gems-kernel/source/THIRDPARTY/xnu/bsd/skywalk/namespace/netns.c
2024-06-03 11:29:39 -05:00

2185 lines
57 KiB
C

/*
* Copyright (c) 2016-2021 Apple Inc. All rights reserved.
*
* @APPLE_OSREFERENCE_LICENSE_HEADER_START@
*
* This file contains Original Code and/or Modifications of Original Code
* as defined in and that are subject to the Apple Public Source License
* Version 2.0 (the 'License'). You may not use this file except in
* compliance with the License. The rights granted to you under the License
* may not be used to create, or enable the creation or redistribution of,
* unlawful or unlicensed copies of an Apple operating system, or to
* circumvent, violate, or enable the circumvention or violation of, any
* terms of an Apple operating system software license agreement.
*
* Please obtain a copy of the License at
* http://www.opensource.apple.com/apsl/ and read it before using this file.
*
* The Original Code and all software distributed under the License are
* distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
* EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
* INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
* Please see the License for the specific language governing rights and
* limitations under the License.
*
* @APPLE_OSREFERENCE_LICENSE_HEADER_END@
*/
#include <kern/assert.h>
#include <kern/locks.h>
#include <kern/zalloc.h>
#include <libkern/tree.h>
#include <sys/kernel.h>
#include <sys/sysctl.h>
#include <sys/bitstring.h>
#include <net/if.h>
#include <net/kpi_interface.h>
#include <net/restricted_in_port.h>
#include <netinet/in.h>
#include <netinet/in_pcb.h>
#include <netinet/tcp_fsm.h>
#include <netinet/tcp_var.h>
#include <netinet6/in6_var.h>
#include <string.h>
#include <skywalk/os_skywalk.h>
#include <skywalk/os_skywalk_private.h>
#include <skywalk/os_stats_private.h>
#include <skywalk/nexus/flowswitch/flow/flow_var.h>
#include <skywalk/nexus/flowswitch/nx_flowswitch.h>
#include <net/if_ports_used.h>
static int __netns_inited = 0;
/*
* Logging
*/
#define NS_VERB_PROTO(proto) ((proto == IPPROTO_TCP) ? SK_VERB_NS_TCP : \
SK_VERB_NS_UDP)
#define NS_VERB_IP(addr_len) ((addr_len == sizeof (struct in_addr)) ? \
SK_VERB_NS_IPV4 : SK_VERB_NS_IPV6)
#define PROTO_STR(proto) ((proto == IPPROTO_TCP) ? "tcp" : "udp")
#define LEN_TO_AF(len) (((len == sizeof (struct in_addr)) ? \
AF_INET : AF_INET6))
/*
* Locking
* Netns is currently protected by a global mutex, NETNS_LOCK. This lock is
* aquired at the entry of every kernel-facing function, and released at the
* end. Data within netns_token structures is also protected under this lock.
*/
#define NETNS_LOCK() \
lck_mtx_lock(&netns_lock)
#define NETNS_LOCK_SPIN() \
lck_mtx_lock_spin(&netns_lock)
#define NETNS_LOCK_CONVERT() do { \
NETNS_LOCK_ASSERT_HELD(); \
lck_mtx_convert_spin(&netns_lock); \
} while (0)
#define NETNS_UNLOCK() \
lck_mtx_unlock(&netns_lock)
#define NETNS_LOCK_ASSERT_HELD() \
LCK_MTX_ASSERT(&netns_lock, LCK_MTX_ASSERT_OWNED)
#define NETNS_LOCK_ASSERT_NOTHELD() \
LCK_MTX_ASSERT(&netns_lock, LCK_MTX_ASSERT_NOTOWNED)
static LCK_GRP_DECLARE(netns_lock_group, "netns_lock");
static LCK_MTX_DECLARE(netns_lock, &netns_lock_group);
/*
* Internal data structures and parameters
*/
/*
* Local ports are kept track of by reference counts kept in a tree specific to
* an <IP, protocol> tuple (see struct ns).
*
* Note: port numbers are stored in host byte order.
*/
struct ns_reservation {
RB_ENTRY(ns_reservation) nsr_link;
uint32_t nsr_refs[NETNS_OWNER_MAX + 1];
in_port_t nsr_port;
bool nsr_reuseport:1;
};
#define NETNS_REF_COUNT(nsr, flags) \
(nsr)->nsr_refs[((flags) & NETNS_OWNER_MASK)]
static inline int nsr_cmp(const struct ns_reservation *,
const struct ns_reservation *);
RB_HEAD(ns_reservation_tree, ns_reservation);
RB_PROTOTYPE(ns_reservation_tree, ns_reservation, nsr_link, nsr_cmp);
RB_GENERATE(ns_reservation_tree, ns_reservation, nsr_link, nsr_cmp);
static inline struct ns_reservation *ns_reservation_tree_find(
struct ns_reservation_tree *, const in_port_t);
/*
* A namespace keeps track of the local port numbers in use for a given
* <IP, protocol> tuple. There are also global namespaces for each
* protocol to accomodate INADDR_ANY behavior and diagnostics.
*/
struct ns {
RB_ENTRY(ns) ns_link;
void *ns_addr_key;
union {
uint32_t ns_addr[4];
struct in_addr ns_inaddr;
struct in6_addr ns_in6addr;
};
uint8_t ns_addr_len;
uint8_t ns_proto;
in_port_t ns_last_ephemeral_port_down;
in_port_t ns_last_ephemeral_port_up;
uint8_t ns_is_freeable;
uint32_t ns_n_reservations;
struct ns_reservation_tree ns_reservations;
};
static uint32_t netns_n_namespaces;
static inline int ns_cmp(const struct ns *, const struct ns *);
RB_HEAD(netns_namespaces_tree, ns) netns_namespaces =
RB_INITIALIZER(netns_namespaces);
RB_PROTOTYPE_PREV(netns_namespaces_tree, ns, ns_link, ns_cmp);
RB_GENERATE_PREV(netns_namespaces_tree, ns, ns_link, ns_cmp);
/*
* Declare pointers to global namespaces for each protocol.
* All non-wildcard reservations will have an entry here.
*/
#define NETNS_N_GLOBAL 4
static struct ns *netns_global_non_wild[NETNS_N_GLOBAL];
static struct ns *netns_global_wild[NETNS_N_GLOBAL];
#define NETNS_ADDRLEN_V4 (sizeof(struct in_addr))
#define NETNS_ADDRLEN_V6 (sizeof(struct in6_addr))
#define NETNS_NS_TCP 0
#define NETNS_NS_UDP 1
#define NETNS_NS_V4 0
#define NETNS_NS_V6 2
#define NETNS_NS_GLOBAL_IDX(proto, addrlen) \
((((proto) == IPPROTO_TCP) ? NETNS_NS_TCP : NETNS_NS_UDP) | \
(((addrlen) == NETNS_ADDRLEN_V4) ? NETNS_NS_V4 : NETNS_NS_V6))
#define NETNS_NS_UDP_EPHEMERAL_RESERVE 4096
/*
* Internal token structure
*
* Note: port numbers are stored in host byte order.
*/
struct ns_token {
/* Reservation state */
ifnet_t nt_ifp;
SLIST_ENTRY(ns_token) nt_ifp_link;
SLIST_ENTRY(ns_token) nt_all_link;
uint32_t nt_state; /* NETNS_STATE_* */
/* Reservation context */
union {
uint32_t nt_addr[4];
struct in_addr nt_inaddr;
struct in6_addr nt_in6addr;
};
uint8_t nt_addr_len;
uint8_t nt_proto;
in_port_t nt_port;
uint32_t nt_flags;
/* Optional information about the flow */
struct ns_flow_info *nt_flow_info;
};
/* Valid values for nt_state */
#define NETNS_STATE_HALFCLOSED 0x1 /* half closed */
#define NETNS_STATE_WITHDRAWN 0x2 /* withdrawn; not offloadable */
#define NETNS_STATE_BITS "\020\01HALFCLOSED\02WITHDRAWN"
/* List of tokens not bound to an ifnet */
SLIST_HEAD(, ns_token) netns_unbound_tokens = SLIST_HEAD_INITIALIZER(
netns_unbound_tokens);
/* List of all tokens currently allocated in the system */
SLIST_HEAD(, ns_token) netns_all_tokens = SLIST_HEAD_INITIALIZER(
netns_all_tokens);
/*
* Memory management
*/
static SKMEM_TYPE_DEFINE(netns_ns_zone, struct ns);
#define NETNS_NS_TOKEN_ZONE_NAME "netns.ns_token"
static unsigned int netns_ns_token_size; /* size of zone element */
static struct skmem_cache *netns_ns_token_cache; /* for ns_token */
#define NETNS_NS_FLOW_INFO_ZONE_NAME "netns.ns_flow_info"
static unsigned int netns_ns_flow_info_size; /* size of zone element */
static struct skmem_cache *netns_ns_flow_info_cache; /* for ns_flow_info */
#define NETNS_NS_RESERVATION_ZONE_NAME "netns.ns_reservation"
static unsigned int netns_ns_reservation_size; /* size of zone element */
static struct skmem_cache *netns_ns_reservation_cache; /* for ns_reservation */
static struct ns_reservation *netns_ns_reservation_alloc(in_port_t, uint32_t);
static void netns_ns_reservation_free(struct ns_reservation *);
static struct ns *netns_ns_alloc(zalloc_flags_t);
static void netns_ns_free(struct ns *);
static void netns_ns_cleanup(struct ns *);
static struct ns_token *netns_ns_token_alloc(boolean_t);
static void netns_ns_token_free(struct ns_token *);
/*
* Utility/internal code
*/
static struct ns *_netns_get_ns(uint32_t *, uint8_t, uint8_t, bool);
static inline boolean_t _netns_is_wildcard_addr(const uint32_t *, uint8_t);
static int _netns_reserve_common(struct ns *, in_port_t, uint32_t);
static void _netns_release_common(struct ns *, in_port_t, uint32_t);
static inline void netns_clear_ifnet(struct ns_token *);
static int _netns_reserve_kpi_common(struct ns *, netns_token *, uint32_t *,
uint8_t, uint8_t, in_port_t *, uint32_t, struct ns_flow_info *);
static void _netns_set_ifnet_internal(struct ns_token *, struct ifnet *);
static struct ns_reservation *
netns_ns_reservation_alloc(in_port_t port, uint32_t flags)
{
struct ns_reservation *res;
VERIFY(port != 0);
res = skmem_cache_alloc(netns_ns_reservation_cache, SKMEM_SLEEP);
ASSERT(res != NULL);
bzero(res, netns_ns_reservation_size);
res->nsr_port = port;
res->nsr_reuseport = ((flags & NETNS_REUSEPORT) != 0);
return res;
}
static void
netns_ns_reservation_free(struct ns_reservation *res)
{
skmem_cache_free(netns_ns_reservation_cache, res);
}
static struct ns *
netns_ns_alloc(zalloc_flags_t how)
{
struct ns *namespace;
in_port_t first = (in_port_t)ipport_firstauto;
in_port_t last = (in_port_t)ipport_lastauto;
in_port_t rand_port;
namespace = zalloc_flags(netns_ns_zone, how | Z_ZERO);
if (namespace == NULL) {
return NULL;
}
namespace->ns_is_freeable = 1;
RB_INIT(&namespace->ns_reservations);
/*
* Randomize the initial ephemeral port starting point, just in case
* this namespace is for an ipv6 address which gets brought up and
* down often.
