diff options
Diffstat (limited to 'src/network/netdev')
37 files changed, 5671 insertions, 0 deletions
diff --git a/src/network/netdev/bond.c b/src/network/netdev/bond.c new file mode 100644 index 0000000..550a7f8 --- /dev/null +++ b/src/network/netdev/bond.c @@ -0,0 +1,561 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ + +#include <netinet/ether.h> +#include <linux/if_bonding.h> + +#include "sd-netlink.h" + +#include "alloc-util.h" +#include "conf-parser.h" +#include "ether-addr-util.h" +#include "extract-word.h" +#include "missing.h" +#include "netdev/bond.h" +#include "string-table.h" +#include "string-util.h" + +/* + * Number of seconds between instances where the bonding + * driver sends learning packets to each slaves peer switch + */ +#define LEARNING_PACKETS_INTERVAL_MIN_SEC (1 * USEC_PER_SEC) +#define LEARNING_PACKETS_INTERVAL_MAX_SEC (0x7fffffff * USEC_PER_SEC) + +/* Number of IGMP membership reports to be issued after + * a failover event. + */ +#define RESEND_IGMP_MIN 0 +#define RESEND_IGMP_MAX 255 +#define RESEND_IGMP_DEFAULT 1 + +/* + * Number of packets to transmit through a slave before + * moving to the next one. + */ +#define PACKETS_PER_SLAVE_MIN 0 +#define PACKETS_PER_SLAVE_MAX 65535 +#define PACKETS_PER_SLAVE_DEFAULT 1 + +/* + * Number of peer notifications (gratuitous ARPs and + * unsolicited IPv6 Neighbor Advertisements) to be issued after a + * failover event. + */ +#define GRATUITOUS_ARP_MIN 0 +#define GRATUITOUS_ARP_MAX 255 +#define GRATUITOUS_ARP_DEFAULT 1 + +static const char* const bond_mode_table[_NETDEV_BOND_MODE_MAX] = { + [NETDEV_BOND_MODE_BALANCE_RR] = "balance-rr", + [NETDEV_BOND_MODE_ACTIVE_BACKUP] = "active-backup", + [NETDEV_BOND_MODE_BALANCE_XOR] = "balance-xor", + [NETDEV_BOND_MODE_BROADCAST] = "broadcast", + [NETDEV_BOND_MODE_802_3AD] = "802.3ad", + [NETDEV_BOND_MODE_BALANCE_TLB] = "balance-tlb", + [NETDEV_BOND_MODE_BALANCE_ALB] = "balance-alb", +}; + +DEFINE_STRING_TABLE_LOOKUP(bond_mode, BondMode); +DEFINE_CONFIG_PARSE_ENUM(config_parse_bond_mode, bond_mode, BondMode, "Failed to parse bond mode"); + +static const char* const bond_xmit_hash_policy_table[_NETDEV_BOND_XMIT_HASH_POLICY_MAX] = { + [NETDEV_BOND_XMIT_HASH_POLICY_LAYER2] = "layer2", + [NETDEV_BOND_XMIT_HASH_POLICY_LAYER34] = "layer3+4", + [NETDEV_BOND_XMIT_HASH_POLICY_LAYER23] = "layer2+3", + [NETDEV_BOND_XMIT_HASH_POLICY_ENCAP23] = "encap2+3", + [NETDEV_BOND_XMIT_HASH_POLICY_ENCAP34] = "encap3+4", +}; + +DEFINE_STRING_TABLE_LOOKUP(bond_xmit_hash_policy, BondXmitHashPolicy); +DEFINE_CONFIG_PARSE_ENUM(config_parse_bond_xmit_hash_policy, + bond_xmit_hash_policy, + BondXmitHashPolicy, + "Failed to parse bond transmit hash policy") + +static const char* const bond_lacp_rate_table[_NETDEV_BOND_LACP_RATE_MAX] = { + [NETDEV_BOND_LACP_RATE_SLOW] = "slow", + [NETDEV_BOND_LACP_RATE_FAST] = "fast", +}; + +DEFINE_STRING_TABLE_LOOKUP(bond_lacp_rate, BondLacpRate); +DEFINE_CONFIG_PARSE_ENUM(config_parse_bond_lacp_rate, bond_lacp_rate, BondLacpRate, "Failed to parse bond lacp rate") + +static const char* const bond_ad_select_table[_NETDEV_BOND_AD_SELECT_MAX] = { + [NETDEV_BOND_AD_SELECT_STABLE] = "stable", + [NETDEV_BOND_AD_SELECT_BANDWIDTH] = "bandwidth", + [NETDEV_BOND_AD_SELECT_COUNT] = "count", +}; + +DEFINE_STRING_TABLE_LOOKUP(bond_ad_select, BondAdSelect); +DEFINE_CONFIG_PARSE_ENUM(config_parse_bond_ad_select, bond_ad_select, BondAdSelect, "Failed to parse bond AD select"); + +static const char* const bond_fail_over_mac_table[_NETDEV_BOND_FAIL_OVER_MAC_MAX] = { + [NETDEV_BOND_FAIL_OVER_MAC_NONE] = "none", + [NETDEV_BOND_FAIL_OVER_MAC_ACTIVE] = "active", + [NETDEV_BOND_FAIL_OVER_MAC_FOLLOW] = "follow", +}; + +DEFINE_STRING_TABLE_LOOKUP(bond_fail_over_mac, BondFailOverMac); +DEFINE_CONFIG_PARSE_ENUM(config_parse_bond_fail_over_mac, bond_fail_over_mac, BondFailOverMac, "Failed to parse bond fail over MAC"); + +static const char *const bond_arp_validate_table[_NETDEV_BOND_ARP_VALIDATE_MAX] = { + [NETDEV_BOND_ARP_VALIDATE_NONE] = "none", + [NETDEV_BOND_ARP_VALIDATE_ACTIVE]= "active", + [NETDEV_BOND_ARP_VALIDATE_BACKUP]= "backup", + [NETDEV_BOND_ARP_VALIDATE_ALL]= "all", +}; + +DEFINE_STRING_TABLE_LOOKUP(bond_arp_validate, BondArpValidate); +DEFINE_CONFIG_PARSE_ENUM(config_parse_bond_arp_validate, bond_arp_validate, BondArpValidate, "Failed to parse bond arp validate"); + +static const char *const bond_arp_all_targets_table[_NETDEV_BOND_ARP_ALL_TARGETS_MAX] = { + [NETDEV_BOND_ARP_ALL_TARGETS_ANY] = "any", + [NETDEV_BOND_ARP_ALL_TARGETS_ALL] = "all", +}; + +DEFINE_STRING_TABLE_LOOKUP(bond_arp_all_targets, BondArpAllTargets); +DEFINE_CONFIG_PARSE_ENUM(config_parse_bond_arp_all_targets, bond_arp_all_targets, BondArpAllTargets, "Failed to parse bond Arp all targets"); + +static const char *bond_primary_reselect_table[_NETDEV_BOND_PRIMARY_RESELECT_MAX] = { + [NETDEV_BOND_PRIMARY_RESELECT_ALWAYS] = "always", + [NETDEV_BOND_PRIMARY_RESELECT_BETTER]= "better", + [NETDEV_BOND_PRIMARY_RESELECT_FAILURE]= "failure", +}; + +DEFINE_STRING_TABLE_LOOKUP(bond_primary_reselect, BondPrimaryReselect); +DEFINE_CONFIG_PARSE_ENUM(config_parse_bond_primary_reselect, bond_primary_reselect, BondPrimaryReselect, "Failed to parse bond primary reselect"); + +static uint8_t bond_mode_to_kernel(BondMode mode) { + switch (mode) { + case NETDEV_BOND_MODE_BALANCE_RR: + return BOND_MODE_ROUNDROBIN; + case NETDEV_BOND_MODE_ACTIVE_BACKUP: + return BOND_MODE_ACTIVEBACKUP; + case NETDEV_BOND_MODE_BALANCE_XOR: + return BOND_MODE_XOR; + case NETDEV_BOND_MODE_BROADCAST: + return BOND_MODE_BROADCAST; + case NETDEV_BOND_MODE_802_3AD: + return BOND_MODE_8023AD; + case NETDEV_BOND_MODE_BALANCE_TLB: + return BOND_MODE_TLB; + case NETDEV_BOND_MODE_BALANCE_ALB: + return BOND_MODE_ALB; + default: + return (uint8_t) -1; + } +} + +static uint8_t bond_xmit_hash_policy_to_kernel(BondXmitHashPolicy policy) { + switch (policy) { + case NETDEV_BOND_XMIT_HASH_POLICY_LAYER2: + return BOND_XMIT_POLICY_LAYER2; + case NETDEV_BOND_XMIT_HASH_POLICY_LAYER34: + return BOND_XMIT_POLICY_LAYER34; + case NETDEV_BOND_XMIT_HASH_POLICY_LAYER23: + return BOND_XMIT_POLICY_LAYER23; + case NETDEV_BOND_XMIT_HASH_POLICY_ENCAP23: + return BOND_XMIT_POLICY_ENCAP23; + case NETDEV_BOND_XMIT_HASH_POLICY_ENCAP34: + return BOND_XMIT_POLICY_ENCAP34; + default: + return (uint8_t) -1; + } +} + +static int netdev_bond_fill_message_create(NetDev *netdev, Link *link, sd_netlink_message *m) { + Bond *b; + ArpIpTarget *target = NULL; + int r, i = 0; + + assert(netdev); + assert(!link); + assert(m); + + b = BOND(netdev); + + assert(b); + + if (b->mode != _NETDEV_BOND_MODE_INVALID) { + r = sd_netlink_message_append_u8(m, IFLA_BOND_MODE, + bond_mode_to_kernel(b->mode)); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BOND_MODE attribute: %m"); + } + + if (b->xmit_hash_policy != _NETDEV_BOND_XMIT_HASH_POLICY_INVALID) { + r = sd_netlink_message_append_u8(m, IFLA_BOND_XMIT_HASH_POLICY, + bond_xmit_hash_policy_to_kernel(b->xmit_hash_policy)); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BOND_XMIT_HASH_POLICY attribute: %m"); + } + + if (b->lacp_rate != _NETDEV_BOND_LACP_RATE_INVALID && + b->mode == NETDEV_BOND_MODE_802_3AD) { + r = sd_netlink_message_append_u8(m, IFLA_BOND_AD_LACP_RATE, b->lacp_rate ); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BOND_AD_LACP_RATE attribute: %m"); + } + + if (b->miimon != 0) { + r = sd_netlink_message_append_u32(m, IFLA_BOND_MIIMON, b->miimon / USEC_PER_MSEC); + if (r < 0) + log_netdev_error_errno(netdev, r, "Could not append IFLA_BOND_BOND_MIIMON attribute: %m"); + } + + if (b->downdelay != 0) { + r = sd_netlink_message_append_u32(m, IFLA_BOND_DOWNDELAY, b->downdelay / USEC_PER_MSEC); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BOND_DOWNDELAY attribute: %m"); + } + + if (b->updelay != 0) { + r = sd_netlink_message_append_u32(m, IFLA_BOND_UPDELAY, b->updelay / USEC_PER_MSEC); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BOND_UPDELAY attribute: %m"); + } + + if (b->arp_interval != 0) { + r = sd_netlink_message_append_u32(m, IFLA_BOND_ARP_INTERVAL, b->arp_interval / USEC_PER_MSEC); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BOND_ARP_INTERVAL attribute: %m"); + + if ((b->lp_interval >= LEARNING_PACKETS_INTERVAL_MIN_SEC) && + (b->lp_interval <= LEARNING_PACKETS_INTERVAL_MAX_SEC)) { + r = sd_netlink_message_append_u32(m, IFLA_BOND_LP_INTERVAL, b->lp_interval / USEC_PER_SEC); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BOND_LP_INTERVAL attribute: %m"); + } + } + + if (b->ad_select != _NETDEV_BOND_AD_SELECT_INVALID && + b->mode == NETDEV_BOND_MODE_802_3AD) { + r = sd_netlink_message_append_u8(m, IFLA_BOND_AD_SELECT, b->ad_select); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BOND_AD_SELECT attribute: %m"); + } + + if (b->fail_over_mac != _NETDEV_BOND_FAIL_OVER_MAC_INVALID && + b->mode == NETDEV_BOND_MODE_ACTIVE_BACKUP) { + r = sd_netlink_message_append_u8(m, IFLA_BOND_FAIL_OVER_MAC, b->fail_over_mac); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BOND_FAIL_OVER_MAC attribute: %m"); + } + + if (b->arp_validate != _NETDEV_BOND_ARP_VALIDATE_INVALID) { + r = sd_netlink_message_append_u32(m, IFLA_BOND_ARP_VALIDATE, b->arp_validate); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BOND_ARP_VALIDATE attribute: %m"); + } + + if (b->arp_all_targets != _NETDEV_BOND_ARP_ALL_TARGETS_INVALID) { + r = sd_netlink_message_append_u32(m, IFLA_BOND_ARP_ALL_TARGETS, b->arp_all_targets); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BOND_ARP_ALL_TARGETS attribute: %m"); + } + + if (b->primary_reselect != _NETDEV_BOND_PRIMARY_RESELECT_INVALID) { + r = sd_netlink_message_append_u8(m, IFLA_BOND_PRIMARY_RESELECT, b->primary_reselect); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BOND_PRIMARY_RESELECT attribute: %m"); + } + + if (b->resend_igmp <= RESEND_IGMP_MAX) { + r = sd_netlink_message_append_u32(m, IFLA_BOND_RESEND_IGMP, b->resend_igmp); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BOND_RESEND_IGMP attribute: %m"); + } + + if (b->packets_per_slave <= PACKETS_PER_SLAVE_MAX && + b->mode == NETDEV_BOND_MODE_BALANCE_RR) { + r = sd_netlink_message_append_u32(m, IFLA_BOND_PACKETS_PER_SLAVE, b->packets_per_slave); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BOND_PACKETS_PER_SLAVE attribute: %m"); + } + + if (b->num_grat_arp <= GRATUITOUS_ARP_MAX) { + r = sd_netlink_message_append_u8(m, IFLA_BOND_NUM_PEER_NOTIF, b->num_grat_arp); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BOND_NUM_PEER_NOTIF attribute: %m"); + } + + if (b->min_links != 0) { + r = sd_netlink_message_append_u32(m, IFLA_BOND_MIN_LINKS, b->min_links); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BOND_MIN_LINKS attribute: %m"); + } + + if (b->ad_actor_sys_prio != 0) { + r = sd_netlink_message_append_u16(m, IFLA_BOND_AD_ACTOR_SYS_PRIO, b->ad_actor_sys_prio); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BOND_AD_ACTOR_SYS_PRIO attribute: %m"); + } + + if (b->ad_user_port_key != 0) { + r = sd_netlink_message_append_u16(m, IFLA_BOND_AD_USER_PORT_KEY, b->ad_user_port_key); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BOND_AD_USER_PORT_KEY attribute: %m"); + } + + if (b->ad_actor_system) { + r = sd_netlink_message_append_ether_addr(m, IFLA_BOND_AD_ACTOR_SYSTEM, b->ad_actor_system); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BOND_AD_ACTOR_SYSTEM attribute: %m"); + } + + r = sd_netlink_message_append_u8(m, IFLA_BOND_ALL_SLAVES_ACTIVE, b->all_slaves_active); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BOND_ALL_SLAVES_ACTIVE attribute: %m"); + + if (b->tlb_dynamic_lb >= 0) { + r = sd_netlink_message_append_u8(m, IFLA_BOND_TLB_DYNAMIC_LB, b->tlb_dynamic_lb); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BOND_TLB_DYNAMIC_LB attribute: %m"); + } + + if (b->arp_interval > 0) { + if (b->n_arp_ip_targets > 0) { + + r = sd_netlink_message_open_container(m, IFLA_BOND_ARP_IP_TARGET); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not open contaniner IFLA_BOND_ARP_IP_TARGET : %m"); + + LIST_FOREACH(arp_ip_target, target, b->arp_ip_targets) { + r = sd_netlink_message_append_u32(m, i++, target->ip.in.s_addr); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BOND_ARP_ALL_TARGETS attribute: %m"); + } + + r = sd_netlink_message_close_container(m); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not close contaniner IFLA_BOND_ARP_IP_TARGET : %m"); + } + } + + return 0; +} + +int config_parse_arp_ip_target_address(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + Bond *b = userdata; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + for (;;) { + _cleanup_free_ ArpIpTarget *buffer = NULL; + _cleanup_free_ char *n = NULL; + int f; + + r = extract_first_word(&rvalue, &n, NULL, 0); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse Bond ARP ip target address, ignoring assignment: %s", rvalue); + return 0; + } + + if (r == 0) + break; + + buffer = new0(ArpIpTarget, 1); + if (!buffer) + return -ENOMEM; + + r = in_addr_from_string_auto(n, &f, &buffer->ip); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Bond ARP ip target address is invalid, ignoring assignment: %s", n); + return 0; + } + + if (f != AF_INET) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Bond ARP ip target address is invalid, ignoring assignment: %s", n); + return 0; + } + + LIST_PREPEND(arp_ip_target, b->arp_ip_targets, TAKE_PTR(buffer)); + b->n_arp_ip_targets++; + } + + if (b->n_arp_ip_targets > NETDEV_BOND_ARP_TARGETS_MAX) + log_syntax(unit, LOG_WARNING, filename, line, 0, + "More than the maximum number of kernel-supported ARP ip targets specified: %d > %d", + b->n_arp_ip_targets, NETDEV_BOND_ARP_TARGETS_MAX); + + return 0; +} + +int config_parse_ad_actor_sys_prio(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + Bond *b = userdata; + uint16_t v; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + r = safe_atou16(rvalue, &v); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse actor system priority '%s', ignoring: %m", rvalue); + return 0; + } + + if (v == 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse actor system priority '%s'. Range is [1,65535], ignoring.", rvalue); + return 0; + } + + b->ad_actor_sys_prio = v; + + return 0; +} + +int config_parse_ad_user_port_key(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + Bond *b = userdata; + uint16_t v; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + r = safe_atou16(rvalue, &v); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse user port key '%s', ignoring: %m", rvalue); + return 0; + } + + if (v > 1023) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse user port key '%s'. Range is [0,1023], ignoring.", rvalue); + return 0; + } + + b->ad_user_port_key = v; + + return 0; +} + +int config_parse_ad_actor_system(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + Bond *b = userdata; + _cleanup_free_ struct ether_addr *n = NULL; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + n = new0(struct ether_addr, 1); + if (!n) + return log_oom(); + + r = ether_addr_from_string(rvalue, n); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Not a valid MAC address %s. Ignoring assignment: %m", rvalue); + return 0; + } + + if (ether_addr_is_null(n) || (n->ether_addr_octet[0] & 0x01)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Not a valid MAC address %s, can not be null or multicast. Ignoring assignment.", rvalue); + return 0; + } + + free_and_replace(b->ad_actor_system, n); + + return 0; +} + +static void bond_done(NetDev *netdev) { + ArpIpTarget *t = NULL, *n = NULL; + Bond *b; + + assert(netdev); + + b = BOND(netdev); + + assert(b); + + free(b->ad_actor_system); + + LIST_FOREACH_SAFE(arp_ip_target, t, n, b->arp_ip_targets) + free(t); + + b->arp_ip_targets = NULL; +} + +static void bond_init(NetDev *netdev) { + Bond *b; + + assert(netdev); + + b = BOND(netdev); + + assert(b); + + b->mode = _NETDEV_BOND_MODE_INVALID; + b->xmit_hash_policy = _NETDEV_BOND_XMIT_HASH_POLICY_INVALID; + b->lacp_rate = _NETDEV_BOND_LACP_RATE_INVALID; + b->ad_select = _NETDEV_BOND_AD_SELECT_INVALID; + b->fail_over_mac = _NETDEV_BOND_FAIL_OVER_MAC_INVALID; + b->arp_validate = _NETDEV_BOND_ARP_VALIDATE_INVALID; + b->arp_all_targets = _NETDEV_BOND_ARP_ALL_TARGETS_INVALID; + b->primary_reselect = _NETDEV_BOND_PRIMARY_RESELECT_INVALID; + + b->all_slaves_active = false; + b->tlb_dynamic_lb = -1; + + b->resend_igmp = RESEND_IGMP_DEFAULT; + b->packets_per_slave = PACKETS_PER_SLAVE_DEFAULT; + b->num_grat_arp = GRATUITOUS_ARP_DEFAULT; + b->lp_interval = LEARNING_PACKETS_INTERVAL_MIN_SEC; + + LIST_HEAD_INIT(b->arp_ip_targets); + b->n_arp_ip_targets = 0; +} + +const NetDevVTable bond_vtable = { + .object_size = sizeof(Bond), + .init = bond_init, + .done = bond_done, + .sections = "Match\0NetDev\0Bond\0", + .fill_message_create = netdev_bond_fill_message_create, + .create_type = NETDEV_CREATE_MASTER, +}; diff --git a/src/network/netdev/bond.h b/src/network/netdev/bond.h new file mode 100644 index 0000000..31b922b --- /dev/null +++ b/src/network/netdev/bond.h @@ -0,0 +1,163 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ +#pragma once + +#include "in-addr-util.h" +#include "list.h" + +#include "netdev/netdev.h" + +/* + * Maximum number of targets supported by the kernel for a single + * bond netdev. + */ +#define NETDEV_BOND_ARP_TARGETS_MAX 16 + +typedef enum BondMode { + NETDEV_BOND_MODE_BALANCE_RR, + NETDEV_BOND_MODE_ACTIVE_BACKUP, + NETDEV_BOND_MODE_BALANCE_XOR, + NETDEV_BOND_MODE_BROADCAST, + NETDEV_BOND_MODE_802_3AD, + NETDEV_BOND_MODE_BALANCE_TLB, + NETDEV_BOND_MODE_BALANCE_ALB, + _NETDEV_BOND_MODE_MAX, + _NETDEV_BOND_MODE_INVALID = -1 +} BondMode; + +typedef enum BondXmitHashPolicy { + NETDEV_BOND_XMIT_HASH_POLICY_LAYER2, + NETDEV_BOND_XMIT_HASH_POLICY_LAYER34, + NETDEV_BOND_XMIT_HASH_POLICY_LAYER23, + NETDEV_BOND_XMIT_HASH_POLICY_ENCAP23, + NETDEV_BOND_XMIT_HASH_POLICY_ENCAP34, + _NETDEV_BOND_XMIT_HASH_POLICY_MAX, + _NETDEV_BOND_XMIT_HASH_POLICY_INVALID = -1 +} BondXmitHashPolicy; + +typedef enum BondLacpRate { + NETDEV_BOND_LACP_RATE_SLOW, + NETDEV_BOND_LACP_RATE_FAST, + _NETDEV_BOND_LACP_RATE_MAX, + _NETDEV_BOND_LACP_RATE_INVALID = -1, +} BondLacpRate; + +typedef enum BondAdSelect { + NETDEV_BOND_AD_SELECT_STABLE, + NETDEV_BOND_AD_SELECT_BANDWIDTH, + NETDEV_BOND_AD_SELECT_COUNT, + _NETDEV_BOND_AD_SELECT_MAX, + _NETDEV_BOND_AD_SELECT_INVALID = -1, +} BondAdSelect; + +typedef enum BondFailOverMac { + NETDEV_BOND_FAIL_OVER_MAC_NONE, + NETDEV_BOND_FAIL_OVER_MAC_ACTIVE, + NETDEV_BOND_FAIL_OVER_MAC_FOLLOW, + _NETDEV_BOND_FAIL_OVER_MAC_MAX, + _NETDEV_BOND_FAIL_OVER_MAC_INVALID = -1, +} BondFailOverMac; + +typedef enum BondArpValidate { + NETDEV_BOND_ARP_VALIDATE_NONE, + NETDEV_BOND_ARP_VALIDATE_ACTIVE, + NETDEV_BOND_ARP_VALIDATE_BACKUP, + NETDEV_BOND_ARP_VALIDATE_ALL, + _NETDEV_BOND_ARP_VALIDATE_MAX, + _NETDEV_BOND_ARP_VALIDATE_INVALID = -1, +} BondArpValidate; + +typedef enum BondArpAllTargets { + NETDEV_BOND_ARP_ALL_TARGETS_ANY, + NETDEV_BOND_ARP_ALL_TARGETS_ALL, + _NETDEV_BOND_ARP_ALL_TARGETS_MAX, + _NETDEV_BOND_ARP_ALL_TARGETS_INVALID = -1, +} BondArpAllTargets; + +typedef enum BondPrimaryReselect { + NETDEV_BOND_PRIMARY_RESELECT_ALWAYS, + NETDEV_BOND_PRIMARY_RESELECT_BETTER, + NETDEV_BOND_PRIMARY_RESELECT_FAILURE, + _NETDEV_BOND_PRIMARY_RESELECT_MAX, + _NETDEV_BOND_PRIMARY_RESELECT_INVALID = -1, +} BondPrimaryReselect; + +typedef struct ArpIpTarget { + union in_addr_union ip; + + LIST_FIELDS(struct ArpIpTarget, arp_ip_target); +} ArpIpTarget; + +typedef struct Bond { + NetDev meta; + + BondMode mode; + BondXmitHashPolicy xmit_hash_policy; + BondLacpRate lacp_rate; + BondAdSelect ad_select; + BondFailOverMac fail_over_mac; + BondArpValidate arp_validate; + BondArpAllTargets arp_all_targets; + BondPrimaryReselect primary_reselect; + + int tlb_dynamic_lb; + + bool all_slaves_active; + + unsigned resend_igmp; + unsigned packets_per_slave; + unsigned num_grat_arp; + unsigned min_links; + + uint16_t ad_actor_sys_prio; + uint16_t ad_user_port_key; + struct ether_addr *ad_actor_system; + + usec_t miimon; + usec_t updelay; + usec_t downdelay; + usec_t arp_interval; + usec_t lp_interval; + + int n_arp_ip_targets; + ArpIpTarget *arp_ip_targets; +} Bond; + +DEFINE_NETDEV_CAST(BOND, Bond); +extern const NetDevVTable bond_vtable; + +const char *bond_mode_to_string(BondMode d) _const_; +BondMode bond_mode_from_string(const char *d) _pure_; + +const char *bond_xmit_hash_policy_to_string(BondXmitHashPolicy d) _const_; +BondXmitHashPolicy bond_xmit_hash_policy_from_string(const char *d) _pure_; + +const char *bond_lacp_rate_to_string(BondLacpRate d) _const_; +BondLacpRate bond_lacp_rate_from_string(const char *d) _pure_; + +const char *bond_fail_over_mac_to_string(BondFailOverMac d) _const_; +BondFailOverMac bond_fail_over_mac_from_string(const char *d) _pure_; + +const char *bond_ad_select_to_string(BondAdSelect d) _const_; +BondAdSelect bond_ad_select_from_string(const char *d) _pure_; + +const char *bond_arp_validate_to_string(BondArpValidate d) _const_; +BondArpValidate bond_arp_validate_from_string(const char *d) _pure_; + +const char *bond_arp_all_targets_to_string(BondArpAllTargets d) _const_; +BondArpAllTargets bond_arp_all_targets_from_string(const char *d) _pure_; + +const char *bond_primary_reselect_to_string(BondPrimaryReselect d) _const_; +BondPrimaryReselect bond_primary_reselect_from_string(const char *d) _pure_; + +CONFIG_PARSER_PROTOTYPE(config_parse_bond_mode); +CONFIG_PARSER_PROTOTYPE(config_parse_bond_xmit_hash_policy); +CONFIG_PARSER_PROTOTYPE(config_parse_bond_lacp_rate); +CONFIG_PARSER_PROTOTYPE(config_parse_bond_ad_select); +CONFIG_PARSER_PROTOTYPE(config_parse_bond_fail_over_mac); +CONFIG_PARSER_PROTOTYPE(config_parse_bond_arp_validate); +CONFIG_PARSER_PROTOTYPE(config_parse_bond_arp_all_targets); +CONFIG_PARSER_PROTOTYPE(config_parse_bond_primary_reselect); +CONFIG_PARSER_PROTOTYPE(config_parse_arp_ip_target_address); +CONFIG_PARSER_PROTOTYPE(config_parse_ad_actor_sys_prio); +CONFIG_PARSER_PROTOTYPE(config_parse_ad_user_port_key); +CONFIG_PARSER_PROTOTYPE(config_parse_ad_actor_system); diff --git a/src/network/netdev/bridge.c b/src/network/netdev/bridge.c new file mode 100644 index 0000000..aadb3ab --- /dev/null +++ b/src/network/netdev/bridge.c @@ -0,0 +1,163 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ + +#include <net/if.h> + +#include "missing.h" +#include "netlink-util.h" +#include "netdev/bridge.h" +#include "networkd-manager.h" +#include "vlan-util.h" + +/* callback for brige netdev's parameter set */ +static int netdev_bridge_set_handler(sd_netlink *rtnl, sd_netlink_message *m, NetDev *netdev) { + int r; + + assert(netdev); + assert(m); + + r = sd_netlink_message_get_errno(m); + if (r < 0) { + log_netdev_warning_errno(netdev, r, "Bridge parameters could not be set: %m"); + return 1; + } + + log_netdev_debug(netdev, "Bridge parameters set success"); + + return 1; +} + +static int netdev_bridge_post_create(NetDev *netdev, Link *link, sd_netlink_message *m) { + _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *req = NULL; + Bridge *b; + int r; + + assert(netdev); + + b = BRIDGE(netdev); + + assert(b); + + r = sd_rtnl_message_new_link(netdev->manager->rtnl, &req, RTM_NEWLINK, netdev->ifindex); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not allocate RTM_SETLINK message: %m"); + + r = sd_netlink_message_set_flags(req, NLM_F_REQUEST | NLM_F_ACK); + if (r < 0) + return log_link_error_errno(link, r, "Could not set netlink flags: %m"); + + r = sd_netlink_message_open_container(req, IFLA_LINKINFO); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_LINKINFO attribute: %m"); + + r = sd_netlink_message_open_container_union(req, IFLA_INFO_DATA, netdev_kind_to_string(netdev->kind)); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_INFO_DATA attribute: %m"); + + /* convert to jiffes */ + if (b->forward_delay != USEC_INFINITY) { + r = sd_netlink_message_append_u32(req, IFLA_BR_FORWARD_DELAY, usec_to_jiffies(b->forward_delay)); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BR_FORWARD_DELAY attribute: %m"); + } + + if (b->hello_time > 0) { + r = sd_netlink_message_append_u32(req, IFLA_BR_HELLO_TIME, usec_to_jiffies(b->hello_time)); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BR_HELLO_TIME attribute: %m"); + } + + if (b->max_age > 0) { + r = sd_netlink_message_append_u32(req, IFLA_BR_MAX_AGE, usec_to_jiffies(b->max_age)); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BR_MAX_AGE attribute: %m"); + } + + if (b->ageing_time != USEC_INFINITY) { + r = sd_netlink_message_append_u32(req, IFLA_BR_AGEING_TIME, usec_to_jiffies(b->ageing_time)); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BR_AGEING_TIME attribute: %m"); + } + + if (b->priority > 0) { + r = sd_netlink_message_append_u16(req, IFLA_BR_PRIORITY, b->priority); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BR_PRIORITY attribute: %m"); + } + + if (b->group_fwd_mask > 0) { + r = sd_netlink_message_append_u16(req, IFLA_BR_GROUP_FWD_MASK, b->group_fwd_mask); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BR_GROUP_FWD_MASK attribute: %m"); + } + + if (b->default_pvid != VLANID_INVALID) { + r = sd_netlink_message_append_u16(req, IFLA_BR_VLAN_DEFAULT_PVID, b->default_pvid); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BR_VLAN_DEFAULT_PVID attribute: %m"); + } + + if (b->mcast_querier >= 0) { + r = sd_netlink_message_append_u8(req, IFLA_BR_MCAST_QUERIER, b->mcast_querier); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BR_MCAST_QUERIER attribute: %m"); + } + + if (b->mcast_snooping >= 0) { + r = sd_netlink_message_append_u8(req, IFLA_BR_MCAST_SNOOPING, b->mcast_snooping); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BR_MCAST_SNOOPING attribute: %m"); + } + + if (b->vlan_filtering >= 0) { + r = sd_netlink_message_append_u8(req, IFLA_BR_VLAN_FILTERING, b->vlan_filtering); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BR_VLAN_FILTERING attribute: %m"); + } + + if (b->stp >= 0) { + r = sd_netlink_message_append_u32(req, IFLA_BR_STP_STATE, b->stp); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_BR_STP_STATE attribute: %m"); + } + + r = sd_netlink_message_close_container(req); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_LINKINFO attribute: %m"); + + r = sd_netlink_message_close_container(req); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_INFO_DATA attribute: %m"); + + r = netlink_call_async(netdev->manager->rtnl, NULL, req, netdev_bridge_set_handler, + netdev_destroy_callback, netdev); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not send rtnetlink message: %m"); + + netdev_ref(netdev); + + return r; +} + +static void bridge_init(NetDev *n) { + Bridge *b; + + b = BRIDGE(n); + + assert(b); + + b->mcast_querier = -1; + b->mcast_snooping = -1; + b->vlan_filtering = -1; + b->stp = -1; + b->default_pvid = VLANID_INVALID; + b->forward_delay = USEC_INFINITY; + b->ageing_time = USEC_INFINITY; +} + +const NetDevVTable bridge_vtable = { + .object_size = sizeof(Bridge), + .init = bridge_init, + .sections = "Match\0NetDev\0Bridge\0", + .post_create = netdev_bridge_post_create, + .create_type = NETDEV_CREATE_MASTER, +}; diff --git a/src/network/netdev/bridge.h b/src/network/netdev/bridge.h new file mode 100644 index 0000000..3edc93a --- /dev/null +++ b/src/network/netdev/bridge.h @@ -0,0 +1,24 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ +#pragma once + +#include "netdev/netdev.h" + +typedef struct Bridge { + NetDev meta; + + int mcast_querier; + int mcast_snooping; + int vlan_filtering; + int stp; + uint16_t priority; + uint16_t group_fwd_mask; + uint16_t default_pvid; + + usec_t forward_delay; + usec_t hello_time; + usec_t max_age; + usec_t ageing_time; +} Bridge; + +DEFINE_NETDEV_CAST(BRIDGE, Bridge); +extern const NetDevVTable bridge_vtable; diff --git a/src/network/netdev/dummy.c b/src/network/netdev/dummy.c new file mode 100644 index 0000000..aded1c5 --- /dev/null +++ b/src/network/netdev/dummy.c @@ -0,0 +1,9 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ + +#include "netdev/dummy.h" + +const NetDevVTable dummy_vtable = { + .object_size = sizeof(Dummy), + .sections = "Match\0NetDev\0", + .create_type = NETDEV_CREATE_INDEPENDENT, +}; diff --git a/src/network/netdev/dummy.h b/src/network/netdev/dummy.h new file mode 100644 index 0000000..93e0651 --- /dev/null +++ b/src/network/netdev/dummy.h @@ -0,0 +1,11 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ +#pragma once + +#include "netdev/netdev.h" + +typedef struct Dummy { + NetDev meta; +} Dummy; + +DEFINE_NETDEV_CAST(DUMMY, Dummy); +extern const NetDevVTable dummy_vtable; diff --git a/src/network/netdev/fou-tunnel.c b/src/network/netdev/fou-tunnel.c new file mode 100644 index 0000000..65dad38 --- /dev/null +++ b/src/network/netdev/fou-tunnel.c @@ -0,0 +1,129 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ + +#include <arpa/inet.h> +#include <net/if.h> +#include <linux/ip.h> + +#include "conf-parser.h" +#include "missing.h" +#include "netdev/fou-tunnel.h" +#include "networkd-link.h" +#include "networkd-manager.h" +#include "parse-util.h" +#include "sd-netlink.h" +#include "string-table.h" +#include "string-util.h" +#include "util.h" + +static const char* const fou_encap_type_table[_NETDEV_FOO_OVER_UDP_ENCAP_MAX] = { + [NETDEV_FOO_OVER_UDP_ENCAP_DIRECT] = "FooOverUDP", + [NETDEV_FOO_OVER_UDP_ENCAP_GUE] = "GenericUDPEncapsulation", +}; + +DEFINE_STRING_TABLE_LOOKUP(fou_encap_type, FooOverUDPEncapType); +DEFINE_CONFIG_PARSE_ENUM(config_parse_fou_encap_type, fou_encap_type, FooOverUDPEncapType, + "Failed to parse Encapsulation="); + +static int netdev_fill_fou_tunnel_message(NetDev *netdev, sd_netlink_message **ret) { + _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *m = NULL; + FouTunnel *t; + int r; + + assert(netdev); + + t = FOU(netdev); + + assert(t); + + r = sd_genl_message_new(netdev->manager->genl, SD_GENL_FOU, FOU_CMD_ADD, &m); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Failed to allocate generic netlink message: %m"); + + r = sd_netlink_message_append_u16(m, FOU_ATTR_PORT, htobe16(t->port)); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append FOU_ATTR_PORT attribute: %m"); + + r = sd_netlink_message_append_u8(m, FOU_ATTR_TYPE, FOU_ENCAP_GUE); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append FOU_ATTR_TYPE attribute: %m"); + + r = sd_netlink_message_append_u8(m, FOU_ATTR_AF, AF_INET); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append FOU_ATTR_AF attribute: %m"); + + r = sd_netlink_message_append_u8(m, FOU_ATTR_IPPROTO, t->fou_protocol); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append FOU_ATTR_IPPROTO attribute: %m"); + + *ret = m; + m = NULL; + + return 0; +} + +static int netdev_fou_tunnel_create(NetDev *netdev) { + _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *m = NULL; + uint32_t serial; + FouTunnel *t; + int r; + + assert(netdev); + + t = FOU(netdev); + + assert(t); + + r = netdev_fill_fou_tunnel_message(netdev, &m); + if (r < 0) + return r; + + r = sd_netlink_send(netdev->manager->genl, m, &serial); + if (r < 0 && r != -EADDRINUSE) + return log_netdev_error_errno(netdev, r, "Failed to add FooOverUDP tunnel: %m"); + + return 0; +} + +static int netdev_fou_tunnel_verify(NetDev *netdev, const char *filename) { + FouTunnel *t; + + assert(netdev); + assert(filename); + + t = FOU(netdev); + + assert(t); + + if (t->fou_encap_type == NETDEV_FOO_OVER_UDP_ENCAP_DIRECT && t->fou_protocol <= 0) { + log_netdev_error(netdev, "FooOverUDP protocol not configured in %s. Rejecting configuration.", filename); + return -EINVAL; + } + + if (t->fou_encap_type == NETDEV_FOO_OVER_UDP_ENCAP_GUE && t->fou_protocol > 0) { + log_netdev_error(netdev, "FooOverUDP GUE can't be set with protocol configured in %s. Rejecting configuration.", filename); + return -EINVAL; + } + + return 0; +} + +static void fou_tunnel_init(NetDev *netdev) { + FouTunnel *t; + + assert(netdev); + + t = FOU(netdev); + + assert(t); + + t->fou_encap_type = NETDEV_FOO_OVER_UDP_ENCAP_DIRECT; +} + +const NetDevVTable foutnl_vtable = { + .object_size = sizeof(FouTunnel), + .init = fou_tunnel_init, + .sections = "Match\0NetDev\0FooOverUDP\0", + .create = netdev_fou_tunnel_create, + .create_type = NETDEV_CREATE_INDEPENDENT, + .config_verify = netdev_fou_tunnel_verify, +}; diff --git a/src/network/netdev/fou-tunnel.h b/src/network/netdev/fou-tunnel.h new file