must be
 *      equal to the object it is being compared to.
 *
 *      The definition of a objects conists of the type (meta type), a
 *      identifier (meta id) and additional type specific information.
 *      The meta id is either TCF_META_TYPE_VALUE for values provided by
 *      userspace or a index to the meta operations table consisting of
 *      function pointers to type specific meta data collectors returning
 *      the value of the requested meta value.
 *
 *               lvalue                                   rvalue
 *            +-----------+                           +-----------+
 *            | type: INT |                           | type: INT |
 *       def  | id: DEV   |                           | id: VALUE |
 *            | data:     |                           | data: 3   |
 *            +-----------+                           +-----------+
 *                  |                                       |
 *                  ---> meta_ops[INT][DEV](...)            |
 *                            |                             |
 *                  -----------                             |
 *                  V                                       V
 *            +-----------+                           +-----------+
 *            | type: INT |                           | type: INT |
 *       obj  | id: DEV |                             | id: VALUE |
 *            | data: 2   |<--data got filled out     | data: 3   |
 *            +-----------+                           +-----------+
 *                  |                                         |
 *                  --------------> 2  equals 3 <--------------
 *
 *      This is a simplified schema, the complexity varies depending
 *      on the meta type. Obviously, the length of the data must also
 *      be provided for non-numeric types.
 *
 *      Additionally, type dependent modifiers such as shift operators
 *      or mask may be applied to extend the functionaliy. As of now,
 *      the variable length type supports shifting the byte string to
 *      the right, eating up any number of octets and thus supporting
 *      wildcard interface name comparisons such as "ppp%" matching
 *      ppp0..9.
 *
 *      NOTE: Certain meta values depend on other subsystems and are
 *            only available if that subsystem is enabled in the kernel.
 */

#include <linux/slab.h>
#include <linux/module.h>
#include <linux/types.h>
#include <linux/kernel.h>
#include <linux/sched.h>
#include <linux/sched/loadavg.h>
#include <linux/string.h>
#include <linux/skbuff.h>
#include <linux/random.h>
#include <linux/if_vlan.h>
#include <linux/tc_ematch/tc_em_meta.h>
#include <net/dst.h>
#include <net/route.h>
#include <net/pkt_cls.h>
#include <net/sock.h>


struct meta_obj {
	
unsigned long		value;
	
unsigned int		len;
};


struct meta_value {
	
struct tcf_meta_val	hdr;
	
unsigned long		val;
	
unsigned int		len;
};


struct meta_match {
	
struct meta_value	lvalue;
	
struct meta_value	rvalue;
};




static inline int meta_id(struct meta_value *v)
{
	return TCF_META_ID(v->hdr.kind);
}


Contributors
PersonTokensPropCommitsCommitProp
David S. Miller22100.00%1100.00%


Total22100.00%1100.00%






static inline int meta_type(struct meta_value *v)
{
	return TCF_META_TYPE(v->hdr.kind);
}


Contributors
PersonTokensPropCommitsCommitProp
David S. Miller22100.00%1100.00%


Total22100.00%1100.00%




#define META_COLLECTOR(FUNC) static void meta_##FUNC(struct sk_buff *skb, \
        struct tcf_pkt_info *info, struct meta_value *v, \
        struct meta_obj *dst, int *err)

/**************************************************************************
 * System status & misc
 **************************************************************************/


META_COLLECTOR(int_random)
{
	get_random_bytes(&dst->value, sizeof(dst->value));
}




static inline unsigned long fixed_loadavg(int load)
{
	int rnd_load = load + (FIXED_1/200);
	int rnd_frac = ((rnd_load & (FIXED_1-1)) * 100) >> FSHIFT;

	return ((rnd_load >> FSHIFT) * 100) + rnd_frac;
}


Contributors
PersonTokensPropCommitsCommitProp
David S. Miller54100.00%1100.00%


Total54100.00%1100.00%




META_COLLECTOR(int_loadavg_0)
{
	dst->value = fixed_loadavg(avenrun[0]);
}


META_COLLECTOR(int_loadavg_1)
{
	dst->value = fixed_loadavg(avenrun[1]);
}


META_COLLECTOR(int_loadavg_2)
{
	dst->value = fixed_loadavg(avenrun[2]);
}

/**************************************************************************
 * Device names & indices
 **************************************************************************/




static inline int int_dev(struct net_device *dev, struct meta_obj *dst)
{
	if (unlikely(dev == NULL))
		return -1;

	dst->value = dev->ifindex;
	return 0;
}


Contributors
PersonTokensPropCommitsCommitProp
David S. Miller41100.00%1100.00%


Total41100.00%1100.00%






static inline int var_dev(struct net_device *dev, struct meta_obj *dst)
{
	if (unlikely(dev == NULL))
		return -1;

	dst->value = (unsigned long) dev->name;
	dst->len = strlen(dev->name);
	return 0;
}


