int tca_build_msg(struct rtnl_tc *tca, int type, int flags,
struct nl_msg **result)
{
struct nl_msg *msg;
struct tcmsg tchdr = {
.tcm_family = AF_UNSPEC,
.tcm_ifindex = tca->tc_ifindex,
.tcm_handle = tca->tc_handle,
.tcm_parent = tca->tc_parent,
};
msg = nlmsg_alloc_simple(type, flags);
if (!msg)
return -NLE_NOMEM;
if (nlmsg_append(msg, &tchdr, sizeof(tchdr), NLMSG_ALIGNTO) < 0)
goto nla_put_failure;
if (tca->ce_mask & TCA_ATTR_KIND)
NLA_PUT_STRING(msg, TCA_KIND, tca->tc_kind);
*result = msg;
return 0;
nla_put_failure:
nlmsg_free(msg);
return -NLE_MSGSIZE;
}
void tca_set_kind(struct rtnl_tc *t, const char *kind)
{
strncpy(t->tc_kind, kind, sizeof(t->tc_kind) - 1);
t->ce_mask |= TCA_ATTR_KIND;
}
static int veth_put_attrs(struct nl_msg *msg, struct rtnl_link *link)
{
struct rtnl_link *peer = link->l_info;
struct ifinfomsg ifi;
struct nlattr *data, *info_peer;
memset(&ifi, 0, sizeof ifi);
ifi.ifi_family = peer->l_family;
ifi.ifi_type = peer->l_arptype;
ifi.ifi_index = peer->l_index;
ifi.ifi_flags = peer->l_flags;
ifi.ifi_change = peer->l_change;
if (!(data = nla_nest_start(msg, IFLA_INFO_DATA)))
return -NLE_MSGSIZE;
if (!(info_peer = nla_nest_start(msg, VETH_INFO_PEER)))
return -NLE_MSGSIZE;
if (nlmsg_append(msg, &ifi, sizeof(ifi), NLMSG_ALIGNTO) < 0)
return -NLE_MSGSIZE;
rtnl_link_fill_info(msg, peer);
nla_nest_end(msg, info_peer);
nla_nest_end(msg, data);
return 0;
}
static struct nl_msg *sfq_get_opts(struct rtnl_qdisc *qdisc)
{
struct rtnl_sfq *sfq;
struct tc_sfq_qopt opts;
struct nl_msg *msg;
sfq = sfq_qdisc(qdisc);
if (!sfq)
return NULL;
msg = nlmsg_alloc();
if (!msg)
goto errout;
memset(&opts, 0, sizeof(opts));
opts.quantum = sfq->qs_quantum;
opts.perturb_period = sfq->qs_perturb;
opts.limit = sfq->qs_limit;
if (nlmsg_append(msg, &opts, sizeof(opts), NL_DONTPAD) < 0)
goto errout;
return msg;
errout:
nlmsg_free(msg);
return NULL;
}
static struct nl_msg *fifo_get_opts(struct rtnl_qdisc *qdisc)
{
struct rtnl_fifo *fifo;
struct tc_fifo_qopt opts;
struct nl_msg *msg;
fifo = fifo_qdisc(qdisc);
if (!fifo || !(fifo->qf_mask & SCH_FIFO_ATTR_LIMIT))
return NULL;
msg = nlmsg_build_no_hdr();
if (!msg)
goto errout;
memset(&opts, 0, sizeof(opts));
opts.limit = fifo->qf_limit;
if (nlmsg_append(msg, &opts, sizeof(opts), 0) < 0)
goto errout;
return msg;
errout:
nlmsg_free(msg);
return NULL;
}
static struct nl_msg *build_rule_msg(struct rtnl_rule *tmpl, int cmd, int flags)
{
struct nl_msg *msg;
struct rtmsg rtm = {
.rtm_type = RTN_UNSPEC
};
if (cmd == RTM_NEWRULE)
rtm.rtm_type = RTN_UNICAST;
if (tmpl->ce_mask & RULE_ATTR_FAMILY)
rtm.rtm_family = tmpl->r_family;
if (tmpl->ce_mask & RULE_ATTR_TABLE)
rtm.rtm_table = tmpl->r_table;
if (tmpl->ce_mask & RULE_ATTR_DSFIELD)
rtm.rtm_tos = tmpl->r_dsfield;
if (tmpl->ce_mask & RULE_ATTR_TYPE)
rtm.rtm_type = tmpl->r_type;
if (tmpl->ce_mask & RULE_ATTR_SRC_LEN)
rtm.rtm_src_len = tmpl->r_src_len;
if (tmpl->ce_mask & RULE_ATTR_DST_LEN)
rtm.rtm_dst_len = tmpl->r_dst_len;
msg = nlmsg_alloc_simple(cmd, flags);
if (!msg)
goto nla_put_failure;
if (nlmsg_append(msg, &rtm, sizeof(rtm), NLMSG_ALIGNTO) < 0)
goto nla_put_failure;
if (tmpl->ce_mask & RULE_ATTR_SRC)
NLA_PUT_ADDR(msg, RTA_SRC, tmpl->r_src);
if (tmpl->ce_mask & RULE_ATTR_DST)
NLA_PUT_ADDR(msg, RTA_DST, tmpl->r_dst);
if (tmpl->ce_mask & RULE_ATTR_PRIO)
NLA_PUT_U32(msg, RTA_PRIORITY, tmpl->r_prio);
if (tmpl->ce_mask & RULE_ATTR_MARK)
NLA_PUT_U32(msg, RTA_PROTOINFO, tmpl->r_mark);
if (tmpl->ce_mask & RULE_ATTR_REALMS)
NLA_PUT_U32(msg, RTA_FLOW, tmpl->r_realms);
if (tmpl->ce_mask & RULE_ATTR_IIF)
NLA_PUT_STRING(msg, RTA_IIF, tmpl->r_iif);
return msg;
nla_put_failure:
nlmsg_free(msg);
return NULL;
}
/* return after DAD finishes for all known IPv6 addresses or an error */
int
virNetDevIPWaitDadFinish(virSocketAddrPtr *addrs, size_t count)
{
struct nl_msg *nlmsg = NULL;
struct ifaddrmsg ifa;
struct nlmsghdr *resp = NULL;
unsigned int recvbuflen;
int ret = -1;
bool dad = true;
time_t max_time = time(NULL) + VIR_DAD_WAIT_TIMEOUT;
if (!(nlmsg = nlmsg_alloc_simple(RTM_GETADDR,
NLM_F_REQUEST | NLM_F_DUMP))) {
virReportOOMError();
return -1;
}
memset(&ifa, 0, sizeof(ifa));
/* DAD is for IPv6 adresses only. */
ifa.ifa_family = AF_INET6;
if (nlmsg_append(nlmsg, &ifa, sizeof(ifa), NLMSG_ALIGNTO) < 0) {
virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("allocated netlink buffer is too small"));
goto cleanup;
}
/* Periodically query netlink until DAD finishes on all known addresses. */
while (dad && time(NULL) < max_time) {
if (virNetlinkCommand(nlmsg, &resp, &recvbuflen, 0, 0,
NETLINK_ROUTE, 0) < 0)
goto cleanup;
if (virNetlinkGetErrorCode(resp, recvbuflen) < 0) {
virReportError(VIR_ERR_SYSTEM_ERROR, "%s",
_("error reading DAD state information"));
goto cleanup;
}
/* Parse response. */
dad = virNetDevIPParseDadStatus(resp, recvbuflen, addrs, count);
if (dad)
usleep(1000 * 10);
VIR_FREE(resp);
}
/* Check timeout. */
if (dad) {
virReportError(VIR_ERR_SYSTEM_ERROR,
_("Duplicate Address Detection "
"not finished in %d seconds"), VIR_DAD_WAIT_TIMEOUT);
} else {
ret = 0;
}
cleanup:
VIR_FREE(resp);
nlmsg_free(nlmsg);
return ret;
}
int rtnl_tc_msg_build(struct rtnl_tc *tc, int type, int flags,
struct nl_msg **result)
{
struct nl_msg *msg;
struct rtnl_tc_ops *ops;
struct tcmsg tchdr = {
.tcm_family = AF_UNSPEC,
.tcm_ifindex = tc->tc_ifindex,
.tcm_handle = tc->tc_handle,
.tcm_parent = tc->tc_parent,
};
int err = -NLE_MSGSIZE;
msg = nlmsg_alloc_simple(type, flags);
if (!msg)
return -NLE_NOMEM;
if (nlmsg_append(msg, &tchdr, sizeof(tchdr), NLMSG_ALIGNTO) < 0)
goto nla_put_failure;
if (tc->ce_mask & TCA_ATTR_KIND)
NLA_PUT_STRING(msg, TCA_KIND, tc->tc_kind);
ops = rtnl_tc_get_ops(tc);
if (ops && (ops->to_msg_fill || ops->to_msg_fill_raw)) {
struct nlattr *opts;
void *data = rtnl_tc_data(tc);
if (ops->to_msg_fill) {
if (!(opts = nla_nest_start(msg, TCA_OPTIONS)))
goto nla_put_failure;
if ((err = ops->to_msg_fill(tc, data, msg)) < 0)
goto nla_put_failure;
nla_nest_end(msg, opts);
} else if ((err = ops->to_msg_fill_raw(tc, data, msg)) < 0)
goto nla_put_failure;
}
*result = msg;
return 0;
nla_put_failure:
nlmsg_free(msg);
return err;
}
void tca_set_kind(struct rtnl_tc *t, const char *kind)
{
strncpy(t->tc_kind, kind, sizeof(t->tc_kind) - 1);
t->ce_mask |= TCA_ATTR_KIND;
}
TError TNlCgFilter::Create(const TNlLink &link) {
TError error = TError::Success();
struct nl_msg *msg;
int ret;
struct tcmsg tchdr;
tchdr.tcm_family = AF_UNSPEC;
tchdr.tcm_ifindex = link.GetIndex();
tchdr.tcm_handle = Handle;
tchdr.tcm_parent = Parent;
tchdr.tcm_info = TC_H_MAKE(FilterPrio << 16, htons(ETH_P_ALL));
msg = nlmsg_alloc_simple(RTM_NEWTFILTER, NLM_F_EXCL|NLM_F_CREATE);
if (!msg)
return TError(EError::Unknown, "Unable to add filter: no memory");
ret = nlmsg_append(msg, &tchdr, sizeof(tchdr), NLMSG_ALIGNTO);
if (ret < 0) {
error = TError(EError::Unknown, std::string("Unable to add filter: ") + nl_geterror(ret));
goto free_msg;
}
ret = nla_put(msg, TCA_KIND, strlen(FilterType) + 1, FilterType);
if (ret < 0) {
error = TError(EError::Unknown, std::string("Unable to add filter: ") + nl_geterror(ret));
goto free_msg;
}
ret = nla_put(msg, TCA_OPTIONS, 0, NULL);
if (ret < 0) {
error = TError(EError::Unknown, std::string("Unable to add filter: ") + nl_geterror(ret));
goto free_msg;
}
L() << "netlink " << link.GetDesc()
<< ": add tfilter id 0x" << std::hex << Handle
<< " parent 0x" << Parent << std::dec << std::endl;
ret = nl_send_sync(link.GetSock(), msg);
if (ret)
error = TError(EError::Unknown, std::string("Unable to add filter: ") + nl_geterror(ret));
if (!Exists(link))
error = TError(EError::Unknown, "BUG: created filter doesn't exist");
return error;
free_msg:
nlmsg_free(msg);
return error;
}
int nl_dect_cell_build_msg(struct nl_msg *msg, struct nl_dect_cell *cell)
{
struct dectmsg dm = {
.dm_index = cell->c_index,
};
if (nlmsg_append(msg, &dm, sizeof(dm), NLMSG_ALIGNTO) < 0)
goto nla_put_failure;
if (cell->ce_mask & CELL_ATTR_NAME)
NLA_PUT_STRING(msg, DECTA_CELL_NAME, cell->c_name);
if (cell->ce_mask & CELL_ATTR_FLAGS)
NLA_PUT_U32(msg, DECTA_CELL_FLAGS, cell->c_flags);
if (cell->ce_mask & CELL_ATTR_LINK)
NLA_PUT_U8(msg, DECTA_CELL_CLUSTER, cell->c_link);
return 0;
nla_put_failure:
return -NLE_MSGSIZE;
}
static struct trans_tbl cell_flags[] = {
__ADD(DECT_CELL_CCP, ccp)
__ADD(DECT_CELL_SLAVE, slave)
__ADD(DECT_CELL_MONITOR, monitor)
};
char *nl_dect_cell_flags2str(uint32_t flags, char *buf, size_t len)
{
return __flags2str(flags, buf, len, cell_flags, ARRAY_SIZE(cell_flags));
}
uint32_t nl_dect_cell_str2flags(const char *str)
{
return __str2flags(str, cell_flags, ARRAY_SIZE(cell_flags));
}
/** @cond SKIP */
struct nl_object_ops nl_dect_cell_obj_ops = {
.oo_name = "nl_dect/cell",
.oo_size = sizeof(struct nl_dect_cell),
.oo_free_data = cell_free_data,
.oo_dump = {
[NL_DUMP_LINE] = cell_dump,
},
.oo_id_attrs = CELL_ATTR_NAME,
static struct nl_msg * prepare_tcmsg(int type, int flags, uint32_t parent, uint32_t handle, uint32_t info) {
struct nl_msg *msg = nlmsg_alloc_simple(type, flags);
if (!msg)
exit_errno("nlmsg_alloc_simple");
struct tcmsg tcmsg;
memset(&tcmsg, 0, sizeof(tcmsg));
tcmsg.tcm_family = AF_UNSPEC;
tcmsg.tcm_ifindex = ifindex;
tcmsg.tcm_parent = parent;
tcmsg.tcm_handle = handle;
tcmsg.tcm_info = info;
nlmsg_append(msg, &tcmsg, sizeof(tcmsg), NLMSG_ALIGNTO);
return msg;
}
static int red_msg_fill(struct rtnl_tc *tc, void *data, struct nl_msg *msg)
{
struct rtnl_red *red = data;
if (!red)
BUG();
#if 0
memset(&opts, 0, sizeof(opts));
opts.quantum = sfq->qs_quantum;
opts.perturb_period = sfq->qs_perturb;
opts.limit = sfq->qs_limit;
if (nlmsg_append(msg, &opts, sizeof(opts), NL_DONTPAD) < 0)
goto errout;
#endif
return -NLE_OPNOTSUPP;
}
/**
* Builds a netlink request message to do a lookup
* @arg req Requested match.