*/
if (first == last) {
rand_port = first;
} else {
read_frandom(&rand_port, sizeof(rand_port));
if (first > last) {
rand_port = last + (rand_port % (first - last));
} else {
rand_port = first + (rand_port % (last - first));
}
}
namespace->ns_last_ephemeral_port_down = rand_port;
namespace->ns_last_ephemeral_port_up = rand_port;
return namespace;
}
static void
netns_ns_free(struct ns *namespace)
{
struct ns_reservation *res;
struct ns_reservation *tmp_res;
#if SK_LOG
char tmp_ip_str[MAX_IPv6_STR_LEN];
#endif /* SK_LOG */
SK_DF(NS_VERB_IP(namespace->ns_addr_len) |
NS_VERB_PROTO(namespace->ns_proto),
"freeing %s ns for IP %s",
PROTO_STR(namespace->ns_proto),
inet_ntop(LEN_TO_AF(namespace->ns_addr_len),
namespace->ns_addr, tmp_ip_str, sizeof(tmp_ip_str)));
RB_FOREACH_SAFE(res, ns_reservation_tree, &namespace->ns_reservations,
tmp_res) {
netns_ns_reservation_free(res);
namespace->ns_n_reservations--;
RB_REMOVE(ns_reservation_tree, &namespace->ns_reservations,
res);
}
VERIFY(RB_EMPTY(&namespace->ns_reservations));
if (netns_global_wild[NETNS_NS_GLOBAL_IDX(namespace->ns_proto,
namespace->ns_addr_len)] == namespace) {
netns_global_wild[NETNS_NS_GLOBAL_IDX(namespace->ns_proto,
namespace->ns_addr_len)] = NULL;
}
if (netns_global_non_wild[NETNS_NS_GLOBAL_IDX(namespace->ns_proto,
namespace->ns_addr_len)] == namespace) {
netns_global_non_wild[NETNS_NS_GLOBAL_IDX(namespace->ns_proto,
namespace->ns_addr_len)] = NULL;
}
zfree(netns_ns_zone, namespace);
}
static void
netns_ns_cleanup(struct ns *namespace)
{
if (namespace->ns_is_freeable &&
RB_EMPTY(&namespace->ns_reservations)) {
RB_REMOVE(netns_namespaces_tree, &netns_namespaces, namespace);
netns_n_namespaces--;
netns_ns_free(namespace);
}
}
static struct ns_token *
netns_ns_token_alloc(boolean_t with_nfi)
{
struct ns_token *token;
NETNS_LOCK_ASSERT_HELD();
NETNS_LOCK_CONVERT();
token = skmem_cache_alloc(netns_ns_token_cache, SKMEM_SLEEP);
ASSERT(token != NULL);
bzero(token, netns_ns_token_size);
if (with_nfi) {
token->nt_flow_info = skmem_cache_alloc(netns_ns_flow_info_cache,
SKMEM_SLEEP);
ASSERT(token->nt_flow_info != NULL);
}
SLIST_INSERT_HEAD(&netns_all_tokens, token, nt_all_link);
return token;
}
static void
netns_ns_token_free(struct ns_token *token)
{
NETNS_LOCK_ASSERT_HELD();
NETNS_LOCK_CONVERT();
SLIST_REMOVE(&netns_all_tokens, token, ns_token, nt_all_link);
if (token->nt_flow_info != NULL) {
skmem_cache_free(netns_ns_flow_info_cache, token->nt_flow_info);
}
skmem_cache_free(netns_ns_token_cache, token);
}
__attribute__((always_inline))
static inline int
nsr_cmp(const struct ns_reservation *nsr1, const struct ns_reservation *nsr2)
{
#define NSR_COMPARE(r1, r2) ((int)(r1)->nsr_port - (int)(r2)->nsr_port)
return NSR_COMPARE(nsr1, nsr2);
}
__attribute__((always_inline))
static inline int
ns_cmp(const struct ns *a, const struct ns *b)
{
int d;
if ((d = (a->ns_addr_len - b->ns_addr_len)) != 0) {
return d;
}
if ((d = (a->ns_proto - b->ns_proto)) != 0) {
return d;
}
if ((d = flow_ip_cmp(a->ns_addr_key, b->ns_addr_key,
b->ns_addr_len)) != 0) {
return d;
}
return 0;
}
/*
* Common routine to look up a reservation.
*
* NOTE: Assumes the caller holds the NETNS global lock
*/
__attribute__((always_inline))
static inline struct ns_reservation *
ns_reservation_tree_find(struct ns_reservation_tree *tree, const in_port_t port)
{
struct ns_reservation res;
res.nsr_port = port;
return RB_FIND(ns_reservation_tree, tree, &res);
}
/*
* Retrieve the namespace for the supplied <address, protocol> tuple.
* If create is set and such a namespace doesn't already exist, one will be
* created.
*/
static struct ns *
_netns_get_ns(uint32_t *addr, uint8_t addr_len, uint8_t proto, bool create)
{
struct ns *namespace = NULL;
struct ns find = {
.ns_addr_key = addr,
.ns_addr_len = addr_len,
.ns_proto = proto,
};
#if SK_LOG
char tmp_ip_str[MAX_IPv6_STR_LEN];
#endif /* SK_LOG */
VERIFY(addr_len == sizeof(struct in_addr) ||
addr_len == sizeof(struct in6_addr));
NETNS_LOCK_ASSERT_HELD();
namespace = RB_FIND(netns_namespaces_tree, &netns_namespaces, &find);
if (create && namespace == NULL) {
SK_DF(NS_VERB_IP(addr_len) | NS_VERB_PROTO(proto),
"allocating %s ns for IP %s",
PROTO_STR(proto), inet_ntop(LEN_TO_AF(addr_len), addr,
tmp_ip_str, sizeof(tmp_ip_str)));
NETNS_LOCK_CONVERT();
namespace = netns_ns_alloc(Z_WAITOK | Z_NOFAIL);
__builtin_assume(namespace != NULL);
memcpy(namespace->ns_addr, addr, addr_len);
namespace->ns_addr_key = &namespace->ns_addr;
namespace->ns_addr_len = addr_len;
namespace->ns_proto = proto;
RB_INSERT(netns_namespaces_tree, &netns_namespaces, namespace);
netns_n_namespaces++;
if (_netns_is_wildcard_addr(addr, addr_len) &&
netns_global_wild[NETNS_NS_GLOBAL_IDX(proto,
addr_len)] == NULL) {
netns_global_wild[NETNS_NS_GLOBAL_IDX(proto,
addr_len)] = namespace;
}
}
return namespace;
}
/*
* Return true if the supplied address is a wildcard (INADDR_ANY)
*/
__attribute__((always_inline))
static boolean_t
_netns_is_wildcard_addr(const uint32_t *addr, uint8_t addr_len)
{
boolean_t wildcard;
switch (addr_len) {
case sizeof(struct in_addr):
wildcard = (addr[0] == 0);
break;
case sizeof(struct in6_addr):
wildcard = (addr[0] == 0 && addr[1] == 0 &&
addr[2] == 0 && addr[3] == 0);
break;
default:
wildcard = FALSE;
break;
}
return wildcard;
}
__attribute__((always_inline))
static boolean_t
_netns_is_port_used(struct ns * gns, struct ns_reservation *curr_res, in_port_t port)
{
struct ns_reservation *res = NULL;
if (gns == NULL) {
return FALSE;
}
res = ns_reservation_tree_find(&gns->ns_reservations, port);
if (res != NULL && res != curr_res) {
if (!res->nsr_reuseport) {
return TRUE;
}
}
return FALSE;
}
/*
* Internal shared code to reserve ports within a specific namespace.
