int
pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sns,
struct pf_pool *rpool, enum pf_sn_types type)
{
unsigned char hash[16];
struct pf_addr faddr;
struct pf_addr *raddr = &rpool->addr.v.a.addr;
struct pf_addr *rmask = &rpool->addr.v.a.mask;
u_int64_t states;
u_int16_t weight;
u_int64_t load;
u_int64_t cload;
if (sns[type] == NULL && rpool->opts & PF_POOL_STICKYADDR &&
(rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE &&
pf_map_addr_sticky(af, r, saddr, naddr, sns, rpool, type) == 0)
return (0);
if (rpool->addr.type == PF_ADDR_NOROUTE)
return (1);
if (rpool->addr.type == PF_ADDR_DYNIFTL) {
switch (af) {
#ifdef INET
case AF_INET:
if (rpool->addr.p.dyn->pfid_acnt4 < 1 &&
((rpool->opts & PF_POOL_TYPEMASK) !=
PF_POOL_ROUNDROBIN) &&
((rpool->opts & PF_POOL_TYPEMASK) !=
PF_POOL_LEASTSTATES))
return (1);
raddr = &rpool->addr.p.dyn->pfid_addr4;
rmask = &rpool->addr.p.dyn->pfid_mask4;
break;
#endif /* INET */
#ifdef INET6
case AF_INET6:
if (rpool->addr.p.dyn->pfid_acnt6 < 1 &&
((rpool->opts & PF_POOL_TYPEMASK) !=
PF_POOL_ROUNDROBIN) &&
((rpool->opts & PF_POOL_TYPEMASK) !=
PF_POOL_LEASTSTATES))
return (1);
raddr = &rpool->addr.p.dyn->pfid_addr6;
rmask = &rpool->addr.p.dyn->pfid_mask6;
break;
#endif /* INET6 */
}
} else if (rpool->addr.type == PF_ADDR_TABLE) {
if (((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN) &&
((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_LEASTSTATES))
return (1); /* unsupported */
} else {
raddr = &rpool->addr.v.a.addr;
rmask = &rpool->addr.v.a.mask;
}
switch (rpool->opts & PF_POOL_TYPEMASK) {
case PF_POOL_NONE:
PF_ACPY(naddr, raddr, af);
break;
case PF_POOL_BITMASK:
PF_POOLMASK(naddr, raddr, rmask, saddr, af);
break;
case PF_POOL_RANDOM:
if (init_addr != NULL && PF_AZERO(init_addr, af)) {
switch (af) {
#ifdef INET
case AF_INET:
rpool->counter.addr32[0] = htonl(arc4random());
break;
#endif /* INET */
#ifdef INET6
case AF_INET6:
if (rmask->addr32[3] != 0xffffffff)
rpool->counter.addr32[3] =
htonl(arc4random());
else
break;
if (rmask->addr32[2] != 0xffffffff)
rpool->counter.addr32[2] =
htonl(arc4random());
else
break;
if (rmask->addr32[1] != 0xffffffff)
rpool->counter.addr32[1] =
htonl(arc4random());
else
break;
if (rmask->addr32[0] != 0xffffffff)
rpool->counter.addr32[0] =
htonl(arc4random());
break;
#endif /* INET6 */
}
PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
PF_ACPY(init_addr, naddr, af);
} else {
PF_AINC(&rpool->counter, af);
PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
}
break;
case PF_POOL_SRCHASH:
pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
break;
case PF_POOL_ROUNDROBIN:
if (rpool->addr.type == PF_ADDR_TABLE ||
rpool->addr.type == PF_ADDR_DYNIFTL) {
if (pfr_pool_get(rpool, &raddr, &rmask, af)) {
/*
* reset counter in case its value
* has been removed from the pool.