* @arg flags additional netlink message flags
* @arg result Result pointer
*
* Builds a new netlink message requesting a change of link attributes.
* The netlink message header isn't fully equipped with all relevant
* fields and must be sent out via nl_send_auto_complete() or
* supplemented as needed.
* \a old must point to a link currently configured in the kernel
* and \a tmpl must contain the attributes to be changed set via
* \c rtnl_link_set_* functions.
*
* @return 0 on success or a negative error code.
*/
int flnl_lookup_build_request(struct flnl_request *req, int flags,
struct nl_msg **result)
{
struct nl_msg *msg;
struct nl_addr *addr;
uint64_t fwmark;
int tos, scope, table;
struct fib_result_nl fr = {0};
fwmark = flnl_request_get_fwmark(req);
tos = flnl_request_get_tos(req);
scope = flnl_request_get_scope(req);
table = flnl_request_get_table(req);
fr.fl_fwmark = fwmark != UINT_LEAST64_MAX ? fwmark : 0;
fr.fl_tos = tos >= 0 ? tos : 0;
fr.fl_scope = scope >= 0 ? scope : RT_SCOPE_UNIVERSE;
fr.tb_id_in = table >= 0 ? table : RT_TABLE_UNSPEC;
addr = flnl_request_get_addr(req);
if (!addr)
return -NLE_MISSING_ATTR;
fr.fl_addr = *(uint32_t *) nl_addr_get_binary_addr(addr);
msg = nlmsg_alloc_simple(0, flags);
if (!msg)
return -NLE_NOMEM;
if (nlmsg_append(msg, &fr, sizeof(fr), NLMSG_ALIGNTO) < 0)
goto errout;
*result = msg;
return 0;
errout:
nlmsg_free(msg);
return -NLE_MSGSIZE;
}
/**
* Builds a netlink request message to do a lookup
* @arg req Requested match.
* @arg flags additional netlink message flags
*
* Builds a new netlink message requesting a change of link attributes.
* The netlink message header isn't fully equipped with all relevant
* fields and must be sent out via nl_send_auto_complete() or
* supplemented as needed.
* \a old must point to a link currently configured in the kernel
* and \a tmpl must contain the attributes to be changed set via
* \c rtnl_link_set_* functions.
*
* @return New netlink message
* @note Not all attributes can be changed, see
* \ref link_changeable "Changeable Attributes" for more details.
*/
struct nl_msg *flnl_lookup_build_request(struct flnl_request *req, int flags)
{
struct nl_msg *msg;
struct nl_addr *addr;
uint64_t fwmark;
int tos, scope, table;
struct fib_result_nl fr = {0};
fwmark = flnl_request_get_fwmark(req);
tos = flnl_request_get_tos(req);
scope = flnl_request_get_scope(req);
table = flnl_request_get_table(req);
fr.fl_fwmark = fwmark != UINT_LEAST64_MAX ? fwmark : 0;
fr.fl_tos = tos >= 0 ? tos : 0;
fr.fl_scope = scope >= 0 ? scope : RT_SCOPE_UNIVERSE;
fr.tb_id_in = table >= 0 ? table : RT_TABLE_UNSPEC;
addr = flnl_request_get_addr(req);
if (!addr) {
nl_error(EINVAL, "Request must specify the address");
return NULL;
}
fr.fl_addr = *(uint32_t *) nl_addr_get_binary_addr(addr);
msg = nlmsg_alloc_simple(0, flags);
if (!msg)
goto errout;
if (nlmsg_append(msg, &fr, sizeof(fr), NLMSG_ALIGNTO) < 0)
goto errout;
return msg;
errout:
nlmsg_free(msg);
return NULL;
}
/**
* Construct and transmit a Netlink message
* @arg sk Netlink socket (required)
* @arg type Netlink message type (required)
* @arg flags Netlink message flags (optional)
* @arg buf Data buffer (optional)
* @arg size Size of data buffer (optional)
*
* Allocates a new Netlink message based on `type` and `flags`. If `buf`
* points to payload of length `size` that payload will be appended to the
* message.
*
* Sends out the message using `nl_send_auto()` and frees the message
* afterwards.
*
* @see nl_send_auto()
*
* @return Number of characters sent on success or a negative error code.
* @retval -NLE_NOMEM Unable to allocate Netlink message
*/
int nl_send_simple(struct nl_sock *sk, int type, int flags, void *buf,
size_t size)
{
int err;
struct nl_msg *msg;
msg = nlmsg_alloc_simple(type, flags);
if (!msg)
return -NLE_NOMEM;
if (buf && size) {
err = nlmsg_append(msg, buf, size, NLMSG_ALIGNTO);
if (err < 0)
goto errout;
}
err = nl_send_auto(sk, msg);
errout:
nlmsg_free(msg);
return err;
}
int xfrmnl_ae_build_get_request(struct nl_addr* daddr, unsigned int spi, unsigned int protocol,
unsigned int mark_mask, unsigned int mark_value, struct nl_msg **result)
{
struct nl_msg *msg;
struct xfrm_aevent_id ae_id;
struct xfrmnl_mark mark;
if (!daddr || !spi)
{
fprintf(stderr, "APPLICATION BUG: %s:%d:%s: A valid destination address, spi must be specified\n",
__FILE__, __LINE__, __PRETTY_FUNCTION__);
assert(0);
return -NLE_MISSING_ATTR;
}
memset(&ae_id, 0, sizeof(ae_id));
memcpy (&ae_id.sa_id.daddr, nl_addr_get_binary_addr (daddr), sizeof (uint8_t) * nl_addr_get_len (daddr));
ae_id.sa_id.spi = htonl(spi);
ae_id.sa_id.family = nl_addr_get_family (daddr);
ae_id.sa_id.proto = protocol;
if (!(msg = nlmsg_alloc_simple(XFRM_MSG_GETAE, 0)))
return -NLE_NOMEM;
if (nlmsg_append(msg, &ae_id, sizeof(ae_id), NLMSG_ALIGNTO) < 0)
goto nla_put_failure;
mark.m = mark_mask;
mark.v = mark_value;
NLA_PUT (msg, XFRMA_MARK, sizeof (struct xfrmnl_mark), &mark);
*result = msg;
return 0;
nla_put_failure:
nlmsg_free(msg);
return -NLE_MSGSIZE;
}
static struct nl_msg *build_addr_msg(struct rtnl_addr *tmpl, int cmd, int flags)
{
struct nl_msg *msg;
struct ifaddrmsg am = {
.ifa_family = tmpl->a_family,
.ifa_index = tmpl->a_ifindex,
.ifa_prefixlen = tmpl->a_prefixlen,
};
if (tmpl->a_mask & ADDR_ATTR_FLAGS)
am.ifa_flags = tmpl->a_flags;
if (tmpl->a_mask & ADDR_ATTR_SCOPE)
am.ifa_scope = tmpl->a_scope;
else {
/* compatibility hack */
if (tmpl->a_family == AF_INET &&
tmpl->a_mask & ADDR_ATTR_LOCAL &&
*((char *) nl_addr_get_binary_addr(tmpl->a_local)) == 127)
am.ifa_scope = RT_SCOPE_HOST;
else
am.ifa_scope = RT_SCOPE_UNIVERSE;
}
msg = nlmsg_build_simple(cmd, flags);
if (!msg)
goto nla_put_failure;
if (nlmsg_append(msg, &am, sizeof(am), 1) < 0)
goto nla_put_failure;
if (tmpl->a_mask & ADDR_ATTR_LOCAL)
NLA_PUT_ADDR(msg, IFA_LOCAL, tmpl->a_local);
if (tmpl->a_mask & ADDR_ATTR_PEER)
NLA_PUT_ADDR(msg, IFA_ADDRESS, tmpl->a_peer);
else
NLA_PUT_ADDR(msg, IFA_ADDRESS, tmpl->a_local);
if (tmpl->a_mask & ADDR_ATTR_LABEL)
NLA_PUT_STRING(msg, IFA_LABEL, tmpl->a_label);
if (tmpl->a_mask & ADDR_ATTR_BROADCAST)
NLA_PUT_ADDR(msg, IFA_BROADCAST, tmpl->a_bcast);
if (tmpl->a_mask & ADDR_ATTR_ANYCAST)
NLA_PUT_ADDR(msg, IFA_ANYCAST, tmpl->a_anycast);
return msg;
nla_put_failure:
nlmsg_free(msg);
return NULL;
}
/**
* @name Address Addition
* @{
*/
/**
* Build netlink request message to request addition of new address
* @arg addr Address object representing the new address.
* @arg flags Additional netlink message flags.
*
* Builds a new netlink message requesting the addition of a new
* address. The netlink message header isn't fully equipped with
* all relevant fields and must thus be sent out via nl_send_auto_complete()
* or supplemented as needed.
*
* Minimal required attributes:
* - interface index (rtnl_addr_set_ifindex())
* - local address (rtnl_addr_set_local())
*
* The scope will default to universe except for loopback addresses in
* which case a host scope is used if not specified otherwise.
*
* @note Free the memory after usage using nlmsg_free().
* @return Newly allocated netlink message or NULL if an error occured.
*/
struct nl_msg *rtnl_addr_build_add_request(struct rtnl_addr *addr, int flags)
{
int required = ADDR_ATTR_IFINDEX | ADDR_ATTR_FAMILY |
ADDR_ATTR_PREFIXLEN | ADDR_ATTR_LOCAL;
if ((addr->a_mask & required) != required) {
nl_error(EINVAL, "Missing mandatory attributes, required are: "
"ifindex, family, prefixlen, local address.");
return NULL;
}
return build_addr_msg(addr, RTM_NEWADDR, NLM_F_CREATE | flags);
}
/**
* virNetlinkCommand:
* @nlmsg: pointer to netlink message
* @respbuf: pointer to pointer where response buffer will be allocated
* @respbuflen: pointer to integer holding the size of the response buffer
* on return of the function.
* @src_pid: the pid of the process to send a message
* @dst_pid: the pid of the process to talk to, i.e., pid = 0 for kernel
* @protocol: netlink protocol
* @groups: the group identifier
*
* Send the given message to the netlink layer and receive response.
* Returns 0 on success, -1 on error. In case of error, no response
* buffer will be returned.
*/
int virNetlinkCommand(struct nl_msg *nl_msg,
struct nlmsghdr **resp, unsigned int *respbuflen,
uint32_t src_pid, uint32_t dst_pid,
unsigned int protocol, unsigned int groups)
{
int ret = -1;
struct sockaddr_nl nladdr = {
.nl_family = AF_NETLINK,
.nl_pid = dst_pid,
.nl_groups = 0,
};
ssize_t nbytes;
struct pollfd fds[1];
int fd;
int n;
struct nlmsghdr *nlmsg = nlmsg_hdr(nl_msg);
virNetlinkHandle *nlhandle = NULL;
int len = 0;
if (protocol >= MAX_LINKS) {
virReportSystemError(EINVAL,
_("invalid protocol argument: %d"), protocol);
goto cleanup;
}
if (!(nlhandle = virNetlinkCreateSocket(protocol)))
goto cleanup;
fd = nl_socket_get_fd(nlhandle);
if (fd < 0) {
virReportSystemError(errno,
"%s", _("cannot get netlink socket fd"));
goto cleanup;
}
if (groups && nl_socket_add_membership(nlhandle, groups) < 0) {
virReportSystemError(errno,
"%s", _("cannot add netlink membership"));
goto cleanup;
}
nlmsg_set_dst(nl_msg, &nladdr);
nlmsg->nlmsg_pid = src_pid ? src_pid : getpid();
nbytes = nl_send_auto_complete(nlhandle, nl_msg);
if (nbytes < 0) {
virReportSystemError(errno,
"%s", _("cannot send to netlink socket"));
goto cleanup;
}
memset(fds, 0, sizeof(fds));
fds[0].fd = fd;
fds[0].events = POLLIN;
n = poll(fds, ARRAY_CARDINALITY(fds), NETLINK_ACK_TIMEOUT_S);
if (n <= 0) {
if (n < 0)
virReportSystemError(errno, "%s",
_("error in poll call"));
if (n == 0)
virReportSystemError(ETIMEDOUT, "%s",
_("no valid netlink response was received"));
goto cleanup;
}
len = nl_recv(nlhandle, &nladdr, (unsigned char **)resp, NULL);
if (len == 0) {
virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("nl_recv failed - returned 0 bytes"));
goto cleanup;
}
if (len < 0) {
virReportSystemError(errno, "%s", _("nl_recv failed"));
goto cleanup;
}
ret = 0;
*respbuflen = len;
cleanup:
if (ret < 0) {
*resp = NULL;
*respbuflen = 0;
}
virNetlinkFree(nlhandle);
return ret;
}
/**
* virNetlinkDumpLink:
*
* @ifname: The name of the interface; only use if ifindex <= 0
* @ifindex: The interface index; may be <= 0 if ifname is given
* @data: Gets a pointer to the raw data from netlink.