*
* Note: port numbers are in host byte-order here.
*/
static int
_netns_reserve_common(struct ns *namespace, in_port_t port, uint32_t flags)
{
struct ns_reservation *res = NULL, *exist = NULL;
uint8_t proto, addr_len;
int err = 0;
#if SK_LOG
char tmp_ip_str[MAX_IPv6_STR_LEN];
#endif /* SK_LOG */
VERIFY(port != 0);
proto = namespace->ns_proto;
addr_len = namespace->ns_addr_len;
NETNS_LOCK_CONVERT();
res = netns_ns_reservation_alloc(port, flags);
if (res == NULL) {
SK_DF(NS_VERB_IP(addr_len) | NS_VERB_PROTO(proto),
"ERROR %s:%s:%d // flags 0x%x // OUT OF MEMORY",
inet_ntop(LEN_TO_AF(namespace->ns_addr_len),
namespace->ns_addr, tmp_ip_str,
sizeof(tmp_ip_str)), PROTO_STR(proto), port, flags);
return ENOMEM;
}
exist = RB_INSERT(ns_reservation_tree, &namespace->ns_reservations,
res);
if (__probable(exist == NULL)) {
namespace->ns_n_reservations++;
} else {
netns_ns_reservation_free(res);
res = exist;
}
SK_DF(NS_VERB_IP(addr_len) | NS_VERB_PROTO(proto),
"pre: %s:%s:%d // flags 0x%x // refs %d sky, %d ls, "
"%d bsd %d pf", inet_ntop(LEN_TO_AF(namespace->ns_addr_len),
namespace->ns_addr, tmp_ip_str, sizeof(tmp_ip_str)),
PROTO_STR(proto), port, flags,
NETNS_REF_COUNT(res, NETNS_SKYWALK),
NETNS_REF_COUNT(res, NETNS_LISTENER),
NETNS_REF_COUNT(res, NETNS_BSD),
NETNS_REF_COUNT(res, NETNS_PF));
/* Make reservation */
/*
* Bypass collision detection for reservations in the global non-wild
* namespace. We use that namespace for reference counts only.
*/
if (namespace !=
netns_global_non_wild[NETNS_NS_GLOBAL_IDX(proto, addr_len)]) {
struct ns_reservation *skres;
boolean_t is_wild = _netns_is_wildcard_addr(namespace->ns_addr,
addr_len);
struct ns *gns =
netns_global_wild[NETNS_NS_GLOBAL_IDX(proto, addr_len)];
if (NETNS_IS_SKYWALK(flags)) {
if ((!is_wild || exist != NULL) && gns != NULL &&
(skres = ns_reservation_tree_find(
&gns->ns_reservations, port)) != NULL &&
NETNS_REF_COUNT(skres, NETNS_LISTENER) == 0) {
/*
* The mere existence of any non-skywalk
* listener wildcard entry for this
* protocol/port number means this must fail.
*/
SK_ERR("ADDRINUSE: Duplicate wildcard");
err = EADDRINUSE;
goto done;
}
if (is_wild) {
gns = netns_global_non_wild[
NETNS_NS_GLOBAL_IDX(proto, addr_len)];
VERIFY(gns != NULL);
if (_netns_is_port_used(netns_global_non_wild[
NETNS_NS_GLOBAL_IDX(proto, NETNS_ADDRLEN_V4)], res, port) ||
_netns_is_port_used(netns_global_non_wild[
NETNS_NS_GLOBAL_IDX(proto, NETNS_ADDRLEN_V6)], res, port)) {
/*
* If Skywalk is trying to reserve a
* wildcard, then the mere existance of
* any entry in either v4/v6 non-wild
* namespace for this port means this
* must fail.
*/
SK_ERR("ADDRINUSE: Wildcard with non-wild.");
err = EADDRINUSE;
goto done;
}
}
} else {
/*
* Check if Skywalk has reserved a wildcard entry.
* Note that the arithmetic OR here is intentional.
*/
if ((!is_wild || exist != NULL) && gns != NULL &&
(skres = ns_reservation_tree_find(
&gns->ns_reservations, port)) != NULL &&
(NETNS_REF_COUNT(skres, NETNS_SKYWALK) |
NETNS_REF_COUNT(skres, NETNS_LISTENER)) != 0) {
/*
* BSD is trying to reserve a proto/port for
* which Skywalk already has a wildcard
* reservation.
*/
SK_ERR("ADDRINUSE: BSD requesting Skywalk port");
err = EADDRINUSE;
goto done;
}
/*
* If BSD is trying to reserve a wildcard,
* ensure Skywalk has not already reserved
* a non-wildcard.
*/
if (is_wild) {
gns = netns_global_non_wild[
NETNS_NS_GLOBAL_IDX(proto, addr_len)];
VERIFY(gns != NULL);
/*
* Note that the arithmetic OR here is
* intentional.
*/
if ((skres = ns_reservation_tree_find(
&gns->ns_reservations, port)) != NULL &&
(NETNS_REF_COUNT(skres, NETNS_SKYWALK) |
NETNS_REF_COUNT(skres,
NETNS_LISTENER)) != 0) {
SK_ERR("ADDRINUSE: BSD wildcard with non-wild.");
err = EADDRINUSE;
goto done;
}
}
}
switch (flags & NETNS_OWNER_MASK) {
case NETNS_SKYWALK:
/* check collision w/ BSD */
if (NETNS_REF_COUNT(res, NETNS_BSD) > 0 ||
NETNS_REF_COUNT(res, NETNS_PF) > 0) {
SK_ERR("ERROR - Skywalk got ADDRINUSE (w/ BSD)");
err = EADDRINUSE;
goto done;
}
/* BEGIN CSTYLED */
/*
* Scenarios with new Skywalk connected flow:
* 1. With existing Skywalk connected flow,
* NETNS_REF_COUNT(res, NETNS_LISTENER) == 0 &&
* NETNS_REF_COUNT(res, NETNS_SKYWALK) == 1
* reject by failing the wild gns lookup below.
* 2. With existing Skywalk 3-tuple listener,
* NETNS_REF_COUNT(res, NETNS_LISTENER) == 1
* bypass the check below.
* 3. With existing Skywalk 2-tuple listener,
* NETNS_REF_COUNT(res, NETNS_LISTENER) == 0 &&
* NETNS_REF_COUNT(res, NETNS_SKYWALK) == 0
* pass with successful wild gns lookup.
*/
/* END CSTYLED */
if (NETNS_REF_COUNT(res, NETNS_LISTENER) == 0 &&
NETNS_REF_COUNT(res, NETNS_SKYWALK) > 0) {
/* check if covered by wild Skywalk listener */
gns = netns_global_wild[
NETNS_NS_GLOBAL_IDX(proto, addr_len)];
if (gns != NULL &&
(skres = ns_reservation_tree_find(
&gns->ns_reservations, port)) != NULL &&
NETNS_REF_COUNT(skres, NETNS_LISTENER)
!= 0) {
err = 0;
goto done;
}
if (addr_len == sizeof(struct in_addr)) {
/* If address is IPv4, also check for wild IPv6 registration */
gns = netns_global_wild[
NETNS_NS_GLOBAL_IDX(proto, NETNS_ADDRLEN_V6)];
if (gns != NULL &&
(skres = ns_reservation_tree_find(
&gns->ns_reservations, port)) != NULL &&
NETNS_REF_COUNT(skres, NETNS_LISTENER)
!= 0) {
err = 0;
goto done;
}
}
SK_ERR("ERROR - Skywalk got ADDRINUSE (w/ SK connected flow)");
err = EADDRINUSE;
}
/*
* XXX: Duplicate 5-tuple flows under a Skywalk
* listener are currently detected by flow manager,
* till we implement 5-tuple-aware netns.
*/
break;
case NETNS_LISTENER:
if (NETNS_REF_COUNT(res, NETNS_BSD) > 0 ||
NETNS_REF_COUNT(res, NETNS_PF) > 0 ||
NETNS_REF_COUNT(res, NETNS_LISTENER) > 0 ||
_netns_is_port_used(netns_global_wild[
NETNS_NS_GLOBAL_IDX(proto, NETNS_ADDRLEN_V4)], res, port) ||
_netns_is_port_used(netns_global_wild[
NETNS_NS_GLOBAL_IDX(proto, NETNS_ADDRLEN_V6)], res, port) ||
_netns_is_port_used(netns_global_non_wild[
NETNS_NS_GLOBAL_IDX(proto, NETNS_ADDRLEN_V4)], res, port) ||
_netns_is_port_used(netns_global_non_wild[
NETNS_NS_GLOBAL_IDX(proto, NETNS_ADDRLEN_V6)], res, port)) {
SK_ERR("ERROR - Listener got ADDRINUSE");
err = EADDRINUSE;
}
break;
case NETNS_BSD:
case NETNS_PF:
if (NETNS_REF_COUNT(res, NETNS_SKYWALK) > 0 ||
NETNS_REF_COUNT(res, NETNS_LISTENER) > 0) {
SK_ERR("ERROR - %s got ADDRINUSE",
((flags & NETNS_OWNER_MASK) == NETNS_PF) ?
