mirror of
https://github.com/samba-team/samba.git
synced 2024-12-25 23:21:54 +03:00
79fe4bbf73
ctdb_private.h already defines set_close_on_exec. Autobuild-User: Volker Lendecke <vlendec@samba.org> Autobuild-Date: Sat Dec 17 18:41:39 CET 2011 on sn-devel-104
1036 lines
24 KiB
C
1036 lines
24 KiB
C
/*
|
|
Unix SMB/CIFS implementation.
|
|
|
|
Socket IPv4/IPv6 functions
|
|
|
|
Copyright (C) Stefan Metzmacher 2004
|
|
Copyright (C) Andrew Tridgell 2004-2005
|
|
Copyright (C) Jelmer Vernooij 2004
|
|
|
|
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 3 of the License, or
|
|
(at your option) any later version.
|
|
|
|
This program is distributed in the hope that it will be useful,
|
|
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
GNU General Public License for more details.
|
|
|
|
You should have received a copy of the GNU General Public License
|
|
along with this program. If not, see <http://www.gnu.org/licenses/>.
|
|
*/
|
|
|
|
#include "includes.h"
|
|
#include "system/filesys.h"
|
|
#include "lib/socket/socket.h"
|
|
#include "system/network.h"
|
|
#include "lib/util/util_net.h"
|
|
|
|
_PUBLIC_ const struct socket_ops *socket_ipv4_ops(enum socket_type type);
|
|
_PUBLIC_ const struct socket_ops *socket_ipv6_ops(enum socket_type type);
|
|
|
|
static NTSTATUS ipv4_init(struct socket_context *sock)
|
|
{
|
|
int type;
|
|
|
|
switch (sock->type) {
|
|
case SOCKET_TYPE_STREAM:
|
|
type = SOCK_STREAM;
|
|
break;
|
|
case SOCKET_TYPE_DGRAM:
|
|
type = SOCK_DGRAM;
|
|
break;
|
|
default:
|
|
return NT_STATUS_INVALID_PARAMETER;
|
|
}
|
|
|
|
sock->fd = socket(PF_INET, type, 0);
|
|
if (sock->fd == -1) {
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
|
|
smb_set_close_on_exec(sock->fd);
|
|
|
|
sock->backend_name = "ipv4";
|
|
sock->family = AF_INET;
|
|
|
|
return NT_STATUS_OK;
|
|
}
|
|
|
|
static void ip_close(struct socket_context *sock)
|
|
{
|
|
if (sock->fd != -1) {
|
|
close(sock->fd);
|
|
sock->fd = -1;
|
|
}
|
|
}
|
|
|
|
static NTSTATUS ip_connect_complete(struct socket_context *sock, uint32_t flags)
|
|
{
|
|
int error=0, ret;
|
|
socklen_t len = sizeof(error);
|
|
|
|
/* check for any errors that may have occurred - this is needed
|
|
for non-blocking connect */
|
|
ret = getsockopt(sock->fd, SOL_SOCKET, SO_ERROR, &error, &len);
|
|
if (ret == -1) {
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
if (error != 0) {
|
|
return map_nt_error_from_unix_common(error);
|
|
}
|
|
|
|
if (!(flags & SOCKET_FLAG_BLOCK)) {
|
|
ret = set_blocking(sock->fd, false);
|
|
if (ret == -1) {
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
}
|
|
|
|
sock->state = SOCKET_STATE_CLIENT_CONNECTED;
|
|
|
|
return NT_STATUS_OK;
|
|
}
|
|
|
|
|
|
static NTSTATUS ipv4_connect(struct socket_context *sock,
|
|
const struct socket_address *my_address,
|
|
const struct socket_address *srv_address,
|
|
uint32_t flags)
|
|
{
|
|
struct sockaddr_in srv_addr;
|
|
struct in_addr my_ip;
|
|
struct in_addr srv_ip;
|
|
int ret;
|
|
|
|
if (my_address && my_address->sockaddr) {
|
|
ret = bind(sock->fd, my_address->sockaddr, my_address->sockaddrlen);
|
|
if (ret == -1) {
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
} else if (my_address) {
|
|
my_ip = interpret_addr2(my_address->addr);
|
|
|
|
if (my_ip.s_addr != 0 || my_address->port != 0) {
|
|
struct sockaddr_in my_addr;