mode 100644 index 0000000..b8abed1 --- /dev/null +++ b/src/network/netdev/fou-tunnel.h @@ -0,0 +1,36 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ +#pragma once + +#if HAVE_LINUX_FOU_H +#include <linux/fou.h> +#endif + +#include "in-addr-util.h" +#include "missing_fou.h" +#include "netdev/netdev.h" + +typedef enum FooOverUDPEncapType { + NETDEV_FOO_OVER_UDP_ENCAP_UNSPEC = FOU_ENCAP_UNSPEC, + NETDEV_FOO_OVER_UDP_ENCAP_DIRECT = FOU_ENCAP_DIRECT, + NETDEV_FOO_OVER_UDP_ENCAP_GUE = FOU_ENCAP_GUE, + _NETDEV_FOO_OVER_UDP_ENCAP_MAX, + _NETDEV_FOO_OVER_UDP_ENCAP_INVALID = -1, +} FooOverUDPEncapType; + +typedef struct FouTunnel { + NetDev meta; + + uint8_t fou_protocol; + + uint16_t port; + + FooOverUDPEncapType fou_encap_type; +} FouTunnel; + +DEFINE_NETDEV_CAST(FOU, FouTunnel); +extern const NetDevVTable foutnl_vtable; + +const char *fou_encap_type_to_string(FooOverUDPEncapType d) _const_; +FooOverUDPEncapType fou_encap_type_from_string(const char *d) _pure_; + +CONFIG_PARSER_PROTOTYPE(config_parse_fou_encap_type); diff --git a/src/network/netdev/geneve.c b/src/network/netdev/geneve.c new file mode 100644 index 0000000..089bbfe --- /dev/null +++ b/src/network/netdev/geneve.c @@ -0,0 +1,299 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ + +#include <net/if.h> + +#include "sd-netlink.h" + +#include "alloc-util.h" +#include "conf-parser.h" +#include "extract-word.h" +#include "geneve.h" +#include "netlink-util.h" +#include "parse-util.h" +#include "string-util.h" +#include "strv.h" +#include "missing.h" +#include "networkd-manager.h" + +#define GENEVE_FLOW_LABEL_MAX_MASK 0xFFFFFU +#define DEFAULT_GENEVE_DESTINATION_PORT 6081 + +/* callback for geneve netdev's created without a backing Link */ +static int geneve_netdev_create_handler(sd_netlink *rtnl, sd_netlink_message *m, NetDev *netdev) { + int r; + + assert(netdev); + assert(netdev->state != _NETDEV_STATE_INVALID); + + r = sd_netlink_message_get_errno(m); + if (r == -EEXIST) + log_netdev_info(netdev, "Geneve netdev exists, using existing without changing its parameters"); + else if (r < 0) { + log_netdev_warning_errno(netdev, r, "Geneve netdev could not be created: %m"); + netdev_drop(netdev); + + return 1; + } + + log_netdev_debug(netdev, "Geneve created"); + + return 1; +} + +static int netdev_geneve_create(NetDev *netdev) { + _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *m = NULL; + Geneve *v; + int r; + + assert(netdev); + + v = GENEVE(netdev); + + r = sd_rtnl_message_new_link(netdev->manager->rtnl, &m, RTM_NEWLINK, 0); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not allocate RTM_NEWLINK message: %m"); + + r = sd_netlink_message_append_string(m, IFLA_IFNAME, netdev->ifname); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IFNAME, attribute: %m"); + + if (netdev->mac) { + r = sd_netlink_message_append_ether_addr(m, IFLA_ADDRESS, netdev->mac); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_ADDRESS attribute: %m"); + } + + if (netdev->mtu != 0) { + r = sd_netlink_message_append_u32(m, IFLA_MTU, netdev->mtu); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_MTU attribute: %m"); + } + + r = sd_netlink_message_open_container(m, IFLA_LINKINFO); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_LINKINFO attribute: %m"); + + r = sd_netlink_message_open_container_union(m, IFLA_INFO_DATA, netdev_kind_to_string(netdev->kind)); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_INFO_DATA attribute: %m"); + + if (v->id <= GENEVE_VID_MAX) { + r = sd_netlink_message_append_u32(m, IFLA_GENEVE_ID, v->id); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_GENEVE_ID attribute: %m"); + } + + if (!in_addr_is_null(v->remote_family, &v->remote)) { + + if (v->remote_family == AF_INET) + r = sd_netlink_message_append_in_addr(m, IFLA_GENEVE_REMOTE, &v->remote.in); + else + r = sd_netlink_message_append_in6_addr(m, IFLA_GENEVE_REMOTE6, &v->remote.in6); + + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_GENEVE_GROUP attribute: %m"); + + } + + if (v->ttl) { + r = sd_netlink_message_append_u8(m, IFLA_GENEVE_TTL, v->ttl); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_GENEVE_TTL attribute: %m"); + } + + r = sd_netlink_message_append_u8(m, IFLA_GENEVE_TOS, v->tos); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_GENEVE_TOS attribute: %m"); + + r = sd_netlink_message_append_u8(m, IFLA_GENEVE_UDP_CSUM, v->udpcsum); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_GENEVE_UDP_CSUM attribute: %m"); + + r = sd_netlink_message_append_u8(m, IFLA_GENEVE_UDP_ZERO_CSUM6_TX, v->udp6zerocsumtx); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_GENEVE_UDP_ZERO_CSUM6_TX attribute: %m"); + + r = sd_netlink_message_append_u8(m, IFLA_GENEVE_UDP_ZERO_CSUM6_RX, v->udp6zerocsumrx); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_GENEVE_UDP_ZERO_CSUM6_RX attribute: %m"); + + if (v->dest_port != DEFAULT_GENEVE_DESTINATION_PORT) { + r = sd_netlink_message_append_u16(m, IFLA_GENEVE_PORT, htobe16(v->dest_port)); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_GENEVE_PORT attribute: %m"); + } + + if (v->flow_label > 0) { + r = sd_netlink_message_append_u32(m, IFLA_GENEVE_LABEL, htobe32(v->flow_label)); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_GENEVE_LABEL attribute: %m"); + } + + r = sd_netlink_message_close_container(m); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_INFO_DATA attribute: %m"); + + r = sd_netlink_message_close_container(m); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_LINKINFO attribute: %m"); + + r = netlink_call_async(netdev->manager->rtnl, NULL, m, geneve_netdev_create_handler, + netdev_destroy_callback, netdev); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not send rtnetlink message: %m"); + + netdev_ref(netdev); + netdev->state = NETDEV_STATE_CREATING; + + log_netdev_debug(netdev, "Creating"); + + return r; +} + +int config_parse_geneve_vni(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + Geneve *v = userdata; + uint32_t f; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + r = safe_atou32(rvalue, &f); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse Geneve VNI '%s'.", rvalue); + return 0; + } + + if (f > GENEVE_VID_MAX){ + log_syntax(unit, LOG_ERR, filename, line, r, "Geneve VNI out is of range '%s'.", rvalue); + return 0; + } + + v->id = f; + + return 0; +} + +int config_parse_geneve_address(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + Geneve *v = userdata; + union in_addr_union *addr = data, buffer; + int r, f; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + r = in_addr_from_string_auto(rvalue, &f, &buffer); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "geneve '%s' address is invalid, ignoring assignment: %s", lvalue, rvalue); + return 0; + } + + r = in_addr_is_multicast(f, &buffer); + if (r > 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "geneve invalid multicast '%s' address, ignoring assignment: %s", lvalue, rvalue); + return 0; + } + + v->remote_family = f; + *addr = buffer; + + return 0; +} + +int config_parse_geneve_flow_label(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + Geneve *v = userdata; + uint32_t f; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + r = safe_atou32(rvalue, &f); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse Geneve flow label '%s'.", rvalue); + return 0; + } + + if (f & ~GENEVE_FLOW_LABEL_MAX_MASK) { + log_syntax(unit, LOG_ERR, filename, line, r, + "Geneve flow label '%s' not valid. Flow label range should be [0-1048575].", rvalue); + return 0; + } + + v->flow_label = f; + + return 0; +} + +static int netdev_geneve_verify(NetDev *netdev, const char *filename) { + Geneve *v = GENEVE(netdev); + + assert(netdev); + assert(v); + assert(filename); + + if (v->ttl == 0) { + log_warning("Invalid Geneve TTL value '0' configured in '%s'. Ignoring", filename); + return -EINVAL; + } + + return 0; +} + +static void geneve_init(NetDev *netdev) { + Geneve *v; + + assert(netdev); + + v = GENEVE(netdev); + + assert(v); + + v->id = GENEVE_VID_MAX + 1; + v->dest_port = DEFAULT_GENEVE_DESTINATION_PORT; + v->udpcsum = false; + v->udp6zerocsumtx = false; + v->udp6zerocsumrx = false; +} + +const NetDevVTable geneve_vtable = { + .object_size = sizeof(Geneve), + .init = geneve_init, + .sections = "Match\0NetDev\0GENEVE\0", + .create = netdev_geneve_create, + .create_type = NETDEV_CREATE_INDEPENDENT, + .config_verify = netdev_geneve_verify, +}; diff --git a/src/network/netdev/geneve.h b/src/network/netdev/geneve.h new file mode 100644 index 0000000..c201981 --- /dev/null +++ b/src/network/netdev/geneve.h @@ -0,0 +1,38 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ +#pragma once + +typedef struct Geneve Geneve; + +#include "in-addr-util.h" +#include "netdev.h" +#include "networkd-link.h" +#include "networkd-network.h" + +#define GENEVE_VID_MAX (1u << 24) - 1 + +struct Geneve { + NetDev meta; + + uint32_t id; + uint32_t flow_label; + + int remote_family; + + uint8_t tos; + uint8_t ttl; + + uint16_t dest_port; + + bool udpcsum; + bool udp6zerocsumtx; + bool udp6zerocsumrx; + + union in_addr_union remote; +}; + +DEFINE_NETDEV_CAST(GENEVE, Geneve); +extern const NetDevVTable geneve_vtable; + +CONFIG_PARSER_PROTOTYPE(config_parse_geneve_vni); +CONFIG_PARSER_PROTOTYPE(config_parse_geneve_address); +CONFIG_PARSER_PROTOTYPE(config_parse_geneve_flow_label); diff --git a/src/network/netdev/ipvlan.c b/src/network/netdev/ipvlan.c new file mode 100644 index 0000000..5bb6a5b --- /dev/null +++ b/src/network/netdev/ipvlan.c @@ -0,0 +1,73 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ + +#include <net/if.h> + +#include "conf-parser.h" +#include "netdev/ipvlan.h" +#include "string-table.h" + +static const char* const ipvlan_mode_table[_NETDEV_IPVLAN_MODE_MAX] = { + [NETDEV_IPVLAN_MODE_L2] = "L2", + [NETDEV_IPVLAN_MODE_L3] = "L3", + [NETDEV_IPVLAN_MODE_L3S] = "L3S", +}; + +DEFINE_STRING_TABLE_LOOKUP(ipvlan_mode, IPVlanMode); +DEFINE_CONFIG_PARSE_ENUM(config_parse_ipvlan_mode, ipvlan_mode, IPVlanMode, "Failed to parse ipvlan mode"); + +static const char* const ipvlan_flags_table[_NETDEV_IPVLAN_FLAGS_MAX] = { + [NETDEV_IPVLAN_FLAGS_BRIGDE] = "bridge", + [NETDEV_IPVLAN_FLAGS_PRIVATE] = "private", + [NETDEV_IPVLAN_FLAGS_VEPA] = "vepa", +}; + +DEFINE_STRING_TABLE_LOOKUP(ipvlan_flags, IPVlanFlags); +DEFINE_CONFIG_PARSE_ENUM(config_parse_ipvlan_flags, ipvlan_flags, IPVlanFlags, "Failed to parse ipvlan flags"); + +static int netdev_ipvlan_fill_message_create(NetDev *netdev, Link *link, sd_netlink_message *req) { + IPVlan *m; + int r; + + assert(netdev); + assert(link); + assert(netdev->ifname); + + m = IPVLAN(netdev); + + assert(m); + + if (m->mode != _NETDEV_IPVLAN_MODE_INVALID) { + r = sd_netlink_message_append_u16(req, IFLA_IPVLAN_MODE, m->mode); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPVLAN_MODE attribute: %m"); + } + + if (m->flags != _NETDEV_IPVLAN_FLAGS_INVALID) { + r = sd_netlink_message_append_u16(req, IFLA_IPVLAN_FLAGS, m->flags); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPVLAN_FLAGS attribute: %m"); + } + + return 0; +} + +static void ipvlan_init(NetDev *n) { + IPVlan *m; + + assert(n); + + m = IPVLAN(n); + + assert(m); + + m->mode = _NETDEV_IPVLAN_MODE_INVALID; + m->flags = _NETDEV_IPVLAN_FLAGS_INVALID; +} + +const NetDevVTable ipvlan_vtable = { + .object_size = sizeof(IPVlan), + .init = ipvlan_init, + .sections = "Match\0NetDev\0IPVLAN\0", + .fill_message_create = netdev_ipvlan_fill_message_create, + .create_type = NETDEV_CREATE_STACKED, +}; diff --git a/src/network/netdev/ipvlan.h b/src/network/netdev/ipvlan.h new file mode 100644 index 0000000..fb426d3 --- /dev/null +++ b/src/network/netdev/ipvlan.h @@ -0,0 +1,42 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ +#pragma once + +#include <linux/if_link.h> + +#include "missing_if_link.h" +#include "netdev/netdev.h" + +typedef enum IPVlanMode { + NETDEV_IPVLAN_MODE_L2 = IPVLAN_MODE_L2, + NETDEV_IPVLAN_MODE_L3 = IPVLAN_MODE_L3, + NETDEV_IPVLAN_MODE_L3S = IPVLAN_MODE_L3S, + _NETDEV_IPVLAN_MODE_MAX, + _NETDEV_IPVLAN_MODE_INVALID = -1 +} IPVlanMode; + +typedef enum IPVlanFlags { + NETDEV_IPVLAN_FLAGS_BRIGDE, + NETDEV_IPVLAN_FLAGS_PRIVATE = IPVLAN_F_PRIVATE, + NETDEV_IPVLAN_FLAGS_VEPA = IPVLAN_F_VEPA, + _NETDEV_IPVLAN_FLAGS_MAX, + _NETDEV_IPVLAN_FLAGS_INVALID = -1 +} IPVlanFlags; + +typedef struct IPVlan { + NetDev meta; + + IPVlanMode mode; + IPVlanFlags flags; +} IPVlan; + +DEFINE_NETDEV_CAST(IPVLAN, IPVlan); +extern const NetDevVTable ipvlan_vtable; + +const char *ipvlan_mode_to_string(IPVlanMode d) _const_; +IPVlanMode ipvlan_mode_from_string(const char *d) _pure_; + +const char *ipvlan_flags_to_string(IPVlanFlags d) _const_; +IPVlanFlags ipvlan_flags_from_string(const char *d) _pure_; + +CONFIG_PARSER_PROTOTYPE(config_parse_ipvlan_mode); +CONFIG_PARSER_PROTOTYPE(config_parse_ipvlan_flags); diff --git a/src/network/netdev/macvlan.c b/src/network/netdev/macvlan.c new file mode 100644 index 0000000..871f020 --- /dev/null +++ b/src/network/netdev/macvlan.c @@ -0,0 +1,72 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ + +#include <net/if.h> + +#include "conf-parser.h" +#include "netdev/macvlan.h" +#include "string-table.h" + +static const char* const macvlan_mode_table[_NETDEV_MACVLAN_MODE_MAX] = { + [NETDEV_MACVLAN_MODE_PRIVATE] = "private", + [NETDEV_MACVLAN_MODE_VEPA] = "vepa", + [NETDEV_MACVLAN_MODE_BRIDGE] = "bridge", + [NETDEV_MACVLAN_MODE_PASSTHRU] = "passthru", +}; + +DEFINE_STRING_TABLE_LOOKUP(macvlan_mode, MacVlanMode); +DEFINE_CONFIG_PARSE_ENUM(config_parse_macvlan_mode, macvlan_mode, MacVlanMode, "Failed to parse macvlan mode"); + +static int netdev_macvlan_fill_message_create(NetDev *netdev, Link *link, sd_netlink_message *req) { + MacVlan *m; + int r; + + assert(netdev); + assert(link); + assert(netdev->ifname); + + if (netdev->kind == NETDEV_KIND_MACVLAN) + m = MACVLAN(netdev); + else + m = MACVTAP(netdev); + + assert(m); + + if (m->mode != _NETDEV_MACVLAN_MODE_INVALID) { + r = sd_netlink_message_append_u32(req, IFLA_MACVLAN_MODE, m->mode); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_MACVLAN_MODE attribute: %m"); + } + + return 0; +} + +static void macvlan_init(NetDev *n) { + MacVlan *m; + + assert(n); + + if (n->kind == NETDEV_KIND_MACVLAN) + m = MACVLAN(n); + else + m = MACVTAP(n); + + assert(m); + + m->mode = _NETDEV_MACVLAN_MODE_INVALID; +} + +const NetDevVTable macvtap_vtable = { + .object_size = sizeof(MacVlan), + .init = macvlan_init, + .sections = "Match\0NetDev\0MACVTAP\0", + .fill_message_create = netdev_macvlan_fill_message_create, + .create_type = NETDEV_CREATE_STACKED, +}; + +const NetDevVTable macvlan_vtable = { + .object_size = sizeof(MacVlan), + .init = macvlan_init, + .sections = "Match\0NetDev\0MACVLAN\0", + .fill_message_create = netdev_macvlan_fill_message_create, + .create_type = NETDEV_CREATE_STACKED, +}; diff --git a/src/network/netdev/macvlan.h b/src/network/netdev/macvlan.h new file mode 100644 index 0000000..b473f1e --- /dev/null +++ b/src/network/netdev/macvlan.h @@ -0,0 +1,31 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ +#pragma once + +typedef struct MacVlan MacVlan; + +#include "netdev/netdev.h" + +typedef enum MacVlanMode { + NETDEV_MACVLAN_MODE_PRIVATE = MACVLAN_MODE_PRIVATE, + NETDEV_MACVLAN_MODE_VEPA = MACVLAN_MODE_VEPA, + NETDEV_MACVLAN_MODE_BRIDGE = MACVLAN_MODE_BRIDGE, + NETDEV_MACVLAN_MODE_PASSTHRU = MACVLAN_MODE_PASSTHRU, + _NETDEV_MACVLAN_MODE_MAX, + _NETDEV_MACVLAN_MODE_INVALID = -1 +} MacVlanMode; + +struct MacVlan { + NetDev meta; + + MacVlanMode mode; +}; + +DEFINE_NETDEV_CAST(MACVLAN, MacVlan); +DEFINE_NETDEV_CAST(MACVTAP, MacVlan); +extern const NetDevVTable macvlan_vtable; +extern const NetDevVTable macvtap_vtable; + +const char *macvlan_mode_to_string(MacVlanMode d) _const_; +MacVlanMode macvlan_mode_from_string(const char *d) _pure_; + +CONFIG_PARSER_PROTOTYPE(config_parse_macvlan_mode); diff --git a/src/network/netdev/netdev-gperf.gperf b/src/network/netdev/netdev-gperf.gperf new file mode 100644 index 0000000..f7ca98f --- /dev/null +++ b/src/network/netdev/netdev-gperf.gperf @@ -0,0 +1,174 @@ +%{ +#if __GNUC__ >= 7 +_Pragma("GCC diagnostic ignored \"-Wimplicit-fallthrough\"") +#endif +#include <stddef.h> +#include "conf-parser.h" +#include "network-internal.h" +#include "netdev/bond.h" +#include "netdev/bridge.h" +#include "netdev/geneve.h" +#include "netdev/ipvlan.h" +#include "netdev/macvlan.h" +#include "netdev/tunnel.h" +#include "netdev/tuntap.h" +#include "netdev/veth.h" +#include "netdev/vlan.h" +#include "netdev/vxlan.h" +#include "netdev/vrf.h" +#include "netdev/netdev.h" +#include "netdev/vxcan.h" +#include "netdev/wireguard.h" +#include "netdev/fou-tunnel.h" +#include "vlan-util.h" +%} +struct ConfigPerfItem; +%null_strings +%language=ANSI-C +%define slot-name section_and_lvalue +%define hash-function-name network_netdev_gperf_hash +%define lookup-function-name network_netdev_gperf_lookup +%readonly-tables +%omit-struct-type +%struct-type +%includes +%% +Match.Host, config_parse_net_condition, CONDITION_HOST, offsetof(NetDev, match_host) +Match.Virtualization, config_parse_net_condition, CONDITION_VIRTUALIZATION, offsetof(NetDev, match_virt) +Match.KernelCommandLine, config_parse_net_condition, CONDITION_KERNEL_COMMAND_LINE, offsetof(NetDev, match_kernel_cmdline) +Match.KernelVersion, config_parse_net_condition, CONDITION_KERNEL_VERSION, offsetof(NetDev, match_kernel_version) +Match.Architecture, config_parse_net_condition, CONDITION_ARCHITECTURE, offsetof(NetDev, match_arch) +NetDev.Description, config_parse_string, 0, offsetof(NetDev, description) +NetDev.Name, config_parse_ifname, 0, offsetof(NetDev, ifname) +NetDev.Kind, config_parse_netdev_kind, 0, offsetof(NetDev, kind) +NetDev.MTUBytes, config_parse_mtu, AF_UNSPEC, offsetof(NetDev, mtu) +NetDev.MACAddress, config_parse_hwaddr, 0, offsetof(NetDev, mac) +VLAN.Id, config_parse_vlanid, 0, offsetof(VLan, id) +VLAN.GVRP, config_parse_tristate, 0, offsetof(VLan, gvrp) +VLAN.MVRP, config_parse_tristate, 0, offsetof(VLan, mvrp) +VLAN.LooseBinding, config_parse_tristate, 0, offsetof(VLan, loose_binding) +VLAN.ReorderHeader, config_parse_tristate, 0, offsetof(VLan, reorder_hdr) +MACVLAN.Mode, config_parse_macvlan_mode, 0, offsetof(MacVlan, mode) +MACVTAP.Mode, config_parse_macvlan_mode, 0, offsetof(MacVlan, mode) +IPVLAN.Mode, config_parse_ipvlan_mode, 0, offsetof(IPVlan, mode) +IPVLAN.Flags, config_parse_ipvlan_flags, 0, offsetof(IPVlan, flags) +Tunnel.Local, config_parse_tunnel_address, 0, offsetof(Tunnel, local) +Tunnel.Remote, config_parse_tunnel_address, 0, offsetof(Tunnel, remote) +Tunnel.TOS, config_parse_unsigned, 0, offsetof(Tunnel, tos) +Tunnel.TTL, config_parse_unsigned, 0, offsetof(Tunnel, ttl) +Tunnel.Key, config_parse_tunnel_key, 0, offsetof(Tunnel, key) +Tunnel.InputKey, config_parse_tunnel_key, 0, offsetof(Tunnel, ikey) +Tunnel.OutputKey, config_parse_tunnel_key, 0, offsetof(Tunnel, okey) +Tunnel.DiscoverPathMTU, config_parse_bool, 0, offsetof(Tunnel, pmtudisc) +Tunnel.Mode, config_parse_ip6tnl_mode, 0, offsetof(Tunnel, ip6tnl_mode) +Tunnel.IPv6FlowLabel, config_parse_ipv6_flowlabel, 0, offsetof(Tunnel, ipv6_flowlabel) +Tunnel.CopyDSCP, config_parse_bool, 0, offsetof(Tunnel, copy_dscp) +Tunnel.EncapsulationLimit, config_parse_encap_limit, 0, offsetof(Tunnel, encap_limit) +Tunnel.Independent, config_parse_bool, 0, offsetof(Tunnel, independent) +Tunnel.AllowLocalRemote, config_parse_tristate, 0, offsetof(Tunnel, allow_localremote) +Tunnel.FooOverUDP, config_parse_bool, 0, offsetof(Tunnel, fou_tunnel) +Tunnel.FOUDestinationPort, config_parse_ip_port, 0, offsetof(Tunnel, fou_destination_port) +Tunnel.FOUSourcePort, config_parse_ip_port, 0, offsetof(Tunnel, encap_src_port) +Tunnel.Encapsulation, config_parse_fou_encap_type, 0, offsetof(Tunnel, fou_encap_type) +Tunnel.IPv6RapidDeploymentPrefix, config_parse_6rd_prefix, 0, 0 +Tunnel.ERSPANIndex, config_parse_uint32, 0, offsetof(Tunnel, erspan_index) +Tunnel.SerializeTunneledPackets, config_parse_tristate, 0, offsetof(Tunnel, erspan_sequence) +Tunnel.ISATAP, config_parse_tristate, 0, offsetof(Tunnel, isatap) +FooOverUDP.Protocol, config_parse_uint8, 0, offsetof(FouTunnel, fou_protocol) +FooOverUDP.Encapsulation, config_parse_fou_encap_type, 0, offsetof(FouTunnel, fou_encap_type) +FooOverUDP.Port, config_parse_ip_port, 0, offsetof(FouTunnel, port) +Peer.Name, config_parse_ifname, 0, offsetof(Veth, ifname_peer) +Peer.MACAddress, config_parse_hwaddr, 0, offsetof(Veth, mac_peer) +VXCAN.Peer, config_parse_ifname, 0, offsetof(VxCan, ifname_peer) +VXLAN.Id, config_parse_uint64, 0, offsetof(VxLan, id) +VXLAN.Group, config_parse_vxlan_address, 0, offsetof(VxLan, remote) +VXLAN.Local, config_parse_vxlan_address, 0, offsetof(VxLan, local) +VXLAN.Remote, config_parse_vxlan_address, 0, offsetof(VxLan, remote) +VXLAN.TOS, config_parse_unsigned, 0, offsetof(VxLan, tos) +VXLAN.TTL, config_parse_unsigned, 0, offsetof(VxLan, ttl) +VXLAN.MacLearning, config_parse_bool, 0, offsetof(VxLan, learning) +VXLAN.ARPProxy, config_parse_bool, 0, offsetof(VxLan, arp_proxy) +VXLAN.ReduceARPProxy, config_parse_bool, 0, offsetof(VxLan, arp_proxy) +VXLAN.L2MissNotification, config_parse_bool, 0, offsetof(VxLan, l2miss) +VXLAN.L3MissNotification, config_parse_bool, 0, offsetof(VxLan, l3miss) +VXLAN.RouteShortCircuit, config_parse_bool, 0, offsetof(VxLan, route_short_circuit) +VXLAN.UDPCheckSum, config_parse_bool, 0, offsetof(VxLan, udpcsum) +VXLAN.UDPChecksum, config_parse_bool, 0, offsetof(VxLan, udpcsum) +VXLAN.UDP6ZeroCheckSumRx, config_parse_bool, 0, offsetof(VxLan, udp6zerocsumrx) +VXLAN.UDP6ZeroChecksumRx, config_parse_bool, 0, offsetof(VxLan, udp6zerocsumrx) +VXLAN.UDP6ZeroCheckSumTx, config_parse_bool, 0, offsetof(VxLan, udp6zerocsumtx) +VXLAN.UDP6ZeroChecksumTx, config_parse_bool, 0, offsetof(VxLan, udp6zerocsumtx) +VXLAN.RemoteChecksumTx, config_parse_bool, 0, offsetof(VxLan, remote_csum_tx) +VXLAN.RemoteChecksumRx, config_parse_bool, 0, offsetof(VxLan, remote_csum_rx) +VXLAN.FDBAgeingSec, config_parse_sec, 0, offsetof(VxLan, fdb_ageing) +VXLAN.GroupPolicyExtension, config_parse_bool, 0, offsetof(VxLan, group_policy) +VXLAN.MaximumFDBEntries, config_parse_unsigned, 0, offsetof(VxLan, max_fdb) +VXLAN.PortRange, config_parse_port_range, 0, 0 +VXLAN.DestinationPort, config_parse_ip_port, 0, offsetof(VxLan, dest_port) +VXLAN.FlowLabel, config_parse_flow_label, 0, 0 +GENEVE.Id, config_parse_geneve_vni, 0, offsetof(Geneve, id) +GENEVE.Remote, config_parse_geneve_address, 0, offsetof(Geneve, remote) +GENEVE.TOS, config_parse_uint8, 0, offsetof(Geneve, tos) +GENEVE.TTL, config_parse_uint8, 0, offsetof(Geneve, ttl) +GENEVE.UDPChecksum, config_parse_bool, 0, offsetof(Geneve, udpcsum) +GENEVE.UDP6ZeroCheckSumRx, config_parse_bool, 0, offsetof(Geneve, udp6zerocsumrx) +GENEVE.UDP6ZeroChecksumRx, config_parse_bool, 0, offsetof(Geneve, udp6zerocsumrx) +GENEVE.UDP6ZeroCheckSumTx, config_parse_bool, 0, offsetof(Geneve, udp6zerocsumtx) +GENEVE.UDP6ZeroChecksumTx, config_parse_bool, 0, offsetof(Geneve, udp6zerocsumtx) +GENEVE.DestinationPort, config_parse_ip_port, 0, offsetof(Geneve, dest_port) +GENEVE.FlowLabel, config_parse_geneve_flow_label, 0, 0 +Tun.OneQueue, config_parse_bool, 0, offsetof(TunTap, one_queue) +Tun.MultiQueue, config_parse_bool, 0, offsetof(TunTap, multi_queue) +Tun.PacketInfo, config_parse_bool, 0, offsetof(TunTap, packet_info) +Tun.User, config_parse_string, 0, offsetof(TunTap, user_name) +Tun.Group, config_parse_string, 0, offsetof(TunTap, group_name) +Tap.OneQueue, config_parse_bool, 0, offsetof(TunTap, one_queue) +Tap.MultiQueue, config_parse_bool, 0, offsetof(TunTap, multi_queue) +Tap.PacketInfo, config_parse_bool, 0, offsetof(TunTap, packet_info) +Tap.VNetHeader, config_parse_bool, 0, offsetof(TunTap, vnet_hdr) +Tap.User, config_parse_string, 0, offsetof(TunTap, user_name) +Tap.Group, config_parse_string, 0, offsetof(TunTap, group_name) +Bond.Mode, config_parse_bond_mode, 0, offsetof(Bond, mode) +Bond.TransmitHashPolicy, config_parse_bond_xmit_hash_policy, 0, offsetof(Bond, xmit_hash_policy) +Bond.LACPTransmitRate, config_parse_bond_lacp_rate, 0, offsetof(Bond, lacp_rate) +Bond.AdSelect, config_parse_bond_ad_select, 0, offsetof(Bond, ad_select) +Bond.FailOverMACPolicy, config_parse_bond_fail_over_mac, 0, offsetof(Bond, fail_over_mac) +Bond.ARPIPTargets, config_parse_arp_ip_target_address, 0, 0 +Bond.ARPValidate, config_parse_bond_arp_validate, 0, offsetof(Bond, arp_validate) +Bond.ARPAllTargets, config_parse_bond_arp_all_targets, 0, offsetof(Bond, arp_all_targets) +Bond.PrimaryReselectPolicy, config_parse_bond_primary_reselect, 0, offsetof(Bond, primary_reselect) +Bond.ResendIGMP, config_parse_unsigned, 0, offsetof(Bond, resend_igmp) +Bond.PacketsPerSlave, config_parse_unsigned, 0, offsetof(Bond, packets_per_slave) +Bond.GratuitousARP, config_parse_unsigned, 0, offsetof(Bond, num_grat_arp) +Bond.AllSlavesActive, config_parse_bool, 0, offsetof(Bond, all_slaves_active) +Bond.DynamicTransmitLoadBalancing, config_parse_tristate, 0, offsetof(Bond, tlb_dynamic_lb) +Bond.MinLinks, config_parse_unsigned, 0, offsetof(Bond, min_links) +Bond.MIIMonitorSec, config_parse_sec, 0, offsetof(Bond, miimon) +Bond.UpDelaySec, config_parse_sec, 0, offsetof(Bond, updelay) +Bond.DownDelaySec, config_parse_sec, 0, offsetof(Bond, downdelay) +Bond.ARPIntervalSec, config_parse_sec, 0, offsetof(Bond, arp_interval) +Bond.LearnPacketIntervalSec, config_parse_sec, 0, offsetof(Bond, lp_interval) +Bond.AdActorSystemPriority, config_parse_ad_actor_sys_prio, 0, offsetof(Bond, ad_actor_sys_prio) +Bond.AdUserPortKey, config_parse_ad_user_port_key, 0, offsetof(Bond, ad_user_port_key) +Bond.AdActorSystem, config_parse_ad_actor_system, 0, offsetof(Bond, ad_actor_system) +Bridge.HelloTimeSec, config_parse_sec, 0, offsetof(Bridge, hello_time) +Bridge.MaxAgeSec, config_parse_sec, 0, offsetof(Bridge, max_age) +Bridge.AgeingTimeSec, config_parse_sec, 0, offsetof(Bridge, ageing_time) +Bridge.ForwardDelaySec, config_parse_sec, 0, offsetof(Bridge, forward_delay) +Bridge.Priority, config_parse_uint16, 0, offsetof(Bridge, priority) +Bridge.GroupForwardMask, config_parse_uint16, 0, offsetof(Bridge, group_fwd_mask) +Bridge.DefaultPVID, config_parse_default_port_vlanid, 0, offsetof(Bridge, default_pvid) +Bridge.MulticastQuerier, config_parse_tristate, 0, offsetof(Bridge, mcast_querier) +Bridge.MulticastSnooping, config_parse_tristate, 0, offsetof(Bridge, mcast_snooping) +Bridge.VLANFiltering, config_parse_tristate, 0, offsetof(Bridge, vlan_filtering) +Bridge.STP, config_parse_tristate, 0, offsetof(Bridge, stp) +VRF.TableId, config_parse_uint32, 0, offsetof(Vrf, table) /* deprecated */ +VRF.Table, config_parse_uint32, 0, offsetof(Vrf, table) +WireGuard.FwMark, config_parse_unsigned, 0, offsetof(Wireguard, fwmark) +WireGuard.ListenPort, config_parse_wireguard_listen_port, 0, offsetof(Wireguard, port) +WireGuard.PrivateKey, config_parse_wireguard_private_key, 0, 0 +WireGuardPeer.AllowedIPs, config_parse_wireguard_allowed_ips, 0, 0 +WireGuardPeer.Endpoint, config_parse_wireguard_endpoint, 0, 0 +WireGuardPeer.PublicKey, config_parse_wireguard_public_key, 0, 0 +WireGuardPeer.PresharedKey, config_parse_wireguard_preshared_key, 0, 0 +WireGuardPeer.PersistentKeepalive, config_parse_wireguard_keepalive, 0, 0 diff --git a/src/network/netdev/netdev.c b/src/network/netdev/netdev.c new file mode 100644 index 0000000..0263917 --- /dev/null +++ b/src/network/netdev/netdev.c @@ -0,0 +1,816 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ + +#include <net/if.h> + +#include "alloc-util.h" +#include "conf-files.h" +#include "conf-parser.h" +#include "fd-util.h" +#include "list.h" +#include "netlink-util.h" +#include "network-internal.h" +#include "netdev/netdev.h" +#include "networkd-manager.h" +#include "networkd-link.h" +#include "siphash24.h" +#include "stat-util.h" +#include "string-table.h" +#include "string-util.h" +#include "strv.h" + +#include "netdev/bridge.h" +#include "netdev/bond.h" +#include "netdev/geneve.h" +#include "netdev/vlan.h" +#include "netdev/macvlan.h" +#include "netdev/ipvlan.h" +#include "netdev/vxlan.h" +#include "netdev/tunnel.h" +#include "netdev/tuntap.h" +#include "netdev/veth.h" +#include "netdev/dummy.h" +#include "netdev/vrf.h" +#include "netdev/vcan.h" +#include "netdev/vxcan.h" +#include "netdev/wireguard.h" +#include "netdev/netdevsim.h" +#include "netdev/fou-tunnel.h" + +const NetDevVTable * const netdev_vtable[_NETDEV_KIND_MAX] = { + [NETDEV_KIND_BRIDGE] = &bridge_vtable, + [NETDEV_KIND_BOND] = &bond_vtable, + [NETDEV_KIND_VLAN] = &vlan_vtable, + [NETDEV_KIND_MACVLAN] = &macvlan_vtable, + [NETDEV_KIND_MACVTAP] = &macvtap_vtable, + [NETDEV_KIND_IPVLAN] = &ipvlan_vtable, + [NETDEV_KIND_VXLAN] = &vxlan_vtable, + [NETDEV_KIND_IPIP] = &ipip_vtable, + [NETDEV_KIND_GRE] = &gre_vtable, + [NETDEV_KIND_GRETAP] = &gretap_vtable, + [NETDEV_KIND_IP6GRE] = &ip6gre_vtable, + [NETDEV_KIND_IP6GRETAP] = &ip6gretap_vtable, + [NETDEV_KIND_SIT] = &sit_vtable, + [NETDEV_KIND_VTI] = &vti_vtable, + [NETDEV_KIND_VTI6] = &vti6_vtable, + [NETDEV_KIND_VETH] = &veth_vtable, + [NETDEV_KIND_DUMMY] = &dummy_vtable, + [NETDEV_KIND_TUN] = &tun_vtable, + [NETDEV_KIND_TAP] = &tap_vtable, + [NETDEV_KIND_IP6TNL] = &ip6tnl_vtable, + [NETDEV_KIND_VRF] = &vrf_vtable, + [NETDEV_KIND_VCAN] = &vcan_vtable, + [NETDEV_KIND_GENEVE] = &geneve_vtable, + [NETDEV_KIND_VXCAN] = &vxcan_vtable, + [NETDEV_KIND_WIREGUARD] = &wireguard_vtable, + [NETDEV_KIND_NETDEVSIM] = &netdevsim_vtable, + [NETDEV_KIND_FOU] = &foutnl_vtable, + [NETDEV_KIND_ERSPAN] = &erspan_vtable, +}; + +static const char* const netdev_kind_table[_NETDEV_KIND_MAX] = { + [NETDEV_KIND_BRIDGE] = "bridge", + [NETDEV_KIND_BOND] = "bond", + [NETDEV_KIND_VLAN] = "vlan", + [NETDEV_KIND_MACVLAN] = "macvlan", + [NETDEV_KIND_MACVTAP] = "macvtap", + [NETDEV_KIND_IPVLAN] = "ipvlan", + [NETDEV_KIND_VXLAN] = "vxlan", + [NETDEV_KIND_IPIP] = "ipip", + [NETDEV_KIND_GRE] = "gre", + [NETDEV_KIND_GRETAP] = "gretap", + [NETDEV_KIND_IP6GRE] = "ip6gre", + [NETDEV_KIND_IP6GRETAP] = "ip6gretap", + [NETDEV_KIND_SIT] = "sit", + [NETDEV_KIND_VETH] = "veth", + [NETDEV_KIND_VTI] = "vti", + [NETDEV_KIND_VTI6] = "vti6", + [NETDEV_KIND_DUMMY] = "dummy", + [NETDEV_KIND_TUN] = "tun", + [NETDEV_KIND_TAP] = "tap", + [NETDEV_KIND_IP6TNL] = "ip6tnl", + [NETDEV_KIND_VRF] = "vrf", + [NETDEV_KIND_VCAN] = "vcan", + [NETDEV_KIND_GENEVE] = "geneve", + [NETDEV_KIND_VXCAN] = "vxcan", + [NETDEV_KIND_WIREGUARD] = "wireguard", + [NETDEV_KIND_NETDEVSIM] = "netdevsim", + [NETDEV_KIND_FOU] = "fou", + [NETDEV_KIND_ERSPAN] = "erspan", +}; + +DEFINE_STRING_TABLE_LOOKUP(netdev_kind, NetDevKind); + +int config_parse_netdev_kind( + const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + + NetDevKind k, *kind = data; + + assert(rvalue); + assert(data); + + k = netdev_kind_from_string(rvalue); + if (k < 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse netdev kind, ignoring assignment: %s", rvalue); + return 0; + } + + if (*kind != _NETDEV_KIND_INVALID && *kind != k) { + log_syntax(unit, LOG_ERR, filename, line, 0, + "Specified netdev kind is different from the previous value '%s', ignoring assignment: %s", + netdev_kind_to_string(*kind), rvalue); + return 0; + } + + *kind = k; + + return 0; +} + +static void netdev_callbacks_clear(NetDev *netdev) { + netdev_join_callback *callback; + + if (!netdev) + return; + + while ((callback = netdev->callbacks)) { + LIST_REMOVE(callbacks, netdev->callbacks, callback); + link_unref(callback->link); + free(callback); + } +} + +bool netdev_is_managed(NetDev *netdev) { + if (!netdev || !netdev->manager || !netdev->ifname) + return false; + + return hashmap_get(netdev->manager->netdevs, netdev->ifname) == netdev; +} + +static void netdev_detach_from_manager(NetDev *netdev) { + if (netdev->ifname && netdev->manager) + hashmap_remove(netdev->manager->netdevs, netdev->ifname); +} + +static NetDev *netdev_free(NetDev *netdev) { + assert(netdev); + + netdev_callbacks_clear(netdev); + + netdev_detach_from_manager(netdev); + + free(netdev->filename); + + free(netdev->description); + free(netdev->ifname); + free(netdev->mac); + + condition_free_list(netdev->match_host); + condition_free_list(netdev->match_virt); + condition_free_list(netdev->match_kernel_cmdline); + condition_free_list(netdev->match_kernel_version); + condition_free_list(netdev->match_arch); + + /* Invoke the per-kind done() destructor, but