"PF" : "BSD");
err = EADDRINUSE;
}
break;
default:
panic("_netns_reserve_common: invalid owner 0x%x",
flags & NETNS_OWNER_MASK);
/* NOTREACHED */
__builtin_unreachable();
}
}
done:
ASSERT(res != NULL);
if (__probable(err == 0)) {
NETNS_REF_COUNT(res, flags)++;
/* Check for wrap around */
VERIFY(NETNS_REF_COUNT(res, flags) != 0);
SK_DF(NS_VERB_IP(namespace->ns_addr_len) |
NS_VERB_PROTO(namespace->ns_proto),
"post: %s:%s:%d err %d // flags 0x%x // refs %d sky, "
"%d ls, %d bsd %d pf",
inet_ntop(LEN_TO_AF(namespace->ns_addr_len),
namespace->ns_addr, tmp_ip_str, sizeof(tmp_ip_str)),
PROTO_STR(namespace->ns_proto), port, err, flags,
NETNS_REF_COUNT(res, NETNS_SKYWALK),
NETNS_REF_COUNT(res, NETNS_LISTENER),
NETNS_REF_COUNT(res, NETNS_BSD),
NETNS_REF_COUNT(res, NETNS_PF));
} else {
if (exist == NULL) {
RB_REMOVE(ns_reservation_tree,
&namespace->ns_reservations, res);
namespace->ns_n_reservations--;
netns_ns_reservation_free(res);
}
}
return err;
}
/*
* Internal shared code to release ports within a specific namespace.
*/
static void
_netns_release_common(struct ns *namespace, in_port_t port, uint32_t flags)
{
struct ns_reservation *res;
uint32_t refs;
int i;
#if SK_LOG
char tmp_ip_str[MAX_IPv6_STR_LEN];
#endif /* SK_LOG */
NETNS_LOCK_ASSERT_HELD();
res = ns_reservation_tree_find(&namespace->ns_reservations, port);
if (res == NULL) {
SK_DF(NS_VERB_IP(namespace->ns_addr_len) |
NS_VERB_PROTO(namespace->ns_proto),
"ERROR %s:%s:%d // flags 0x%x // not found",
inet_ntop(LEN_TO_AF(namespace->ns_addr_len),
namespace->ns_addr, tmp_ip_str, sizeof(tmp_ip_str)),
PROTO_STR(namespace->ns_proto), port, flags);
VERIFY(res != NULL);
}
SK_DF(NS_VERB_IP(namespace->ns_addr_len) |
NS_VERB_PROTO(namespace->ns_proto),
"%s:%s:%d // flags 0x%x // refs %d sky, %d ls, %d bsd, %d pf",
inet_ntop(LEN_TO_AF(namespace->ns_addr_len),
namespace->ns_addr, tmp_ip_str, sizeof(tmp_ip_str)),
PROTO_STR(namespace->ns_proto), port, flags,
NETNS_REF_COUNT(res, NETNS_SKYWALK),
NETNS_REF_COUNT(res, NETNS_LISTENER),
NETNS_REF_COUNT(res, NETNS_BSD),
NETNS_REF_COUNT(res, NETNS_PF));
/* Release reservation */
VERIFY(NETNS_REF_COUNT(res, flags) > 0);
NETNS_REF_COUNT(res, flags) -= 1;
/* Clean up memory, if appropriate */
for (i = 0, refs = 0; i <= NETNS_OWNER_MAX && refs == 0; i++) {
refs |= res->nsr_refs[i];
}
if (refs == 0) {
RB_REMOVE(ns_reservation_tree, &namespace->ns_reservations,
res);
namespace->ns_n_reservations--;
NETNS_LOCK_CONVERT();
netns_ns_reservation_free(res);
netns_ns_cleanup(namespace);
}
}
__attribute__((always_inline))
static inline void
netns_init_global_ns(struct ns **global_ptr, uint8_t proto, uint8_t addrlen)
{
struct ns *namespace;
namespace = *global_ptr = netns_ns_alloc(Z_WAITOK);
memset(namespace->ns_addr, 0xFF, addrlen);
namespace->ns_addr_len = addrlen;
namespace->ns_proto = proto;
namespace->ns_is_freeable = 0;
}
__attribute__((always_inline))
static inline void
netns_clear_ifnet(struct ns_token *nstoken)
{
#if SK_LOG
char tmp_ip_str[MAX_IPv6_STR_LEN];
#endif /* SK_LOG */
NETNS_LOCK_ASSERT_HELD();
if (nstoken->nt_ifp != NULL) {
SLIST_REMOVE(&nstoken->nt_ifp->if_netns_tokens, nstoken,
ns_token, nt_ifp_link);
SK_DF(NS_VERB_IP(nstoken->nt_addr_len) |
NS_VERB_PROTO(nstoken->nt_proto),
"%s:%s:%d // removed from ifnet %d",
inet_ntop(LEN_TO_AF(nstoken->nt_addr_len),
nstoken->nt_addr, tmp_ip_str, sizeof(tmp_ip_str)),
PROTO_STR(nstoken->nt_proto), nstoken->nt_port,
nstoken->nt_ifp->if_index);
NETNS_LOCK_CONVERT();
ifnet_decr_iorefcnt(nstoken->nt_ifp);
nstoken->nt_ifp = NULL;
} else {
SLIST_REMOVE(&netns_unbound_tokens, nstoken, ns_token,
nt_ifp_link);
}
}
/*
* Internal shared code to perform a port[-range] reservation, along with all
* the boilerplate and sanity checks expected for a call coming in from the
* surrounding kernel code.
*/
static int
_netns_reserve_kpi_common(struct ns *ns, netns_token *token, uint32_t *addr,
uint8_t addr_len, uint8_t proto, in_port_t *port, uint32_t flags,
struct ns_flow_info *nfi)
{
boolean_t ns_want_cleanup = (ns == NULL);
struct ns_token *nt;
int err = 0;
in_port_t hport;
#if SK_LOG
char tmp_ip_str[MAX_IPv6_STR_LEN];
#endif /* SK_LOG */
struct ifnet *ifp = (nfi != NULL) ? nfi->nfi_ifp : NULL;
NETNS_LOCK_ASSERT_HELD();
hport = ntohs(*port);
VERIFY((flags & NETNS_OWNER_MASK) <= NETNS_OWNER_MAX);
VERIFY(addr_len == sizeof(struct in_addr) ||
addr_len == sizeof(struct in6_addr));
VERIFY(proto == IPPROTO_TCP || proto == IPPROTO_UDP);
VERIFY(hport != 0);
SK_DF(NS_VERB_IP(addr_len) | NS_VERB_PROTO(proto),
"reserving %s:%s:%d // flags 0x%x // token %svalid",
inet_ntop(LEN_TO_AF(addr_len), addr, tmp_ip_str,
sizeof(tmp_ip_str)), PROTO_STR(proto), hport, flags,
NETNS_TOKEN_VALID(token) ? "" : "in");
/*
* See the documentation for NETNS_PRERESERVED in netns.h for an
* explanation of this block.
*/
if (NETNS_TOKEN_VALID(token)) {
if (flags & NETNS_PRERESERVED) {
nt = *token;
VERIFY(nt->nt_addr_len == addr_len);
VERIFY(memcmp(nt->nt_addr, addr, addr_len) == 0);
VERIFY(nt->nt_proto == proto);
VERIFY(nt->nt_port == hport);
VERIFY((nt->nt_flags &
NETNS_RESERVATION_FLAGS | NETNS_PRERESERVED) ==
(flags & NETNS_RESERVATION_FLAGS));
if ((nt->nt_flags & NETNS_CONFIGURATION_FLAGS) ==
(flags & NETNS_CONFIGURATION_FLAGS)) {
SK_DF(NS_VERB_IP(nt->nt_addr_len) |
NS_VERB_PROTO(nt->nt_proto),
"%s:%s:%d // flags 0x%x -> 0x%x",
inet_ntop(LEN_TO_AF(nt->nt_addr_len),
nt->nt_addr, tmp_ip_str,
sizeof(tmp_ip_str)),
PROTO_STR(nt->nt_proto),
nt->nt_port, nt->nt_flags, flags);
nt->nt_flags &= ~NETNS_CONFIGURATION_FLAGS;
nt->nt_flags |=
flags & NETNS_CONFIGURATION_FLAGS;
}
SK_DF(NS_VERB_IP(addr_len) | NS_VERB_PROTO(proto),
"token was prereserved");
goto done;
} else {
panic("Request to overwrite valid netns token");
/* NOTREACHED */
__builtin_unreachable();
}
}
/*
* TODO: Check range against bitmap
*/
if (hport == 0) {
/*
* Caller request an arbitrary range of ports
* TODO: Need to figure out how to allocate
* emphemeral ports only.
*/
SK_DF(NS_VERB_IP(addr_len) | NS_VERB_PROTO(proto),
"ERROR - wildcard port not yet supported");
err = ENOMEM;
goto done;
}
/*
* Fetch namespace for the specified address/protocol, creating
* a new namespace if necessary.
*/
if (ns == NULL) {
ASSERT(ns_want_cleanup);
ns = _netns_get_ns(addr, addr_len, proto, true);
}
if (__improbable(ns == NULL)) {
SK_DF(NS_VERB_IP(addr_len) | NS_VERB_PROTO(proto),
"ERROR - couldn't create namespace");
err = ENOMEM;
goto done;
}
/*
* Make a reservation in the namespace
* This will return an error if an incompatible reservation
* already exists.