|
|
ZERO_STRUCT(my_addr);
|
|
#ifdef HAVE_SOCK_SIN_LEN
|
|
my_addr.sin_len = sizeof(my_addr);
|
|
#endif
|
|
my_addr.sin_addr.s_addr = my_ip.s_addr;
|
|
my_addr.sin_port = htons(my_address->port);
|
|
my_addr.sin_family = PF_INET;
|
|
|
|
ret = bind(sock->fd, (struct sockaddr *)&my_addr, sizeof(my_addr));
|
|
if (ret == -1) {
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
}
|
|
}
|
|
|
|
if (srv_address->sockaddr) {
|
|
ret = connect(sock->fd, srv_address->sockaddr, srv_address->sockaddrlen);
|
|
if (ret == -1) {
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
} else {
|
|
srv_ip = interpret_addr2(srv_address->addr);
|
|
if (!srv_ip.s_addr) {
|
|
return NT_STATUS_BAD_NETWORK_NAME;
|
|
}
|
|
|
|
SMB_ASSERT(srv_address->port != 0);
|
|
|
|
ZERO_STRUCT(srv_addr);
|
|
#ifdef HAVE_SOCK_SIN_LEN
|
|
srv_addr.sin_len = sizeof(srv_addr);
|
|
#endif
|
|
srv_addr.sin_addr.s_addr= srv_ip.s_addr;
|
|
srv_addr.sin_port = htons(srv_address->port);
|
|
srv_addr.sin_family = PF_INET;
|
|
|
|
ret = connect(sock->fd, (const struct sockaddr *)&srv_addr, sizeof(srv_addr));
|
|
if (ret == -1) {
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
}
|
|
|
|
return ip_connect_complete(sock, flags);
|
|
}
|
|
|
|
|
|
/*
|
|
note that for simplicity of the API, socket_listen() is also
|
|
use for DGRAM sockets, but in reality only a bind() is done
|
|
*/
|
|
static NTSTATUS ipv4_listen(struct socket_context *sock,
|
|
const struct socket_address *my_address,
|
|
int queue_size, uint32_t flags)
|
|
{
|
|
struct sockaddr_in my_addr;
|
|
struct in_addr ip_addr;
|
|
int ret;
|
|
|
|
socket_set_option(sock, "SO_REUSEADDR=1", NULL);
|
|
|
|
if (my_address->sockaddr) {
|
|
ret = bind(sock->fd, my_address->sockaddr, my_address->sockaddrlen);
|
|
} else {
|
|
ip_addr = interpret_addr2(my_address->addr);
|
|
|
|
ZERO_STRUCT(my_addr);
|
|
#ifdef HAVE_SOCK_SIN_LEN
|
|
my_addr.sin_len = sizeof(my_addr);
|
|
#endif
|
|
my_addr.sin_addr.s_addr = ip_addr.s_addr;
|
|
my_addr.sin_port = htons(my_address->port);
|
|
my_addr.sin_family = PF_INET;
|
|
|
|
ret = bind(sock->fd, (struct sockaddr *)&my_addr, sizeof(my_addr));
|
|
}
|
|
|
|
if (ret == -1) {
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
|
|
if (sock->type == SOCKET_TYPE_STREAM) {
|
|
ret = listen(sock->fd, queue_size);
|
|
if (ret == -1) {
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
}
|
|
|
|
if (!(flags & SOCKET_FLAG_BLOCK)) {
|
|
ret = set_blocking(sock->fd, false);
|
|
if (ret == -1) {
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
}
|
|
|
|
sock->state= SOCKET_STATE_SERVER_LISTEN;
|
|
|
|
return NT_STATUS_OK;
|
|
}
|
|
|
|
static NTSTATUS ipv4_accept(struct socket_context *sock, struct socket_context **new_sock)
|
|
{
|
|
struct sockaddr_in cli_addr;
|
|
socklen_t cli_addr_len = sizeof(cli_addr);
|
|
int new_fd;
|
|
|
|
if (sock->type != SOCKET_TYPE_STREAM) {
|
|
return NT_STATUS_INVALID_PARAMETER;
|
|
}
|
|
|
|
new_fd = accept(sock->fd, (struct sockaddr *)&cli_addr, &cli_addr_len);
|
|
if (new_fd == -1) {
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
|
|
if (!(sock->flags & SOCKET_FLAG_BLOCK)) {
|
|
int ret = set_blocking(new_fd, false);
|
|
if (ret == -1) {
|
|
close(new_fd);
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
}
|
|
|
|
/* TODO: we could add a 'accept_check' hook here
|
|