MUST BE FREED BY CALLER!
* @nlattr: Pointer to a pointer of netlink attributes that will contain
* the results
* @src_pid: pid used for nl_pid of the local end of the netlink message
* (0 == "use getpid()")
* @dst_pid: pid of destination nl_pid if the kernel
* is not the target of the netlink message but it is to be
* sent to another process (0 if sending to the kernel)
*
* Get information from netlink about an interface given its name or index.
*
* Returns 0 on success, -1 on fatal error.
*/
int
virNetlinkDumpLink(const char *ifname, int ifindex,
void **nlData, struct nlattr **tb,
uint32_t src_pid, uint32_t dst_pid)
{
int rc = -1;
struct nlmsghdr *resp = NULL;
struct nlmsgerr *err;
struct ifinfomsg ifinfo = {
.ifi_family = AF_UNSPEC,
.ifi_index = ifindex
};
unsigned int recvbuflen;
struct nl_msg *nl_msg;
if (ifname && ifindex <= 0 && virNetDevGetIndex(ifname, &ifindex) < 0)
return -1;
ifinfo.ifi_index = ifindex;
nl_msg = nlmsg_alloc_simple(RTM_GETLINK, NLM_F_REQUEST);
if (!nl_msg) {
virReportOOMError();
return -1;
}
if (nlmsg_append(nl_msg, &ifinfo, sizeof(ifinfo), NLMSG_ALIGNTO) < 0)
goto buffer_too_small;
if (ifname) {
if (nla_put(nl_msg, IFLA_IFNAME, strlen(ifname)+1, ifname) < 0)
goto buffer_too_small;
}
# ifdef RTEXT_FILTER_VF
/* if this filter exists in the kernel's netlink implementation,
* we need to set it, otherwise the response message will not
* contain the IFLA_VFINFO_LIST that we're looking for.
*/
{
uint32_t ifla_ext_mask = RTEXT_FILTER_VF;
if (nla_put(nl_msg, IFLA_EXT_MASK,
sizeof(ifla_ext_mask), &ifla_ext_mask) < 0) {
goto buffer_too_small;
}
}
# endif
if (virNetlinkCommand(nl_msg, &resp, &recvbuflen,
src_pid, dst_pid, NETLINK_ROUTE, 0) < 0)
goto cleanup;
if (recvbuflen < NLMSG_LENGTH(0) || resp == NULL)
goto malformed_resp;
switch (resp->nlmsg_type) {
case NLMSG_ERROR:
err = (struct nlmsgerr *)NLMSG_DATA(resp);
if (resp->nlmsg_len < NLMSG_LENGTH(sizeof(*err)))
goto malformed_resp;
if (err->error) {
virReportSystemError(-err->error,
_("error dumping %s (%d) interface"),
ifname, ifindex);
goto cleanup;
}
break;
case GENL_ID_CTRL:
case NLMSG_DONE:
rc = nlmsg_parse(resp, sizeof(struct ifinfomsg),
tb, IFLA_MAX, NULL);
if (rc < 0)
goto malformed_resp;
break;
default:
goto malformed_resp;
}
rc = 0;
cleanup:
nlmsg_free(nl_msg);
if (rc < 0)
VIR_FREE(resp);
*nlData = resp;
return rc;
malformed_resp:
virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("malformed netlink response message"));
goto cleanup;
buffer_too_small:
virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("allocated netlink buffer is too small"));
goto cleanup;
}
int ompi_btl_usnic_nl_ip_rt_lookup(struct usnic_rtnl_sk *unlsk,
const char *src_ifname,
uint32_t src_addr,
uint32_t dst_addr, int *metric)
{
struct nl_msg *nlm;
struct rtmsg rmsg;
struct nl_lookup_arg arg;
int msg_cnt;
int err;
int oif;
oif = if_nametoindex(src_ifname);
if (0 == oif) {
return errno;
}
arg.nh_addr = 0;
arg.oif = oif;
arg.found = 0;
arg.replied = 0;
arg.unlsk = unlsk;
arg.msg_count = msg_cnt = 0;
memset(&rmsg, 0, sizeof(rmsg));
rmsg.rtm_family = AF_INET;
rmsg.rtm_dst_len = sizeof(dst_addr)*8;
rmsg.rtm_src_len = sizeof(src_addr)*8;
nlm = nlmsg_alloc_simple(RTM_GETROUTE, 0);
nlmsg_append(nlm, &rmsg, sizeof(rmsg), NLMSG_ALIGNTO);
nla_put_u32(nlm, RTA_DST, dst_addr);
nla_put_u32(nlm, RTA_SRC, src_addr);
err = rtnl_send_ack_disable(unlsk, nlm);
nlmsg_free(nlm);
if (err < 0) {
usnic_err("Failed to send rtnl query %s\n", nl_geterror(err));
return err;
}
err = nl_socket_modify_cb(unlsk->sock, NL_CB_MSG_IN, NL_CB_CUSTOM,
rtnl_raw_parse_cb, &arg);
if (err != 0) {
usnic_err("Failed to setup callback function, error %s\n",
nl_geterror(err));
return err;
}
while (!arg.replied) {
err = nl_recvmsgs_default(unlsk->sock);
if (err < 0) {
/* err will be returned as -NLE_AGAIN if the socket times out */
usnic_err("Failed to receive rtnl query results %s\n",
nl_geterror(err));
return err;
}
}
if (arg.found) {
if (metric != NULL) {
*metric = arg.metric;
}
return 0;
}
else {
return -1;
}
}
static struct nl_msg *
virNetDevCreateNetlinkAddressMessage(int messageType,
const char *ifname,
virSocketAddr *addr,
unsigned int prefix,
virSocketAddr *broadcast,
virSocketAddr *peer)
{
struct nl_msg *nlmsg = NULL;
struct ifaddrmsg ifa;
unsigned int ifindex;
void *addrData = NULL;
void *peerData = NULL;
void *broadcastData = NULL;
size_t addrDataLen;
if (virNetDevGetIPAddressBinary(addr, &addrData, &addrDataLen) < 0)
return NULL;
if (peer && VIR_SOCKET_ADDR_VALID(peer)) {
if (virNetDevGetIPAddressBinary(peer, &peerData, &addrDataLen) < 0)
return NULL;
} else if (broadcast) {
if (virNetDevGetIPAddressBinary(broadcast, &broadcastData,
&addrDataLen) < 0)
return NULL;
}
/* Get the interface index */
if ((ifindex = if_nametoindex(ifname)) == 0)
return NULL;
if (!(nlmsg = nlmsg_alloc_simple(messageType,
NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL))) {
virReportOOMError();
return NULL;
}
memset(&ifa, 0, sizeof(ifa));
ifa.ifa_prefixlen = prefix;
ifa.ifa_family = VIR_SOCKET_ADDR_FAMILY(addr);
ifa.ifa_index = ifindex;
ifa.ifa_scope = 0;
if (nlmsg_append(nlmsg, &ifa, sizeof(ifa), NLMSG_ALIGNTO) < 0)
goto buffer_too_small;
if (nla_put(nlmsg, IFA_LOCAL, addrDataLen, addrData) < 0)
goto buffer_too_small;
if (peerData) {
if (nla_put(nlmsg, IFA_ADDRESS, addrDataLen, peerData) < 0)
goto buffer_too_small;
}
if (broadcastData) {
if (nla_put(nlmsg, IFA_BROADCAST, addrDataLen, broadcastData) < 0)
goto buffer_too_small;
}
return nlmsg;
buffer_too_small:
virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("allocated netlink buffer is too small"));
nlmsg_free(nlmsg);
return NULL;
}
/**
* virNetDevIPRouteAdd:
* @ifname: the interface name
* @addr: the IP network address (IPv4 or IPv6)
* @prefix: number of 1 bits in the netmask
* @gateway: via address for route (same as @addr)
*
* Add a route for a network IP address to an interface. This function
* *does not* remove any previously added IP static routes.
*
* Returns 0 in case of success or -1 in case of error.
*/
int
virNetDevIPRouteAdd(const char *ifname,
virSocketAddrPtr addr,
unsigned int prefix,
virSocketAddrPtr gateway,
unsigned int metric)
{
int ret = -1;
struct nl_msg *nlmsg = NULL;
struct nlmsghdr *resp = NULL;
unsigned int recvbuflen;
unsigned int ifindex;
struct rtmsg rtmsg;
void *gatewayData = NULL;
void *addrData = NULL;
size_t addrDataLen;
int errCode;
virSocketAddr defaultAddr;
virSocketAddrPtr actualAddr;
char *toStr = NULL;
char *viaStr = NULL;
actualAddr = addr;
/* If we have no valid network address, then use the default one */
if (!addr || !VIR_SOCKET_ADDR_VALID(addr)) {
VIR_DEBUG("computing default address");
int family = VIR_SOCKET_ADDR_FAMILY(gateway);
if (family == AF_INET) {
if (virSocketAddrParseIPv4(&defaultAddr, VIR_SOCKET_ADDR_IPV4_ALL) < 0)
goto cleanup;
} else {
if (virSocketAddrParseIPv6(&defaultAddr, VIR_SOCKET_ADDR_IPV6_ALL) < 0)
goto cleanup;
}
actualAddr = &defaultAddr;
}
toStr = virSocketAddrFormat(actualAddr);
viaStr = virSocketAddrFormat(gateway);
VIR_DEBUG("Adding route %s/%d via %s", toStr, prefix, viaStr);
if (virNetDevGetIPAddressBinary(actualAddr, &addrData, &addrDataLen) < 0 ||
virNetDevGetIPAddressBinary(gateway, &gatewayData, &addrDataLen) < 0)
goto cleanup;
/* Get the interface index */
if ((ifindex = if_nametoindex(ifname)) == 0)
goto cleanup;
if (!(nlmsg = nlmsg_alloc_simple(RTM_NEWROUTE,
NLM_F_REQUEST | NLM_F_CREATE |
NLM_F_EXCL))) {
virReportOOMError();
goto cleanup;
}
memset(&rtmsg, 0, sizeof(rtmsg));
rtmsg.rtm_family = VIR_SOCKET_ADDR_FAMILY(gateway);
rtmsg.rtm_table = RT_TABLE_MAIN;
rtmsg.rtm_scope = RT_SCOPE_UNIVERSE;
rtmsg.rtm_protocol = RTPROT_BOOT;
rtmsg.rtm_type = RTN_UNICAST;
rtmsg.rtm_dst_len = prefix;
if (nlmsg_append(nlmsg, &rtmsg, sizeof(rtmsg), NLMSG_ALIGNTO) < 0)
goto buffer_too_small;
if (prefix > 0 && nla_put(nlmsg, RTA_DST, addrDataLen, addrData) < 0)
goto buffer_too_small;
if (nla_put(nlmsg, RTA_GATEWAY, addrDataLen, gatewayData) < 0)
goto buffer_too_small;
if (nla_put_u32(nlmsg, RTA_OIF, ifindex) < 0)
goto buffer_too_small;
if (metric > 0 && nla_put_u32(nlmsg, RTA_PRIORITY, metric) < 0)
goto buffer_too_small;
if (virNetlinkCommand(nlmsg, &resp, &recvbuflen, 0, 0,
NETLINK_ROUTE, 0) < 0)
goto cleanup;
if ((errCode = virNetlinkGetErrorCode(resp, recvbuflen)) < 0) {
virReportSystemError(errCode, _("Error adding route to %s"), ifname);
goto cleanup;
}
ret = 0;
cleanup:
VIR_FREE(toStr);
VIR_FREE(viaStr);
nlmsg_free(nlmsg);
return ret;
buffer_too_small:
virReportError(VIR_ERR_INTERNAL_ERROR, "%s",
_("allocated netlink buffer is too small"));
goto cleanup;
}
static void copy_cacheinfo_into_route(struct rta_cacheinfo *ci,
struct rtnl_route *route)
{
struct rtnl_rtcacheinfo nci = {
.rtci_clntref = ci->rta_clntref,
.rtci_last_use = ci->rta_lastuse,
.rtci_expires = ci->rta_expires,
.rtci_error = ci->rta_error,
.rtci_used = ci->rta_used,
.rtci_id = ci->rta_id,
.rtci_ts = ci->rta_ts,
.rtci_tsage = ci->rta_tsage,
};
rtnl_route_set_cacheinfo(route, &nci);
}
static int route_msg_parser(struct nl_cache_ops *ops, struct sockaddr_nl *who,
struct nlmsghdr *nlh, struct nl_parser_param *pp)
{
struct rtmsg *rtm;
struct rtnl_route *route;
struct nlattr *tb[RTA_MAX + 1];
struct nl_addr *src = NULL, *dst = NULL, *addr;
int err;
route = rtnl_route_alloc();
if (!route) {
err = nl_errno(ENOMEM);
goto errout;
}
route->ce_msgtype = nlh->nlmsg_type;
err = nlmsg_parse(nlh, sizeof(struct rtmsg), tb, RTA_MAX,
route_policy);
if (err < 0)
goto errout;
rtm = nlmsg_data(nlh);
rtnl_route_set_family(route, rtm->rtm_family);
rtnl_route_set_tos(route, rtm->rtm_tos);
rtnl_route_set_table(route, rtm->rtm_table);
rtnl_route_set_type(route, rtm->rtm_type);
rtnl_route_set_scope(route, rtm->rtm_scope);
rtnl_route_set_protocol(route, rtm->rtm_protocol);
rtnl_route_set_flags(route, rtm->rtm_flags);
if (tb[RTA_DST]) {
dst = nla_get_addr(tb[RTA_DST], rtm->rtm_family);
if (dst == NULL)
goto errout_errno;
} else {
dst = nl_addr_alloc(0);
nl_addr_set_family(dst, rtm->rtm_family);
}
nl_addr_set_prefixlen(dst, rtm->rtm_dst_len);
err = rtnl_route_set_dst(route, dst);
if (err < 0)
goto errout;
nl_addr_put(dst);
if (tb[RTA_SRC]) {
src = nla_get_addr(tb[RTA_SRC], rtm->rtm_family);
if (src == NULL)
goto errout_errno;
} else if (rtm->rtm_src_len)
src = nl_addr_alloc(0);
if (src) {
nl_addr_set_prefixlen(src, rtm->rtm_src_len);