*/
err = _netns_reserve_common(ns, hport, flags);
if (__improbable(err != 0)) {
NETNS_LOCK_CONVERT();
if (ns_want_cleanup) {
netns_ns_cleanup(ns);
}
SK_DF(NS_VERB_IP(addr_len) | NS_VERB_PROTO(proto),
"ERROR - reservation collision");
goto done;
}
if (!_netns_is_wildcard_addr(ns->ns_addr, addr_len)) {
/* Record the reservation in the non-wild namespace */
struct ns *nwns;
nwns = netns_global_non_wild[NETNS_NS_GLOBAL_IDX(proto,
addr_len)];
err = _netns_reserve_common(nwns, hport, flags);
if (__improbable(err != 0)) {
/* Need to free the specific namespace entry */
NETNS_LOCK_CONVERT();
_netns_release_common(ns, hport, flags);
if (ns_want_cleanup) {
netns_ns_cleanup(ns);
}
SK_DF(NS_VERB_IP(addr_len) | NS_VERB_PROTO(proto),
"ERROR - reservation collision");
goto done;
}
}
nt = netns_ns_token_alloc(nfi != NULL ? true : false);
ASSERT(nt->nt_ifp == NULL);
_netns_set_ifnet_internal(nt, ifp);
memcpy(nt->nt_addr, addr, addr_len);
nt->nt_addr_len = addr_len;
nt->nt_proto = proto;
nt->nt_port = hport;
nt->nt_flags = flags;
if (nfi != NULL) {
VERIFY(nt->nt_flow_info != NULL);
memcpy(nt->nt_flow_info, nfi, sizeof(struct ns_flow_info));
/*
* The local port is passed as a separate argument
*/
if (nfi->nfi_laddr.sa.sa_family == AF_INET) {
nt->nt_flow_info->nfi_laddr.sin.sin_port = *port;
} else if (nfi->nfi_laddr.sa.sa_family == AF_INET6) {
nt->nt_flow_info->nfi_laddr.sin6.sin6_port = *port;
}
}
*token = nt;
done:
return err;
}
/*
* Kernel-facing functions
*/
int
netns_init(void)
{
VERIFY(__netns_inited == 0);
netns_ns_reservation_size = sizeof(struct ns_reservation);
netns_ns_reservation_cache = skmem_cache_create(NETNS_NS_RESERVATION_ZONE_NAME,
netns_ns_reservation_size, sizeof(uint64_t), NULL, NULL, NULL,
NULL, NULL, 0);
if (netns_ns_reservation_cache == NULL) {
panic("%s: skmem_cache create failed (%s)", __func__,
NETNS_NS_RESERVATION_ZONE_NAME);
/* NOTREACHED */
__builtin_unreachable();
}
netns_ns_token_size = sizeof(struct ns_token);
netns_ns_token_cache = skmem_cache_create(NETNS_NS_TOKEN_ZONE_NAME,
netns_ns_token_size, sizeof(uint64_t), NULL, NULL, NULL, NULL,
NULL, 0);
if (netns_ns_token_cache == NULL) {
panic("%s: skmem_cache create failed (%s)", __func__,
NETNS_NS_TOKEN_ZONE_NAME);
/* NOTREACHED */
__builtin_unreachable();
}
netns_ns_flow_info_size = sizeof(struct ns_flow_info);
netns_ns_flow_info_cache = skmem_cache_create(NETNS_NS_FLOW_INFO_ZONE_NAME,
netns_ns_flow_info_size, sizeof(uint64_t), NULL, NULL, NULL,
NULL, NULL, 0);
if (netns_ns_flow_info_cache == NULL) {
panic("%s: skmem_cache create failed (%s)", __func__,
NETNS_NS_FLOW_INFO_ZONE_NAME);
/* NOTREACHED */
__builtin_unreachable();
}
SLIST_INIT(&netns_unbound_tokens);
SLIST_INIT(&netns_all_tokens);
netns_n_namespaces = 0;
RB_INIT(&netns_namespaces);
SK_D("initializing global namespaces");
netns_init_global_ns(
&netns_global_non_wild[NETNS_NS_GLOBAL_IDX(IPPROTO_TCP,
NETNS_ADDRLEN_V4)], IPPROTO_TCP, sizeof(struct in_addr));
netns_init_global_ns(
&netns_global_non_wild[NETNS_NS_GLOBAL_IDX(IPPROTO_UDP,
NETNS_ADDRLEN_V4)], IPPROTO_UDP, sizeof(struct in_addr));
netns_init_global_ns(
&netns_global_non_wild[NETNS_NS_GLOBAL_IDX(IPPROTO_TCP,
NETNS_ADDRLEN_V6)], IPPROTO_TCP, sizeof(struct in6_addr));
netns_init_global_ns(
&netns_global_non_wild[NETNS_NS_GLOBAL_IDX(IPPROTO_UDP,
NETNS_ADDRLEN_V6)], IPPROTO_UDP, sizeof(struct in6_addr));
/* Done */
__netns_inited = 1;
sk_features |= SK_FEATURE_NETNS;
SK_D("initialized netns");
return 0;
}
void
netns_uninit(void)
{
if (__netns_inited == 1) {
struct ns *namespace;
struct ns *temp_namespace;
int i;
RB_FOREACH_SAFE(namespace, netns_namespaces_tree,
&netns_namespaces, temp_namespace) {
RB_REMOVE(netns_namespaces_tree, &netns_namespaces,
namespace);
netns_n_namespaces--;
netns_ns_free(namespace);
}
for (i = 0; i < NETNS_N_GLOBAL; i++) {
netns_ns_free(netns_global_non_wild[i]);
}
if (netns_ns_flow_info_cache != NULL) {
skmem_cache_destroy(netns_ns_flow_info_cache);
netns_ns_flow_info_cache = NULL;
}
if (netns_ns_token_cache != NULL) {
skmem_cache_destroy(netns_ns_token_cache);
netns_ns_token_cache = NULL;
}
if (netns_ns_reservation_cache != NULL) {
skmem_cache_destroy(netns_ns_reservation_cache);
netns_ns_reservation_cache = NULL;
}
__netns_inited = 0;
sk_features &= ~SK_FEATURE_NETNS;
SK_D("uninitialized netns");
}
}
void
netns_reap_caches(boolean_t purge)
{
/* these aren't created unless netns is enabled */
if (netns_ns_token_cache != NULL) {
skmem_cache_reap_now(netns_ns_token_cache, purge);
}
if (netns_ns_reservation_cache != NULL) {
skmem_cache_reap_now(netns_ns_reservation_cache, purge);
}
if (netns_ns_flow_info_cache != NULL) {
skmem_cache_reap_now(netns_ns_flow_info_cache, purge);
}
}
boolean_t
netns_is_enabled(void)
{
return __netns_inited == 1;
}
int
netns_reserve(netns_token *token, uint32_t *addr, uint8_t addr_len,
uint8_t proto, in_port_t port, uint32_t flags, struct ns_flow_info *nfi)
{
int err = 0;
#if SK_LOG
char tmp_ip_str[MAX_IPv6_STR_LEN];
#endif /* SK_LOG */
if (__netns_inited == 0) {
*token = NULL;
return err;
}
if (proto != IPPROTO_TCP && proto != IPPROTO_UDP) {
SK_ERR("netns doesn't support non TCP/UDP protocol");
return ENOTSUP;
}
SK_DF(NS_VERB_IP(addr_len) | NS_VERB_PROTO(proto),
"%s:%s:%d // flags 0x%x", inet_ntop(LEN_TO_AF(addr_len), addr,
tmp_ip_str, sizeof(tmp_ip_str)), PROTO_STR(proto), ntohs(port),
flags);
/*
* Check wether the process is allowed to bind to a restricted port
*/
if (!current_task_can_use_restricted_in_port(port,
proto, flags)) {
*token = NULL;
return EADDRINUSE;
}
NETNS_LOCK_SPIN();
err = _netns_reserve_kpi_common(NULL, token, addr, addr_len,
proto, &port, flags, nfi);
NETNS_UNLOCK();
return err;
}
/* Import net.inet.{tcp,udp}.randomize_ports sysctls */
extern int udp_use_randomport;
extern int tcp_use_randomport;
int
netns_reserve_ephemeral(netns_token *token, uint32_t *addr, uint8_t addr_len,
uint8_t proto, in_port_t *port, uint32_t flags, struct ns_flow_info *nfi)
{
int err = 0;
in_port_t first = (in_port_t)ipport_firstauto;
in_port_t last = (in_port_t)ipport_lastauto;
in_port_t rand_port;
in_port_t last_port;
in_port_t n_last_port;
struct ns *namespace;
boolean_t count_up = true;
boolean_t use_randomport = (proto == IPPROTO_TCP) ?