* which get the black/white lists via socket_set_accept_filter()
|
|
* or something like that
|
|
* --metze
|
|
*/
|
|
|
|
(*new_sock) = talloc(NULL, struct socket_context);
|
|
if (!(*new_sock)) {
|
|
close(new_fd);
|
|
return NT_STATUS_NO_MEMORY;
|
|
}
|
|
|
|
/* copy the socket_context */
|
|
(*new_sock)->type = sock->type;
|
|
(*new_sock)->state = SOCKET_STATE_SERVER_CONNECTED;
|
|
(*new_sock)->flags = sock->flags;
|
|
|
|
(*new_sock)->fd = new_fd;
|
|
|
|
(*new_sock)->private_data = NULL;
|
|
(*new_sock)->ops = sock->ops;
|
|
(*new_sock)->backend_name = sock->backend_name;
|
|
|
|
return NT_STATUS_OK;
|
|
}
|
|
|
|
static NTSTATUS ip_recv(struct socket_context *sock, void *buf,
|
|
size_t wantlen, size_t *nread)
|
|
{
|
|
ssize_t gotlen;
|
|
|
|
*nread = 0;
|
|
|
|
gotlen = recv(sock->fd, buf, wantlen, 0);
|
|
if (gotlen == 0) {
|
|
return NT_STATUS_END_OF_FILE;
|
|
} else if (gotlen == -1) {
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
|
|
*nread = gotlen;
|
|
|
|
return NT_STATUS_OK;
|
|
}
|
|
|
|
|
|
static NTSTATUS ipv4_recvfrom(struct socket_context *sock, void *buf,
|
|
size_t wantlen, size_t *nread,
|
|
TALLOC_CTX *addr_ctx, struct socket_address **_src)
|
|
{
|
|
ssize_t gotlen;
|
|
struct sockaddr_in *from_addr;
|
|
socklen_t from_len = sizeof(*from_addr);
|
|
struct socket_address *src;
|
|
char addrstring[INET_ADDRSTRLEN];
|
|
|
|
src = talloc(addr_ctx, struct socket_address);
|
|
if (!src) {
|
|
return NT_STATUS_NO_MEMORY;
|
|
}
|
|
|
|
src->family = sock->backend_name;
|
|
|
|
from_addr = talloc(src, struct sockaddr_in);
|
|
if (!from_addr) {
|
|
talloc_free(src);
|
|
return NT_STATUS_NO_MEMORY;
|
|
}
|
|
|
|
src->sockaddr = (struct sockaddr *)from_addr;
|
|
|
|
*nread = 0;
|
|
|
|
gotlen = recvfrom(sock->fd, buf, wantlen, 0,
|
|
src->sockaddr, &from_len);
|
|
if (gotlen == 0) {
|
|
talloc_free(src);
|
|
return NT_STATUS_END_OF_FILE;
|
|
} else if (gotlen == -1) {
|
|
talloc_free(src);
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
|
|
src->sockaddrlen = from_len;
|
|
|
|
if (inet_ntop(AF_INET, &from_addr->sin_addr, addrstring,
|
|
sizeof(addrstring)) == NULL) {
|
|
talloc_free(src);
|
|
return NT_STATUS_INTERNAL_ERROR;
|
|
}
|
|
src->addr = talloc_strdup(src, addrstring);
|
|
if (src->addr == NULL) {
|
|
talloc_free(src);
|
|
return NT_STATUS_NO_MEMORY;
|
|
}
|
|
src->port = ntohs(from_addr->sin_port);
|
|
|
|
*nread = gotlen;
|
|
*_src = src;
|
|
return NT_STATUS_OK;
|
|
}
|
|
|
|
static NTSTATUS ip_send(struct socket_context *sock,
|
|
const DATA_BLOB *blob, size_t *sendlen)
|
|
{
|
|
ssize_t len;
|
|
|
|
*sendlen = 0;
|
|
|
|
len = send(sock->fd, blob->data, blob->length, 0);
|
|
if (len == -1) {
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
|
|
*sendlen = len;
|
|
|
|
return NT_STATUS_OK;
|
|
}
|
|
|
|
static NTSTATUS ipv4_sendto(struct socket_context *sock,
|
|
const DATA_BLOB *blob, size_t *sendlen,
|
|
const struct socket_address *dest_addr)
|
|
{
|
|
ssize_t len;
|
|
|
|
if (dest_addr->sockaddr) {
|
|
len = sendto(sock->fd, blob->data, blob->length, 0,
|
|
dest_addr->sockaddr, dest_addr->sockaddrlen);
|
|
} else {
|
|
struct sockaddr_in srv_addr;
|
|
struct in_addr addr;
|
|
|
|
SMB_ASSERT(dest_addr->port != 0);
|
|
|
|
ZERO_STRUCT(srv_addr);
|
|
#ifdef HAVE_SOCK_SIN_LEN
|
|
srv_addr.sin_len = sizeof(srv_addr);
|
|
#endif