rtnl_route_set_src(route, src);
nl_addr_put(src);
}
if (tb[RTA_IIF])
rtnl_route_set_iif(route, nla_get_string(tb[RTA_IIF]));
if (tb[RTA_OIF])
rtnl_route_set_oif(route, nla_get_u32(tb[RTA_OIF]));
if (tb[RTA_GATEWAY]) {
addr = nla_get_addr(tb[RTA_GATEWAY], route->rt_family);
if (addr == NULL)
goto errout_errno;
rtnl_route_set_gateway(route, addr);
nl_addr_put(addr);
}
if (tb[RTA_PRIORITY])
rtnl_route_set_prio(route, nla_get_u32(tb[RTA_PRIORITY]));
if (tb[RTA_PREFSRC]) {
addr = nla_get_addr(tb[RTA_PREFSRC], route->rt_family);
if (addr == NULL)
goto errout_errno;
rtnl_route_set_pref_src(route, addr);
nl_addr_put(addr);
}
if (tb[RTA_METRICS]) {
struct nlattr *mtb[RTAX_MAX + 1];
int i;
err = nla_parse_nested(mtb, RTAX_MAX, tb[RTA_METRICS], NULL);
if (err < 0)
goto errout;
for (i = 1; i <= RTAX_MAX; i++) {
if (mtb[i] && nla_len(mtb[i]) >= sizeof(uint32_t)) {
uint32_t m = nla_get_u32(mtb[i]);
if (rtnl_route_set_metric(route, i, m) < 0)
goto errout_errno;
}
}
}
if (tb[RTA_MULTIPATH]) {
struct rtnl_nexthop *nh;
struct rtnexthop *rtnh = nla_data(tb[RTA_MULTIPATH]);
size_t tlen = nla_len(tb[RTA_MULTIPATH]);
while (tlen >= sizeof(*rtnh) && tlen >= rtnh->rtnh_len) {
nh = rtnl_route_nh_alloc();
if (!nh)
goto errout;
rtnl_route_nh_set_weight(nh, rtnh->rtnh_hops);
rtnl_route_nh_set_ifindex(nh, rtnh->rtnh_ifindex);
rtnl_route_nh_set_flags(nh, rtnh->rtnh_flags);
if (rtnh->rtnh_len > sizeof(*rtnh)) {
struct nlattr *ntb[RTA_MAX + 1];
nla_parse(ntb, RTA_MAX, (struct nlattr *)
RTNH_DATA(rtnh),
rtnh->rtnh_len - sizeof(*rtnh),
route_policy);
if (ntb[RTA_GATEWAY]) {
nh->rtnh_gateway = nla_get_addr(
ntb[RTA_GATEWAY],
route->rt_family);
nh->rtnh_mask = NEXTHOP_HAS_GATEWAY;
}
}
rtnl_route_add_nexthop(route, nh);
tlen -= RTNH_ALIGN(rtnh->rtnh_len);
rtnh = RTNH_NEXT(rtnh);
}
}
if (tb[RTA_FLOW])
rtnl_route_set_realms(route, nla_get_u32(tb[RTA_FLOW]));
if (tb[RTA_CACHEINFO])
copy_cacheinfo_into_route(nla_data(tb[RTA_CACHEINFO]), route);
if (tb[RTA_MP_ALGO])
rtnl_route_set_mp_algo(route, nla_get_u32(tb[RTA_MP_ALGO]));
err = pp->pp_cb((struct nl_object *) route, pp);
if (err < 0)
goto errout;
err = P_ACCEPT;
errout:
rtnl_route_put(route);
return err;
errout_errno:
err = nl_get_errno();
goto errout;
}
static int route_request_update(struct nl_cache *c, struct nl_handle *h)
{
return nl_rtgen_request(h, RTM_GETROUTE, AF_UNSPEC, NLM_F_DUMP);
}
/**
* @name Cache Management
* @{
*/
/**
* Build a route cache holding all routes currently configured in the kernel
* @arg handle netlink handle
*
* Allocates a new cache, initializes it properly and updates it to
* contain all routes currently configured in the kernel.
*
* @note The caller is responsible for destroying and freeing the
* cache after using it.
* @return The cache or NULL if an error has occured.
*/
struct nl_cache *rtnl_route_alloc_cache(struct nl_handle *handle)
{
struct nl_cache *cache;
cache = nl_cache_alloc(&rtnl_route_ops);
if (!cache)
return NULL;
if (handle && nl_cache_refill(handle, cache) < 0) {
free(cache);
return NULL;
}
return cache;
}
/** @} */
/**
* @name Route Addition
* @{
*/
static struct nl_msg *build_route_msg(struct rtnl_route *tmpl, int cmd,
int flags)
{
struct nl_msg *msg;
struct nl_addr *addr;
int scope, i, oif, nmetrics = 0;
struct nlattr *metrics;
struct rtmsg rtmsg = {
.rtm_family = rtnl_route_get_family(tmpl),
.rtm_dst_len = rtnl_route_get_dst_len(tmpl),
.rtm_src_len = rtnl_route_get_src_len(tmpl),
.rtm_tos = rtnl_route_get_tos(tmpl),
.rtm_table = rtnl_route_get_table(tmpl),
.rtm_type = rtnl_route_get_type(tmpl),
.rtm_protocol = rtnl_route_get_protocol(tmpl),
.rtm_flags = rtnl_route_get_flags(tmpl),
};
if (rtmsg.rtm_family == AF_UNSPEC) {
nl_error(EINVAL, "Cannot build route message, address " \
"family is unknown.");
return NULL;
}
scope = rtnl_route_get_scope(tmpl);
if (scope == RT_SCOPE_NOWHERE) {
if (rtmsg.rtm_type == RTN_LOCAL)
scope = RT_SCOPE_HOST;
else {
/* XXX Change to UNIVERSE if gw || nexthops */
scope = RT_SCOPE_LINK;
}
}
rtmsg.rtm_scope = scope;
msg = nlmsg_alloc_simple(cmd, flags);
if (msg == NULL)
return NULL;
if (nlmsg_append(msg, &rtmsg, sizeof(rtmsg), NLMSG_ALIGNTO) < 0)
goto nla_put_failure;
addr = rtnl_route_get_dst(tmpl);
if (addr)
NLA_PUT_ADDR(msg, RTA_DST, addr);
addr = rtnl_route_get_src(tmpl);
if (addr)
NLA_PUT_ADDR(msg, RTA_SRC, addr);
addr = rtnl_route_get_gateway(tmpl);
if (addr)
NLA_PUT_ADDR(msg, RTA_GATEWAY, addr);
addr = rtnl_route_get_pref_src(tmpl);
if (addr)
NLA_PUT_ADDR(msg, RTA_PREFSRC, addr);
NLA_PUT_U32(msg, RTA_PRIORITY, rtnl_route_get_prio(tmpl));
oif = rtnl_route_get_oif(tmpl);
if (oif != RTNL_LINK_NOT_FOUND)
NLA_PUT_U32(msg, RTA_OIF, oif);
for (i = 1; i <= RTAX_MAX; i++)
if (rtnl_route_get_metric(tmpl, i) != UINT_MAX)
nmetrics++;
if (nmetrics > 0) {
unsigned int val;
metrics = nla_nest_start(msg, RTA_METRICS);
if (metrics == NULL)
goto nla_put_failure;
for (i = 1; i <= RTAX_MAX; i++) {
val = rtnl_route_get_metric(tmpl, i);
if (val != UINT_MAX)
NLA_PUT_U32(msg, i, val);
}
nla_nest_end(msg, metrics);
}
#if 0
RTA_IIF,
RTA_MULTIPATH,
RTA_PROTOINFO,
RTA_FLOW,
RTA_CACHEINFO,
RTA_SESSION,
RTA_MP_ALGO,
#endif
return msg;
nla_put_failure:
nlmsg_free(msg);
return NULL;
}
struct nl_msg *rtnl_route_build_add_request(struct rtnl_route *tmpl, int flags)
{
return build_route_msg(tmpl, RTM_NEWROUTE, NLM_F_CREATE | flags);
}
int rtnl_route_add(struct nl_handle *handle, struct rtnl_route *route,
int flags)
{
struct nl_msg *msg;
int err;
msg = rtnl_route_build_add_request(route, flags);
if (!msg)
return nl_get_errno();
err = nl_send_auto_complete(handle, msg);
nlmsg_free(msg);
if (err < 0)
return err;
return nl_wait_for_ack(handle);
}
struct nl_msg *rtnl_route_build_del_request(struct rtnl_route *tmpl, int flags)
{
return build_route_msg(tmpl, RTM_DELROUTE, flags);
}
int rtnl_route_del(struct nl_handle *handle, struct rtnl_route *route,
int flags)
{
struct nl_msg *msg;
int err;
msg = rtnl_route_build_del_request(route, flags);
if (!msg)
return nl_get_errno();
err = nl_send_auto_complete(handle, msg);
nlmsg_free(msg);
if (err < 0)
return err;
return nl_wait_for_ack(handle);
}
/** @} */
static struct nl_af_group route_groups[] = {
{ AF_INET, RTNLGRP_IPV4_ROUTE },
{ AF_INET6, RTNLGRP_IPV6_ROUTE },
{ AF_DECnet, RTNLGRP_DECnet_ROUTE },
{ END_OF_GROUP_LIST },
};
static struct nl_cache_ops rtnl_route_ops = {
.co_name = "route/route",
.co_hdrsize = sizeof(struct rtmsg),
.co_msgtypes = {
{ RTM_NEWROUTE, NL_ACT_NEW, "new" },
{ RTM_DELROUTE, NL_ACT_DEL, "del" },
{ RTM_GETROUTE, NL_ACT_GET, "get" },
END_OF_MSGTYPES_LIST,
},
.co_protocol = NETLINK_ROUTE,
.co_groups = route_groups,
.co_request_update = route_request_update,
.co_msg_parser = route_msg_parser,
.co_obj_ops = &route_obj_ops,
};
static void __init route_init(void)
{
nl_cache_mngt_register(&rtnl_route_ops);
}
static void __exit route_exit(void)
{
nl_cache_mngt_unregister(&rtnl_route_ops);
}
/**
* virNetDevMacVLanDelete:
*
* @ifname: Name of the interface
*
* Tear down an interface with the given name.
*
* Returns 0 on success, -1 on fatal error.
*/
int virNetDevMacVLanDelete(const char *ifname)
{
int rc = -1;
struct nlmsghdr *resp;
struct nlmsgerr *err;
struct ifinfomsg ifinfo = { .ifi_family = AF_UNSPEC };
unsigned char *recvbuf = NULL;
unsigned int recvbuflen;
struct nl_msg *nl_msg;
nl_msg = nlmsg_alloc_simple(RTM_DELLINK,
NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
if (!nl_msg) {
virReportOOMError();
return -1;
}
if (nlmsg_append(nl_msg, &ifinfo, sizeof(ifinfo), NLMSG_ALIGNTO) < 0)
goto buffer_too_small;
if (nla_put(nl_msg, IFLA_IFNAME, strlen(ifname)+1, ifname) < 0)
goto buffer_too_small;
if (virNetlinkCommand(nl_msg, &recvbuf, &recvbuflen, 0) < 0) {
goto cleanup;
}
if (recvbuflen < NLMSG_LENGTH(0) || recvbuf == NULL)
goto malformed_resp;
resp = (struct nlmsghdr *)recvbuf;
switch (resp->nlmsg_type) {
case NLMSG_ERROR:
err = (struct nlmsgerr *)NLMSG_DATA(resp);
if (resp->nlmsg_len < NLMSG_LENGTH(sizeof(*err)))
goto malformed_resp;
if (err->error) {
virReportSystemError(-err->error,
_("error destroying %s interface"),
ifname);
goto cleanup;
}
break;
case NLMSG_DONE:
break;
default:
goto malformed_resp;
}
rc = 0;
cleanup:
nlmsg_free(nl_msg);
VIR_FREE(recvbuf);
return rc;
malformed_resp:
virNetDevError(VIR_ERR_INTERNAL_ERROR, "%s",
_("malformed netlink response message"));
goto cleanup;
buffer_too_small:
virNetDevError(VIR_ERR_INTERNAL_ERROR, "%s",
_("allocated netlink buffer is too small"));
goto cleanup;
}
static int build_rule_msg(struct rtnl_rule *tmpl, int cmd, int flags,
struct nl_msg **result)
{
struct nl_msg *msg;
struct fib_rule_hdr frh = {
.family = tmpl->r_family,
.table = tmpl->r_table,
.action = tmpl->r_action,
.flags = tmpl->r_flags,
.tos = tmpl->r_dsfield,
};
if (!(tmpl->ce_mask & RULE_ATTR_FAMILY))
return -NLE_MISSING_ATTR;
msg = nlmsg_alloc_simple(cmd, flags);
if (!msg)
return -NLE_NOMEM;
if (nlmsg_append(msg, &frh, sizeof(frh), NLMSG_ALIGNTO) < 0)
goto nla_put_failure;
if (tmpl->ce_mask & RULE_ATTR_SRC) {
frh.src_len = nl_addr_get_prefixlen(tmpl->r_src);
NLA_PUT_ADDR(msg, FRA_SRC, tmpl->r_src);
}
if (tmpl->ce_mask & RULE_ATTR_DST) {
frh.dst_len = nl_addr_get_prefixlen(tmpl->r_dst);
NLA_PUT_ADDR(msg, FRA_DST, tmpl->r_dst);
}
if (tmpl->ce_mask & RULE_ATTR_IIFNAME)
NLA_PUT_STRING(msg, FRA_IIFNAME, tmpl->r_iifname);
if (tmpl->ce_mask & RULE_ATTR_OIFNAME)
NLA_PUT_STRING(msg, FRA_OIFNAME, tmpl->r_oifname);
if (tmpl->ce_mask & RULE_ATTR_PRIO)
NLA_PUT_U32(msg, FRA_PRIORITY, tmpl->r_prio);
if (tmpl->ce_mask & RULE_ATTR_MARK)
NLA_PUT_U32(msg, FRA_FWMARK, tmpl->r_mark);
if (tmpl->ce_mask & RULE_ATTR_MASK)
NLA_PUT_U32(msg, FRA_FWMASK, tmpl->r_mask);
if (tmpl->ce_mask & RULE_ATTR_GOTO)
NLA_PUT_U32(msg, FRA_GOTO, tmpl->r_goto);
if (tmpl->ce_mask & RULE_ATTR_FLOW)
NLA_PUT_U32(msg, FRA_FLOW, tmpl->r_flow);
*result = msg;
return 0;
nla_put_failure:
nlmsg_free(msg);
return -NLE_MSGSIZE;
}
/**
* Build netlink request message to add a new rule
* @arg tmpl template with data of new rule
* @arg flags additional netlink message flags
*
* Builds a new netlink message requesting a addition of a new
* rule. The netlink message header isn't fully equipped with
* all relevant fields and must thus be sent out via nl_send_auto_complete()
* or supplemented as needed. \a tmpl must contain the attributes of the new
* address set via \c rtnl_rule_set_* functions.