*/
bzero(&rpool->counter, sizeof(rpool->counter));
if (pfr_pool_get(rpool, &raddr, &rmask, af))
return (1);
}
} else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
return (1);
/* iterate over table if it contains entries which are weighted */
if ((rpool->addr.type == PF_ADDR_TABLE &&
rpool->addr.p.tbl->pfrkt_refcntcost > 0) ||
(rpool->addr.type == PF_ADDR_DYNIFTL &&
rpool->addr.p.dyn->pfid_kt->pfrkt_refcntcost > 0)) {
do {
if (rpool->addr.type == PF_ADDR_TABLE ||
rpool->addr.type == PF_ADDR_DYNIFTL) {
if (pfr_pool_get(rpool,
&raddr, &rmask, af))
return (1);
} else {
log(LOG_ERR, "pf: pf_map_addr: "
"weighted RR failure");
return (1);
}
if (rpool->weight >= rpool->curweight)
break;
PF_AINC(&rpool->counter, af);
} while (1);
weight = rpool->weight;
}
PF_ACPY(naddr, &rpool->counter, af);
if (init_addr != NULL && PF_AZERO(init_addr, af))
PF_ACPY(init_addr, naddr, af);
PF_AINC(&rpool->counter, af);
break;
case PF_POOL_LEASTSTATES:
/* retrieve an address first */
if (rpool->addr.type == PF_ADDR_TABLE ||
rpool->addr.type == PF_ADDR_DYNIFTL) {
if (pfr_pool_get(rpool, &raddr, &rmask, af)) {
/* see PF_POOL_ROUNDROBIN */
bzero(&rpool->counter, sizeof(rpool->counter));
if (pfr_pool_get(rpool, &raddr, &rmask, af))
return (1);
}
} else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
return (1);
states = rpool->states;
weight = rpool->weight;
if ((rpool->addr.type == PF_ADDR_TABLE &&
rpool->addr.p.tbl->pfrkt_refcntcost > 0) ||
(rpool->addr.type == PF_ADDR_DYNIFTL &&
rpool->addr.p.dyn->pfid_kt->pfrkt_refcntcost > 0))
load = ((UINT16_MAX * rpool->states) / rpool->weight);
else
load = states;
PF_ACPY(&faddr, &rpool->counter, af);
PF_ACPY(naddr, &rpool->counter, af);
if (init_addr != NULL && PF_AZERO(init_addr, af))
PF_ACPY(init_addr, naddr, af);
/*
* iterate *once* over whole table and find destination with
* least connection
*/
do {
PF_AINC(&rpool->counter, af);
if (rpool->addr.type == PF_ADDR_TABLE ||
rpool->addr.type == PF_ADDR_DYNIFTL) {
if (pfr_pool_get(rpool, &raddr, &rmask, af))
return (1);
} else if (pf_match_addr(0, raddr, rmask,
&rpool->counter, af))
return (1);
if ((rpool->addr.type == PF_ADDR_TABLE &&
rpool->addr.p.tbl->pfrkt_refcntcost > 0) ||
(rpool->addr.type == PF_ADDR_DYNIFTL &&
rpool->addr.p.dyn->pfid_kt->pfrkt_refcntcost > 0))
cload = ((UINT16_MAX * rpool->states)
/ rpool->weight);
else
cload = rpool->states;
/* find lc minimum */
if (cload < load) {
states = rpool->states;
weight = rpool->weight;
load = cload;
PF_ACPY(naddr, &rpool->counter, af);
if (init_addr != NULL &&
PF_AZERO(init_addr, af))
PF_ACPY(init_addr, naddr, af);
}
} while (pf_match_addr(1, &faddr, rmask, &rpool->counter, af) &&
(states > 0));
if (rpool->addr.type == PF_ADDR_TABLE) {
if (pfr_states_increase(rpool->addr.p.tbl,
naddr, af) == -1) {
if (pf_status.debug >= LOG_DEBUG) {
log(LOG_DEBUG,"pf: pf_map_addr: "
"selected address ");
pf_print_host(naddr, 0, af);
addlog(". Failed to increase count!\n");
}
return (1);
}
} else if (rpool->addr.type == PF_ADDR_DYNIFTL) {
if (pfr_states_increase(rpool->addr.p.dyn->pfid_kt,
naddr, af) == -1) {
if (pf_status.debug >= LOG_DEBUG) {
log(LOG_DEBUG, "pf: pf_map_addr: "
"selected address ");
pf_print_host(naddr, 0, af);
addlog(". Failed to increase count!\n");
}
return (1);
}
}
break;
}
if (rpool->opts & PF_POOL_STICKYADDR) {
if (sns[type] != NULL) {
pf_remove_src_node(sns[type]);
sns[type] = NULL;
}
if (pf_insert_src_node(&sns[type], r, type, af, saddr, naddr,
0))
return (1);
}
if (pf_status.debug >= LOG_NOTICE &&
(rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
log(LOG_NOTICE, "pf: pf_map_addr: selected address ");
pf_print_host(naddr, 0, af);
if ((rpool->opts & PF_POOL_TYPEMASK) ==
PF_POOL_LEASTSTATES)
addlog(" with state count %llu", states);