|
|
addr = interpret_addr2(dest_addr->addr);
|
|
if (addr.s_addr == 0) {
|
|
return NT_STATUS_HOST_UNREACHABLE;
|
|
}
|
|
srv_addr.sin_addr.s_addr = addr.s_addr;
|
|
srv_addr.sin_port = htons(dest_addr->port);
|
|
srv_addr.sin_family = PF_INET;
|
|
|
|
*sendlen = 0;
|
|
|
|
len = sendto(sock->fd, blob->data, blob->length, 0,
|
|
(struct sockaddr *)&srv_addr, sizeof(srv_addr));
|
|
}
|
|
if (len == -1) {
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
|
|
*sendlen = len;
|
|
|
|
return NT_STATUS_OK;
|
|
}
|
|
|
|
static NTSTATUS ipv4_set_option(struct socket_context *sock, const char *option, const char *val)
|
|
{
|
|
set_socket_options(sock->fd, option);
|
|
return NT_STATUS_OK;
|
|
}
|
|
|
|
static char *ipv4_get_peer_name(struct socket_context *sock, TALLOC_CTX *mem_ctx)
|
|
{
|
|
struct sockaddr_in peer_addr;
|
|
socklen_t len = sizeof(peer_addr);
|
|
struct hostent *he;
|
|
int ret;
|
|
|
|
ret = getpeername(sock->fd, (struct sockaddr *)&peer_addr, &len);
|
|
if (ret == -1) {
|
|
return NULL;
|
|
}
|
|
|
|
he = gethostbyaddr((char *)&peer_addr.sin_addr, sizeof(peer_addr.sin_addr), AF_INET);
|
|
if (he == NULL) {
|
|
return NULL;
|
|
}
|
|
|
|
return talloc_strdup(mem_ctx, he->h_name);
|
|
}
|
|
|
|
static struct socket_address *ipv4_get_peer_addr(struct socket_context *sock, TALLOC_CTX *mem_ctx)
|
|
{
|
|
struct sockaddr_in *peer_addr;
|
|
socklen_t len = sizeof(*peer_addr);
|
|
struct socket_address *peer;
|
|
char addrstring[INET_ADDRSTRLEN];
|
|
int ret;
|
|
|
|
peer = talloc(mem_ctx, struct socket_address);
|
|
if (!peer) {
|
|
return NULL;
|
|
}
|
|
|
|
peer->family = sock->backend_name;
|
|
peer_addr = talloc(peer, struct sockaddr_in);
|
|
if (!peer_addr) {
|
|
talloc_free(peer);
|
|
return NULL;
|
|
}
|
|
|
|
peer->sockaddr = (struct sockaddr *)peer_addr;
|
|
|
|
ret = getpeername(sock->fd, peer->sockaddr, &len);
|
|
if (ret == -1) {
|
|
talloc_free(peer);
|
|
return NULL;
|
|
}
|
|
|
|
peer->sockaddrlen = len;
|
|
|
|
if (inet_ntop(AF_INET, &peer_addr->sin_addr, addrstring,
|
|
sizeof(addrstring)) == NULL) {
|
|
talloc_free(peer);
|
|
return NULL;
|
|
}
|
|
peer->addr = talloc_strdup(peer, addrstring);
|
|
if (!peer->addr) {
|
|
talloc_free(peer);
|
|
return NULL;
|
|
}
|
|
peer->port = ntohs(peer_addr->sin_port);
|
|
|
|
return peer;
|
|
}
|
|
|
|
static struct socket_address *ipv4_get_my_addr(struct socket_context *sock, TALLOC_CTX *mem_ctx)
|
|
{
|
|
struct sockaddr_in *local_addr;
|
|
socklen_t len = sizeof(*local_addr);
|
|
struct socket_address *local;
|
|
char addrstring[INET_ADDRSTRLEN];
|
|
int ret;
|
|
|
|
local = talloc(mem_ctx, struct socket_address);
|
|
if (!local) {
|
|
return NULL;
|
|
}
|
|
|
|
local->family = sock->backend_name;
|
|
local_addr = talloc(local, struct sockaddr_in);
|
|
if (!local_addr) {
|
|
talloc_free(local);
|
|
return NULL;
|
|
}
|
|
|
|
local->sockaddr = (struct sockaddr *)local_addr;
|
|
|
|
ret = getsockname(sock->fd, local->sockaddr, &len);
|
|
if (ret == -1) {
|
|
talloc_free(local);
|
|
return NULL;
|
|
}
|
|
|
|
local->sockaddrlen = len;
|
|
|
|
if (inet_ntop(AF_INET, &local_addr->sin_addr, addrstring,
|
|
sizeof(addrstring)) == NULL) {
|
|
talloc_free(local);
|
|
return NULL;
|
|
}
|
|
local->addr = talloc_strdup(local, addrstring);
|
|
if (!local->addr) {
|
|
talloc_free(local);
|
|
return NULL;
|
|
}
|
|