tcp_use_randomport : udp_use_randomport;
#if SK_LOG
char tmp_ip_str[MAX_IPv6_STR_LEN];
#endif /* SK_LOG */
if (__netns_inited == 0) {
*token = NULL;
return err;
}
if (proto != IPPROTO_TCP && proto != IPPROTO_UDP) {
SK_ERR("netns doesn't support non TCP/UDP protocol");
return ENOTSUP;
}
SK_DF(NS_VERB_IP(addr_len) | NS_VERB_PROTO(proto),
"%s:%s:%d // flags 0x%x", inet_ntop(LEN_TO_AF(addr_len), addr,
tmp_ip_str, sizeof(tmp_ip_str)), PROTO_STR(proto), ntohs(*port),
flags);
NETNS_LOCK_SPIN();
namespace = _netns_get_ns(addr, addr_len, proto, true);
if (namespace == NULL) {
err = ENOMEM;
NETNS_UNLOCK();
return err;
}
if (proto == IPPROTO_UDP) {
if (UINT16_MAX - namespace->ns_n_reservations <
NETNS_NS_UDP_EPHEMERAL_RESERVE) {
SK_ERR("UDP ephemeral port not available"
"(less than 4096 UDP ports left)");
err = EADDRNOTAVAIL;
NETNS_UNLOCK();
return err;
}
}
if (first == last) {
rand_port = first;
} else {
if (use_randomport) {
NETNS_LOCK_CONVERT();
read_frandom(&rand_port, sizeof(rand_port));
if (first > last) {
rand_port = last + (rand_port %
(first - last));
count_up = false;
} else {
rand_port = first + (rand_port %
(last - first));
}
} else {
if (first > last) {
rand_port =
namespace->ns_last_ephemeral_port_down - 1;
if (rand_port < last || rand_port > first) {
rand_port = last;
}
count_up = false;
} else {
rand_port =
namespace->ns_last_ephemeral_port_up + 1;
if (rand_port < first || rand_port > last) {
rand_port = first;
}
}
}
}
last_port = rand_port;
n_last_port = htons(last_port);
while (true) {
if (n_last_port == 0) {
SK_ERR("ephemeral port search range includes 0");
err = EINVAL;
break;
}
/*
* Skip if this is a restricted port as we do not want to
* restricted ports as ephemeral
*/
if (!IS_RESTRICTED_IN_PORT(n_last_port)) {
err = _netns_reserve_kpi_common(namespace, token, addr,
addr_len, proto, &n_last_port, flags, nfi);
if (err == 0 || err != EADDRINUSE) {
break;
}
}
if (count_up) {
last_port++;
if (last_port < first || last_port > last) {
last_port = first;
}
} else {
last_port--;
if (last_port < last || last_port > first) {
last_port = last;
}
}
n_last_port = htons(last_port);
if (last_port == rand_port || first == last) {
SK_ERR("couldn't find free ephemeral port");
err = EADDRNOTAVAIL;
break;
}
}
if (err == 0) {
*port = n_last_port;
if (count_up) {
namespace->ns_last_ephemeral_port_up = last_port;
} else {
namespace->ns_last_ephemeral_port_down = last_port;
}
} else {
netns_ns_cleanup(namespace);
}
NETNS_UNLOCK();
return err;
}
void
netns_release(netns_token *token)
{
struct ns *ns;
struct ns_token *nt;
uint8_t proto, addr_len;
#if SK_LOG
char tmp_ip_str[MAX_IPv6_STR_LEN];
#endif /* SK_LOG */
if (!NETNS_TOKEN_VALID(token)) {
return;
}
if (__netns_inited == 0) {
*token = NULL;
return;
}
NETNS_LOCK_SPIN();
nt = *token;
*token = NULL;
VERIFY((nt->nt_flags & NETNS_OWNER_MASK) <= NETNS_OWNER_MAX);
VERIFY(nt->nt_addr_len == sizeof(struct in_addr) ||
nt->nt_addr_len == sizeof(struct in6_addr));
VERIFY(nt->nt_proto == IPPROTO_TCP || nt->nt_proto == IPPROTO_UDP);
addr_len = nt->nt_addr_len;
proto = nt->nt_proto;
SK_DF(NS_VERB_IP(addr_len) | NS_VERB_PROTO(proto),
"releasing %s:%s:%d",
inet_ntop(LEN_TO_AF(nt->nt_addr_len), nt->nt_addr,
tmp_ip_str, sizeof(tmp_ip_str)), PROTO_STR(proto),
nt->nt_port);
if (!_netns_is_wildcard_addr(nt->nt_addr, addr_len)) {
/* Remove from global non-wild namespace */
ns = netns_global_non_wild[NETNS_NS_GLOBAL_IDX(proto,
addr_len)];
VERIFY(ns != NULL);
_netns_release_common(ns, nt->nt_port, nt->nt_flags);
}
ns = _netns_get_ns(nt->nt_addr, addr_len, proto, false);
VERIFY(ns != NULL);
_netns_release_common(ns, nt->nt_port, nt->nt_flags);
netns_clear_ifnet(nt);
netns_ns_token_free(nt);
NETNS_UNLOCK();
}
int
netns_change_addr(netns_token *token, uint32_t *addr, uint8_t addr_len)
{
int err = 0;
struct ns *old_namespace;
struct ns *new_namespace;
struct ns *global_namespace;
struct ns_token *nt;
uint8_t proto;
#if SK_LOG
char tmp_ip_str_1[MAX_IPv6_STR_LEN];
char tmp_ip_str_2[MAX_IPv6_STR_LEN];
#endif /* SK_LOG */
if (__netns_inited == 0) {
return 0;
}
NETNS_LOCK();
VERIFY(NETNS_TOKEN_VALID(token));
nt = *token;
VERIFY((nt->nt_flags & NETNS_OWNER_MASK) == NETNS_BSD);
VERIFY(nt->nt_addr_len == sizeof(struct in_addr) ||
nt->nt_addr_len == sizeof(struct in6_addr));
VERIFY(nt->nt_proto == IPPROTO_TCP || nt->nt_proto == IPPROTO_UDP);
proto = nt->nt_proto;
#if SK_LOG
inet_ntop(LEN_TO_AF(nt->nt_addr_len), nt->nt_addr,
tmp_ip_str_1, sizeof(tmp_ip_str_1));
inet_ntop(LEN_TO_AF(addr_len), addr, tmp_ip_str_2,
sizeof(tmp_ip_str_2));
#endif /* SK_LOG */
SK_DF(NS_VERB_IP(addr_len) | NS_VERB_PROTO(proto),
"changing address for %s:%d from %s to %s",
PROTO_STR(proto), nt->nt_port, tmp_ip_str_1,
tmp_ip_str_2);
if (nt->nt_addr_len == addr_len &&
memcmp(nt->nt_addr, addr, nt->nt_addr_len) == 0) {
SK_DF(NS_VERB_IP(addr_len) | NS_VERB_PROTO(proto),
"address didn't change, exiting early");
goto done;
}
old_namespace = _netns_get_ns(nt->nt_addr, nt->nt_addr_len, proto,
false);
VERIFY(old_namespace != NULL);
new_namespace = _netns_get_ns(addr, addr_len, proto, true);
if (new_namespace == NULL) {
err = ENOMEM;
goto done;
}
/* Acquire reservation in new namespace */
if ((err = _netns_reserve_common(new_namespace, nt->nt_port,
nt->nt_flags))) {
NETNS_LOCK_CONVERT();
netns_ns_cleanup(new_namespace);
SK_ERR("ERROR - reservation collision under new namespace");
goto done;
}
/* Release from old namespace */
_netns_release_common(old_namespace, nt->nt_port, nt->nt_flags);
if (!_netns_is_wildcard_addr(nt->nt_addr, nt->nt_addr_len)) {
/*
* Old address is non-wildcard.
* Remove old reservation from global non-wild namespace
*/
global_namespace = netns_global_non_wild[
NETNS_NS_GLOBAL_IDX(proto, nt->nt_addr_len)];
VERIFY(global_namespace != NULL);
_netns_release_common(global_namespace, nt->nt_port,
nt->nt_flags);
}
if (!_netns_is_wildcard_addr(addr, addr_len)) {
/*
* New address is non-wildcard.