if ((rpool->addr.type == PF_ADDR_TABLE &&
rpool->addr.p.tbl->pfrkt_refcntcost > 0) ||
(rpool->addr.type == PF_ADDR_DYNIFTL &&
rpool->addr.p.dyn->pfid_kt->pfrkt_refcntcost > 0))
addlog(" with weight %u", weight);
addlog("\n");
}
return (0);
}
int
pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sn)
{
struct pf_pool *rpool = &r->rpool;
struct pf_addr *raddr = NULL, *rmask = NULL;
/* Try to find a src_node if none was given and this
is a sticky-address rule. */
if (*sn == NULL && r->rpool.opts & PF_POOL_STICKYADDR &&
(r->rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_NONE)
*sn = pf_find_src_node(saddr, r, af, 0);
/* If a src_node was found or explicitly given and it has a non-zero
route address, use this address. A zeroed address is found if the
src node was created just a moment ago in pf_create_state and it
needs to be filled in with routing decision calculated here. */
if (*sn != NULL && !PF_AZERO(&(*sn)->raddr, af)) {
PF_ACPY(naddr, &(*sn)->raddr, af);
if (V_pf_status.debug >= PF_DEBUG_MISC) {
printf("pf_map_addr: src tracking maps ");
pf_print_host(saddr, 0, af);
printf(" to ");
pf_print_host(naddr, 0, af);
printf("\n");
}
return (0);
}
/* Find the route using chosen algorithm. Store the found route
in src_node if it was given or found. */
if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
return (1);
if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
switch (af) {
#ifdef INET
case AF_INET:
if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
(rpool->opts & PF_POOL_TYPEMASK) !=
PF_POOL_ROUNDROBIN)
return (1);
raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
break;
#endif /* INET */
#ifdef INET6
case AF_INET6:
if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
(rpool->opts & PF_POOL_TYPEMASK) !=
PF_POOL_ROUNDROBIN)
return (1);
raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
break;
#endif /* INET6 */
}
} else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
return (1); /* unsupported */
} else {
raddr = &rpool->cur->addr.v.a.addr;
rmask = &rpool->cur->addr.v.a.mask;
}
switch (rpool->opts & PF_POOL_TYPEMASK) {
case PF_POOL_NONE:
PF_ACPY(naddr, raddr, af);
break;
case PF_POOL_BITMASK:
PF_POOLMASK(naddr, raddr, rmask, saddr, af);
break;
case PF_POOL_RANDOM:
if (init_addr != NULL && PF_AZERO(init_addr, af)) {
switch (af) {
#ifdef INET
case AF_INET:
rpool->counter.addr32[0] = htonl(arc4random());
break;
#endif /* INET */
#ifdef INET6
case AF_INET6:
if (rmask->addr32[3] != 0xffffffff)
rpool->counter.addr32[3] =
htonl(arc4random());
else
break;
if (rmask->addr32[2] != 0xffffffff)
rpool->counter.addr32[2] =
htonl(arc4random());
else
break;
if (rmask->addr32[1] != 0xffffffff)
rpool->counter.addr32[1] =
htonl(arc4random());
else
break;
if (rmask->addr32[0] != 0xffffffff)
rpool->counter.addr32[0] =
htonl(arc4random());
break;
#endif /* INET6 */
}
PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
PF_ACPY(init_addr, naddr, af);
} else {
PF_AINC(&rpool->counter, af);
PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
}
break;
case PF_POOL_SRCHASH:
{
unsigned char hash[16];
pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
break;
}
case PF_POOL_ROUNDROBIN:
{
struct pf_pooladdr *acur = rpool->cur;
/*
* XXXGL: in the round-robin case we need to store
* the round-robin machine state in the rule, thus
* forwarding thread needs to modify rule.
*
* This is done w/o locking, because performance is assumed
* more important than round-robin precision.
*
* In the simpliest case we just update the "rpool->cur"
* pointer. However, if pool contains tables or dynamic
* addresses, then "tblidx" is also used to store machine
* state. Since "tblidx" is int, concurrent access to it can't
* lead to inconsistence, only to lost of precision.