local->port = ntohs(local_addr->sin_port);
|
|
|
|
return local;
|
|
}
|
|
static int ip_get_fd(struct socket_context *sock)
|
|
{
|
|
return sock->fd;
|
|
}
|
|
|
|
static NTSTATUS ip_pending(struct socket_context *sock, size_t *npending)
|
|
{
|
|
int value = 0;
|
|
if (ioctl(sock->fd, FIONREAD, &value) == 0) {
|
|
*npending = value;
|
|
return NT_STATUS_OK;
|
|
}
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
|
|
static const struct socket_ops ipv4_ops = {
|
|
.name = "ipv4",
|
|
.fn_init = ipv4_init,
|
|
.fn_connect = ipv4_connect,
|
|
.fn_connect_complete = ip_connect_complete,
|
|
.fn_listen = ipv4_listen,
|
|
.fn_accept = ipv4_accept,
|
|
.fn_recv = ip_recv,
|
|
.fn_recvfrom = ipv4_recvfrom,
|
|
.fn_send = ip_send,
|
|
.fn_sendto = ipv4_sendto,
|
|
.fn_pending = ip_pending,
|
|
.fn_close = ip_close,
|
|
|
|
.fn_set_option = ipv4_set_option,
|
|
|
|
.fn_get_peer_name = ipv4_get_peer_name,
|
|
.fn_get_peer_addr = ipv4_get_peer_addr,
|
|
.fn_get_my_addr = ipv4_get_my_addr,
|
|
|
|
.fn_get_fd = ip_get_fd
|
|
};
|
|
|
|
_PUBLIC_ const struct socket_ops *socket_ipv4_ops(enum socket_type type)
|
|
{
|
|
return &ipv4_ops;
|
|
}
|
|
|
|
#if HAVE_IPV6
|
|
|
|
static struct in6_addr interpret_addr6(const char *name)
|
|
{
|
|
char addr[INET6_ADDRSTRLEN];
|
|
struct in6_addr dest6;
|
|
const char *sp = name;
|
|
char *p;
|
|
int ret;
|
|
|
|
if (sp == NULL) return in6addr_any;
|
|
|
|
p = strchr_m(sp, '%');
|
|
|
|
if (strcasecmp(sp, "localhost") == 0) {
|
|
sp = "::1";
|
|
}
|
|
|
|
/*
|
|
* Cope with link-local.
|
|
* This is IP:v6:addr%ifname.
|
|
*/
|
|
|
|
if (p && (p > sp) && (if_nametoindex(p+1) != 0)) {
|
|
strlcpy(addr, sp,
|
|
MIN(PTR_DIFF(p,sp)+1,
|
|
sizeof(addr)));
|
|
sp = addr;
|
|
}
|
|
|
|
ret = inet_pton(AF_INET6, sp, &dest6);
|
|
if (ret > 0) {
|
|
return dest6;
|
|
}
|
|
|
|
return in6addr_any;
|
|
}
|
|
|
|
static NTSTATUS ipv6_init(struct socket_context *sock)
|
|
{
|
|
int type;
|
|
|
|
switch (sock->type) {
|
|
case SOCKET_TYPE_STREAM:
|
|
type = SOCK_STREAM;
|
|
break;
|
|
case SOCKET_TYPE_DGRAM:
|
|
type = SOCK_DGRAM;
|
|
break;
|
|
default:
|
|
return NT_STATUS_INVALID_PARAMETER;
|
|
}
|
|
|
|
sock->fd = socket(PF_INET6, type, 0);
|
|
if (sock->fd == -1) {
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
|
|
smb_set_close_on_exec(sock->fd);
|
|
|
|
sock->backend_name = "ipv6";
|
|
sock->family = AF_INET6;
|
|
|
|
return NT_STATUS_OK;
|
|
}
|
|
|
|
static NTSTATUS ipv6_tcp_connect(struct socket_context *sock,
|
|
const struct socket_address *my_address,
|
|
const struct socket_address *srv_address,
|
|
uint32_t flags)
|
|
{
|
|
int ret;
|
|
|
|
if (my_address && my_address->sockaddr) {
|
|
ret = bind(sock->fd, my_address->sockaddr, my_address->sockaddrlen);
|
|
if (ret == -1) {
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
} else if (my_address) {
|
|
struct in6_addr my_ip;
|
|
my_ip = interpret_addr6(my_address->addr);
|
|
|
|
if (memcmp(&my_ip, &in6addr_any, sizeof(my_ip)) || my_address->port != 0) {
|
|
struct sockaddr_in6 my_addr;
|
|
ZERO_STRUCT(my_addr);
|
|
my_addr.sin6_addr = my_ip;
|
|
my_addr.sin6_port = htons(my_address->port);
|
|
my_addr.sin6_family = PF_INET6;
|
|
|
|
ret = bind(sock->fd, (struct sockaddr *)&my_addr, sizeof(my_addr));
|
|
if (ret == -1) {
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
}
|
|
}
|
|