* Record new reservation in global non-wild namespace
*/
global_namespace = netns_global_non_wild[
NETNS_NS_GLOBAL_IDX(proto, addr_len)];
VERIFY(global_namespace != NULL);
if ((err = _netns_reserve_common(global_namespace,
nt->nt_port, nt->nt_flags)) != 0) {
SK_ERR("ERROR - reservation collision under new global namespace");
/* XXX: Should not fail. Maybe assert instead */
goto done;
}
}
memcpy(nt->nt_addr, addr, addr_len);
nt->nt_addr_len = addr_len;
done:
NETNS_UNLOCK();
return err;
}
static void
_netns_set_ifnet_internal(struct ns_token *nt, struct ifnet *ifp)
{
#if SK_LOG
char tmp_ip_str[MAX_IPv6_STR_LEN];
#endif /* SK_LOG */
NETNS_LOCK_ASSERT_HELD();
if (ifp != NULL && ifnet_is_attached(ifp, 1)) {
nt->nt_ifp = ifp;
SLIST_INSERT_HEAD(&ifp->if_netns_tokens, nt, nt_ifp_link);
SK_DF(NS_VERB_IP(nt->nt_addr_len) | NS_VERB_PROTO(nt->nt_proto),
"%s:%s:%d // added to ifnet %d",
inet_ntop(LEN_TO_AF(nt->nt_addr_len),
nt->nt_addr, tmp_ip_str, sizeof(tmp_ip_str)),
PROTO_STR(nt->nt_proto), nt->nt_port,
ifp->if_index);
} else {
SLIST_INSERT_HEAD(&netns_unbound_tokens, nt, nt_ifp_link);
}
}
void
netns_set_ifnet(netns_token *token, ifnet_t ifp)
{
struct ns_token *nt;
#if SK_LOG
char tmp_ip_str[MAX_IPv6_STR_LEN];
#endif /* SK_LOG */
if (__netns_inited == 0) {
return;
}
NETNS_LOCK();
VERIFY(NETNS_TOKEN_VALID(token));
nt = *token;
if (nt->nt_ifp == ifp) {
SK_DF(NS_VERB_IP(nt->nt_addr_len) | NS_VERB_PROTO(nt->nt_proto),
"%s:%s:%d // ifnet already %d, exiting early",
inet_ntop(LEN_TO_AF(nt->nt_addr_len),
nt->nt_addr, tmp_ip_str, sizeof(tmp_ip_str)),
PROTO_STR(nt->nt_proto), nt->nt_port,
ifp ? ifp->if_index : -1);
NETNS_UNLOCK();
return;
}
netns_clear_ifnet(nt);
_netns_set_ifnet_internal(nt, ifp);
NETNS_UNLOCK();
}
void
netns_ifnet_detach(ifnet_t ifp)
{
struct ns_token *token, *tmp_token;
if (__netns_inited == 0) {
return;
}
NETNS_LOCK();
SLIST_FOREACH_SAFE(token, &ifp->if_netns_tokens, nt_ifp_link,
tmp_token) {
netns_clear_ifnet(token);
SLIST_INSERT_HEAD(&netns_unbound_tokens, token, nt_ifp_link);
}
NETNS_UNLOCK();
}
static void
_netns_set_state(netns_token *token, uint32_t state)
{
struct ns_token *nt;
#if SK_LOG
char tmp_ip_str[MAX_IPv6_STR_LEN];
#endif /* SK_LOG */
if (__netns_inited == 0) {
return;
}
NETNS_LOCK();
VERIFY(NETNS_TOKEN_VALID(token));
nt = *token;
nt->nt_state |= state;
SK_DF(NS_VERB_IP(nt->nt_addr_len) | NS_VERB_PROTO(nt->nt_proto),
"%s:%s:%d // state 0x%b",
inet_ntop(LEN_TO_AF(nt->nt_addr_len), nt->nt_addr,
tmp_ip_str, sizeof(tmp_ip_str)),
PROTO_STR(nt->nt_proto), nt->nt_port, state, NETNS_STATE_BITS);
NETNS_UNLOCK();
}
void
netns_half_close(netns_token *token)
{
_netns_set_state(token, NETNS_STATE_HALFCLOSED);
}
void
netns_withdraw(netns_token *token)
{
_netns_set_state(token, NETNS_STATE_WITHDRAWN);
}
int
netns_get_flow_info(netns_token *token,
struct ns_flow_info *nfi)
{
if (__netns_inited == 0) {
return ENOTSUP;
}
NETNS_LOCK();
if (!NETNS_TOKEN_VALID(token) ||
nfi == NULL) {
NETNS_UNLOCK();
return EINVAL;
}
struct ns_token *nt = *token;
if (nt->nt_flow_info == NULL) {
NETNS_UNLOCK();
return ENOENT;
}
memcpy(nfi, nt->nt_flow_info, sizeof(struct ns_flow_info));
NETNS_UNLOCK();
return 0;
}
void
netns_change_flags(netns_token *token, uint32_t set_flags,
uint32_t clear_flags)
{
struct ns_token *nt;
#if SK_LOG
char tmp_ip_str[MAX_IPv6_STR_LEN];
#endif /* SK_LOG */
if (__netns_inited == 0) {
return;
}
NETNS_LOCK();
VERIFY(NETNS_TOKEN_VALID(token));
nt = *token;
VERIFY(!((set_flags | clear_flags) & NETNS_RESERVATION_FLAGS));
/* TODO: verify set and clear flags don't overlap? */
SK_DF(NS_VERB_IP(nt->nt_addr_len) | NS_VERB_PROTO(nt->nt_proto),
"%s:%s:%d // flags 0x%x -> 0x%x",
inet_ntop(LEN_TO_AF(nt->nt_addr_len), nt->nt_addr,
tmp_ip_str, sizeof(tmp_ip_str)),
PROTO_STR(nt->nt_proto), nt->nt_port, nt->nt_flags,
nt->nt_flags | set_flags & ~clear_flags);
nt->nt_flags |= set_flags;
nt->nt_flags &= ~clear_flags;
NETNS_UNLOCK();
}
/*
* Port offloading KPI
*/
static inline void
netns_local_port_scan_flow_entry(struct flow_entry *fe, protocol_family_t protocol,
u_int32_t flags, u_int8_t *bitfield)
{
struct ns_token *token;
boolean_t iswildcard = false;
if (fe == NULL) {
return;
}
if (fe->fe_flags & FLOWENTF_EXTRL_PORT) {
return;
}
token = fe->fe_port_reservation;
if (token == NULL) {
return;
}
/*
* We are only interested in active flows over skywalk channels
*/
if ((token->nt_flags & NETNS_OWNER_MASK) != NETNS_SKYWALK) {
return;
}
if (token->nt_state & NETNS_STATE_WITHDRAWN) {
return;
}
if (!(flags & IFNET_GET_LOCAL_PORTS_ANYTCPSTATEOK) &&
(flags & IFNET_GET_LOCAL_PORTS_ACTIVEONLY) &&
(token->nt_state & NETNS_STATE_HALFCLOSED)) {
return;
}
VERIFY(token->nt_addr_len == sizeof(struct in_addr) ||
token->nt_addr_len == sizeof(struct in6_addr));
if (token->nt_addr_len == sizeof(struct in_addr)) {
if (protocol == PF_INET6) {
return;
}
iswildcard = token->nt_inaddr.s_addr == INADDR_ANY;
} else if (token->nt_addr_len == sizeof(struct in6_addr)) {
if (protocol == PF_INET) {
return;
}
iswildcard = IN6_IS_ADDR_UNSPECIFIED(
&token->nt_in6addr);
}
if (!(flags & IFNET_GET_LOCAL_PORTS_WILDCARDOK) && iswildcard) {
return;
}
if ((flags & IFNET_GET_LOCAL_PORTS_TCPONLY) &&
token->nt_proto == IPPROTO_UDP) {
return;
}
if ((flags & IFNET_GET_LOCAL_PORTS_UDPONLY) &&
token->nt_proto == IPPROTO_TCP) {
return;
}
if ((flags & IFNET_GET_LOCAL_PORTS_RECVANYIFONLY) &&
!(token->nt_flags & NETNS_RECVANYIF)) {
return;
}
if ((flags & IFNET_GET_LOCAL_PORTS_EXTBGIDLEONLY) &&
!(token->nt_flags & NETNS_EXTBGIDLE)) {
return;
}
if (token->nt_ifp != NULL && (token->nt_ifp->if_eflags & IFEF_AWDL) != 0) {
struct flow_route *fr = fe->fe_route;
if (fr == NULL || fr->fr_rt_dst == NULL ||
(fr->fr_rt_dst->rt_flags & (RTF_UP | RTF_CONDEMNED)) != RTF_UP) {
#if DEBUG || DEVELOPMENT
char lbuf[MAX_IPv6_STR_LEN + 6] = {};
char fbuf[MAX_IPv6_STR_LEN + 6] = {};
in_port_t lport;
in_port_t fport;
char pname[MAXCOMLEN + 1];
const struct ns_flow_info *nfi = token->nt_flow_info;
proc_name(nfi->nfi_owner_pid, pname, sizeof(pname));
if (protocol == PF_INET) {
inet_ntop(PF_INET, &nfi->nfi_laddr.sin.sin_addr,
lbuf, sizeof(lbuf));
inet_ntop(PF_INET, &nfi->nfi_faddr.sin.sin_addr,
fbuf, sizeof(fbuf));
lport = nfi->nfi_laddr.sin.sin_port;
fport = nfi->nfi_faddr.sin.sin_port;
} else {
inet_ntop(PF_INET6, &nfi->nfi_laddr.sin6.sin6_addr.s6_addr,
lbuf, sizeof(lbuf));
inet_ntop(PF_INET6, &nfi->nfi_faddr.sin6.sin6_addr,
fbuf, sizeof(fbuf));
lport = nfi->nfi_laddr.sin6.sin6_port;
fport = nfi->nfi_faddr.sin6.sin6_port;
}
os_log(OS_LOG_DEFAULT,
"netns_local_port_scan_flow_entry: route is down %s %s:%u %s:%u ifp %s proc %s:%d",
token->nt_proto == IPPROTO_TCP ? "tcp" : "udp",
lbuf, ntohs(lport), fbuf, ntohs(fport),
token->nt_ifp->if_xname, pname, nfi->nfi_owner_pid);
#endif /* DEBUG || DEVELOPMENT */
return;
}
}
#if DEBUG || DEVELOPMENT
if (!(flags & IFNET_GET_LOCAL_PORTS_NOWAKEUPOK) &&
(token->nt_flags & NETNS_NOWAKEFROMSLEEP)) {
char lbuf[MAX_IPv6_STR_LEN + 6] = {};
char fbuf[MAX_IPv6_STR_LEN + 6] = {};
in_port_t lport;
in_port_t fport;
char pname[MAXCOMLEN + 1];
const struct ns_flow_info *nfi = token->nt_flow_info;
proc_name(nfi->nfi_owner_pid, pname, sizeof(pname));
if (protocol == PF_INET) {
inet_ntop(PF_INET, &nfi->nfi_laddr.sin.sin_addr,
lbuf, sizeof(lbuf));
inet_ntop(PF_INET, &nfi->nfi_faddr.sin.sin_addr,
fbuf, sizeof(fbuf));
lport = nfi->nfi_laddr.sin.sin_port;
fport = nfi->nfi_faddr.sin.sin_port;
} else {
inet_ntop(PF_INET6, &nfi->nfi_laddr.sin6.sin6_addr.s6_addr,
lbuf, sizeof(lbuf));
inet_ntop(PF_INET6, &nfi->nfi_faddr.sin6.sin6_addr,
fbuf, sizeof(fbuf));
lport = nfi->nfi_laddr.sin6.sin6_port;
fport = nfi->nfi_faddr.sin6.sin6_port;
}
os_log(OS_LOG_DEFAULT,
"netns_local_port_scan_flow_entry: no wake from sleep %s %s:%u %s:%u ifp %s proc %s:%d",
token->nt_proto == IPPROTO_TCP ? "tcp" : "udp",
lbuf, ntohs(lport), fbuf, ntohs(fport),
token->nt_ifp != NULL ? token->nt_ifp->if_xname : "",
pname, nfi->nfi_owner_pid);
}
#endif /* DEBUG || DEVELOPMENT */
if (token->nt_ifp != NULL && token->nt_flow_info != NULL) {
/*
* When the flow has "no wake from sleep" option, do not set the port in the bitmap
* except if explicetely requested by the driver.