Contributors
PersonTokensPropCommitsCommitProp
David S. Miller56100.00%1100.00%


Total56100.00%1100.00%




META_COLLECTOR(int_dev)
{
	*err = int_dev(skb->dev, dst);
}


META_COLLECTOR(var_dev)
{
	*err = var_dev(skb->dev, dst);
}

/**************************************************************************
 * vlan tag
 **************************************************************************/


META_COLLECTOR(int_vlan_tag)
{
	unsigned short tag;

	if (skb_vlan_tag_present(skb))
		dst->value = skb_vlan_tag_get(skb);
	else if (!__vlan_get_tag(skb, &tag))
		dst->value = tag;
	else
		*err = -1;
}



/**************************************************************************
 * skb attributes
 **************************************************************************/


META_COLLECTOR(int_priority)
{
	dst->value = skb->priority;
}


META_COLLECTOR(int_protocol)
{
	/* Let userspace take care of the byte ordering */
	dst->value = tc_skb_protocol(skb);
}


META_COLLECTOR(int_pkttype)
{
	dst->value = skb->pkt_type;
}


META_COLLECTOR(int_pktlen)
{
	dst->value = skb->len;
}


META_COLLECTOR(int_datalen)
{
	dst->value = skb->data_len;
}


META_COLLECTOR(int_maclen)
{
	dst->value = skb->mac_len;
}


META_COLLECTOR(int_rxhash)
{
	dst->value = skb_get_hash(skb);
}

/**************************************************************************
 * Netfilter
 **************************************************************************/


META_COLLECTOR(int_mark)
{
	dst->value = skb->mark;
}

/**************************************************************************
 * Traffic Control
 **************************************************************************/


META_COLLECTOR(int_tcindex)
{
	dst->value = skb->tc_index;
}

/**************************************************************************
 * Routing
 **************************************************************************/


META_COLLECTOR(int_rtclassid)
{
	if (unlikely(skb_dst(skb) == NULL))
		*err = -1;
	else
#ifdef CONFIG_IP_ROUTE_CLASSID
		dst->value = skb_dst(skb)->tclassid;
#else
		dst->value = 0;
#endif
}


META_COLLECTOR(int_rtiif)
{
	if (unlikely(skb_rtable(skb) == NULL))
		*err = -1;
	else
		dst->value = inet_iif(skb);
}

/**************************************************************************
 * Socket Attributes
 **************************************************************************/


#define skip_nonlocal(skb) \
	(unlikely(skb->sk == NULL))


META_COLLECTOR(int_sk_family)
{
	if (skip_nonlocal(skb)) {
		*err = -1;
		return;
	}
	dst->value = skb->sk->sk_family;
}


META_COLLECTOR(int_sk_state)
{
	if (skip_nonlocal(skb)) {
		*err = -1;
		return;
	}
	dst->value = skb->sk->sk_state;
}


META_COLLECTOR(int_sk_reuse)
{
	if (skip_nonlocal(skb)) {
		*err = -1;
		return;
	}
	dst->value = skb->sk->sk_reuse;
}


META_COLLECTOR(int_sk_bound_if)
{
	if (skip_nonlocal(skb)) {
		*err = -1;
		return;
	}
	/* No error if bound_dev_if is 0, legal userspace check */
	dst->value = skb->sk->sk_bound_dev_if;
}


META_COLLECTOR(var_sk_bound_if)
{
	if (skip_nonlocal(skb)) {
		*err = -1;
		return;
	}

	if (skb->sk->sk_bound_dev_if == 0) {
		dst->value = (unsigned long) "any";
		dst->len = 3;
	} else {
		struct net_device *dev;

		rcu_read_lock();
		dev = dev_get_by_index_rcu(sock_net(skb->sk),
					   skb->sk->sk_bound_dev_if);
		*err = var_dev(dev, dst);
		rcu_read_unlock();
	}
}


META_COLLECTOR(int_sk_refcnt)
{
	if (skip_nonlocal(skb)) {
		*err = -1;
		return;
	}
	dst->value = atomic_read(&skb->sk->sk_refcnt);
}