|
|
if (srv_address->sockaddr) {
|
|
ret = connect(sock->fd, srv_address->sockaddr, srv_address->sockaddrlen);
|
|
} else {
|
|
struct in6_addr srv_ip;
|
|
struct sockaddr_in6 srv_addr;
|
|
srv_ip = interpret_addr6(srv_address->addr);
|
|
if (memcmp(&srv_ip, &in6addr_any, sizeof(srv_ip)) == 0) {
|
|
return NT_STATUS_BAD_NETWORK_NAME;
|
|
}
|
|
|
|
ZERO_STRUCT(srv_addr);
|
|
srv_addr.sin6_addr = srv_ip;
|
|
srv_addr.sin6_port = htons(srv_address->port);
|
|
srv_addr.sin6_family = PF_INET6;
|
|
|
|
ret = connect(sock->fd, (const struct sockaddr *)&srv_addr, sizeof(srv_addr));
|
|
}
|
|
if (ret == -1) {
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
|
|
return ip_connect_complete(sock, flags);
|
|
}
|
|
|
|
/*
|
|
fix the sin6_scope_id based on the address interface
|
|
*/
|
|
static void fix_scope_id(struct sockaddr_in6 *in6,
|
|
const char *address)
|
|
{
|
|
const char *p = strchr(address, '%');
|
|
if (p != NULL) {
|
|
in6->sin6_scope_id = if_nametoindex(p+1);
|
|
}
|
|
}
|
|
|
|
|
|
static NTSTATUS ipv6_listen(struct socket_context *sock,
|
|
const struct socket_address *my_address,
|
|
int queue_size, uint32_t flags)
|
|
{
|
|
struct sockaddr_in6 my_addr;
|
|
struct in6_addr ip_addr;
|
|
int ret;
|
|
|
|
socket_set_option(sock, "SO_REUSEADDR=1", NULL);
|
|
|
|
if (my_address->sockaddr) {
|
|
ret = bind(sock->fd, my_address->sockaddr, my_address->sockaddrlen);
|
|
} else {
|
|
int one = 1;
|
|
ip_addr = interpret_addr6(my_address->addr);
|
|
|
|
ZERO_STRUCT(my_addr);
|
|
my_addr.sin6_addr = ip_addr;
|
|
my_addr.sin6_port = htons(my_address->port);
|
|
my_addr.sin6_family = PF_INET6;
|
|
fix_scope_id(&my_addr, my_address->addr);
|
|
|
|
/* when binding on ipv6 we always want to only bind on v6 */
|
|
ret = setsockopt(sock->fd, IPPROTO_IPV6, IPV6_V6ONLY,
|
|
(const void *)&one, sizeof(one));
|
|
if (ret != -1) {
|
|
ret = bind(sock->fd, (struct sockaddr *)&my_addr, sizeof(my_addr));
|
|
}
|
|
}
|
|
|
|
if (ret == -1) {
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
|
|
if (sock->type == SOCKET_TYPE_STREAM) {
|
|
ret = listen(sock->fd, queue_size);
|
|
if (ret == -1) {
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
}
|
|
|
|
if (!(flags & SOCKET_FLAG_BLOCK)) {
|
|
ret = set_blocking(sock->fd, false);
|
|
if (ret == -1) {
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
}
|
|
|
|
sock->state= SOCKET_STATE_SERVER_LISTEN;
|
|
|
|
return NT_STATUS_OK;
|
|
}
|
|
|
|
static NTSTATUS ipv6_tcp_accept(struct socket_context *sock, struct socket_context **new_sock)
|
|
{
|
|
struct sockaddr_in6 cli_addr;
|
|
socklen_t cli_addr_len = sizeof(cli_addr);
|
|
int new_fd;
|
|
|
|
if (sock->type != SOCKET_TYPE_STREAM) {
|
|
return NT_STATUS_INVALID_PARAMETER;
|
|
}
|
|
|
|
new_fd = accept(sock->fd, (struct sockaddr *)&cli_addr, &cli_addr_len);
|
|
if (new_fd == -1) {
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
|
|
if (!(sock->flags & SOCKET_FLAG_BLOCK)) {
|
|
int ret = set_blocking(new_fd, false);
|
|
if (ret == -1) {
|
|
close(new_fd);
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
}
|
|
|
|
/* TODO: we could add a 'accept_check' hook here
|
|
* which get the black/white lists via socket_set_accept_filter()
|
|
* or something like that
|
|
* --metze
|
|
*/
|
|
|
|
(*new_sock) = talloc(NULL, struct socket_context);
|
|
if (!(*new_sock)) {
|
|