* We always add the flow to the list of port in order to report spurious wakes
*/
if ((flags & IFNET_GET_LOCAL_PORTS_NOWAKEUPOK) ||
(token->nt_flags & NETNS_NOWAKEFROMSLEEP) == 0) {
bitstr_set(bitfield, token->nt_port);
}
(void) if_ports_used_add_flow_entry(fe, token->nt_ifp->if_index,
token->nt_flow_info, token->nt_flags);
} else {
SK_ERR("%s: unknown owner port %u"
" nt_flags 0x%x ifindex %u nt_flow_info %p\n",
__func__, token->nt_port,
token->nt_flags,
token->nt_ifp != NULL ? token->nt_ifp->if_index : 0,
token->nt_flow_info);
}
}
static void
netns_get_if_local_ports(ifnet_t ifp, protocol_family_t protocol,
u_int32_t flags, u_int8_t *bitfield)
{
struct nx_flowswitch *fsw = NULL;
if (ifp == NULL || ifp->if_na == NULL) {
return;
}
/* Ensure that the interface is attached and won't detach */
if (!ifnet_is_attached(ifp, 1)) {
return;
}
fsw = fsw_ifp_to_fsw(ifp);
if (fsw == NULL) {
goto done;
}
FSW_RLOCK(fsw);
NETNS_LOCK();
flow_mgr_foreach_flow(fsw->fsw_flow_mgr, ^(struct flow_entry *_fe) {
netns_local_port_scan_flow_entry(_fe, protocol, flags,
bitfield);
});
NETNS_UNLOCK();
FSW_UNLOCK(fsw);
done:
ifnet_decr_iorefcnt(ifp);
}
errno_t
netns_get_local_ports(ifnet_t ifp, protocol_family_t protocol,
u_int32_t flags, u_int8_t *bitfield)
{
if (__netns_inited == 0) {
return 0;
}
if (ifp != NULL) {
netns_get_if_local_ports(ifp, protocol, flags, bitfield);
} else {
errno_t error;
ifnet_t *ifp_list;
uint32_t count, i;
error = ifnet_list_get_all(IFNET_FAMILY_ANY, &ifp_list, &count);
if (error != 0) {
os_log_error(OS_LOG_DEFAULT,
"%s: ifnet_list_get_all() failed %d",
__func__, error);
return error;
}
for (i = 0; i < count; i++) {
if (TAILQ_EMPTY(&ifp_list[i]->if_addrhead)) {
continue;
}
netns_get_if_local_ports(ifp_list[i], protocol, flags,
bitfield);
}
ifnet_list_free(ifp_list);
}
return 0;
}
uint32_t
netns_find_anyres_byaddr(struct ifaddr *ifa, uint8_t proto)
{
int result = 0;
int ifa_addr_len;
struct ns_token *token;
struct ifnet *ifp = ifa->ifa_ifp;
struct sockaddr *ifa_addr = ifa->ifa_addr;
if (__netns_inited == 0) {
return ENOTSUP;
}
if ((ifa_addr->sa_family != AF_INET) &&
(ifa_addr->sa_family != AF_INET6)) {
return 0;
}
ifa_addr_len = (ifa_addr->sa_family == AF_INET) ?
sizeof(struct in_addr) : sizeof(struct in6_addr);
NETNS_LOCK();
SLIST_FOREACH(token, &ifp->if_netns_tokens, nt_ifp_link) {
if ((token->nt_flags & NETNS_OWNER_MASK) == NETNS_PF) {
continue;
}
if (token->nt_addr_len != ifa_addr_len) {
continue;
}
if (token->nt_proto != proto) {
continue;
}
if (ifa_addr->sa_family == AF_INET) {
if (token->nt_inaddr.s_addr ==
(satosin(ifa->ifa_addr))->sin_addr.s_addr) {
result = 1;
break;
}
} else if (ifa_addr->sa_family == AF_INET6) {
if (IN6_ARE_ADDR_EQUAL(IFA_IN6(ifa),
&token->nt_in6addr)) {
result = 1;
break;
}
}
}
NETNS_UNLOCK();
return result;
}
static uint32_t
_netns_lookup_ns_n_reservations(uint32_t *addr, uint8_t addr_len, uint8_t proto)
{
uint32_t ns_n_reservations = 0;
NETNS_LOCK_SPIN();
struct ns *namespace = _netns_get_ns(addr, addr_len, proto, true);
if (namespace != NULL) {
ns_n_reservations = namespace->ns_n_reservations;
}
NETNS_UNLOCK();
return ns_n_reservations;
}
uint32_t
netns_lookup_reservations_count_in(struct in_addr addr, uint8_t proto)
{
return _netns_lookup_ns_n_reservations(&addr.s_addr, sizeof(struct in_addr), proto);
}
uint32_t
netns_lookup_reservations_count_in6(struct in6_addr addr, uint8_t proto)
{
if (IN6_IS_SCOPE_EMBED(&addr)) {
addr.s6_addr16[1] = 0;
}
return _netns_lookup_ns_n_reservations(&addr.s6_addr32[0], sizeof(struct in6_addr), proto);
}
/*
* Sysctl interface
*/
static int netns_ctl_dump_all SYSCTL_HANDLER_ARGS;
SYSCTL_NODE(_kern_skywalk, OID_AUTO, netns, CTLFLAG_RW | CTLFLAG_LOCKED,
0, "Netns interface");
SYSCTL_PROC(_kern_skywalk_stats, OID_AUTO, netns,
CTLTYPE_STRUCT | CTLFLAG_RW | CTLFLAG_LOCKED,
0, 0, netns_ctl_dump_all, "-",
"Namespace contents (struct netns_ctl_dump_header, "
"skywalk/os_stats_private.h)");
static int
netns_ctl_write_ns(struct sysctl_req *req, struct ns *namespace,
boolean_t is_global)
{
struct ns_reservation *res;
struct netns_ctl_dump_header response_header;
struct netns_ctl_dump_record response_record;
int err;
/* Fill out header */
memset(&response_header, 0, sizeof(response_header));
response_header.ncdh_n_records = namespace->ns_n_reservations;
response_header.ncdh_proto = namespace->ns_proto;
if (is_global) {
response_header.ncdh_addr_len = 0;
} else {
response_header.ncdh_addr_len = namespace->ns_addr_len;
}
memcpy(response_header.ncdh_addr, namespace->ns_addr,
namespace->ns_addr_len);
err = SYSCTL_OUT(req, &response_header, sizeof(response_header));
if (err) {
return err;
}
/* Fill out records */
RB_FOREACH(res, ns_reservation_tree, &namespace->ns_reservations) {
memset(&response_record, 0, sizeof(response_record));
response_record.ncdr_port = res->nsr_port;
response_record.ncdr_port_end = 0;
response_record.ncdr_listener_refs =
NETNS_REF_COUNT(res, NETNS_LISTENER);
response_record.ncdr_skywalk_refs =
NETNS_REF_COUNT(res, NETNS_SKYWALK);
response_record.ncdr_bsd_refs =
NETNS_REF_COUNT(res, NETNS_BSD);
response_record.ncdr_pf_refs =
NETNS_REF_COUNT(res, NETNS_PF);
err = SYSCTL_OUT(req, &response_record,
sizeof(response_record));
if (err) {
return err;
}
}
return 0;
}
static int
netns_ctl_dump_all SYSCTL_HANDLER_ARGS
{
#pragma unused(oidp, arg1, arg2)
struct ns *namespace;
int i, err = 0;
if (!kauth_cred_issuser(kauth_cred_get())) {
return EPERM;
}
if (__netns_inited == 0) {
return ENOTSUP;
}
NETNS_LOCK();
for (i = 0; i < NETNS_N_GLOBAL; i++) {
err = netns_ctl_write_ns(req, netns_global_non_wild[i], true);
if (err) {
goto done;
}
}
RB_FOREACH(namespace, netns_namespaces_tree, &netns_namespaces) {
err = netns_ctl_write_ns(req, namespace, false);
if (err) {
goto done;
}
}
/*
* If this is just a request for length, add slop because
* this is dynamically changing data
*/
if (req->oldptr == USER_ADDR_NULL) {
req->oldidx += 20 * sizeof(struct netns_ctl_dump_record);
}
done:
NETNS_UNLOCK();
return err;
}
/* CSTYLED */