*
* @return The netlink message
*/
int rtnl_rule_build_add_request(struct rtnl_rule *tmpl, int flags,
struct nl_msg **result)
{
return build_rule_msg(tmpl, RTM_NEWRULE, NLM_F_CREATE | flags,
result);
}
/**
* Add a new rule
* @arg sk Netlink socket.
* @arg tmpl template with requested changes
* @arg flags additional netlink message flags
*
* Builds a netlink message by calling rtnl_rule_build_add_request(),
* sends the request to the kernel and waits for the next ACK to be
* received and thus blocks until the request has been fullfilled.
*
* @return 0 on sucess or a negative error if an error occured.
*/
int rtnl_rule_add(struct nl_sock *sk, struct rtnl_rule *tmpl, int flags)
{
struct nl_msg *msg;
int err;
if ((err = rtnl_rule_build_add_request(tmpl, flags, &msg)) < 0)
return err;
err = nl_send_auto_complete(sk, msg);
nlmsg_free(msg);
if (err < 0)
return err;
return wait_for_ack(sk);
}
/** @} */
/**
* @name Rule Deletion
* @{
*/
/**
* Build a netlink request message to delete a rule
* @arg rule rule to delete
* @arg flags additional netlink message flags
*
* Builds a new netlink message requesting a deletion of a rule.
* The netlink message header isn't fully equipped with all relevant
* fields and must thus be sent out via nl_send_auto_complete()
* or supplemented as needed. \a rule must point to an existing
* address.
*
* @return The netlink message
*/
int rtnl_rule_build_delete_request(struct rtnl_rule *rule, int flags,
struct nl_msg **result)
{
return build_rule_msg(rule, RTM_DELRULE, flags, result);
}
/**
* Delete a rule
* @arg sk Netlink socket.
* @arg rule rule to delete
* @arg flags additional netlink message flags
*
* Builds a netlink message by calling rtnl_rule_build_delete_request(),
* sends the request to the kernel and waits for the next ACK to be
* received and thus blocks until the request has been fullfilled.
*
* @return 0 on sucess or a negative error if an error occured.
*/
int rtnl_rule_delete(struct nl_sock *sk, struct rtnl_rule *rule, int flags)
{
struct nl_msg *msg;
int err;
if ((err = rtnl_rule_build_delete_request(rule, flags, &msg)) < 0)
return err;
err = nl_send_auto_complete(sk, msg);
nlmsg_free(msg);
if (err < 0)
return err;
return wait_for_ack(sk);
}
/** @} */
/**
* @name Attribute Modification
* @{
*/
void rtnl_rule_set_family(struct rtnl_rule *rule, int family)
{
rule->r_family = family;
rule->ce_mask |= RULE_ATTR_FAMILY;
}
TError TNlLink::AddXVlan(const std::string &vlantype,
const std::string &master,
uint32_t type,
const std::string &hw,
int mtu) {
TError error = TError::Success();
int ret;
uint32_t masterIdx;
struct nl_msg *msg;
struct nlattr *linkinfo, *infodata;
struct ifinfomsg ifi = { 0 };
struct ether_addr *ea = nullptr;
auto Name = GetName();
if (hw.length()) {
// FIXME THREADS
ea = ether_aton(hw.c_str());
if (!ea)
return TError(EError::Unknown, "Invalid " + vlantype + " mac address " + hw);
}
TNlLink masterLink(Nl, master);
error = masterLink.Load();
if (error)
return error;
masterIdx = masterLink.GetIndex();
msg = nlmsg_alloc_simple(RTM_NEWLINK, NLM_F_CREATE);
if (!msg)
return TError(EError::Unknown, "Unable to add " + vlantype + ": no memory");
ret = nlmsg_append(msg, &ifi, sizeof(ifi), NLMSG_ALIGNTO);
if (ret < 0) {
error = TError(EError::Unknown, "Unable to add " + vlantype + ": " + nl_geterror(ret));
goto free_msg;
}
/* link configuration */
ret = nla_put(msg, IFLA_LINK, sizeof(uint32_t), &masterIdx);
if (ret < 0) {
error = TError(EError::Unknown, std::string("Unable to put IFLA_LINK: ") + nl_geterror(ret));
goto free_msg;
}
ret = nla_put(msg, IFLA_IFNAME, Name.length() + 1, Name.c_str());
if (ret < 0) {
error = TError(EError::Unknown, std::string("Unable to put IFLA_IFNAME: ") + nl_geterror(ret));
goto free_msg;
}
if (mtu > 0) {
ret = nla_put(msg, IFLA_MTU, sizeof(int), &mtu);
if (ret < 0) {
error = TError(EError::Unknown, std::string("Unable to put IFLA_MTU: ") + nl_geterror(ret));
goto free_msg;
}
}
if (ea) {
struct nl_addr *addr = nl_addr_build(AF_LLC, ea, ETH_ALEN);
ret = nla_put(msg, IFLA_ADDRESS, nl_addr_get_len(addr), nl_addr_get_binary_addr(addr));
if (ret < 0) {
error = TError(EError::Unknown, std::string("Unable to put IFLA_ADDRESS: ") + nl_geterror(ret));
goto free_msg;
}
nl_addr_put(addr);
}
/* link type */
linkinfo = nla_nest_start(msg, IFLA_LINKINFO);
if (!linkinfo) {
error = TError(EError::Unknown, "Unable to add " + vlantype + ": can't nest IFLA_LINKINFO");
goto free_msg;
}
ret = nla_put(msg, IFLA_INFO_KIND, vlantype.length() + 1, vlantype.c_str());
if (ret < 0) {
error = TError(EError::Unknown, std::string("Unable to put IFLA_INFO_KIND: ") + nl_geterror(ret));
goto free_msg;
}
/* xvlan specific */
infodata = nla_nest_start(msg, IFLA_INFO_DATA);
if (!infodata) {
error = TError(EError::Unknown, "Unable to add " + vlantype + ": can't nest IFLA_INFO_DATA");
goto free_msg;
}
if (vlantype == "macvlan") {
ret = nla_put(msg, IFLA_MACVLAN_MODE, sizeof(uint32_t), &type);
if (ret < 0) {
error = TError(EError::Unknown, std::string("Unable to put IFLA_MACVLAN_MODE: ") + nl_geterror(ret));
goto free_msg;
}
#ifdef IFLA_IPVLAN_MAX
} else if (vlantype == "ipvlan") {
uint16_t mode = type;
ret = nla_put(msg, IFLA_IPVLAN_MODE, sizeof(uint16_t), &mode);
if (ret < 0) {
error = TError(EError::Unknown, std::string("Unable to put IFLA_IPVLAN_MODE: ") + nl_geterror(ret));
goto free_msg;
}
#endif
}
nla_nest_end(msg, infodata);
nla_nest_end(msg, linkinfo);
L() << "netlink: add " << vlantype << " " << Name << " master " << master
<< " type " << type << " hw " << hw << " mtu " << mtu << std::endl;
ret = nl_send_sync(GetSock(), msg);
if (ret)
return Error(ret, "Cannot add " + vlantype);
return Load();
free_msg:
nlmsg_free(msg);
return error;
}
int opal_btl_usnic_nl_ip_rt_lookup(struct usnic_rtnl_sk *unlsk,
const char *src_ifname,
uint32_t src_addr,
uint32_t dst_addr, int *metric)
{
struct nl_msg *nlm;
struct rtmsg rmsg;
struct nl_lookup_arg arg;
int msg_cnt;
int err;
int oif;
oif = if_nametoindex(src_ifname);
if (0 == oif) {
return errno;
}
arg.nh_addr = 0;
arg.oif = oif;
arg.found = 0;
arg.replied = 0;
arg.unlsk = unlsk;
arg.msg_count = msg_cnt = 0;
memset(&rmsg, 0, sizeof(rmsg));
rmsg.rtm_family = AF_INET;
rmsg.rtm_dst_len = sizeof(dst_addr)*8;
rmsg.rtm_src_len = sizeof(src_addr)*8;
nlm = nlmsg_alloc_simple(RTM_GETROUTE, 0);
nlmsg_append(nlm, &rmsg, sizeof(rmsg), NLMSG_ALIGNTO);
nla_put_u32(nlm, RTA_DST, dst_addr);
nla_put_u32(nlm, RTA_SRC, src_addr);
err = rtnl_send_ack_disable(unlsk, nlm);
nlmsg_free(nlm);
if (err < 0) {
usnic_err("Failed to send nl route message to kernel, "
"error %s\n", nl_geterror());
return err;
}
err = nl_socket_modify_cb(unlsk->nlh, NL_CB_MSG_IN, NL_CB_CUSTOM,
rtnl_raw_parse_cb, &arg);
if (err != 0) {
usnic_err("Failed to setup callback function, error %s\n", nl_geterror());
return err;
}
while (!arg.replied) {
err = nl_recvmsgs_default(unlsk->nlh);
if (err < 0) {
usnic_err("Failed to receive nl route message from "
"kernel, error %s\n", nl_geterror());
return err;
}
/*
* the return value of nl_recvmsgs_default does not tell
* whether it returns because of successful read or socket
* timeout. So we compare msg count before and after the call
* to decide if no new message arrives. In such case,
* this function needs to terminate to prevent the caller from
* blocking forever
* NL_CB_MSG_IN traps every received message, so
* there should be no premature exit
*/
if (msg_cnt != arg.msg_count)
msg_cnt = arg.msg_count;
else
break;
}
if (arg.found) {
if (metric != NULL) {
*metric = arg.metric;
}
return 0;
}
else {
return -1;
}
}
/**
* Send netlink message.
* @arg sk Netlink socket.
* @arg msg Netlink message to be sent.
* @arg iov iovec to be sent.
* @arg iovlen number of struct iovec to be sent.
* @see nl_sendmsg()
* @return Number of characters sent on success or a negative error code.
*/
int nl_send_iovec(struct nl_sock *sk, struct nl_msg *msg, struct iovec *iov, unsigned iovlen)
{
struct sockaddr_nl *dst;
struct ucred *creds;
struct msghdr hdr = {
.msg_name = (void *) &sk->s_peer,
.msg_namelen = sizeof(struct sockaddr_nl),
.msg_iov = iov,
.msg_iovlen = iovlen,
};
/* Overwrite destination if specified in the message itself, defaults
* to the peer address of the socket.
*/
dst = nlmsg_get_dst(msg);
if (dst->nl_family == AF_NETLINK)
hdr.msg_name = dst;
/* Add credentials if present. */
creds = nlmsg_get_creds(msg);
if (creds != NULL) {
char buf[CMSG_SPACE(sizeof(struct ucred))];
struct cmsghdr *cmsg;
hdr.msg_control = buf;
hdr.msg_controllen = sizeof(buf);
cmsg = CMSG_FIRSTHDR(&hdr);
cmsg->cmsg_level = SOL_SOCKET;
cmsg->cmsg_type = SCM_CREDENTIALS;
cmsg->cmsg_len = CMSG_LEN(sizeof(struct ucred));
memcpy(CMSG_DATA(cmsg), creds, sizeof(struct ucred));
}
return nl_sendmsg(sk, msg, &hdr);
}
/**
* Send netlink message.
* @arg sk Netlink socket.
* @arg msg Netlink message to be sent.
* @see nl_sendmsg()
* @return Number of characters sent on success or a negative error code.