close(new_fd);
|
|
return NT_STATUS_NO_MEMORY;
|
|
}
|
|
|
|
/* copy the socket_context */
|
|
(*new_sock)->type = sock->type;
|
|
(*new_sock)->state = SOCKET_STATE_SERVER_CONNECTED;
|
|
(*new_sock)->flags = sock->flags;
|
|
|
|
(*new_sock)->fd = new_fd;
|
|
|
|
(*new_sock)->private_data = NULL;
|
|
(*new_sock)->ops = sock->ops;
|
|
(*new_sock)->backend_name = sock->backend_name;
|
|
|
|
return NT_STATUS_OK;
|
|
}
|
|
|
|
static NTSTATUS ipv6_recvfrom(struct socket_context *sock, void *buf,
|
|
size_t wantlen, size_t *nread,
|
|
TALLOC_CTX *addr_ctx, struct socket_address **_src)
|
|
{
|
|
ssize_t gotlen;
|
|
struct sockaddr_in6 *from_addr;
|
|
socklen_t from_len = sizeof(*from_addr);
|
|
struct socket_address *src;
|
|
char addrstring[INET6_ADDRSTRLEN];
|
|
|
|
src = talloc(addr_ctx, struct socket_address);
|
|
if (!src) {
|
|
return NT_STATUS_NO_MEMORY;
|
|
}
|
|
|
|
src->family = sock->backend_name;
|
|
|
|
from_addr = talloc(src, struct sockaddr_in6);
|
|
if (!from_addr) {
|
|
talloc_free(src);
|
|
return NT_STATUS_NO_MEMORY;
|
|
}
|
|
|
|
src->sockaddr = (struct sockaddr *)from_addr;
|
|
|
|
*nread = 0;
|
|
|
|
gotlen = recvfrom(sock->fd, buf, wantlen, 0,
|
|
src->sockaddr, &from_len);
|
|
if (gotlen == 0) {
|
|
talloc_free(src);
|
|
return NT_STATUS_END_OF_FILE;
|
|
} else if (gotlen == -1) {
|
|
talloc_free(src);
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
|
|
src->sockaddrlen = from_len;
|
|
|
|
if (inet_ntop(AF_INET6, &from_addr->sin6_addr, addrstring, sizeof(addrstring)) == NULL) {
|
|
DEBUG(0, ("Unable to convert address to string: %s\n", strerror(errno)));
|
|
talloc_free(src);
|
|
return NT_STATUS_INTERNAL_ERROR;
|
|
}
|
|
|
|
src->addr = talloc_strdup(src, addrstring);
|
|
if (src->addr == NULL) {
|
|
talloc_free(src);
|
|
return NT_STATUS_NO_MEMORY;
|
|
}
|
|
src->port = ntohs(from_addr->sin6_port);
|
|
|
|
*nread = gotlen;
|
|
*_src = src;
|
|
return NT_STATUS_OK;
|
|
}
|
|
|
|
static NTSTATUS ipv6_sendto(struct socket_context *sock,
|
|
const DATA_BLOB *blob, size_t *sendlen,
|
|
const struct socket_address *dest_addr)
|
|
{
|
|
ssize_t len;
|
|
|
|
if (dest_addr->sockaddr) {
|
|
len = sendto(sock->fd, blob->data, blob->length, 0,
|
|
dest_addr->sockaddr, dest_addr->sockaddrlen);
|
|
} else {
|
|
struct sockaddr_in6 srv_addr;
|
|
struct in6_addr addr;
|
|
|
|
ZERO_STRUCT(srv_addr);
|
|
addr = interpret_addr6(dest_addr->addr);
|
|
if (addr.s6_addr == 0) {
|
|
return NT_STATUS_HOST_UNREACHABLE;
|
|
}
|
|
srv_addr.sin6_addr = addr;
|
|
srv_addr.sin6_port = htons(dest_addr->port);
|
|
srv_addr.sin6_family = PF_INET6;
|
|
|
|
*sendlen = 0;
|
|
|
|
len = sendto(sock->fd, blob->data, blob->length, 0,
|
|
(struct sockaddr *)&srv_addr, sizeof(srv_addr));
|
|
}
|
|
if (len == -1) {
|
|
return map_nt_error_from_unix_common(errno);
|
|
}
|
|
|
|
*sendlen = len;
|
|
|
|
return NT_STATUS_OK;
|
|
}
|
|
|
|
static NTSTATUS ipv6_set_option(struct socket_context *sock, const char *option, const char *val)
|
|
{
|
|
set_socket_options(sock->fd, option);
|
|
return NT_STATUS_OK;
|
|
}
|
|
|
|
static char *ipv6_tcp_get_peer_name(struct socket_context *sock, TALLOC_CTX *mem_ctx)
|
|
{
|
|
struct sockaddr_in6 peer_addr;
|
|
socklen_t len = sizeof(peer_addr);
|
|
struct hostent *he;
|
|
int ret;
|
|
|
|
ret = getpeername(sock->fd, (struct sockaddr *)&peer_addr, &len);
|
|