META_COLLECTOR(int_sk_rcvbuf)
{
	const struct sock *sk = skb_to_full_sk(skb);

	if (!sk) {
		*err = -1;
		return;
	}
	dst->value = sk->sk_rcvbuf;
}


META_COLLECTOR(int_sk_shutdown)
{
	const struct sock *sk = skb_to_full_sk(skb);

	if (!sk) {
		*err = -1;
		return;
	}
	dst->value = sk->sk_shutdown;
}


META_COLLECTOR(int_sk_proto)
{
	const struct sock *sk = skb_to_full_sk(skb);

	if (!sk) {
		*err = -1;
		return;
	}
	dst->value = sk->sk_protocol;
}


META_COLLECTOR(int_sk_type)
{
	const struct sock *sk = skb_to_full_sk(skb);

	if (!sk) {
		*err = -1;
		return;
	}
	dst->value = sk->sk_type;
}


META_COLLECTOR(int_sk_rmem_alloc)
{
	const struct sock *sk = skb_to_full_sk(skb);

	if (!sk) {
		*err = -1;
		return;
	}
	dst->value = sk_rmem_alloc_get(sk);
}


META_COLLECTOR(int_sk_wmem_alloc)
{
	const struct sock *sk = skb_to_full_sk(skb);

	if (!sk) {
		*err = -1;
		return;
	}
	dst->value = sk_wmem_alloc_get(sk);
}


META_COLLECTOR(int_sk_omem_alloc)
{
	const struct sock *sk = skb_to_full_sk(skb);

	if (!sk) {
		*err = -1;
		return;
	}
	dst->value = atomic_read(&sk->sk_omem_alloc);
}


META_COLLECTOR(int_sk_rcv_qlen)
{
	const struct sock *sk = skb_to_full_sk(skb);

	if (!sk) {
		*err = -1;
		return;
	}
	dst->value = sk->sk_receive_queue.qlen;
}


META_COLLECTOR(int_sk_snd_qlen)
{
	const struct sock *sk = skb_to_full_sk(skb);

	if (!sk) {
		*err = -1;
		return;
	}
	dst->value = sk->sk_write_queue.qlen;
}


META_COLLECTOR(int_sk_wmem_queued)
{
	const struct sock *sk = skb_to_full_sk(skb);

	if (!sk) {
		*err = -1;
		return;
	}
	dst->value = sk->sk_wmem_queued;
}


META_COLLECTOR(int_sk_fwd_alloc)
{
	const struct sock *sk = skb_to_full_sk(skb);

	if (!sk) {
		*err = -1;
		return;
	}
	dst->value = sk->sk_forward_alloc;
}


META_COLLECTOR(int_sk_sndbuf)
{
	const struct sock *sk = skb_to_full_sk(skb);

	if (!sk) {
		*err = -1;
		return;
	}
	dst->value = sk->sk_sndbuf;
}


META_COLLECTOR(int_sk_alloc)
{
	const struct sock *sk = skb_to_full_sk(skb);

	if (!sk) {
		*err = -1;
		return;
	}
	dst->value = (__force int) sk->sk_allocation;
}


META_COLLECTOR(int_sk_hash)
{
	if (skip_nonlocal(skb)) {
		*err = -1;
		return;
	}
	dst->value = skb->sk->sk_hash;
}


META_COLLECTOR(int_sk_lingertime)
{
	const struct sock *sk = skb_to_full_sk(skb);

	if (!sk) {
		*err = -1;
		return;
	}
	dst->value = sk->sk_lingertime / HZ;
}


META_COLLECTOR(int_sk_err_qlen)
{
	const struct sock *sk = skb_to_full_sk(skb);

	if (!sk) {
		*err = -1;
		return;
	}
	dst->value = sk->sk_error_queue.qlen;
}


META_COLLECTOR(int_sk_ack_bl)
{
	const struct sock *sk = skb_to_full_sk(skb);

	if (!sk) {
		*err = -1;
		return;
	}
	dst->value = sk->sk_ack_backlog;
}


META_COLLECTOR(int_sk_max_ack_bl)
{
	const struct sock *sk = skb_to_full_sk(skb);

	if (!sk) {
		*err = -1;
		return;
	}
	dst->value = sk->sk_max_ack_backlog;
}


META_COLLECTOR(int_sk_prio)
{
	const struct sock *sk = skb_to_full_sk(skb);

	if (!sk) {
		*err = -1;
		return;
	}
	dst->value = sk->sk_priority;
}


META_COLLECTOR(int_sk_rcvlowat)
{
	const struct sock *sk = skb_to_full_sk(skb);

	if (!sk) {
		*err = -1;
		return;
	}
	dst->value = sk->sk_rcvlowat;
}


META_COLLECTOR(int_sk_rcvtimeo)
{
	const struct sock *sk = skb_to_full_sk(skb);