*/
int nl_send(struct nl_sock *sk, struct nl_msg *msg)
{
struct iovec iov = {
.iov_base = (void *) nlmsg_hdr(msg),
.iov_len = nlmsg_hdr(msg)->nlmsg_len,
};
return nl_send_iovec(sk, msg, &iov, 1);
}
void nl_auto_complete(struct nl_sock *sk, struct nl_msg *msg)
{
struct nlmsghdr *nlh;
nlh = nlmsg_hdr(msg);
if (nlh->nlmsg_pid == 0)
nlh->nlmsg_pid = sk->s_local.nl_pid;
if (nlh->nlmsg_seq == 0)
nlh->nlmsg_seq = sk->s_seq_next++;
if (msg->nm_protocol == -1)
msg->nm_protocol = sk->s_proto;
nlh->nlmsg_flags |= NLM_F_REQUEST;
if (!(sk->s_flags & NL_NO_AUTO_ACK))
nlh->nlmsg_flags |= NLM_F_ACK;
}
/**
* Send netlink message and check & extend header values as needed.
* @arg sk Netlink socket.
* @arg msg Netlink message to be sent.
*
* Checks the netlink message \c nlh for completness and extends it
* as required before sending it out. Checked fields include pid,
* sequence nr, and flags.
*
* @see nl_send()
* @return Number of characters sent or a negative error code.
*/
int nl_send_auto_complete(struct nl_sock *sk, struct nl_msg *msg)
{
struct nl_cb *cb = sk->s_cb;
nl_auto_complete(sk, msg);
if (cb->cb_send_ow)
return cb->cb_send_ow(sk, msg);
else
return nl_send(sk, msg);
}
/**
* Send simple netlink message using nl_send_auto_complete()
* @arg sk Netlink socket.
* @arg type Netlink message type.
* @arg flags Netlink message flags.
* @arg buf Data buffer.
* @arg size Size of data buffer.
*
* Builds a netlink message with the specified type and flags and
* appends the specified data as payload to the message.
*
* @see nl_send_auto_complete()
* @return Number of characters sent on success or a negative error code.
*/
int nl_send_simple(struct nl_sock *sk, int type, int flags, void *buf,
size_t size)
{
int err;
struct nl_msg *msg;
msg = nlmsg_alloc_simple(type, flags);
if (!msg)
return -NLE_NOMEM;
if (buf && size) {
err = nlmsg_append(msg, buf, size, NLMSG_ALIGNTO);
if (err < 0)
goto errout;
}
err = nl_send_auto_complete(sk, msg);
errout:
nlmsg_free(msg);
return err;
}
/** @} */
/**
* @name Receive
* @{
*/
/**
* Receive data from netlink socket
* @arg sk Netlink socket.
* @arg nla Destination pointer for peer's netlink address.
* @arg buf Destination pointer for message content.
* @arg creds Destination pointer for credentials.
*
* Receives a netlink message, allocates a buffer in \c *buf and
* stores the message content. The peer's netlink address is stored
* in \c *nla. The caller is responsible for freeing the buffer allocated
* in \c *buf if a positive value is returned. Interruped system calls
* are handled by repeating the read. The input buffer size is determined
* by peeking before the actual read is done.
*
* A non-blocking sockets causes the function to return immediately with
* a return value of 0 if no data is available.
*
* @return Number of octets read, 0 on EOF or a negative error code.
*/
int nl_recv(struct nl_sock *sk, struct sockaddr_nl *nla,
unsigned char **buf, struct ucred **creds)
{
int n;
int flags = 0;
static int page_size = 0;
struct iovec iov;
struct msghdr msg = {
.msg_name = (void *) nla,
.msg_namelen = sizeof(struct sockaddr_nl),
.msg_iov = &iov,
.msg_iovlen = 1,
.msg_control = NULL,
.msg_controllen = 0,
.msg_flags = 0,
};
struct cmsghdr *cmsg;
if (sk->s_flags & NL_MSG_PEEK)
flags |= MSG_PEEK;
if (page_size == 0)
page_size = getpagesize();
iov.iov_len = page_size;
iov.iov_base = *buf = malloc(iov.iov_len);
if (sk->s_flags & NL_SOCK_PASSCRED) {
msg.msg_controllen = CMSG_SPACE(sizeof(struct ucred));
msg.msg_control = calloc(1, msg.msg_controllen);
}
retry:
n = recvmsg(sk->s_fd, &msg, flags);
if (!n)
goto abort;
else if (n < 0) {
if (errno == EINTR) {
NL_DBG(3, "recvmsg() returned EINTR, retrying\n");
goto retry;
} else if (errno == EAGAIN) {
NL_DBG(3, "recvmsg() returned EAGAIN, aborting\n");
goto abort;
} else {
free(msg.msg_control);
free(*buf);
return -nl_syserr2nlerr(errno);
}
}
if (iov.iov_len < n ||
msg.msg_flags & MSG_TRUNC) {
/* Provided buffer is not long enough, enlarge it
* and try again. */
iov.iov_len *= 2;
iov.iov_base = *buf = realloc(*buf, iov.iov_len);
goto retry;
} else if (msg.msg_flags & MSG_CTRUNC) {
msg.msg_controllen *= 2;
msg.msg_control = realloc(msg.msg_control, msg.msg_controllen);
goto retry;
} else if (flags != 0) {
/* Buffer is big enough, do the actual reading */
flags = 0;
goto retry;
}
if (msg.msg_namelen != sizeof(struct sockaddr_nl)) {
free(msg.msg_control);
free(*buf);
return -NLE_NOADDR;
}
for (cmsg = CMSG_FIRSTHDR(&msg); cmsg; cmsg = CMSG_NXTHDR(&msg, cmsg)) {
if (cmsg->cmsg_level == SOL_SOCKET &&
cmsg->cmsg_type == SCM_CREDENTIALS) {
*creds = calloc(1, sizeof(struct ucred));
memcpy(*creds, CMSG_DATA(cmsg), sizeof(struct ucred));
break;
}
}
free(msg.msg_control);
return n;
abort:
free(msg.msg_control);
free(*buf);
return 0;
}
#define NL_CB_CALL(cb, type, msg) \
do { \
err = nl_cb_call(cb, type, msg); \
switch (err) { \
case NL_OK: \
err = 0; \
break; \
case NL_SKIP: \
goto skip; \
case NL_STOP: \
goto stop; \
default: \
goto out; \
} \
} while (0)
static int recvmsgs(struct nl_sock *sk, struct nl_cb *cb)
{
int n, err = 0, multipart = 0;
unsigned char *buf = NULL;
struct nlmsghdr *hdr;
struct sockaddr_nl nla = {0};
struct nl_msg *msg = NULL;
struct ucred *creds = NULL;
continue_reading:
NL_DBG(3, "Attempting to read from %p\n", sk);
if (cb->cb_recv_ow)
n = cb->cb_recv_ow(sk, &nla, &buf, &creds);
else
n = nl_recv(sk, &nla, &buf, &creds);
if (n <= 0)
return n;
NL_DBG(3, "recvmsgs(%p): Read %d bytes\n", sk, n);
hdr = (struct nlmsghdr *) buf;
while (nlmsg_ok(hdr, n)) {
NL_DBG(3, "recgmsgs(%p): Processing valid message...\n", sk);
nlmsg_free(msg);
msg = nlmsg_convert(hdr);
if (!msg) {
err = -NLE_NOMEM;
goto out;
}
nlmsg_set_proto(msg, sk->s_proto);
nlmsg_set_src(msg, &nla);
if (creds)
nlmsg_set_creds(msg, creds);
/* Raw callback is the first, it gives the most control
* to the user and he can do his very own parsing. */
if (cb->cb_set[NL_CB_MSG_IN])
NL_CB_CALL(cb, NL_CB_MSG_IN, msg);
/* Sequence number checking. The check may be done by
* the user, otherwise a very simple check is applied
* enforcing strict ordering */
if (cb->cb_set[NL_CB_SEQ_CHECK])
NL_CB_CALL(cb, NL_CB_SEQ_CHECK, msg);
else if (hdr->nlmsg_seq != sk->s_seq_expect) {
if (cb->cb_set[NL_CB_INVALID])
NL_CB_CALL(cb, NL_CB_INVALID, msg);
else {
err = -NLE_SEQ_MISMATCH;
goto out;
}
}
if (hdr->nlmsg_type == NLMSG_DONE ||
hdr->nlmsg_type == NLMSG_ERROR ||
hdr->nlmsg_type == NLMSG_NOOP ||
hdr->nlmsg_type == NLMSG_OVERRUN) {
/* We can't check for !NLM_F_MULTI since some netlink
* users in the kernel are broken. */
sk->s_seq_expect++;
NL_DBG(3, "recvmsgs(%p): Increased expected " \
"sequence number to %d\n",
sk, sk->s_seq_expect);
}
if (hdr->nlmsg_flags & NLM_F_MULTI)
multipart = 1;
/* Other side wishes to see an ack for this message */
if (hdr->nlmsg_flags & NLM_F_ACK) {
if (cb->cb_set[NL_CB_SEND_ACK])
NL_CB_CALL(cb, NL_CB_SEND_ACK, msg);
else {
/* FIXME: implement */
}
}
/* messages terminates a multpart message, this is
* usually the end of a message and therefore we slip
* out of the loop by default. the user may overrule
* this action by skipping this packet. */
if (hdr->nlmsg_type == NLMSG_DONE) {
multipart = 0;
if (cb->cb_set[NL_CB_FINISH])
NL_CB_CALL(cb, NL_CB_FINISH, msg);
}
/* Message to be ignored, the default action is to
* skip this message if no callback is specified. The
* user may overrule this action by returning
* NL_PROCEED. */
else if (hdr->nlmsg_type == NLMSG_NOOP) {
if (cb->cb_set[NL_CB_SKIPPED])
NL_CB_CALL(cb, NL_CB_SKIPPED, msg);
else
goto skip;
}
/* Data got lost, report back to user. The default action is to
* quit parsing. The user may overrule this action by retuning
* NL_SKIP or NL_PROCEED (dangerous) */
else if (hdr->nlmsg_type == NLMSG_OVERRUN) {
if (cb->cb_set[NL_CB_OVERRUN])
NL_CB_CALL(cb, NL_CB_OVERRUN, msg);
else {
err = -NLE_MSG_OVERFLOW;
goto out;
}
}
/* Message carries a nlmsgerr */
else if (hdr->nlmsg_type == NLMSG_ERROR) {
struct nlmsgerr *e = nlmsg_data(hdr);
if (hdr->nlmsg_len < nlmsg_msg_size(sizeof(*e))) {
/* Truncated error message, the default action
* is to stop parsing. The user may overrule
* this action by returning NL_SKIP or
* NL_PROCEED (dangerous) */
if (cb->cb_set[NL_CB_INVALID])
NL_CB_CALL(cb, NL_CB_INVALID, msg);
else {
err = -NLE_MSG_TRUNC;
goto out;
}
} else if (e->error) {
/* Error message reported back from kernel. */
if (cb->cb_err) {
err = cb->cb_err(&nla, e,
cb->cb_err_arg);
if (err < 0)
goto out;
else if (err == NL_SKIP)
goto skip;
else if (err == NL_STOP) {
err = -nl_syserr2nlerr(e->error);
goto out;
}
} else {
err = -nl_syserr2nlerr(e->error);
goto out;
}
} else if (cb->cb_set[NL_CB_ACK])
NL_CB_CALL(cb, NL_CB_ACK, msg);
} else {
/* Valid message (not checking for MULTIPART bit to
* get along with broken kernels. NL_SKIP has no
* effect on this. */
if (cb->cb_set[NL_CB_VALID])
NL_CB_CALL(cb, NL_CB_VALID, msg);
}
skip:
err = 0;
hdr = nlmsg_next(hdr, &n);
}
nlmsg_free(msg);
free(buf);
free(creds);
buf = NULL;
msg = NULL;
creds = NULL;
if (multipart) {
/* Multipart message not yet complete, continue reading */
goto continue_reading;
}
stop:
err = 0;
out:
nlmsg_free(msg);
free(buf);
free(creds);
return err;
}
static int fill_ematch_sequence(struct nl_msg *msg, struct nl_list_head *list)
{
struct rtnl_ematch *e;
nl_list_for_each_entry(e, list, e_list) {
struct tcf_ematch_hdr match = {
.matchid = e->e_id,
.kind = e->e_kind,
.flags = e->e_flags,
};
struct nlattr *attr;
int err = 0;
if (!(attr = nla_nest_start(msg, e->e_index + 1)))
return -NLE_NOMEM;
if (nlmsg_append(msg, &match, sizeof(match), 0) < 0)
return -NLE_NOMEM;
if (e->e_ops->eo_fill)
err = e->e_ops->eo_fill(e, msg);
else if (e->e_flags & TCF_EM_SIMPLE)
err = nlmsg_append(msg, e->e_data, 4, 0);
else if (e->e_datalen > 0)
err = nlmsg_append(msg, e->e_data, e->e_datalen, 0);
NL_DBG(3, "msg %p: added ematch [%d] id=%d kind=%d flags=%d\n",
msg, e->e_index, match.matchid, match.kind, match.flags);
if (err < 0)
return -NLE_NOMEM;
nla_nest_end(msg, attr);
}
nl_list_for_each_entry(e, list, e_list) {
if (e->e_kind == TCF_EM_CONTAINER &&
fill_ematch_sequence(msg, &e->e_childs) < 0)
return -NLE_NOMEM;
}
return 0;
}