*
* Things get worse, if table contains not hosts, but
* prefixes. In this case counter also stores machine state,
* and for IPv6 address, counter can't be updated atomically.
* Probably, using round-robin on a table containing IPv6
* prefixes (or even IPv4) would cause a panic.
*/
if (rpool->cur->addr.type == PF_ADDR_TABLE) {
if (!pfr_pool_get(rpool->cur->addr.p.tbl,
&rpool->tblidx, &rpool->counter, af))
goto get_addr;
} else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
&rpool->tblidx, &rpool->counter, af))
goto get_addr;
} else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
goto get_addr;
try_next:
if (TAILQ_NEXT(rpool->cur, entries) == NULL)
rpool->cur = TAILQ_FIRST(&rpool->list);
else
rpool->cur = TAILQ_NEXT(rpool->cur, entries);
if (rpool->cur->addr.type == PF_ADDR_TABLE) {
rpool->tblidx = -1;
if (pfr_pool_get(rpool->cur->addr.p.tbl,
&rpool->tblidx, &rpool->counter, af)) {
/* table contains no address of type 'af' */
if (rpool->cur != acur)
goto try_next;
return (1);
}
} else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
rpool->tblidx = -1;
if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
&rpool->tblidx, &rpool->counter, af)) {
/* table contains no address of type 'af' */
if (rpool->cur != acur)
goto try_next;
return (1);
}
} else {
raddr = &rpool->cur->addr.v.a.addr;
rmask = &rpool->cur->addr.v.a.mask;
PF_ACPY(&rpool->counter, raddr, af);
}
get_addr:
PF_ACPY(naddr, &rpool->counter, af);
if (init_addr != NULL && PF_AZERO(init_addr, af))
PF_ACPY(init_addr, naddr, af);
PF_AINC(&rpool->counter, af);
break;
}
}
if (*sn != NULL)
PF_ACPY(&(*sn)->raddr, naddr, af);
if (V_pf_status.debug >= PF_DEBUG_MISC &&
(rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
printf("pf_map_addr: selected address ");
pf_print_host(naddr, 0, af);
printf("\n");
}
return (0);
}
int
pf_map_addr(sa_family_t af, struct pf_rule *r, struct pf_addr *saddr,
struct pf_addr *naddr, struct pf_addr *init_addr, struct pf_src_node **sns,
struct pf_pool *rpool, enum pf_sn_types type)
{
unsigned char hash[16];
struct pf_addr *raddr = &rpool->cur->addr.v.a.addr;
struct pf_addr *rmask = &rpool->cur->addr.v.a.mask;
struct pf_pooladdr *acur = rpool->cur;
struct pf_src_node k;
if (sns[type] == NULL && rpool->opts & PF_POOL_STICKYADDR &&
(rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
k.af = af;
k.type = type;
PF_ACPY(&k.addr, saddr, af);
k.rule.ptr = r;
pf_status.scounters[SCNT_SRC_NODE_SEARCH]++;
sns[type] = RB_FIND(pf_src_tree, &tree_src_tracking, &k);
if (sns[type] != NULL) {
if (!PF_AZERO(&(sns[type])->raddr, af))
PF_ACPY(naddr, &(sns[type])->raddr, af);
if (pf_status.debug >= PF_DEBUG_MISC) {
printf("pf_map_addr: src tracking (%u) maps ",
type);
pf_print_host(&k.addr, 0, af);
printf(" to ");
pf_print_host(naddr, 0, af);
printf("\n");
}
return (0);
}
}
if (rpool->cur->addr.type == PF_ADDR_NOROUTE)
return (1);
if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
switch (af) {
#ifdef INET
case AF_INET:
if (rpool->cur->addr.p.dyn->pfid_acnt4 < 1 &&
(rpool->opts & PF_POOL_TYPEMASK) !=
PF_POOL_ROUNDROBIN)
return (1);
raddr = &rpool->cur->addr.p.dyn->pfid_addr4;
rmask = &rpool->cur->addr.p.dyn->pfid_mask4;
break;
#endif /* INET */
#ifdef INET6
case AF_INET6:
if (rpool->cur->addr.p.dyn->pfid_acnt6 < 1 &&
(rpool->opts & PF_POOL_TYPEMASK) !=