	if (!sk) {
		*err = -1;
		return;
	}
	dst->value = sk->sk_rcvtimeo / HZ;
}


META_COLLECTOR(int_sk_sndtimeo)
{
	const struct sock *sk = skb_to_full_sk(skb);

	if (!sk) {
		*err = -1;
		return;
	}
	dst->value = sk->sk_sndtimeo / HZ;
}


META_COLLECTOR(int_sk_sendmsg_off)
{
	const struct sock *sk = skb_to_full_sk(skb);

	if (!sk) {
		*err = -1;
		return;
	}
	dst->value = sk->sk_frag.offset;
}


META_COLLECTOR(int_sk_write_pend)
{
	const struct sock *sk = skb_to_full_sk(skb);

	if (!sk) {
		*err = -1;
		return;
	}
	dst->value = sk->sk_write_pending;
}

/**************************************************************************
 * Meta value collectors assignment table
 **************************************************************************/


struct meta_ops {
	
void		(*get)(struct sk_buff *, struct tcf_pkt_info *,
			       struct meta_value *, struct meta_obj *, int *);
};


#define META_ID(name) TCF_META_ID_##name

#define META_FUNC(name) { .get = meta_##name }

/* Meta value operations table listing all meta value collectors and
 * assigns them to a type and meta id. */

static struct meta_ops __meta_ops[TCF_META_TYPE_MAX + 1][TCF_META_ID_MAX + 1] = {
	[TCF_META_TYPE_VAR] = {
		[META_ID(DEV)]			= META_FUNC(var_dev),
		[META_ID(SK_BOUND_IF)] 		= META_FUNC(var_sk_bound_if),
        },
	[TCF_META_TYPE_INT] = {
		[META_ID(RANDOM)]		= META_FUNC(int_random),
		[META_ID(LOADAVG_0)]		= META_FUNC(int_loadavg_0),
		[META_ID(LOADAVG_1)]		= META_FUNC(int_loadavg_1),
		[META_ID(LOADAVG_2)]		= META_FUNC(int_loadavg_2),
		[META_ID(DEV)]			= META_FUNC(int_dev),
		[META_ID(PRIORITY)]		= META_FUNC(int_priority),
		[META_ID(PROTOCOL)]		= META_FUNC(int_protocol),
		[META_ID(PKTTYPE)]		= META_FUNC(int_pkttype),
		[META_ID(PKTLEN)]		= META_FUNC(int_pktlen),
		[META_ID(DATALEN)]		= META_FUNC(int_datalen),
		[META_ID(MACLEN)]		= META_FUNC(int_maclen),
		[META_ID(NFMARK)]		= META_FUNC(int_mark),
		[META_ID(TCINDEX)]		= META_FUNC(int_tcindex),
		[META_ID(RTCLASSID)]		= META_FUNC(int_rtclassid),
		[META_ID(RTIIF)]		= META_FUNC(int_rtiif),
		[META_ID(SK_FAMILY)]		= META_FUNC(int_sk_family),
		[META_ID(SK_STATE)]		= META_FUNC(int_sk_state),
		[META_ID(SK_REUSE)]		= META_FUNC(int_sk_reuse),
		[META_ID(SK_BOUND_IF)]		= META_FUNC(int_sk_bound_if),
		[META_ID(SK_REFCNT)]		= META_FUNC(int_sk_refcnt),
		[META_ID(SK_RCVBUF)]		= META_FUNC(int_sk_rcvbuf),
		[META_ID(SK_SNDBUF)]		= META_FUNC(int_sk_sndbuf),
		[META_ID(SK_SHUTDOWN)]		= META_FUNC(int_sk_shutdown),
		[META_ID(SK_PROTO)]		= META_FUNC(int_sk_proto),
		[META_ID(SK_TYPE)]		= META_FUNC(int_sk_type),
		[META_ID(SK_RMEM_ALLOC)]	= META_FUNC(int_sk_rmem_alloc),
		[META_ID(SK_WMEM_ALLOC)]	= META_FUNC(int_sk_wmem_alloc),
		[META_ID(SK_OMEM_ALLOC)]	= META_FUNC(int_sk_omem_alloc),
		[META_ID(SK_WMEM_QUEUED)]	= META_FUNC(int_sk_wmem_queued),
		[META_ID(SK_RCV_QLEN)]		= META_FUNC(int_sk_rcv_qlen),
		[META_ID(SK_SND_QLEN)]		= META_FUNC(int_sk_snd_qlen),
		[META_ID(SK_ERR_QLEN)]		= META_FUNC(int_sk_err_qlen),
		[META_ID(SK_FORWARD_ALLOCS)]	= META_FUNC(int_sk_fwd_alloc),
		[META_ID(SK_ALLOCS)]		= META_FUNC(int_sk_alloc),
		[META_ID(SK_HASH)]		= META_FUNC(int_sk_hash),
		[META_ID(SK_LINGERTIME)]	= META_FUNC(int_sk_lingertime),
		[META_ID(SK_ACK_BACKLOG)]	= META_FUNC(int_sk_ack_bl),
		[META_ID(SK_MAX_ACK_BACKLOG)]	= META_FUNC(int_sk_max_ack_bl),
		[META_ID(SK_PRIO)]		= META_FUNC(int_sk_prio),
		[META_ID(SK_RCVLOWAT)]		= META_FUNC(int_sk_rcvlowat),
		[META_ID(SK_RCVTIMEO)]		= META_FUNC(int_sk_rcvtimeo),
		[META_ID(SK_SNDTIMEO)]		= META_FUNC(int_sk_sndtimeo),
		[META_ID(SK_SENDMSG_OFF)]	= META_FUNC(int_sk_sendmsg_off),
		[META_ID(SK_WRITE_PENDING)]	= META_FUNC(int_sk_write_pend),
		[META_ID(VLAN_TAG)]		= META_FUNC(int_vlan_tag),
		[META_ID(RXHASH)]		= META_FUNC(int_rxhash),
        }
};