int rtnl_ematch_fill_attr(struct nl_msg *msg, int attrid,
struct rtnl_ematch_tree *tree)
{
struct tcf_ematch_tree_hdr thdr = {
.progid = tree->et_progid,
};
struct nlattr *list, *topattr;
int err, index = 0;
/* Assign index number to each ematch to allow for references
* to be made while constructing the sequence of matches. */
err = update_container_index(&tree->et_list, &index);
if (err < 0)
return err;
if (!(topattr = nla_nest_start(msg, attrid)))
goto nla_put_failure;
thdr.nmatches = index;
NLA_PUT(msg, TCA_EMATCH_TREE_HDR, sizeof(thdr), &thdr);
if (!(list = nla_nest_start(msg, TCA_EMATCH_TREE_LIST)))
goto nla_put_failure;
if (fill_ematch_sequence(msg, &tree->et_list) < 0)
goto nla_put_failure;
nla_nest_end(msg, list);
nla_nest_end(msg, topattr);
return 0;
nla_put_failure:
return -NLE_NOMEM;
}
/** @} */
extern int ematch_parse(void *, char **, struct nl_list_head *);
int rtnl_ematch_parse_expr(const char *expr, char **errp,
struct rtnl_ematch_tree **result)
{
struct rtnl_ematch_tree *tree;
YY_BUFFER_STATE buf = NULL;
yyscan_t scanner = NULL;
int err;
NL_DBG(2, "Parsing ematch expression \"%s\"\n", expr);
if (!(tree = rtnl_ematch_tree_alloc(RTNL_EMATCH_PROGID)))
return -NLE_FAILURE;
if ((err = ematch_lex_init(&scanner)) < 0) {
err = -NLE_FAILURE;
goto errout;
}
buf = ematch__scan_string(expr, scanner);
if ((err = ematch_parse(scanner, errp, &tree->et_list)) != 0) {
ematch__delete_buffer(buf, scanner);
err = -NLE_PARSE_ERR;
goto errout;
}
ematch_lex_destroy(scanner);
*result = tree;
return 0;
errout:
if (scanner)
ematch_lex_destroy(scanner);
rtnl_ematch_tree_free(tree);
return err;
}
static const char *layer_txt[] = {
[TCF_LAYER_LINK] = "eth",
[TCF_LAYER_NETWORK] = "ip",
[TCF_LAYER_TRANSPORT] = "tcp",
};
char *rtnl_ematch_offset2txt(uint8_t layer, uint16_t offset, char *buf, size_t len)
{
snprintf(buf, len, "%s+%u",
(layer <= TCF_LAYER_MAX) ? layer_txt[layer] : "?",
offset);
return buf;
}
static const char *operand_txt[] = {
[TCF_EM_OPND_EQ] = "=",
[TCF_EM_OPND_LT] = "<",
[TCF_EM_OPND_GT] = ">",
};
char *rtnl_ematch_opnd2txt(uint8_t opnd, char *buf, size_t len)
{
snprintf(buf, len, "%s",
opnd < ARRAY_SIZE(operand_txt) ? operand_txt[opnd] : "?");
return buf;
}
int main(int argc, char *argv[])
{
struct nl_handle *nlh;
struct nl_cache *link_cache, *route_cache;
struct nl_addr *dst;
struct rtnl_route *route;
struct ip_lookup_res res;
struct nl_dump_params params = {
.dp_fd = stdout,
.dp_type = NL_DUMP_FULL
};
int err = 1;
if (argc < 2 || !strcmp(argv[1], "-h"))
print_usage();
if (nltool_init(argc, argv) < 0)
goto errout;
nlh = nltool_alloc_handle();
if (!nlh)
goto errout;
route = rtnl_route_alloc();
if (!route)
goto errout_free_handle;
if (nltool_connect(nlh, NETLINK_ROUTE) < 0)
goto errout_free_route;
link_cache = nltool_alloc_link_cache(nlh);
if (!link_cache)
goto errout_close;
dst = nltool_addr_parse(argv[1]);
if (!dst)
goto errout_link_cache;
route_cache = nltool_alloc_route_cache(nlh);
if (!route_cache)
goto errout_addr_put;
{
struct nl_msg *m;
struct rtmsg rmsg = {
.rtm_family = nl_addr_get_family(dst),
.rtm_dst_len = nl_addr_get_prefixlen(dst),
};
m = nlmsg_alloc_simple(RTM_GETROUTE, 0);
nlmsg_append(m, &rmsg, sizeof(rmsg), NLMSG_ALIGNTO);
nla_put_addr(m, RTA_DST, dst);
if ((err = nl_send_auto_complete(nlh, m)) < 0) {
nlmsg_free(m);
fprintf(stderr, "%s\n", nl_geterror());
goto errout_route_cache;
}
nlmsg_free(m);
nl_socket_modify_cb(nlh, NL_CB_VALID, NL_CB_CUSTOM, cb,
route_cache);
if (nl_recvmsgs_default(nlh) < 0) {
fprintf(stderr, "%s\n", nl_geterror());
goto errout_route_cache;
}
}
rtnl_route_set_dst(route, dst);
nl_cache_dump_filter(route_cache, ¶ms, (struct nl_object *) route);
memset(&res, 0, sizeof(res));
nl_cache_foreach_filter(route_cache, (struct nl_object *) route, route_proc_cb,
&res);
printf("ip lookup result: oif idx: %d oif name %s ",
res.oif, res.oifname);
if (res.nh_addr) {
char buf[INET_ADDRSTRLEN];
inet_ntop(AF_INET, &res.nh_addr, buf, sizeof(buf));
printf("via %s", buf);
}
printf ("\n");
err = 0;
errout_route_cache:
nl_cache_free(route_cache);
errout_addr_put:
nl_addr_put(dst);
errout_link_cache:
nl_cache_free(link_cache);
errout_close:
nl_close(nlh);
errout_free_route:
rtnl_route_put(route);
errout_free_handle:
nl_handle_destroy(nlh);
errout:
return err;
}
/**
* virNetDevMacVLanCreate:
*
* @ifname: The name the interface is supposed to have; optional parameter
* @type: The type of device, i.e., "macvtap", "macvlan"
* @macaddress: The MAC address of the device
* @srcdev: The name of the 'link' device
* @macvlan_mode: The macvlan mode to use
* @retry: Pointer to integer that will be '1' upon return if an interface
* with the same name already exists and it is worth to try
* again with a different name
*
* Create a macvtap device with the given properties.
*
* Returns 0 on success, -1 on fatal error.
*/
int
virNetDevMacVLanCreate(const char *ifname,
const char *type,
const unsigned char *macaddress,
const char *srcdev,
uint32_t macvlan_mode,
int *retry)
{
int rc = -1;
struct nlmsghdr *resp;
struct nlmsgerr *err;
struct ifinfomsg ifinfo = { .ifi_family = AF_UNSPEC };
int ifindex;
unsigned char *recvbuf = NULL;
unsigned int recvbuflen;
struct nl_msg *nl_msg;
struct nlattr *linkinfo, *info_data;
if (virNetDevGetIndex(srcdev, &ifindex) < 0)
return -1;
*retry = 0;
nl_msg = nlmsg_alloc_simple(RTM_NEWLINK,
NLM_F_REQUEST | NLM_F_CREATE | NLM_F_EXCL);
if (!nl_msg) {
virReportOOMError();
return -1;
}
if (nlmsg_append(nl_msg, &ifinfo, sizeof(ifinfo), NLMSG_ALIGNTO) < 0)
goto buffer_too_small;
if (nla_put_u32(nl_msg, IFLA_LINK, ifindex) < 0)
goto buffer_too_small;
if (nla_put(nl_msg, IFLA_ADDRESS, VIR_MAC_BUFLEN, macaddress) < 0)
goto buffer_too_small;
if (ifname &&
nla_put(nl_msg, IFLA_IFNAME, strlen(ifname)+1, ifname) < 0)
goto buffer_too_small;
if (!(linkinfo = nla_nest_start(nl_msg, IFLA_LINKINFO)))
goto buffer_too_small;
if (nla_put(nl_msg, IFLA_INFO_KIND, strlen(type), type) < 0)
goto buffer_too_small;
if (macvlan_mode > 0) {
if (!(info_data = nla_nest_start(nl_msg, IFLA_INFO_DATA)))
goto buffer_too_small;
if (nla_put(nl_msg, IFLA_MACVLAN_MODE, sizeof(macvlan_mode),
&macvlan_mode) < 0)
goto buffer_too_small;
nla_nest_end(nl_msg, info_data);
}
nla_nest_end(nl_msg, linkinfo);
if (virNetlinkCommand(nl_msg, &recvbuf, &recvbuflen, 0) < 0) {
goto cleanup;
}
if (recvbuflen < NLMSG_LENGTH(0) || recvbuf == NULL)
goto malformed_resp;
resp = (struct nlmsghdr *)recvbuf;
switch (resp->nlmsg_type) {
case NLMSG_ERROR:
err = (struct nlmsgerr *)NLMSG_DATA(resp);
if (resp->nlmsg_len < NLMSG_LENGTH(sizeof(*err)))
goto malformed_resp;
switch (err->error) {
case 0:
break;
case -EEXIST:
*retry = 1;
goto cleanup;
default:
virReportSystemError(-err->error,
_("error creating %s type of interface"),
type);
goto cleanup;
}
break;
case NLMSG_DONE:
break;
default:
goto malformed_resp;
}
rc = 0;
cleanup:
nlmsg_free(nl_msg);
VIR_FREE(recvbuf);
return rc;
malformed_resp:
virNetDevError(VIR_ERR_INTERNAL_ERROR, "%s",
_("malformed netlink response message"));
goto cleanup;
buffer_too_small:
virNetDevError(VIR_ERR_INTERNAL_ERROR, "%s",
_("allocated netlink buffer is too small"));
goto cleanup;
}
/**
* Send netlink message with control over sendmsg() message header.
* @arg handle Netlink handle.
* @arg msg Netlink message to be sent.
* @arg hdr Sendmsg() message header.
* @return Number of characters sent on sucess or a negative error code.
*/
int nl_sendmsg(struct nl_handle *handle, struct nl_msg *msg, struct msghdr *hdr)
{
struct nl_cb *cb;
int ret;
struct iovec iov = {
.iov_base = (void *) nlmsg_hdr(msg),
.iov_len = nlmsg_hdr(msg)->nlmsg_len,
};
hdr->msg_iov = &iov;
hdr->msg_iovlen = 1;
nlmsg_set_src(msg, &handle->h_local);
cb = nl_handle_get_cb(handle);
if (cb->cb_set[NL_CB_MSG_OUT])
if (nl_cb_call(cb, NL_CB_MSG_OUT, msg) != NL_PROCEED)
return 0;
ret = sendmsg(handle->h_fd, hdr, 0);
if (ret < 0)
return nl_errno(errno);
return ret;
}
/**
* Send netlink message.
* @arg handle Netlink handle
* @arg msg Netlink message to be sent.
* @see nl_sendmsg()
* @return Number of characters sent on success or a negative error code.
*/
int nl_send(struct nl_handle *handle, struct nl_msg *msg)
{
struct sockaddr_nl *dst;
struct ucred *creds;
struct msghdr hdr = {
.msg_name = (void *) &handle->h_peer,
.msg_namelen = sizeof(struct sockaddr_nl),
};
/* Overwrite destination if specified in the message itself, defaults
* to the peer address of the handle.
*/
dst = nlmsg_get_dst(msg);
if (dst->nl_family == AF_NETLINK)
hdr.msg_name = dst;
/* Add credentials if present. */
creds = nlmsg_get_creds(msg);
if (creds != NULL) {
char buf[CMSG_SPACE(sizeof(struct ucred))];
struct cmsghdr *cmsg;
hdr.msg_control = buf;
hdr.msg_controllen = sizeof(buf);
cmsg = CMSG_FIRSTHDR(&hdr);
cmsg->cmsg_level = SOL_SOCKET;
cmsg->cmsg_type = SCM_CREDENTIALS;
cmsg->cmsg_len = CMSG_LEN(sizeof(struct ucred));
memcpy(CMSG_DATA(cmsg), creds, sizeof(struct ucred));
}
return nl_sendmsg(handle, msg, &hdr);
}
/**
* Send netlink message and check & extend header values as needed.
* @arg handle Netlink handle.
* @arg msg Netlink message to be sent.
*
* Checks the netlink message \c nlh for completness and extends it
* as required before sending it out. Checked fields include pid,
* sequence nr, and flags.
*
* @see nl_send()
* @return Number of characters sent or a negative error code.
*/
int nl_send_auto_complete(struct nl_handle *handle, struct nl_msg *msg)
{
struct nlmsghdr *nlh;
nlh = nlmsg_hdr(msg);
if (nlh->nlmsg_pid == 0)
nlh->nlmsg_pid = handle->h_local.nl_pid;
if (nlh->nlmsg_seq == 0)
nlh->nlmsg_seq = handle->h_seq_next++;
nlh->nlmsg_flags |= (NLM_F_REQUEST | NLM_F_ACK);
if (handle->h_cb->cb_send_ow)
return handle->h_cb->cb_send_ow(handle, msg);
else
return nl_send(handle, msg);
}
/**
* Send simple netlink message using nl_send_auto_complete()
* @arg handle Netlink handle.
* @arg type Netlink message type.
* @arg flags Netlink message flags.
* @arg buf Data buffer.
* @arg size Size of data buffer.
*
* Builds a netlink message with the specified type and flags and
* appends the specified data as payload to the message.
*
* @see nl_send_auto_complete()
* @return Number of characters sent on success or a negative error code.