if (ret == -1) {
|
|
return NULL;
|
|
}
|
|
|
|
he = gethostbyaddr((char *)&peer_addr.sin6_addr, sizeof(peer_addr.sin6_addr), AF_INET6);
|
|
if (he == NULL) {
|
|
return NULL;
|
|
}
|
|
|
|
return talloc_strdup(mem_ctx, he->h_name);
|
|
}
|
|
|
|
static struct socket_address *ipv6_tcp_get_peer_addr(struct socket_context *sock, TALLOC_CTX *mem_ctx)
|
|
{
|
|
struct sockaddr_in6 *peer_addr;
|
|
socklen_t len = sizeof(*peer_addr);
|
|
struct socket_address *peer;
|
|
int ret;
|
|
char addr[128];
|
|
const char *addr_ret;
|
|
|
|
peer = talloc(mem_ctx, struct socket_address);
|
|
if (!peer) {
|
|
return NULL;
|
|
}
|
|
|
|
peer->family = sock->backend_name;
|
|
peer_addr = talloc(peer, struct sockaddr_in6);
|
|
if (!peer_addr) {
|
|
talloc_free(peer);
|
|
return NULL;
|
|
}
|
|
|
|
peer->sockaddr = (struct sockaddr *)peer_addr;
|
|
|
|
ret = getpeername(sock->fd, peer->sockaddr, &len);
|
|
if (ret == -1) {
|
|
talloc_free(peer);
|
|
return NULL;
|
|
}
|
|
|
|
peer->sockaddrlen = len;
|
|
|
|
addr_ret = inet_ntop(AF_INET6, &peer_addr->sin6_addr, addr, sizeof(addr));
|
|
if (addr_ret == NULL) {
|
|
talloc_free(peer);
|
|
return NULL;
|
|
}
|
|
|
|
peer->addr = talloc_strdup(peer, addr_ret);
|
|
if (peer->addr == NULL) {
|
|
talloc_free(peer);
|
|
return NULL;
|
|
}
|
|
|
|
peer->port = ntohs(peer_addr->sin6_port);
|
|
|
|
return peer;
|
|
}
|
|
|
|
static struct socket_address *ipv6_tcp_get_my_addr(struct socket_context *sock, TALLOC_CTX *mem_ctx)
|
|
{
|
|
struct sockaddr_in6 *local_addr;
|
|
socklen_t len = sizeof(*local_addr);
|
|
struct socket_address *local;
|
|
int ret;
|
|
char addrstring[INET6_ADDRSTRLEN];
|
|
|
|
local = talloc(mem_ctx, struct socket_address);
|
|
if (!local) {
|
|
return NULL;
|
|
}
|
|
|
|
local->family = sock->backend_name;
|
|
local_addr = talloc(local, struct sockaddr_in6);
|
|
if (!local_addr) {
|
|
talloc_free(local);
|
|
return NULL;
|
|
}
|
|
|
|
local->sockaddr = (struct sockaddr *)local_addr;
|
|
|
|
ret = getsockname(sock->fd, local->sockaddr, &len);
|
|
if (ret == -1) {
|
|
talloc_free(local);
|
|
return NULL;
|
|
}
|
|
|
|
local->sockaddrlen = len;
|
|
|
|
if (inet_ntop(AF_INET6, &local_addr->sin6_addr, addrstring,
|
|
sizeof(addrstring)) == NULL) {
|
|
DEBUG(0, ("Unable to convert address to string: %s\n",
|
|
strerror(errno)));
|
|
talloc_free(local);
|
|
return NULL;
|
|
}
|
|
|
|
local->addr = talloc_strdup(mem_ctx, addrstring);
|
|
if (!local->addr) {
|
|
talloc_free(local);
|
|
return NULL;
|
|
}
|
|
local->port = ntohs(local_addr->sin6_port);
|
|
|
|
return local;
|
|
}
|
|
|
|
static const struct socket_ops ipv6_tcp_ops = {
|
|
.name = "ipv6",
|
|
.fn_init = ipv6_init,
|
|
.fn_connect = ipv6_tcp_connect,
|
|
.fn_connect_complete = ip_connect_complete,
|
|
.fn_listen = ipv6_listen,
|
|
.fn_accept = ipv6_tcp_accept,
|
|
.fn_recv = ip_recv,
|
|
.fn_recvfrom = ipv6_recvfrom,
|
|
.fn_send = ip_send,
|
|
.fn_sendto = ipv6_sendto,
|
|
.fn_pending = ip_pending,
|
|
.fn_close = ip_close,
|
|
|
|
.fn_set_option = ipv6_set_option,
|
|
|
|
.fn_get_peer_name = ipv6_tcp_get_peer_name,
|
|
.fn_get_peer_addr = ipv6_tcp_get_peer_addr,
|
|
.fn_get_my_addr = ipv6_tcp_get_my_addr,
|
|
|
|
.fn_get_fd = ip_get_fd
|
|
};
|
|
|
|
_PUBLIC_ const struct socket_ops *socket_ipv6_ops(enum socket_type type)
|
|
{
|
|
return &ipv6_tcp_ops;
|
|
}
|
|
|
|
#endif
|