static inline struct meta_ops *meta_ops(struct meta_value *val)
{
	return &__meta_ops[meta_type(val)][meta_id(val)];
}


Contributors
PersonTokensPropCommitsCommitProp
David S. Miller30100.00%1100.00%


Total30100.00%1100.00%



/**************************************************************************
 * Type specific operations for TCF_META_TYPE_VAR
 **************************************************************************/




static int meta_var_compare(struct meta_obj *a, struct meta_obj *b)
{
	int r = a->len - b->len;

	if (r == 0)
		r = memcmp((void *) a->value, (void *) b->value, a->len);

	return r;
}


Contributors
PersonTokensPropCommitsCommitProp
David S. Miller61100.00%1100.00%


Total61100.00%1100.00%






static int meta_var_change(struct meta_value *dst, struct nlattr *nla)
{
	int len = nla_len(nla);

	dst->val = (unsigned long)kmemdup(nla_data(nla), len, GFP_KERNEL);
	if (dst->val == 0UL)
		return -ENOMEM;
	dst->len = len;
	return 0;
}


Contributors
PersonTokensPropCommitsCommitProp
David S. Miller5584.62%133.33%


Patrick McHardy69.23%133.33%


Arnaldo Carvalho de Melo46.15%133.33%


Total65100.00%3100.00%






static void meta_var_destroy(struct meta_value *v)
{
	kfree((void *) v->val);
}


Contributors
PersonTokensPropCommitsCommitProp
David S. Miller22100.00%1100.00%


Total22100.00%1100.00%






static void meta_var_apply_extras(struct meta_value *v,
				  struct meta_obj *dst)
{
	int shift = v->hdr.shift;

	if (shift && shift < dst->len)
		dst->len -= shift;
}


Contributors
PersonTokensPropCommitsCommitProp
David S. Miller41100.00%1100.00%


Total41100.00%1100.00%






static int meta_var_dump(struct sk_buff *skb, struct meta_value *v, int tlv)
{
	if (v->val && v->len &&
	    nla_put(skb, tlv, v->len, (void *) v->val))
		goto nla_put_failure;
	return 0;

nla_put_failure:
	return -1;
}


Contributors
PersonTokensPropCommitsCommitProp
David S. Miller5998.33%266.67%


Patrick McHardy11.67%133.33%


Total60100.00%3100.00%



/**************************************************************************
 * Type specific operations for TCF_META_TYPE_INT
 **************************************************************************/




static int meta_int_compare(struct meta_obj *a, struct meta_obj *b)
{
	/* Let gcc optimize it, the unlikely is not really based on
         * some numbers but jump free code for mismatches seems
         * more logical. */
	if (unlikely(a->value == b->value))
		return 0;
	else if (a->value < b->value)
		return -1;
	else
		return 1;
}


Contributors
PersonTokensPropCommitsCommitProp
David S. Miller4484.62%150.00%


Thomas Graf815.38%150.00%


Total52100.00%2100.00%






static int meta_int_change(struct meta_value *dst, struct nlattr *nla)
{
	if (nla_len(nla) >= sizeof(unsigned long)) {
		dst->val = *(unsigned long *) nla_data(nla);
		dst->len = sizeof(unsigned long);
	} else if (nla_len(nla) == sizeof(u32)) {
		dst->