*/
int nl_send_simple(struct nl_handle *handle, int type, int flags, void *buf,
size_t size)
{
int err;
struct nl_msg *msg;
struct nlmsghdr nlh = {
.nlmsg_len = nlmsg_msg_size(0),
.nlmsg_type = type,
.nlmsg_flags = flags,
};
msg = nlmsg_build(&nlh);
if (!msg)
return nl_errno(ENOMEM);
if (buf && size)
nlmsg_append(msg, buf, size, 1);
err = nl_send_auto_complete(handle, msg);
nlmsg_free(msg);
return err;
}
/** @} */
/**
* @name Receive
* @{
*/
/**
* Receive netlink message from netlink socket.
* @arg handle Netlink handle.
* @arg nla Destination pointer for peer's netlink address.
* @arg buf Destination pointer for message content.
* @arg creds Destination pointer for credentials.
*
* Receives a netlink message, allocates a buffer in \c *buf and
* stores the message content. The peer's netlink address is stored
* in \c *nla. The caller is responsible for freeing the buffer allocated
* in \c *buf if a positive value is returned. Interruped system calls
* are handled by repeating the read. The input buffer size is determined
* by peeking before the actual read is done.
*
* A non-blocking sockets causes the function to return immediately if
* no data is available.
*
* @return Number of octets read, 0 on EOF or a negative error code.
*/
int nl_recv(struct nl_handle *handle, struct sockaddr_nl *nla,
unsigned char **buf, struct ucred **creds)
{
int n;
int flags = MSG_PEEK;
struct iovec iov = {
.iov_len = 4096,
};
struct msghdr msg = {
.msg_name = (void *) nla,
.msg_namelen = sizeof(sizeof(struct sockaddr_nl)),
.msg_iov = &iov,
.msg_iovlen = 1,
.msg_control = NULL,
.msg_controllen = 0,
.msg_flags = 0,
};
struct cmsghdr *cmsg;
iov.iov_base = *buf = calloc(1, iov.iov_len);
if (handle->h_flags & NL_SOCK_PASSCRED) {
msg.msg_controllen = CMSG_SPACE(sizeof(struct ucred));
msg.msg_control = calloc(1, msg.msg_controllen);
}
retry:
if ((n = recvmsg(handle->h_fd, &msg, flags)) <= 0) {
if (!n)
goto abort;
else if (n < 0) {
if (errno == EINTR)
goto retry;
else if (errno == EAGAIN)
goto abort;
else {
free(msg.msg_control);
free(*buf);
return nl_error(errno, "recvmsg failed");
}
}
}
if (iov.iov_len < n) {
/* Provided buffer is not long enough, enlarge it
* and try again. */
iov.iov_len *= 2;
iov.iov_base = *buf = realloc(*buf, iov.iov_len);
goto retry;
} else if (msg.msg_flags & MSG_CTRUNC) {
msg.msg_controllen *= 2;
msg.msg_control = realloc(msg.msg_control, msg.msg_controllen);
goto retry;
} else if (flags != 0) {
/* Buffer is big enough, do the actual reading */
flags = 0;
goto retry;
}
if (msg.msg_namelen != sizeof(struct sockaddr_nl)) {
free(msg.msg_control);
free(*buf);
return nl_error(EADDRNOTAVAIL, "socket address size mismatch");
}
for (cmsg = CMSG_FIRSTHDR(&msg); cmsg; cmsg = CMSG_NXTHDR(&msg, cmsg)) {
if (cmsg->cmsg_level == SOL_SOCKET &&
cmsg->cmsg_type == SCM_CREDENTIALS) {
*creds = calloc(1, sizeof(struct ucred));
memcpy(*creds, CMSG_DATA(cmsg), sizeof(struct ucred));
break;
}
}
free(msg.msg_control);
return n;
abort:
free(msg.msg_control);
free(*buf);
return 0;
}
/**
* Receive a set of messages from a netlink socket.
* @arg handle netlink handle
* @arg cb set of callbacks to control the behaviour.
*
* Repeatedly calls nl_recv() and parses the messages as netlink
* messages. Stops reading if one of the callbacks returns
* NL_EXIT or nl_recv returns either 0 or a negative error code.
*
* A non-blocking sockets causes the function to return immediately if
* no data is available.
*
* @return 0 on success or a negative error code from nl_recv().
*/
int nl_recvmsgs(struct nl_handle *handle, struct nl_cb *cb)
{
int n, err = 0;
unsigned char *buf = NULL;
struct nlmsghdr *hdr;
struct sockaddr_nl nla = {0};
struct nl_msg *msg = NULL;
struct ucred *creds = NULL;
continue_reading:
if (cb->cb_recv_ow)
n = cb->cb_recv_ow(handle, &nla, &buf, &creds);
else
n = nl_recv(handle, &nla, &buf, &creds);
if (n <= 0)
return n;
NL_DBG(3, "recvmsgs(%p): Read %d bytes\n", handle, n);
hdr = (struct nlmsghdr *) buf;
while (nlmsg_ok(hdr, n)) {
NL_DBG(3, "recgmsgs(%p): Processing valid message...\n",
handle);
nlmsg_free(msg);
msg = nlmsg_convert(hdr);
if (!msg) {
err = nl_errno(ENOMEM);
goto out;
}
nlmsg_set_proto(msg, handle->h_proto);
nlmsg_set_src(msg, &nla);
if (creds)
nlmsg_set_creds(msg, creds);
/* Raw callback is the first, it gives the most control
* to the user and he can do his very own parsing. */
if (cb->cb_set[NL_CB_MSG_IN]) {
err = nl_cb_call(cb, NL_CB_MSG_IN, msg);
if (err == NL_SKIP)
goto skip;
else if (err == NL_EXIT || err < 0)
goto out;
}
/* Sequence number checking. The check may be done by
* the user, otherwise a very simple check is applied
* enforcing strict ordering */
if (cb->cb_set[NL_CB_SEQ_CHECK]) {
err = nl_cb_call(cb, NL_CB_SEQ_CHECK, msg);
if (err == NL_SKIP)
goto skip;
else if (err == NL_EXIT || err < 0)
goto out;
} else if (hdr->nlmsg_seq != handle->h_seq_expect) {
if (cb->cb_set[NL_CB_INVALID]) {
err = nl_cb_call(cb, NL_CB_INVALID, msg);
if (err == NL_SKIP)
goto skip;
else if (err == NL_EXIT || err < 0)
goto out;
} else
goto out;
}
if (hdr->nlmsg_type == NLMSG_DONE ||
hdr->nlmsg_type == NLMSG_ERROR ||
hdr->nlmsg_type == NLMSG_NOOP ||
hdr->nlmsg_type == NLMSG_OVERRUN) {
/* We can't check for !NLM_F_MULTI since some netlink
* users in the kernel are broken. */
handle->h_seq_expect++;
NL_DBG(3, "recvmsgs(%p): Increased expected " \
"sequence number to %d\n",
handle, handle->h_seq_expect);
}
/* Other side wishes to see an ack for this message */
if (hdr->nlmsg_flags & NLM_F_ACK) {
if (cb->cb_set[NL_CB_SEND_ACK]) {
err = nl_cb_call(cb, NL_CB_SEND_ACK, msg);
if (err == NL_SKIP)
goto skip;
else if (err == NL_EXIT || err < 0)
goto out;
} else {
/* FIXME: implement */
}
}
/* messages terminates a multpart message, this is
* usually the end of a message and therefore we slip
* out of the loop by default. the user may overrule
* this action by skipping this packet. */
if (hdr->nlmsg_type == NLMSG_DONE) {
if (cb->cb_set[NL_CB_FINISH]) {
err = nl_cb_call(cb, NL_CB_FINISH, msg);
if (err == NL_SKIP)
goto skip;
else if (err == NL_EXIT || err < 0)
goto out;
}
err = 0;
goto out;
}
/* Message to be ignored, the default action is to
* skip this message if no callback is specified. The
* user may overrule this action by returning
* NL_PROCEED. */
else if (hdr->nlmsg_type == NLMSG_NOOP) {
if (cb->cb_set[NL_CB_SKIPPED]) {
err = nl_cb_call(cb, NL_CB_SKIPPED, msg);
if (err == NL_SKIP)
goto skip;
else if (err == NL_EXIT || err < 0)
goto out;
} else
goto skip;
}
/* Data got lost, report back to user. The default action is to
* quit parsing. The user may overrule this action by retuning
* NL_SKIP or NL_PROCEED (dangerous) */
else if (hdr->nlmsg_type == NLMSG_OVERRUN) {
if (cb->cb_set[NL_CB_OVERRUN]) {
err = nl_cb_call(cb, NL_CB_OVERRUN, msg);
if (err == NL_SKIP)
goto skip;
else if (err == NL_EXIT || err < 0)
goto out;
} else
goto out;
}
/* Message carries a nlmsgerr */
else if (hdr->nlmsg_type == NLMSG_ERROR) {
struct nlmsgerr *e = nlmsg_data(hdr);
if (hdr->nlmsg_len < nlmsg_msg_size(sizeof(*e))) {
/* Truncated error message, the default action
* is to stop parsing. The user may overrule
* this action by returning NL_SKIP or
* NL_PROCEED (dangerous) */
if (cb->cb_set[NL_CB_INVALID]) {
err = nl_cb_call(cb, NL_CB_INVALID,
msg);
if (err == NL_SKIP)
goto skip;
else if (err == NL_EXIT || err < 0)
goto out;
} else
goto out;
} else if (e->error) {
/* Error message reported back from kernel. */
if (cb->cb_err) {
err = cb->cb_err(&nla, e,
cb->cb_err_arg);
if (err == NL_SKIP)
goto skip;
else if (err == NL_EXIT || err < 0) {
nl_error(-e->error,
"Netlink Error");
err = e->error;
goto out;
}
} else {
nl_error(-e->error, "Netlink Error");
err = e->error;
goto out;
}
} else if (cb->cb_set[NL_CB_ACK]) {
/* ACK */
err = nl_cb_call(cb, NL_CB_ACK, msg);
if (err == NL_SKIP)
goto skip;
else if (err == NL_EXIT || err < 0)
goto out;
}
} else {
/* Valid message (not checking for MULTIPART bit to
* get along with broken kernels. NL_SKIP has no
* effect on this. */
if (cb->cb_set[NL_CB_VALID]) {
err = nl_cb_call(cb, NL_CB_VALID, msg);
if (err == NL_SKIP)
goto skip;
else if (err == NL_EXIT || err < 0)
goto out;
}
}
skip:
hdr = nlmsg_next(hdr, &n);
}
nlmsg_free(msg);
free(buf);
free(creds);
buf = NULL;
msg = NULL;
creds = NULL;
/* Multipart message not yet complete, continue reading */
goto continue_reading;
out:
nlmsg_free(msg);
free(buf);
free(creds);
return err;
}
/**
* Receive a set of message from a netlink socket using handlers in nl_handle.
* @arg handle netlink handle
*
* Calls nl_recvmsgs() with the handlers configured in the netlink handle.
*/
int nl_recvmsgs_def(struct nl_handle *handle)
{
if (handle->h_cb->cb_recvmsgs_ow)
return handle->h_cb->cb_recvmsgs_ow(handle, handle->h_cb);
else
return nl_recvmsgs(handle, handle->h_cb);
}
static int ack_wait_handler(struct nl_msg *msg, void *arg)
{
return NL_EXIT;
}
/**
* Wait for ACK.
* @arg handle netlink handle
* @pre The netlink socket must be in blocking state.
*
* Waits until an ACK is received for the latest not yet acknowledged
* netlink message.
*/
int nl_wait_for_ack(struct nl_handle *handle)
{
int err;
struct nl_cb *cb = nl_cb_clone(nl_handle_get_cb(handle));
nl_cb_set(cb, NL_CB_ACK, NL_CB_CUSTOM, ack_wait_handler, NULL);
err = nl_recvmsgs(handle, cb);
nl_cb_destroy(cb);
return err;
}
/** @} */
/**
* @name Netlink Family Translations
* @{
*/
static struct trans_tbl nlfamilies[] = {
__ADD(NETLINK_ROUTE,route)
__ADD(NETLINK_W1,w1)
__ADD(NETLINK_USERSOCK,usersock)
__ADD(NETLINK_FIREWALL,firewall)
__ADD(NETLINK_INET_DIAG,inetdiag)
__ADD(NETLINK_NFLOG,nflog)
__ADD(NETLINK_XFRM,xfrm)
__ADD(NETLINK_SELINUX,selinux)
__ADD(NETLINK_ISCSI,iscsi)
__ADD(NETLINK_AUDIT,audit)
__ADD(NETLINK_FIB_LOOKUP,fib_lookup)
__ADD(NETLINK_CONNECTOR,connector)
__ADD(NETLINK_NETFILTER,netfilter)
__ADD(NETLINK_IP6_FW,ip6_fw)
__ADD(NETLINK_DNRTMSG,dnrtmsg)
__ADD(NETLINK_KOBJECT_UEVENT,kobject_uevent)
__ADD(NETLINK_GENERIC,generic)
};
/**
* Convert netlink family to character string.
* @arg family Netlink family.
* @arg buf Destination buffer.
* @arg size Size of destination buffer.
*
* Converts a netlink family to a character string and stores it in
* the specified destination buffer.
*
* @return The destination buffer or the family encoded in hexidecimal
* form if no match was found.
*/
char * nl_nlfamily2str(int family, char *buf, size_t size)
{
return __type2str(family, buf, size, nlfamilies,
ARRAY_SIZE(nlfamilies));
}
/**
* Convert character string to netlink family.
* @arg name Name of netlink family.
*
* Converts the provided character string specifying a netlink
* family to the corresponding numeric value.
*
* @return Numeric netlink family or a negative value if no match was found.
*/
int nl_str2nlfamily(const char *name)
{
return __str2type(name, nlfamilies, ARRAY_SIZE(nlfamilies));
}