PF_POOL_ROUNDROBIN)
return (1);
raddr = &rpool->cur->addr.p.dyn->pfid_addr6;
rmask = &rpool->cur->addr.p.dyn->pfid_mask6;
break;
#endif /* INET6 */
}
} else if (rpool->cur->addr.type == PF_ADDR_TABLE) {
if ((rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN)
return (1); /* unsupported */
} else {
raddr = &rpool->cur->addr.v.a.addr;
rmask = &rpool->cur->addr.v.a.mask;
}
switch (rpool->opts & PF_POOL_TYPEMASK) {
case PF_POOL_NONE:
PF_ACPY(naddr, raddr, af);
break;
case PF_POOL_BITMASK:
PF_POOLMASK(naddr, raddr, rmask, saddr, af);
break;
case PF_POOL_RANDOM:
if (init_addr != NULL && PF_AZERO(init_addr, af)) {
switch (af) {
#ifdef INET
case AF_INET:
rpool->counter.addr32[0] = htonl(arc4random());
break;
#endif /* INET */
#ifdef INET6
case AF_INET6:
if (rmask->addr32[3] != 0xffffffff)
rpool->counter.addr32[3] =
htonl(arc4random());
else
break;
if (rmask->addr32[2] != 0xffffffff)
rpool->counter.addr32[2] =
htonl(arc4random());
else
break;
if (rmask->addr32[1] != 0xffffffff)
rpool->counter.addr32[1] =
htonl(arc4random());
else
break;
if (rmask->addr32[0] != 0xffffffff)
rpool->counter.addr32[0] =
htonl(arc4random());
break;
#endif /* INET6 */
}
PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
PF_ACPY(init_addr, naddr, af);
} else {
PF_AINC(&rpool->counter, af);
PF_POOLMASK(naddr, raddr, rmask, &rpool->counter, af);
}
break;
case PF_POOL_SRCHASH:
pf_hash(saddr, (struct pf_addr *)&hash, &rpool->key, af);
PF_POOLMASK(naddr, raddr, rmask, (struct pf_addr *)&hash, af);
break;
case PF_POOL_ROUNDROBIN:
if (rpool->cur->addr.type == PF_ADDR_TABLE) {
if (!pfr_pool_get(rpool->cur->addr.p.tbl,
&rpool->tblidx, &rpool->counter,
&raddr, &rmask, af))
goto get_addr;
} else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
if (!pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
&rpool->tblidx, &rpool->counter,
&raddr, &rmask, af))
goto get_addr;
} else if (pf_match_addr(0, raddr, rmask, &rpool->counter, af))
goto get_addr;
try_next:
if ((rpool->cur = TAILQ_NEXT(rpool->cur, entries)) == NULL)
rpool->cur = TAILQ_FIRST(&rpool->list);
if (rpool->cur->addr.type == PF_ADDR_TABLE) {
rpool->tblidx = -1;
if (pfr_pool_get(rpool->cur->addr.p.tbl,
&rpool->tblidx, &rpool->counter,
&raddr, &rmask, af)) {
/* table contains no address of type 'af' */
if (rpool->cur != acur)
goto try_next;
return (1);
}
} else if (rpool->cur->addr.type == PF_ADDR_DYNIFTL) {
rpool->tblidx = -1;
if (pfr_pool_get(rpool->cur->addr.p.dyn->pfid_kt,
&rpool->tblidx, &rpool->counter,
&raddr, &rmask, af)) {
/* table contains no address of type 'af' */
if (rpool->cur != acur)
goto try_next;
return (1);
}
} else {
raddr = &rpool->cur->addr.v.a.addr;
rmask = &rpool->cur->addr.v.a.mask;
PF_ACPY(&rpool->counter, raddr, af);
}
get_addr:
PF_ACPY(naddr, &rpool->counter, af);
if (init_addr != NULL && PF_AZERO(init_addr, af))
PF_ACPY(init_addr, naddr, af);
PF_AINC(&rpool->counter, af);
break;
}
if (rpool->opts & PF_POOL_STICKYADDR) {
if (sns[type] != NULL) {
pf_remove_src_node(sns[type]);
sns[type] = NULL;
}
if (pf_insert_src_node(&sns[type], r, type, af, saddr, naddr,
0))
return (1);
}
if (pf_status.debug >= PF_DEBUG_NOISY &&
(rpool->opts & PF_POOL_TYPEMASK) != PF_POOL_NONE) {
printf("pf_map_addr: selected address ");
pf_print_host(naddr, 0, af);
printf("\n");
}
return (0);
}