only if the state field is initialized. We conditionalize that + * because we parse .netdev files twice: once to determine the kind (with a short, minimal NetDev structure + * allocation, with no room for per-kind fields), and once to read the kind's properties (with a full, + * comprehensive NetDev structure allocation with enough space for whatever the specific kind needs). Now, in + * the first case we shouldn't try to destruct the per-kind NetDev fields on destruction, in the second case we + * should. We use the state field to discern the two cases: it's _NETDEV_STATE_INVALID on the first "raw" + * call. */ + if (netdev->state != _NETDEV_STATE_INVALID && + NETDEV_VTABLE(netdev) && + NETDEV_VTABLE(netdev)->done) + NETDEV_VTABLE(netdev)->done(netdev); + + return mfree(netdev); +} + +DEFINE_TRIVIAL_REF_UNREF_FUNC(NetDev, netdev, netdev_free); + +void netdev_drop(NetDev *netdev) { + if (!netdev || netdev->state == NETDEV_STATE_LINGER) + return; + + netdev->state = NETDEV_STATE_LINGER; + + log_netdev_debug(netdev, "netdev removed"); + + netdev_callbacks_clear(netdev); + + netdev_detach_from_manager(netdev); + + netdev_unref(netdev); + + return; +} + +int netdev_get(Manager *manager, const char *name, NetDev **ret) { + NetDev *netdev; + + assert(manager); + assert(name); + assert(ret); + + netdev = hashmap_get(manager->netdevs, name); + if (!netdev) { + *ret = NULL; + return -ENOENT; + } + + *ret = netdev; + + return 0; +} + +static int netdev_enter_failed(NetDev *netdev) { + netdev->state = NETDEV_STATE_FAILED; + + netdev_callbacks_clear(netdev); + + return 0; +} + +static int netdev_enslave_ready(NetDev *netdev, Link* link, link_netlink_message_handler_t callback) { + _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *req = NULL; + int r; + + assert(netdev); + assert(netdev->state == NETDEV_STATE_READY); + assert(netdev->manager); + assert(netdev->manager->rtnl); + assert(IN_SET(netdev->kind, NETDEV_KIND_BRIDGE, NETDEV_KIND_BOND, NETDEV_KIND_VRF)); + assert(link); + assert(callback); + + if (link->flags & IFF_UP && netdev->kind == NETDEV_KIND_BOND) { + log_netdev_debug(netdev, "Link '%s' was up when attempting to enslave it. Bringing link down.", link->ifname); + r = link_down(link); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not bring link down: %m"); + } + + r = sd_rtnl_message_new_link(netdev->manager->rtnl, &req, RTM_SETLINK, link->ifindex); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not allocate RTM_SETLINK message: %m"); + + r = sd_netlink_message_append_u32(req, IFLA_MASTER, netdev->ifindex); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_MASTER attribute: %m"); + + r = netlink_call_async(netdev->manager->rtnl, NULL, req, callback, + link_netlink_destroy_callback, link); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not send rtnetlink message: %m"); + + link_ref(link); + + log_netdev_debug(netdev, "Enslaving link '%s'", link->ifname); + + return 0; +} + +static int netdev_enter_ready(NetDev *netdev) { + netdev_join_callback *callback, *callback_next; + int r; + + assert(netdev); + assert(netdev->ifname); + + if (netdev->state != NETDEV_STATE_CREATING) + return 0; + + netdev->state = NETDEV_STATE_READY; + + log_netdev_info(netdev, "netdev ready"); + + LIST_FOREACH_SAFE(callbacks, callback, callback_next, netdev->callbacks) { + /* enslave the links that were attempted to be enslaved before the + * link was ready */ + r = netdev_enslave_ready(netdev, callback->link, callback->callback); + if (r < 0) + return r; + + LIST_REMOVE(callbacks, netdev->callbacks, callback); + link_unref(callback->link); + free(callback); + } + + if (NETDEV_VTABLE(netdev)->post_create) + NETDEV_VTABLE(netdev)->post_create(netdev, NULL, NULL); + + return 0; +} + +/* callback for netdev's created without a backing Link */ +static int netdev_create_handler(sd_netlink *rtnl, sd_netlink_message *m, NetDev *netdev) { + int r; + + assert(netdev); + assert(netdev->state != _NETDEV_STATE_INVALID); + + r = sd_netlink_message_get_errno(m); + if (r == -EEXIST) + log_netdev_info(netdev, "netdev exists, using existing without changing its parameters"); + else if (r < 0) { + log_netdev_warning_errno(netdev, r, "netdev could not be created: %m"); + netdev_drop(netdev); + + return 1; + } + + log_netdev_debug(netdev, "Created"); + + return 1; +} + +static int netdev_enslave(NetDev *netdev, Link *link, link_netlink_message_handler_t callback) { + int r; + + assert(netdev); + assert(netdev->manager); + assert(netdev->manager->rtnl); + assert(IN_SET(netdev->kind, NETDEV_KIND_BRIDGE, NETDEV_KIND_BOND, NETDEV_KIND_VRF)); + + if (netdev->state == NETDEV_STATE_READY) { + r = netdev_enslave_ready(netdev, link, callback); + if (r < 0) + return r; + } else if (IN_SET(netdev->state, NETDEV_STATE_LINGER, NETDEV_STATE_FAILED)) { + _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *m = NULL; + + r = rtnl_message_new_synthetic_error(netdev->manager->rtnl, -ENODEV, 0, &m); + if (r >= 0) + callback(netdev->manager->rtnl, m, link); + } else { + /* the netdev is not yet read, save this request for when it is */ + netdev_join_callback *cb; + + cb = new(netdev_join_callback, 1); + if (!cb) + return log_oom(); + + *cb = (netdev_join_callback) { + .callback = callback, + .link = link_ref(link), + }; + + LIST_PREPEND(callbacks, netdev->callbacks, cb); + + log_netdev_debug(netdev, "Will enslave '%s', when ready", link->ifname); + } + + return 0; +} + +int netdev_set_ifindex(NetDev *netdev, sd_netlink_message *message) { + uint16_t type; + const char *kind; + const char *received_kind; + const char *received_name; + int r, ifindex; + + assert(netdev); + assert(message); + + r = sd_netlink_message_get_type(message, &type); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not get rtnl message type: %m"); + + if (type != RTM_NEWLINK) { + log_netdev_error(netdev, "Cannot set ifindex from unexpected rtnl message type."); + return -EINVAL; + } + + r = sd_rtnl_message_link_get_ifindex(message, &ifindex); + if (r < 0) { + log_netdev_error_errno(netdev, r, "Could not get ifindex: %m"); + netdev_enter_failed(netdev); + return r; + } else if (ifindex <= 0) { + log_netdev_error(netdev, "Got invalid ifindex: %d", ifindex); + netdev_enter_failed(netdev); + return -EINVAL; + } + + if (netdev->ifindex > 0) { + if (netdev->ifindex != ifindex) { + log_netdev_error(netdev, "Could not set ifindex to %d, already set to %d", + ifindex, netdev->ifindex); + netdev_enter_failed(netdev); + return -EEXIST; + } else + /* ifindex already set to the same for this netdev */ + return 0; + } + + r = sd_netlink_message_read_string(message, IFLA_IFNAME, &received_name); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not get IFNAME: %m"); + + if (!streq(netdev->ifname, received_name)) { + log_netdev_error(netdev, "Received newlink with wrong IFNAME %s", received_name); + netdev_enter_failed(netdev); + return r; + } + + r = sd_netlink_message_enter_container(message, IFLA_LINKINFO); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not get LINKINFO: %m"); + + r = sd_netlink_message_read_string(message, IFLA_INFO_KIND, &received_kind); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not get KIND: %m"); + + r = sd_netlink_message_exit_container(message); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not exit container: %m"); + + if (netdev->kind == NETDEV_KIND_TAP) + /* the kernel does not distinguish between tun and tap */ + kind = "tun"; + else { + kind = netdev_kind_to_string(netdev->kind); + if (!kind) { + log_netdev_error(netdev, "Could not get kind"); + netdev_enter_failed(netdev); + return -EINVAL; + } + } + + if (!streq(kind, received_kind)) { + log_netdev_error(netdev, + "Received newlink with wrong KIND %s, " + "expected %s", received_kind, kind); + netdev_enter_failed(netdev); + return r; + } + + netdev->ifindex = ifindex; + + log_netdev_debug(netdev, "netdev has index %d", netdev->ifindex); + + netdev_enter_ready(netdev); + + return 0; +} + +#define HASH_KEY SD_ID128_MAKE(52,e1,45,bd,00,6f,29,96,21,c6,30,6d,83,71,04,48) + +int netdev_get_mac(const char *ifname, struct ether_addr **ret) { + _cleanup_free_ struct ether_addr *mac = NULL; + uint64_t result; + size_t l, sz; + uint8_t *v; + int r; + + assert(ifname); + assert(ret); + + mac = new0(struct ether_addr, 1); + if (!mac) + return -ENOMEM; + + l = strlen(ifname); + sz = sizeof(sd_id128_t) + l; + v = newa(uint8_t, sz); + + /* fetch some persistent data unique to the machine */ + r = sd_id128_get_machine((sd_id128_t*) v); + if (r < 0) + return r; + + /* combine with some data unique (on this machine) to this + * netdev */ + memcpy(v + sizeof(sd_id128_t), ifname, l); + + /* Let's hash the host machine ID plus the container name. We + * use a fixed, but originally randomly created hash key here. */ + result = siphash24(v, sz, HASH_KEY.bytes); + + assert_cc(ETH_ALEN <= sizeof(result)); + memcpy(mac->ether_addr_octet, &result, ETH_ALEN); + + /* see eth_random_addr in the kernel */ + mac->ether_addr_octet[0] &= 0xfe; /* clear multicast bit */ + mac->ether_addr_octet[0] |= 0x02; /* set local assignment bit (IEEE802) */ + + *ret = TAKE_PTR(mac); + + return 0; +} + +static int netdev_create(NetDev *netdev, Link *link, link_netlink_message_handler_t callback) { + int r; + + assert(netdev); + assert(!link || callback); + + /* create netdev */ + if (NETDEV_VTABLE(netdev)->create) { + assert(!link); + + r = NETDEV_VTABLE(netdev)->create(netdev); + if (r < 0) + return r; + + log_netdev_debug(netdev, "Created"); + } else { + _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *m = NULL; + + r = sd_rtnl_message_new_link(netdev->manager->rtnl, &m, RTM_NEWLINK, 0); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not allocate RTM_NEWLINK message: %m"); + + r = sd_netlink_message_append_string(m, IFLA_IFNAME, netdev->ifname); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IFNAME, attribute: %m"); + + if (netdev->mac) { + r = sd_netlink_message_append_ether_addr(m, IFLA_ADDRESS, netdev->mac); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_ADDRESS attribute: %m"); + } + + if (netdev->mtu) { + r = sd_netlink_message_append_u32(m, IFLA_MTU, netdev->mtu); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_MTU attribute: %m"); + } + + if (link) { + r = sd_netlink_message_append_u32(m, IFLA_LINK, link->ifindex); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_LINK attribute: %m"); + } + + r = sd_netlink_message_open_container(m, IFLA_LINKINFO); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_LINKINFO attribute: %m"); + + r = sd_netlink_message_open_container_union(m, IFLA_INFO_DATA, netdev_kind_to_string(netdev->kind)); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_INFO_DATA attribute: %m"); + + if (NETDEV_VTABLE(netdev)->fill_message_create) { + r = NETDEV_VTABLE(netdev)->fill_message_create(netdev, link, m); + if (r < 0) + return r; + } + + r = sd_netlink_message_close_container(m); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_INFO_DATA attribute: %m"); + + r = sd_netlink_message_close_container(m); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_LINKINFO attribute: %m"); + + if (link) { + r = netlink_call_async(netdev->manager->rtnl, NULL, m, callback, + link_netlink_destroy_callback, link); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not send rtnetlink message: %m"); + + link_ref(link); + } else { + r = netlink_call_async(netdev->manager->rtnl, NULL, m, netdev_create_handler, + netdev_destroy_callback, netdev); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not send rtnetlink message: %m"); + + netdev_ref(netdev); + } + + netdev->state = NETDEV_STATE_CREATING; + + log_netdev_debug(netdev, "Creating"); + } + + return 0; +} + +/* the callback must be called, possibly after a timeout, as otherwise the Link will hang */ +int netdev_join(NetDev *netdev, Link *link, link_netlink_message_handler_t callback) { + int r; + + assert(netdev); + assert(netdev->manager); + assert(netdev->manager->rtnl); + assert(NETDEV_VTABLE(netdev)); + + switch (NETDEV_VTABLE(netdev)->create_type) { + case NETDEV_CREATE_MASTER: + r = netdev_enslave(netdev, link, callback); + if (r < 0) + return r; + + break; + case NETDEV_CREATE_STACKED: + r = netdev_create(netdev, link, callback); + if (r < 0) + return r; + + break; + default: + assert_not_reached("Can not join independent netdev"); + } + + return 0; +} + +int netdev_load_one(Manager *manager, const char *filename) { + _cleanup_(netdev_unrefp) NetDev *netdev_raw = NULL, *netdev = NULL; + _cleanup_fclose_ FILE *file = NULL; + const char *dropin_dirname; + bool independent = false; + int r; + + assert(manager); + assert(filename); + + file = fopen(filename, "re"); + if (!file) { + if (errno == ENOENT) + return 0; + + return -errno; + } + + if (null_or_empty_fd(fileno(file))) { + log_debug("Skipping empty file: %s", filename); + return 0; + } + + netdev_raw = new(NetDev, 1); + if (!netdev_raw) + return log_oom(); + + *netdev_raw = (NetDev) { + .n_ref = 1, + .kind = _NETDEV_KIND_INVALID, + .state = _NETDEV_STATE_INVALID, /* an invalid state means done() of the implementation won't be called on destruction */ + }; + + dropin_dirname = strjoina(basename(filename), ".d"); + r = config_parse_many(filename, network_dirs, dropin_dirname, + "Match\0NetDev\0", + config_item_perf_lookup, network_netdev_gperf_lookup, + CONFIG_PARSE_WARN|CONFIG_PARSE_RELAXED, netdev_raw); + if (r < 0) + return r; + + /* skip out early if configuration does not match the environment */ + if (net_match_config(NULL, NULL, NULL, NULL, NULL, + netdev_raw->match_host, netdev_raw->match_virt, + netdev_raw->match_kernel_cmdline, netdev_raw->match_kernel_version, + netdev_raw->match_arch, + NULL, NULL, NULL, NULL, NULL, NULL) <= 0) + return 0; + + if (netdev_raw->kind == _NETDEV_KIND_INVALID) { + log_warning("NetDev has no Kind configured in %s. Ignoring", filename); + return 0; + } + + if (!netdev_raw->ifname) { + log_warning("NetDev without Name configured in %s. Ignoring", filename); + return 0; + } + + r = fseek(file, 0, SEEK_SET); + if (r < 0) + return -errno; + + netdev = malloc0(NETDEV_VTABLE(netdev_raw)->object_size); + if (!netdev) + return log_oom(); + + netdev->n_ref = 1; + netdev->manager = manager; + netdev->kind = netdev_raw->kind; + netdev->state = NETDEV_STATE_LOADING; /* we initialize the state here for the first time, so that done() will be called on destruction */ + + if (NETDEV_VTABLE(netdev)->init) + NETDEV_VTABLE(netdev)->init(netdev); + + r = config_parse_many(filename, network_dirs, dropin_dirname, + NETDEV_VTABLE(netdev)->sections, + config_item_perf_lookup, network_netdev_gperf_lookup, + CONFIG_PARSE_WARN, netdev); + if (r < 0) + return r; + + /* verify configuration */ + if (NETDEV_VTABLE(netdev)->config_verify) { + r = NETDEV_VTABLE(netdev)->config_verify(netdev, filename); + if (r < 0) + return 0; + } + + netdev->filename = strdup(filename); + if (!netdev->filename) + return log_oom(); + + if (!netdev->mac && netdev->kind != NETDEV_KIND_VLAN) { + r = netdev_get_mac(netdev->ifname, &netdev->mac); + if (r < 0) + return log_error_errno(r, "Failed to generate predictable MAC address for %s: %m", netdev->ifname); + } + + r = hashmap_ensure_allocated(&netdev->manager->netdevs, &string_hash_ops); + if (r < 0) + return r; + + r = hashmap_put(netdev->manager->netdevs, netdev->ifname, netdev); + if (r < 0) + return r; + + LIST_HEAD_INIT(netdev->callbacks); + + log_netdev_debug(netdev, "loaded %s", netdev_kind_to_string(netdev->kind)); + + switch (NETDEV_VTABLE(netdev)->create_type) { + case NETDEV_CREATE_MASTER: + case NETDEV_CREATE_INDEPENDENT: + r = netdev_create(netdev, NULL, NULL); + if (r < 0) + return r; + + break; + default: + break; + } + + switch (netdev->kind) { + case NETDEV_KIND_IPIP: + independent = IPIP(netdev)->independent; + break; + case NETDEV_KIND_GRE: + independent = GRE(netdev)->independent; + break; + case NETDEV_KIND_GRETAP: + independent = GRETAP(netdev)->independent; + break; + case NETDEV_KIND_IP6GRE: + independent = IP6GRE(netdev)->independent; + break; + case NETDEV_KIND_IP6GRETAP: + independent = IP6GRETAP(netdev)->independent; + break; + case NETDEV_KIND_SIT: + independent = SIT(netdev)->independent; + break; + case NETDEV_KIND_VTI: + independent = VTI(netdev)->independent; + break; + case NETDEV_KIND_VTI6: + independent = VTI6(netdev)->independent; + break; + case NETDEV_KIND_IP6TNL: + independent = IP6TNL(netdev)->independent; + break; + default: + break; + } + + if (independent) { + r = netdev_create(netdev, NULL, NULL); + if (r < 0) + return r; + } + + netdev = NULL; + + return 0; +} + +int netdev_load(Manager *manager) { + _cleanup_strv_free_ char **files = NULL; + char **f; + int r; + + assert(manager); + + hashmap_clear_with_destructor(manager->netdevs, netdev_unref); + + r = conf_files_list_strv(&files, ".netdev", NULL, 0, network_dirs); + if (r < 0) + return log_error_errno(r, "Failed to enumerate netdev files: %m"); + + STRV_FOREACH_BACKWARDS(f, files) { + r = netdev_load_one(manager, *f); + if (r < 0) + return r; + } + + return 0; +} diff --git a/src/network/netdev/netdev.h b/src/network/netdev/netdev.h new file mode 100644 index 0000000..d6524da --- /dev/null +++ b/src/network/netdev/netdev.h @@ -0,0 +1,195 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ +#pragma once + +#include "sd-netlink.h" + +#include "conf-parser.h" +#include "list.h" +#include "../networkd-link.h" +#include "time-util.h" + +typedef struct netdev_join_callback netdev_join_callback; + +struct netdev_join_callback { + link_netlink_message_handler_t callback; + Link *link; + + LIST_FIELDS(netdev_join_callback, callbacks); +}; + +typedef enum NetDevKind { + NETDEV_KIND_BRIDGE, + NETDEV_KIND_BOND, + NETDEV_KIND_VLAN, + NETDEV_KIND_MACVLAN, + NETDEV_KIND_MACVTAP, + NETDEV_KIND_IPVLAN, + NETDEV_KIND_VXLAN, + NETDEV_KIND_IPIP, + NETDEV_KIND_GRE, + NETDEV_KIND_GRETAP, + NETDEV_KIND_IP6GRE, + NETDEV_KIND_IP6GRETAP, + NETDEV_KIND_SIT, + NETDEV_KIND_VETH, + NETDEV_KIND_VTI, + NETDEV_KIND_VTI6, + NETDEV_KIND_IP6TNL, + NETDEV_KIND_DUMMY, + NETDEV_KIND_TUN, + NETDEV_KIND_TAP, + NETDEV_KIND_VRF, + NETDEV_KIND_VCAN, + NETDEV_KIND_GENEVE, + NETDEV_KIND_VXCAN, + NETDEV_KIND_WIREGUARD, + NETDEV_KIND_NETDEVSIM, + NETDEV_KIND_FOU, + NETDEV_KIND_ERSPAN, + _NETDEV_KIND_MAX, + _NETDEV_KIND_INVALID = -1 +} NetDevKind; + +typedef enum NetDevState { + NETDEV_STATE_LOADING, + NETDEV_STATE_FAILED, + NETDEV_STATE_CREATING, + NETDEV_STATE_READY, + NETDEV_STATE_LINGER, + _NETDEV_STATE_MAX, + _NETDEV_STATE_INVALID = -1, +} NetDevState; + +typedef enum NetDevCreateType { + NETDEV_CREATE_INDEPENDENT, + NETDEV_CREATE_MASTER, + NETDEV_CREATE_STACKED, + _NETDEV_CREATE_MAX, + _NETDEV_CREATE_INVALID = -1, +} NetDevCreateType; + +typedef struct Manager Manager; +typedef struct Condition Condition; + +typedef struct NetDev { + Manager *manager; + + unsigned n_ref; + + char *filename; + + Condition *match_host; + Condition *match_virt; + Condition *match_kernel_cmdline; + Condition *match_kernel_version; + Condition *match_arch; + + NetDevState state; + NetDevKind kind; + char *description; + char *ifname; + struct ether_addr *mac; + uint32_t mtu; + int ifindex; + + LIST_HEAD(netdev_join_callback, callbacks); +} NetDev; + +typedef struct NetDevVTable { + /* How much memory does an object of this unit type need */ + size_t object_size; + + /* Config file sections this netdev kind understands, separated + * by NUL chars */ + const char *sections; + + /* This should reset all type-specific variables. This should + * not allocate memory, and is called with zero-initialized + * data. It should hence only initialize variables that need + * to be set != 0. */ + void (*init)(NetDev *n); + + /* This should free all kind-specific variables. It should be + * idempotent. */ + void (*done)(NetDev *n); + + /* fill in message to create netdev */ + int (*fill_message_create)(NetDev *netdev, Link *link, sd_netlink_message *message); + + /* specifies if netdev is independent, or a master device or a stacked device */ + NetDevCreateType create_type; + + /* create netdev, if not done via rtnl */ + int (*create)(NetDev *netdev); + + /* perform additional configuration after netdev has been createad */ + int (*post_create)(NetDev *netdev, Link *link, sd_netlink_message *message); + + /* verify that compulsory configuration options were specified */ + int (*config_verify)(NetDev *netdev, const char *filename); +} NetDevVTable; + +extern const NetDevVTable * const netdev_vtable[_NETDEV_KIND_MAX]; + +#define NETDEV_VTABLE(n) ((n)->kind != _NETDEV_KIND_INVALID ? netdev_vtable[(n)->kind] : NULL) + +/* For casting a netdev into the various netdev kinds */ +#define DEFINE_NETDEV_CAST(UPPERCASE, MixedCase) \ + static inline MixedCase* UPPERCASE(NetDev *n) { \ + if (_unlikely_(!n || \ + n->kind != NETDEV_KIND_##UPPERCASE) || \ + n->state == _NETDEV_STATE_INVALID) \ + return NULL; \ + \ + return (MixedCase*) n; \ + } + +/* For casting the various netdev kinds into a netdev */ +#define NETDEV(n) (&(n)->meta) + +int netdev_load(Manager *manager); +int netdev_load_one(Manager *manager, const char *filename); +void netdev_drop(NetDev *netdev); + +NetDev *netdev_unref(NetDev *netdev); +NetDev *netdev_ref(NetDev *netdev); +DEFINE_TRIVIAL_DESTRUCTOR(netdev_destroy_callback, NetDev, netdev_unref); +DEFINE_TRIVIAL_CLEANUP_FUNC(NetDev*, netdev_unref); + +bool netdev_is_managed(NetDev *netdev); +int netdev_get(Manager *manager, const char *name, NetDev **ret); +int netdev_set_ifindex(NetDev *netdev, sd_netlink_message *newlink); +int netdev_get_mac(const char *ifname, struct ether_addr **ret); +int netdev_join(NetDev *netdev, Link *link, link_netlink_message_handler_t cb); + +const char *netdev_kind_to_string(NetDevKind d) _const_; +NetDevKind netdev_kind_from_string(const char *d) _pure_; + +CONFIG_PARSER_PROTOTYPE(config_parse_netdev_kind); + +/* gperf */ +const struct ConfigPerfItem* network_netdev_gperf_lookup(const char *key, GPERF_LEN_TYPE length); + +/* Macros which append INTERFACE= to the message */ + +#define log_netdev_full(netdev, level, error, ...) \ + ({ \ + const NetDev *_n = (netdev); \ + _n ? log_object_internal(level, error, __FILE__, __LINE__, __func__, "INTERFACE=", _n->ifname, NULL, NULL, ##__VA_ARGS__) : \ + log_internal(level, error, __FILE__, __LINE__, __func__, ##__VA_ARGS__); \ + }) + +#define log_netdev_debug(netdev, ...) log_netdev_full(netdev, LOG_DEBUG, 0, ##__VA_ARGS__) +#define log_netdev_info(netdev, ...) log_netdev_full(netdev, LOG_INFO, 0, ##__VA_ARGS__) +#define log_netdev_notice(netdev, ...) log_netdev_full(netdev, LOG_NOTICE, 0, ##__VA_ARGS__) +#define log_netdev_warning(netdev, ...) log_netdev_full(netdev, LOG_WARNING, 0, ## __VA_ARGS__) +#define log_netdev_error(netdev, ...) log_netdev_full(netdev, LOG_ERR, 0, ##__VA_ARGS__) + +#define log_netdev_debug_errno(netdev, error, ...) log_netdev_full(netdev, LOG_DEBUG, error, ##__VA_ARGS__) +#define log_netdev_info_errno(netdev, error, ...) log_netdev_full(netdev, LOG_INFO, error, ##__VA_ARGS__) +#define log_netdev_notice_errno(netdev, error, ...) log_netdev_full(netdev, LOG_NOTICE, error, ##__VA_ARGS__) +#define log_netdev_warning_errno(netdev, error, ...) log_netdev_full(netdev, LOG_WARNING, error, ##__VA_ARGS__) +#define log_netdev_error_errno(netdev, error, ...) log_netdev_full(netdev, LOG_ERR, error, ##__VA_ARGS__) + +#define LOG_NETDEV_MESSAGE(netdev, fmt, ...) "MESSAGE=%s: " fmt, (netdev)->ifname, ##__VA_ARGS__ +#define LOG_NETDEV_INTERFACE(netdev) "INTERFACE=%s", (netdev)->ifname diff --git a/src/network/netdev/netdevsim.c b/src/network/netdev/netdevsim.c new file mode 100644 index 0000000..8caba67 --- /dev/null +++ b/src/network/netdev/netdevsim.c @@ -0,0 +1,10 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ + +#include "netdev/netdevsim.h" +#include "missing.h" + +const NetDevVTable netdevsim_vtable = { + .object_size = sizeof(NetDevSim), + .sections = "Match\0NetDev\0", + .create_type = NETDEV_CREATE_INDEPENDENT, +}; diff --git a/src/network/netdev/netdevsim.h b/src/network/netdev/netdevsim.h new file mode 100644 index 0000000..d3ed0c0 --- /dev/null +++ b/src/network/netdev/netdevsim.h @@ -0,0 +1,13 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ +#pragma once + +typedef struct NetDevSim NetDevSim; + +#include "netdev/netdev.h" + +struct NetDevSim { + NetDev meta; +}; + +DEFINE_NETDEV_CAST(NETDEVSIM, NetDevSim); +extern const NetDevVTable netdevsim_vtable; diff --git a/src/network/netdev/tunnel.c b/src/network/netdev/tunnel.c new file mode 100644 index 0000000..684eddd --- /dev/null +++ b/src/network/netdev/tunnel.c @@ -0,0 +1,936 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ + +#include <arpa/inet.h> +#include <net/if.h> +#include <linux/ip.h> +#include <linux/if_tunnel.h> +#include <linux/ip6_tunnel.h> + +#if HAVE_LINUX_FOU_H +#include <linux/fou.h> +#endif + +#include "sd-netlink.h" + +#include "conf-parser.h" +#include "missing.h" +#include "networkd-link.h" +#include "netdev/tunnel.h" +#include "parse-util.h" +#include "string-table.h" +#include "string-util.h" +#include "util.h" + +#define DEFAULT_TNL_HOP_LIMIT 64 +#define IP6_FLOWINFO_FLOWLABEL htobe32(0x000FFFFF) +#define IP6_TNL_F_ALLOW_LOCAL_REMOTE 0x40 + +static const char* const ip6tnl_mode_table[_NETDEV_IP6_TNL_MODE_MAX] = { + [NETDEV_IP6_TNL_MODE_IP6IP6] = "ip6ip6", + [NETDEV_IP6_TNL_MODE_IPIP6] = "ipip6", + [NETDEV_IP6_TNL_MODE_ANYIP6] = "any", +}; + +DEFINE_STRING_TABLE_LOOKUP(ip6tnl_mode, Ip6TnlMode); +DEFINE_CONFIG_PARSE_ENUM(config_parse_ip6tnl_mode, ip6tnl_mode, Ip6TnlMode, "Failed to parse ip6 tunnel Mode"); + +static int netdev_ipip_fill_message_create(NetDev *netdev, Link *link, sd_netlink_message *m) { + Tunnel *t = IPIP(netdev); + int r; + + assert(netdev); + assert(m); + assert(t); + assert(IN_SET(t->family, AF_INET, AF_UNSPEC)); + + if (link) { + r = sd_netlink_message_append_u32(m, IFLA_IPTUN_LINK, link->ifindex); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_LINK attribute: %m"); + } + + r = sd_netlink_message_append_in_addr(m, IFLA_IPTUN_LOCAL, &t->local.in); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_LOCAL attribute: %m"); + + r = sd_netlink_message_append_in_addr(m, IFLA_IPTUN_REMOTE, &t->remote.in); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_REMOTE attribute: %m"); + + r = sd_netlink_message_append_u8(m, IFLA_IPTUN_TTL, t->ttl); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_TTL attribute: %m"); + + r = sd_netlink_message_append_u8(m, IFLA_IPTUN_PMTUDISC, t->pmtudisc); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_PMTUDISC attribute: %m"); + + if (t->fou_tunnel) { + + r = sd_netlink_message_append_u16(m, IFLA_IPTUN_ENCAP_TYPE, t->fou_encap_type); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_ENCAP_TYPE attribute: %m"); + + r = sd_netlink_message_append_u16(m, IFLA_IPTUN_ENCAP_SPORT, htobe16(t->encap_src_port)); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_ENCAP_SPORT attribute: %m"); + + r = sd_netlink_message_append_u16(m, IFLA_IPTUN_ENCAP_DPORT, htobe16(t->fou_destination_port)); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_ENCAP_DPORT attribute: %m"); + } + + return r; +} + +static int netdev_sit_fill_message_create(NetDev *netdev, Link *link, sd_netlink_message *m) { + Tunnel *t = SIT(netdev); + int r; + + assert(netdev); + assert(m); + assert(t); + assert(IN_SET(t->family, AF_INET, AF_UNSPEC)); + + if (link) { + r = sd_netlink_message_append_u32(m, IFLA_IPTUN_LINK, link->ifindex); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_LINK attribute: %m"); + } + + r = sd_netlink_message_append_in_addr(m, IFLA_IPTUN_LOCAL, &t->local.in); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_LOCAL attribute: %m"); + + r = sd_netlink_message_append_in_addr(m, IFLA_IPTUN_REMOTE, &t->remote.in); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_REMOTE attribute: %m"); + + r = sd_netlink_message_append_u8(m, IFLA_IPTUN_TTL, t->ttl); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_TTL attribute: %m"); + + r = sd_netlink_message_append_u8(m, IFLA_IPTUN_PMTUDISC, t->pmtudisc); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_PMTUDISC attribute: %m"); + + if (t->sixrd_prefixlen > 0) { + r = sd_netlink_message_append_in6_addr(m, IFLA_IPTUN_6RD_PREFIX, &t->sixrd_prefix); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_6RD_PREFIX attribute: %m"); + + /* u16 is deliberate here, even though we're passing a netmask that can never be >128. The kernel is + * expecting to receive the prefixlen as a u16. + */ + r = sd_netlink_message_append_u16(m, IFLA_IPTUN_6RD_PREFIXLEN, t->sixrd_prefixlen); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_6RD_PREFIXLEN attribute: %m"); + } + + if (t->isatap >= 0) { + uint16_t flags = 0; + + SET_FLAG(flags, SIT_ISATAP, t->isatap); + + r = sd_netlink_message_append_u16(m, IFLA_IPTUN_FLAGS, flags); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_FLAGS attribute: %m"); + } + + return r; +} + +static int netdev_gre_fill_message_create(NetDev *netdev, Link *link, sd_netlink_message *m) { + Tunnel *t; + int r; + + assert(netdev); + + if (netdev->kind == NETDEV_KIND_GRE) + t = GRE(netdev); + else + t = GRETAP(netdev); + + assert(t); + assert(IN_SET(t->family, AF_INET, AF_UNSPEC)); + assert(m); + + if (link) { + r = sd_netlink_message_append_u32(m, IFLA_GRE_LINK, link->ifindex); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_GRE_LINK attribute: %m"); + } + + r = sd_netlink_message_append_in_addr(m, IFLA_GRE_LOCAL, &t->local.in); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_GRE_LOCAL attribute: %m"); + + r = sd_netlink_message_append_in_addr(m, IFLA_GRE_REMOTE, &t->remote.in); + if (r < 0) + log_netdev_error_errno(netdev, r, "Could not append IFLA_GRE_REMOTE attribute: %m"); + + r = sd_netlink_message_append_u8(m, IFLA_GRE_TTL, t->ttl); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_GRE_TTL attribute: %m"); + + r = sd_netlink_message_append_u8(m, IFLA_GRE_TOS, t->tos); + if (r < 0) + log_netdev_error_errno(netdev, r, "Could not append IFLA_GRE_TOS attribute: %m"); + + r = sd_netlink_message_append_u8(m, IFLA_GRE_PMTUDISC, t->pmtudisc); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_GRE_PMTUDISC attribute: %m"); + + return r; +} + +static int netdev_erspan_fill_message_create(NetDev *netdev, Link *link, sd_netlink_message *m) { + uint32_t ikey = 0; + uint32_t okey = 0; + uint16_t iflags = 0; + uint16_t oflags = 0; + Tunnel *t; + int r; + + assert(netdev); + + t = ERSPAN(netdev); + + assert(t); + assert(IN_SET(t->family, AF_INET, AF_UNSPEC)); + assert(m); + + r = sd_netlink_message_append_u32(m, IFLA_GRE_ERSPAN_INDEX, t->erspan_index); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_GRE_ERSPAN_INDEX attribute: %m"); + + if (t->key != 0) { + ikey = okey = htobe32(t->key); + iflags |= GRE_KEY; + oflags |= GRE_KEY; + } + + if (t->ikey != 0) { + ikey = htobe32(t->ikey); + iflags |= GRE_KEY; + } + + if (t->okey != 0) { + okey = htobe32(t->okey); + oflags |= GRE_KEY; + } + + if (t->erspan_sequence > 0) { + iflags |= GRE_SEQ; + oflags |= GRE_SEQ; + } else if (t->erspan_sequence == 0) { + iflags &= ~GRE_SEQ; + oflags &= ~GRE_SEQ; + } + + r = sd_netlink_message_append_u32(m, IFLA_GRE_IKEY, ikey); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_GRE_IKEY attribute: %m"); + + r = sd_netlink_message_append_u32(m, IFLA_GRE_OKEY, okey); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_GRE_OKEY attribute: %m"); + + r = sd_netlink_message_append_u16(m, IFLA_GRE_IFLAGS, iflags); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_GRE_IFLAGS attribute: %m"); + + r = sd_netlink_message_append_u16(m, IFLA_GRE_OFLAGS, oflags); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_GRE_OFLAGS, attribute: %m"); + + r = sd_netlink_message_append_in_addr(m, IFLA_GRE_LOCAL, &t->local.in); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_GRE_LOCAL attribute: %m"); + + r = sd_netlink_message_append_in_addr(m, IFLA_GRE_REMOTE, &t->remote.in); + if (r < 0) + log_netdev_error_errno(netdev, r, "Could not append IFLA_GRE_REMOTE attribute: %m"); + + return r; +} + +static int netdev_ip6gre_fill_message_create(NetDev *netdev, Link *link, sd_netlink_message *m) { + Tunnel *t; + int r; + + assert(netdev); + + if (netdev->kind == NETDEV_KIND_IP6GRE) + t = IP6GRE(netdev); + else + t = IP6GRETAP(netdev); + + assert(t); + assert(t->family == AF_INET6); + assert(m); + + if (link) { + r = sd_netlink_message_append_u32(m, IFLA_GRE_LINK, link->ifindex); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_GRE_LINK attribute: %m"); + } + + r = sd_netlink_message_append_in6_addr(m, IFLA_GRE_LOCAL, &t->local.in6); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_GRE_LOCAL attribute: %m"); + + r = sd_netlink_message_append_in6_addr(m, IFLA_GRE_REMOTE, &t->remote.in6); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_GRE_REMOTE attribute: %m"); + + r = sd_netlink_message_append_u8(m, IFLA_GRE_TTL, t->ttl); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_GRE_TTL attribute: %m"); + + if (t->ipv6_flowlabel != _NETDEV_IPV6_FLOWLABEL_INVALID) { + r = sd_netlink_message_append_u32(m, IFLA_GRE_FLOWINFO, t->ipv6_flowlabel); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_GRE_FLOWINFO attribute: %m"); + } + + r = sd_netlink_message_append_u32(m, IFLA_GRE_FLAGS, t->flags); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_GRE_FLAGS attribute: %m"); + + return r; +} + +static int netdev_vti_fill_message_key(NetDev *netdev, Link *link, sd_netlink_message *m) { + uint32_t ikey, okey; + Tunnel *t; + int r; + + assert(m); + + if (netdev->kind == NETDEV_KIND_VTI) + t = VTI(netdev); + else + t = VTI6(netdev); + + assert(t); + + if (t->key != 0) + ikey = okey = htobe32(t->key); + else { + ikey = htobe32(t->ikey); + okey = htobe32(t->okey); + } + + r = sd_netlink_message_append_u32(m, IFLA_VTI_IKEY, ikey); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VTI_IKEY attribute: %m"); + + r = sd_netlink_message_append_u32(m, IFLA_VTI_OKEY, okey); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VTI_OKEY attribute: %m"); + + return 0; +} + +static int netdev_vti_fill_message_create(NetDev *netdev, Link *link, sd_netlink_message *m) { + Tunnel *t = VTI(netdev); + int r; + + assert(netdev); + assert(m); + assert(t); + assert(t->family == AF_INET); + + if (link) { + r = sd_netlink_message_append_u32(m, IFLA_VTI_LINK, link->ifindex); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_LINK attribute: %m"); + } + + r = netdev_vti_fill_message_key(netdev, link, m); + if (r < 0) + return r; + + r = sd_netlink_message_append_in_addr(m, IFLA_VTI_LOCAL, &t->local.in); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_LOCAL attribute: %m"); + + r = sd_netlink_message_append_in_addr(m, IFLA_VTI_REMOTE, &t->remote.in); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_REMOTE attribute: %m"); + + return r; +} + +static int netdev_vti6_fill_message_create(NetDev *netdev, Link *link, sd_netlink_message *m) { + Tunnel *t = VTI6(netdev); + int r; + + assert(netdev); + assert(m); + assert(t); + assert(t->family == AF_INET6); + + if (link) { + r = sd_netlink_message_append_u32(m, IFLA_VTI_LINK, link->ifindex); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_LINK attribute: %m"); + } + + r = netdev_vti_fill_message_key(netdev, link, m); + if (r < 0) + return r; + + r = sd_netlink_message_append_in6_addr(m, IFLA_VTI_LOCAL, &t->local.in6); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_LOCAL attribute: %m"); + + r = sd_netlink_message_append_in6_addr(m, IFLA_VTI_REMOTE, &t->remote.in6); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_REMOTE attribute: %m"); + + return r; +} + +static int netdev_ip6tnl_fill_message_create(NetDev *netdev, Link *link, sd_netlink_message *m) { + Tunnel *t = IP6TNL(netdev); + uint8_t proto; + int r; + + assert(netdev); + assert(m); + assert(t); + assert(t->family == AF_INET6); + + if (link) { + r = sd_netlink_message_append_u32(m, IFLA_IPTUN_LINK, link->ifindex); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_LINK attribute: %m"); + } + + r = sd_netlink_message_append_in6_addr(m, IFLA_IPTUN_LOCAL, &t->local.in6); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_LOCAL attribute: %m"); + + r = sd_netlink_message_append_in6_addr(m, IFLA_IPTUN_REMOTE, &t->remote.in6); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_REMOTE attribute: %m"); + + r = sd_netlink_message_append_u8(m, IFLA_IPTUN_TTL, t->ttl); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_TTL attribute: %m"); + + if (t->ipv6_flowlabel != _NETDEV_IPV6_FLOWLABEL_INVALID) { + r = sd_netlink_message_append_u32(m, IFLA_IPTUN_FLOWINFO, t->ipv6_flowlabel); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_FLOWINFO attribute: %m"); + } + + if (t->copy_dscp) + t->flags |= IP6_TNL_F_RCV_DSCP_COPY; + + if (t->allow_localremote != -1) + SET_FLAG(t->flags, IP6_TNL_F_ALLOW_LOCAL_REMOTE, t->allow_localremote); + + if (t->encap_limit != IPV6_DEFAULT_TNL_ENCAP_LIMIT) { + r = sd_netlink_message_append_u8(m, IFLA_IPTUN_ENCAP_LIMIT, t->encap_limit); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_ENCAP_LIMIT attribute: %m"); + } + + r = sd_netlink_message_append_u32(m, IFLA_IPTUN_FLAGS, t->flags); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_FLAGS attribute: %m"); + + switch (t->ip6tnl_mode) { + case NETDEV_IP6_TNL_MODE_IP6IP6: + proto = IPPROTO_IPV6; + break; + case NETDEV_IP6_TNL_MODE_IPIP6: + proto = IPPROTO_IPIP; + break; + case NETDEV_IP6_TNL_MODE_ANYIP6: + default: + proto = 0; + break; + } + + r = sd_netlink_message_append_u8(m, IFLA_IPTUN_PROTO, proto); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_IPTUN_MODE attribute: %m"); + + return r; +} + +static int netdev_tunnel_verify(NetDev *netdev, const char *filename) { + Tunnel *t = NULL; + + assert(netdev); + assert(filename); + + switch (netdev->kind) { + case NETDEV_KIND_IPIP: + t = IPIP(netdev); + break; + case NETDEV_KIND_SIT: + t = SIT(netdev); + break; + case NETDEV_KIND_GRE: + t = GRE(netdev); + break; + case NETDEV_KIND_GRETAP: + t = GRETAP(netdev); + break; + case NETDEV_KIND_IP6GRE: + t = IP6GRE(netdev); + break; + case NETDEV_KIND_IP6GRETAP: + t = IP6GRETAP(netdev); + break; + case NETDEV_KIND_VTI: + t = VTI(netdev); + break; + case NETDEV_KIND_VTI6: + t = VTI6(netdev); + break; + case NETDEV_KIND_IP6TNL: + t = IP6TNL(netdev); + break; + case NETDEV_KIND_ERSPAN: + t = ERSPAN(netdev); + break; + default: + assert_not_reached("Invalid tunnel kind"); + } + + assert(t); + + if (!IN_SET(t->family, AF_INET, AF_INET6, AF_UNSPEC)) { + log_netdev_error(netdev, + "Tunnel with invalid address family configured in %s. Ignoring", filename); + return -EINVAL; + } + + if (IN_SET(netdev->kind, NETDEV_KIND_VTI, NETDEV_KIND_IPIP, NETDEV_KIND_SIT, NETDEV_KIND_GRE, NETDEV_KIND_GRETAP, NETDEV_KIND_ERSPAN) && + (t->family != AF_INET || in_addr_is_null(t->family, &t->local))) { + log_netdev_error(netdev, + "vti/ipip/sit/gre/gretap/erspan tunnel without a local IPv4 address configured in %s. Ignoring", filename); + return -EINVAL; + } + + if (IN_SET(netdev->kind, NETDEV_KIND_VTI6, NETDEV_KIND_IP6TNL, NETDEV_KIND_IP6GRE, NETDEV_KIND_IP6GRETAP) && + (t->family != AF_INET6 || in_addr_is_null(t->family, &t->local))) { + log_netdev_error(netdev, + "vti6/ip6tnl/ip6gre/ip6gretap tunnel without a local IPv6 address configured in %s. Ignoring", filename); + return -EINVAL; + } + + if (netdev->kind == NETDEV_KIND_IP6TNL && + t->ip6tnl_mode == _NETDEV_IP6_TNL_MODE_INVALID) { + log_netdev_error(netdev, + "ip6tnl without mode configured in %s. Ignoring", filename); + return -EINVAL; + } + + if (t->fou_tunnel && t->fou_destination_port <= 0) { + log_netdev_error(netdev, "FooOverUDP missing port configured in %s. Ignoring", filename); + return -EINVAL; + } + + if (netdev->kind == NETDEV_KIND_ERSPAN && (t->erspan_index >= (1 << 20) || t->erspan_index == 0)) { + log_netdev_error(netdev, "Invalid erspan index %d. Ignoring", t->erspan_index); + return -EINVAL; + } + + return 0; +} + +int config_parse_tunnel_address(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + Tunnel *t = userdata; + union in_addr_union *addr = data, buffer; + int r, f; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + /* This is used to parse addresses on both local and remote ends of the tunnel. + * Address families must match. + * + * "any" is a special value which means that the address is unspecified. + */ + + if (streq(rvalue, "any")) { + *addr = IN_ADDR_NULL; + + /* As a special case, if both the local and remote addresses are + * unspecified, also clear the address family. + */ + if (t->family != AF_UNSPEC && + in_addr_is_null(t->family, &t->local) && + in_addr_is_null(t->family, &t->remote)) + t->family = AF_UNSPEC; + return 0; + } + + r = in_addr_from_string_auto(rvalue, &f, &buffer); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, + "Tunnel address \"%s\" invalid, ignoring assignment: %m", rvalue); + return 0; + } + + if (t->family != AF_UNSPEC && t->family != f) { + log_syntax(unit, LOG_ERR, filename, line, 0, + "Tunnel addresses incompatible, ignoring assignment: %s", rvalue); + return 0; + } + + t->family = f; + *addr = buffer; + return 0; +} + +int config_parse_tunnel_key(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + union in_addr_union buffer; + Tunnel *t = userdata; + uint32_t k; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + r = in_addr_from_string(AF_INET, rvalue, &buffer); + if (r < 0) { + r = safe_atou32(rvalue, &k); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse tunnel key ignoring assignment: %s", rvalue); + return 0; + } + } else + k = be32toh(buffer.in.s_addr); + + if (streq(lvalue, "Key")) + t->key = k; + else if (streq(lvalue, "InputKey")) + t->ikey = k; + else + t->okey = k; + + return 0; +} + +int config_parse_ipv6_flowlabel(const char* unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + IPv6FlowLabel *ipv6_flowlabel = data; + Tunnel *t = userdata; + int k = 0; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(ipv6_flowlabel); + + if (streq(rvalue, "inherit")) { + *ipv6_flowlabel = IP6_FLOWINFO_FLOWLABEL; + t->flags |= IP6_TNL_F_USE_ORIG_FLOWLABEL; + } else { + r = config_parse_int(unit, filename, line, section, section_line, lvalue, ltype, rvalue, &k, userdata); + if (r < 0) + return r; + + if (k > 0xFFFFF) + log_syntax(unit, LOG_ERR, filename, line, 0, "Failed to parse IPv6 flowlabel option, ignoring: %s", rvalue); + else { + *ipv6_flowlabel = htobe32(k) & IP6_FLOWINFO_FLOWLABEL; + t->flags &= ~IP6_TNL_F_USE_ORIG_FLOWLABEL; + } + } + + return 0; +} + +int config_parse_encap_limit(const char* unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + Tunnel *t = userdata; + int k = 0; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + + if (streq(rvalue, "none")) + t->flags |= IP6_TNL_F_IGN_ENCAP_LIMIT; + else { + r = safe_atoi(rvalue, &k); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse Tunnel Encapsulation Limit option, ignoring: %s", rvalue); + return 0; + } + + if (k > 255 || k < 0) + log_syntax(unit, LOG_ERR, filename, line, 0, "Invalid Tunnel Encapsulation value, ignoring: %d", k); + else { + t->encap_limit = k; + t->flags &= ~IP6_TNL_F_IGN_ENCAP_LIMIT; + } + } + + return 0; +} + +int config_parse_6rd_prefix(const char* unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + Tunnel *t = userdata; + + assert(filename); + assert(lvalue); + assert(rvalue); + + union in_addr_union p; + uint8_t l; + int r; + + r = in_addr_prefix_from_string(rvalue, AF_INET6, &p, &l); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse 6rd prefix \"%s\", ignoring: %m", rvalue); + return 0; + } + if (l == 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "6rd prefix length of \"%s\" must be greater than zero, ignoring", rvalue); + return 0; + } + + t->sixrd_prefix = p.in6; + t->sixrd_prefixlen = l; + + return 0; +} + +static void ipip_init(NetDev *n) { + Tunnel *t = IPIP(n); + + assert(n); + assert(t); + + t->pmtudisc = true; + t->fou_encap_type = FOU_ENCAP_DIRECT; +} + +static void sit_init(NetDev *n) { + Tunnel *t = SIT(n); + + assert(n); + assert(t); + + t->pmtudisc = true; + t->isatap = -1; +} + +static void vti_init(NetDev *n) { + Tunnel *t; + + assert(n); + + if (n->kind == NETDEV_KIND_VTI) + t = VTI(n); + else + t = VTI6(n); + + assert(t); + + t->pmtudisc = true; +} + +static void gre_init(NetDev *n) { + Tunnel *t; + + assert(n); + + if (n->kind == NETDEV_KIND_GRE) + t = GRE(n); + else + t = GRETAP(n); + + assert(t); + + t->pmtudisc = true; +} + +static void ip6gre_init(NetDev *n) { + Tunnel *t; + + assert(n); + + if (n->kind == NETDEV_KIND_IP6GRE) + t = IP6GRE(n); + else + t = IP6GRETAP(n); + + assert(t); + + t->ttl = DEFAULT_TNL_HOP_LIMIT; +} + +static void erspan_init(NetDev *n) { + Tunnel *t; + + assert(n); + + t = ERSPAN(n); + + assert(t); + + t->erspan_sequence = -1; +} + +static void ip6tnl_init(NetDev *n) { + Tunnel *t = IP6TNL(n); + + assert(n); + assert(t); + + t->ttl = DEFAULT_TNL_HOP_LIMIT; + t->encap_limit = IPV6_DEFAULT_TNL_ENCAP_LIMIT; + t->ip6tnl_mode = _NETDEV_IP6_TNL_MODE_INVALID; + t->ipv6_flowlabel = _NETDEV_IPV6_FLOWLABEL_INVALID; + t->allow_localremote = -1; +} + +const NetDevVTable ipip_vtable = { + .object_size = sizeof(Tunnel), + .init = ipip_init, + .sections = "Match\0NetDev\0Tunnel\0", + .fill_message_create = netdev_ipip_fill_message_create, + .create_type = NETDEV_CREATE_STACKED, + .config_verify = netdev_tunnel_verify, +}; + +const NetDevVTable sit_vtable = { + .object_size = sizeof(Tunnel), + .init = sit_init, + .sections = "Match\0NetDev\0Tunnel\0", + .fill_message_create = netdev_sit_fill_message_create, + .create_type = NETDEV_CREATE_STACKED, + .config_verify = netdev_tunnel_verify, +}; + +const NetDevVTable vti_vtable = { + .object_size = sizeof(Tunnel), + .init = vti_init, + .sections = "Match\0NetDev\0Tunnel\0", + .fill_message_create = netdev_vti_fill_message_create, + .create_type = NETDEV_CREATE_STACKED, + .config_verify = netdev_tunnel_verify, +}; + +const NetDevVTable vti6_vtable = { + .object_size = sizeof(Tunnel), + .init = vti_init, + .sections = "Match\0NetDev\0Tunnel\0", + .fill_message_create = netdev_vti6_fill_message_create, + .create_type = NETDEV_CREATE_STACKED, + .config_verify = netdev_tunnel_verify, +}; + +const NetDevVTable gre_vtable = { + .object_size = sizeof(Tunnel), + .init = gre_init, + .sections = "Match\0NetDev\0Tunnel\0", + .fill_message_create = netdev_gre_fill_message_create, + .create_type = NETDEV_CREATE_STACKED, + .config_verify = netdev_tunnel_verify, +}; + +const NetDevVTable gretap_vtable = { + .object_size = sizeof(Tunnel), + .init = gre_init, + .sections = "Match\0NetDev\0Tunnel\0", + .fill_message_create = netdev_gre_fill_message_create, + .create_type = NETDEV_CREATE_STACKED, + .config_verify = netdev_tunnel_verify, +}; + +const NetDevVTable ip6gre_vtable = { + .object_size = sizeof(Tunnel), + .init = ip6gre_init, + .sections = "Match\0NetDev\0Tunnel\0", + .fill_message_create = netdev_ip6gre_fill_message_create, + .create_type = NETDEV_CREATE_STACKED, + .config_verify = netdev_tunnel_verify, +}; + +const NetDevVTable ip6gretap_vtable = { + .object_size = sizeof(Tunnel), + .init = ip6gre_init, + .sections = "Match\0NetDev\0Tunnel\0", + .fill_message_create = netdev_ip6gre_fill_message_create, + .create_type = NETDEV_CREATE_STACKED, + .config_verify = netdev_tunnel_verify, +}; + +const NetDevVTable ip6tnl_vtable = { + .object_size = sizeof(Tunnel), + .init = ip6tnl_init, + .sections = "Match\0NetDev\0Tunnel\0", + .fill_message_create = netdev_ip6tnl_fill_message_create, + .create_type = NETDEV_CREATE_STACKED, + .config_verify = netdev_tunnel_verify, +}; + +const NetDevVTable erspan_vtable = { + .object_size = sizeof(Tunnel), + .init = erspan_init, + .sections = "Match\0NetDev\0Tunnel\0", + .fill_message_create = netdev_erspan_fill_message_create, + .create_type = NETDEV_CREATE_INDEPENDENT, + .config_verify = netdev_tunnel_verify, +}; diff --git a/src/network/netdev/tunnel.h b/src/network/netdev/tunnel.h new file mode 100644 index 0000000..8f511dd --- /dev/null +++ b/src/network/netdev/tunnel.h @@ -0,0 +1,91 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ +#pragma once + +#include "in-addr-util.h" + +#include "conf-parser.h" +#include "netdev/netdev.h" +#include "netdev/fou-tunnel.h" + +typedef enum Ip6TnlMode { + NETDEV_IP6_TNL_MODE_IP6IP6, + NETDEV_IP6_TNL_MODE_IPIP6, + NETDEV_IP6_TNL_MODE_ANYIP6, + _NETDEV_IP6_TNL_MODE_MAX, + _NETDEV_IP6_TNL_MODE_INVALID = -1, +} Ip6TnlMode; + +typedef enum IPv6FlowLabel { + NETDEV_IPV6_FLOWLABEL_INHERIT = 0xFFFFF + 1, + _NETDEV_IPV6_FLOWLABEL_MAX, + _NETDEV_IPV6_FLOWLABEL_INVALID = -1, +} IPv6FlowLabel; + +typedef struct Tunnel { + NetDev meta; + + uint8_t encap_limit; + + int family; + int ipv6_flowlabel; + int allow_localremote; + int erspan_sequence; + int isatap; + + unsigned ttl; + unsigned tos; + unsigned flags; + + uint32_t key; + uint32_t ikey; + uint32_t okey; + uint32_t erspan_index; + + union in_addr_union local; + union in_addr_union remote; + + Ip6TnlMode ip6tnl_mode; + FooOverUDPEncapType fou_encap_type; + + bool pmtudisc; + bool copy_dscp; + bool independent; + bool fou_tunnel; + + uint16_t encap_src_port; + uint16_t fou_destination_port; + + struct in6_addr sixrd_prefix; + uint8_t sixrd_prefixlen; +} Tunnel; + +DEFINE_NETDEV_CAST(IPIP, Tunnel); +DEFINE_NETDEV_CAST(GRE, Tunnel); +DEFINE_NETDEV_CAST(GRETAP, Tunnel); +DEFINE_NETDEV_CAST(IP6GRE, Tunnel); +DEFINE_NETDEV_CAST(IP6GRETAP, Tunnel); +DEFINE_NETDEV_CAST(SIT, Tunnel); +DEFINE_NETDEV_CAST(VTI, Tunnel); +DEFINE_NETDEV_CAST(VTI6, Tunnel); +DEFINE_NETDEV_CAST(IP6TNL, Tunnel); +DEFINE_NETDEV_CAST(ERSPAN, Tunnel); +extern const NetDevVTable ipip_vtable; +extern const NetDevVTable sit_vtable; +extern const NetDevVTable vti_vtable; +extern const NetDevVTable vti6_vtable; +extern const NetDevVTable gre_vtable; +extern const NetDevVTable gretap_vtable; +extern const NetDevVTable ip6gre_vtable; +extern const NetDevVTable ip6gretap_vtable; +extern const NetDevVTable ip6tnl_vtable; +extern const NetDevVTable erspan_vtable; + +const char *ip6tnl_mode_to_string(Ip6TnlMode d) _const_; +Ip6TnlMode ip6tnl_mode_from_string(const char *d) _pure_; + +CONFIG_PARSER_PROTOTYPE(config_parse_ip6tnl_mode); +CONFIG_PARSER_PROTOTYPE(config_parse_tunnel_address); +CONFIG_PARSER_PROTOTYPE(config_parse_ipv6_flowlabel); +CONFIG_PARSER_PROTOTYPE(config_parse_encap_limit); +CONFIG_PARSER_PROTOTYPE(config_parse_tunnel_key); +CONFIG_PARSER_PROTOTYPE(config_parse_6rd_prefix); diff --git a/src/network/netdev/tuntap.c b/src/network/netdev/tuntap.c new file mode 100644 index 0000000..951138d --- /dev/null +++ b/src/network/netdev/tuntap.c @@ -0,0 +1,161 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ + +#include <errno.h> +#include <fcntl.h> +#include <linux/if_tun.h> +#include <net/if.h> +#include <netinet/if_ether.h> +#include <sys/ioctl.h> +#include <sys/stat.h> +#include <sys/types.h> + +#include "alloc-util.h" +#include "fd-util.h" +#include "netdev/tuntap.h" +#include "user-util.h" + +#define TUN_DEV "/dev/net/tun" + +static int netdev_fill_tuntap_message(NetDev *netdev, struct ifreq *ifr) { + TunTap *t; + + assert(netdev); + assert(netdev->ifname); + assert(ifr); + + if (netdev->kind == NETDEV_KIND_TAP) { + t = TAP(netdev); + ifr->ifr_flags |= IFF_TAP; + } else { + t = TUN(netdev); + ifr->ifr_flags |= IFF_TUN; + } + + if (!t->packet_info) + ifr->ifr_flags |= IFF_NO_PI; + + if (t->one_queue) + ifr->ifr_flags |= IFF_ONE_QUEUE; + + if (t->multi_queue) + ifr->ifr_flags |= IFF_MULTI_QUEUE; + + if (t->vnet_hdr) + ifr->ifr_flags |= IFF_VNET_HDR; + + strncpy(ifr->ifr_name, netdev->ifname, IFNAMSIZ-1); + + return 0; +} + +static int netdev_tuntap_add(NetDev *netdev, struct ifreq *ifr) { + _cleanup_close_ int fd; + TunTap *t = NULL; + const char *user; + const char *group; + uid_t uid; + gid_t gid; + int r; + + assert(netdev); + assert(ifr); + + fd = open(TUN_DEV, O_RDWR|O_CLOEXEC); + if (fd < 0) + return log_netdev_error_errno(netdev, -errno, "Failed to open tun dev: %m"); + + if (ioctl(fd, TUNSETIFF, ifr) < 0) + return log_netdev_error_errno(netdev, -errno, "TUNSETIFF failed on tun dev: %m"); + + if (netdev->kind == NETDEV_KIND_TAP) + t = TAP(netdev); + else + t = TUN(netdev); + + assert(t); + + if (t->user_name) { + user = t->user_name; + + r = get_user_creds(&user, &uid, NULL, NULL, NULL, USER_CREDS_ALLOW_MISSING); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Cannot resolve user name %s: %m", t->user_name); + + if (ioctl(fd, TUNSETOWNER, uid) < 0) + return log_netdev_error_errno(netdev, -errno, "TUNSETOWNER failed on tun dev: %m"); + } + + if (t->group_name) { + group = t->group_name; + + r = get_group_creds(&group, &gid, USER_CREDS_ALLOW_MISSING); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Cannot resolve group name %s: %m", t->group_name); + + if (ioctl(fd, TUNSETGROUP, gid) < 0) + return log_netdev_error_errno(netdev, -errno, "TUNSETGROUP failed on tun dev: %m"); + + } + + if (ioctl(fd, TUNSETPERSIST, 1) < 0) + return log_netdev_error_errno(netdev, -errno, "TUNSETPERSIST failed on tun dev: %m"); + + return 0; +} + +static int netdev_create_tuntap(NetDev *netdev) { + struct ifreq ifr = {}; + int r; + + r = netdev_fill_tuntap_message(netdev, &ifr); + if (r < 0) + return r; + + return netdev_tuntap_add(netdev, &ifr); +} + +static void tuntap_done(NetDev *netdev) { + TunTap *t = NULL; + + assert(netdev); + + if (netdev->kind == NETDEV_KIND_TUN) + t = TUN(netdev); + else + t = TAP(netdev); + + assert(t); + + t->user_name = mfree(t->user_name); + t->group_name = mfree(t->group_name); +} + +static int tuntap_verify(NetDev *netdev, const char *filename) { + assert(netdev); + + if (netdev->mtu != 0) + log_netdev_warning(netdev, "MTU configured for %s, ignoring", netdev_kind_to_string(netdev->kind)); + + if (netdev->mac) + log_netdev_warning(netdev, "MAC configured for %s, ignoring", netdev_kind_to_string(netdev->kind)); + + return 0; +} + +const NetDevVTable tun_vtable = { + .object_size = sizeof(TunTap), + .sections = "Match\0NetDev\0Tun\0", + .config_verify = tuntap_verify, + .done = tuntap_done, + .create = netdev_create_tuntap, + .create_type = NETDEV_CREATE_INDEPENDENT, +}; + +const NetDevVTable tap_vtable = { + .object_size = sizeof(TunTap), + .sections = "Match\0NetDev\0Tap\0", + .config_verify = tuntap_verify, + .done = tuntap_done, + .create = netdev_create_tuntap, + .create_type = NETDEV_CREATE_INDEPENDENT, +}; diff --git a/src/network/netdev/tuntap.h b/src/network/netdev/tuntap.h new file mode 100644 index 0000000..9d9f992 --- /dev/null +++ b/src/network/netdev/tuntap.h @@ -0,0 +1,22 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ +#pragma once + +typedef struct TunTap TunTap; + +#include "netdev/netdev.h" + +struct TunTap { + NetDev meta; + + char *user_name; + char *group_name; + bool one_queue; + bool multi_queue; + bool packet_info; + bool vnet_hdr; +}; + +DEFINE_NETDEV_CAST(TUN, TunTap); +DEFINE_NETDEV_CAST(TAP, TunTap); +extern const NetDevVTable tun_vtable; +extern const NetDevVTable tap_vtable; diff --git a/src/network/netdev/vcan.c b/src/network/netdev/vcan.c new file mode 100644 index 0000000..574d1ca --- /dev/null +++ b/src/network/netdev/vcan.c @@ -0,0 +1,9 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ + +#include "netdev/vcan.h" + +const NetDevVTable vcan_vtable = { + .object_size = sizeof(VCan), + .sections = "Match\0NetDev\0", + .create_type = NETDEV_CREATE_INDEPENDENT, +}; diff --git a/src/network/netdev/vcan.h b/src/network/netdev/vcan.h new file mode 100644 index 0000000..6f62686 --- /dev/null +++ b/src/network/netdev/vcan.h @@ -0,0 +1,16 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ +#pragma once + +typedef struct VCan VCan; + +#include <linux/can/netlink.h> + +#include "netdev/netdev.h" + +struct VCan { + NetDev meta; +}; + +DEFINE_NETDEV_CAST(VCAN, VCan); + +extern const NetDevVTable vcan_vtable; diff --git a/src/network/netdev/veth.c b/src/network/netdev/veth.c new file mode 100644 index 0000000..3ad95ad --- /dev/null +++ b/src/network/netdev/veth.c @@ -0,0 +1,95 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ + +#include <errno.h> +#include <linux/veth.h> +#include <net/if.h> + +#include "sd-netlink.h" + +#include "netdev/veth.h" + +static int netdev_veth_fill_message_create(NetDev *netdev, Link *link, sd_netlink_message *m) { + Veth *v; + int r; + + assert(netdev); + assert(!link); + assert(m); + + v = VETH(netdev); + + assert(v); + + r = sd_netlink_message_open_container(m, VETH_INFO_PEER); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append VETH_INFO_PEER attribute: %m"); + + if (v->ifname_peer) { + r = sd_netlink_message_append_string(m, IFLA_IFNAME, v->ifname_peer); + if (r < 0) + return log_error_errno(r, "Failed to add netlink interface name: %m"); + } + + if (v->mac_peer) { + r = sd_netlink_message_append_ether_addr(m, IFLA_ADDRESS, v->mac_peer); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_ADDRESS attribute: %m"); + } + + r = sd_netlink_message_close_container(m); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_INFO_DATA attribute: %m"); + + return r; +} + +static int netdev_veth_verify(NetDev *netdev, const char *filename) { + Veth *v; + int r; + + assert(netdev); + assert(filename); + + v = VETH(netdev); + + assert(v); + + if (!v->ifname_peer) { + log_warning("Veth NetDev without peer name configured in %s. Ignoring", + filename); + return -EINVAL; + } + + if (!v->mac_peer) { + r = netdev_get_mac(v->ifname_peer, &v->mac_peer); + if (r < 0) { + log_warning("Failed to generate predictable MAC address for %s. Ignoring", + v->ifname_peer); + return -EINVAL; + } + } + + return 0; +} + +static void veth_done(NetDev *n) { + Veth *v; + + assert(n); + + v = VETH(n); + + assert(v); + + free(v->ifname_peer); + free(v->mac_peer); +} + +const NetDevVTable veth_vtable = { + .object_size = sizeof(Veth), + .sections = "Match\0NetDev\0Peer\0", + .done = veth_done, + .fill_message_create = netdev_veth_fill_message_create, + .create_type = NETDEV_CREATE_INDEPENDENT, + .config_verify = netdev_veth_verify, +}; diff --git a/src/network/netdev/veth.h b/src/network/netdev/veth.h new file mode 100644 index 0000000..0bb9947 --- /dev/null +++ b/src/network/netdev/veth.h @@ -0,0 +1,16 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ +#pragma once + +typedef struct Veth Veth; + +#include "netdev/netdev.h" + +struct Veth { + NetDev meta; + + char *ifname_peer; + struct ether_addr *mac_peer; +}; + +DEFINE_NETDEV_CAST(VETH, Veth); +extern const NetDevVTable veth_vtable; diff --git a/src/network/netdev/vlan.c b/src/network/netdev/vlan.c new file mode 100644 index 0000000..dd548b3 --- /dev/null +++ b/src/network/netdev/vlan.c @@ -0,0 +1,92 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ + +#include <errno.h> +#include <linux/if_vlan.h> +#include <net/if.h> + +#include "netdev/vlan.h" +#include "vlan-util.h" + +static int netdev_vlan_fill_message_create(NetDev *netdev, Link *link, sd_netlink_message *req) { + struct ifla_vlan_flags flags = {}; + VLan *v; + int r; + + assert(netdev); + assert(link); + assert(req); + + v = VLAN(netdev); + + assert(v); + + r = sd_netlink_message_append_u16(req, IFLA_VLAN_ID, v->id); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VLAN_ID attribute: %m"); + + if (v->gvrp != -1) { + flags.mask |= VLAN_FLAG_GVRP; + SET_FLAG(flags.flags, VLAN_FLAG_GVRP, v->gvrp); + } + + if (v->mvrp != -1) { + flags.mask |= VLAN_FLAG_MVRP; + SET_FLAG(flags.flags, VLAN_FLAG_MVRP, v->mvrp); + } + + if (v->reorder_hdr != -1) { + flags.mask |= VLAN_FLAG_REORDER_HDR; + SET_FLAG(flags.flags, VLAN_FLAG_REORDER_HDR, v->reorder_hdr); + } + + if (v->loose_binding != -1) { + flags.mask |= VLAN_FLAG_LOOSE_BINDING; + SET_FLAG(flags.flags, VLAN_FLAG_LOOSE_BINDING, v->loose_binding); + } + + r = sd_netlink_message_append_data(req, IFLA_VLAN_FLAGS, &flags, sizeof(struct ifla_vlan_flags)); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VLAN_FLAGS attribute: %m"); + + return 0; +} + +static int netdev_vlan_verify(NetDev *netdev, const char *filename) { + VLan *v; + + assert(netdev); + assert(filename); + + v = VLAN(netdev); + + assert(v); + + if (v->id == VLANID_INVALID) { + log_warning("VLAN without valid Id (%"PRIu16") configured in %s.", v->id, filename); + return -EINVAL; + } + + return 0; +} + +static void vlan_init(NetDev *netdev) { + VLan *v = VLAN(netdev); + + assert(netdev); + assert(v); + + v->id = VLANID_INVALID; + v->gvrp = -1; + v->mvrp = -1; + v->loose_binding = -1; + v->reorder_hdr = -1; +} + +const NetDevVTable vlan_vtable = { + .object_size = sizeof(VLan), + .init = vlan_init, + .sections = "Match\0NetDev\0VLAN\0", + .fill_message_create = netdev_vlan_fill_message_create, + .create_type = NETDEV_CREATE_STACKED, + .config_verify = netdev_vlan_verify, +}; diff --git a/src/network/netdev/vlan.h b/src/network/netdev/vlan.h new file mode 100644 index 0000000..b815e03 --- /dev/null +++ b/src/network/netdev/vlan.h @@ -0,0 +1,20 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ +#pragma once + +typedef struct VLan VLan; + +#include "netdev/netdev.h" + +struct VLan { + NetDev meta; + + uint16_t id; + + int gvrp; + int mvrp; + int loose_binding; + int reorder_hdr; +}; + +DEFINE_NETDEV_CAST(VLAN, VLan); +extern const NetDevVTable vlan_vtable; diff --git a/src/network/netdev/vrf.c b/src/network/netdev/vrf.c new file mode 100644 index 0000000..5efab53 --- /dev/null +++ b/src/network/netdev/vrf.c @@ -0,0 +1,33 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ + +#include <net/if.h> + +#include "sd-netlink.h" +#include "missing.h" +#include "netdev/vrf.h" + +static int netdev_vrf_fill_message_create(NetDev *netdev, Link *link, sd_netlink_message *m) { + Vrf *v; + int r; + + assert(netdev); + assert(!link); + assert(m); + + v = VRF(netdev); + + assert(v); + + r = sd_netlink_message_append_u32(m, IFLA_VRF_TABLE, v->table); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IPLA_VRF_TABLE attribute: %m"); + + return r; +} + +const NetDevVTable vrf_vtable = { + .object_size = sizeof(Vrf), + .sections = "Match\0NetDev\0VRF\0", + .fill_message_create = netdev_vrf_fill_message_create, + .create_type = NETDEV_CREATE_MASTER, +}; diff --git a/src/network/netdev/vrf.h b/src/network/netdev/vrf.h new file mode 100644 index 0000000..05b3937 --- /dev/null +++ b/src/network/netdev/vrf.h @@ -0,0 +1,15 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ +#pragma once + +typedef struct Vrf Vrf; + +#include "netdev/netdev.h" + +struct Vrf { + NetDev meta; + + uint32_t table; +}; + +DEFINE_NETDEV_CAST(VRF, Vrf); +extern const NetDevVTable vrf_vtable; diff --git a/src/network/netdev/vxcan.c b/src/network/netdev/vxcan.c new file mode 100644 index 0000000..e8ea70a --- /dev/null +++ b/src/network/netdev/vxcan.c @@ -0,0 +1,76 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ + +#if HAVE_LINUX_CAN_VXCAN_H +#include <linux/can/vxcan.h> +#endif + +#include "missing.h" +#include "netdev/vxcan.h" + +static int netdev_vxcan_fill_message_create(NetDev *netdev, Link *link, sd_netlink_message *m) { + VxCan *v; + int r; + + assert(netdev); + assert(!link); + assert(m); + + v = VXCAN(netdev); + + assert(v); + + r = sd_netlink_message_open_container(m, VXCAN_INFO_PEER); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append VXCAN_INFO_PEER attribute: %m"); + + if (v->ifname_peer) { + r = sd_netlink_message_append_string(m, IFLA_IFNAME, v->ifname_peer); + if (r < 0) + return log_error_errno(r, "Failed to add vxcan netlink interface peer name: %m"); + } + + r = sd_netlink_message_close_container(m); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append VXCAN_INFO_PEER attribute: %m"); + + return r; +} + +static int netdev_vxcan_verify(NetDev *netdev, const char *filename) { + VxCan *v; + + assert(netdev); + assert(filename); + + v = VXCAN(netdev); + + assert(v); + + if (!v->ifname_peer) { + log_warning("VxCan NetDev without peer name configured in %s. Ignoring", filename); + return -EINVAL; + } + + return 0; +} + +static void vxcan_done(NetDev *n) { + VxCan *v; + + assert(n); + + v = VXCAN(n); + + assert(v); + + free(v->ifname_peer); +} + +const NetDevVTable vxcan_vtable = { + .object_size = sizeof(VxCan), + .sections = "Match\0NetDev\0VXCAN\0", + .done = vxcan_done, + .fill_message_create = netdev_vxcan_fill_message_create, + .create_type = NETDEV_CREATE_INDEPENDENT, + .config_verify = netdev_vxcan_verify, +}; diff --git a/src/network/netdev/vxcan.h b/src/network/netdev/vxcan.h new file mode 100644 index 0000000..b5de197 --- /dev/null +++ b/src/network/netdev/vxcan.h @@ -0,0 +1,16 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ +#pragma once + +typedef struct VxCan VxCan; + +#include "netdev/netdev.h" + +struct VxCan { + NetDev meta; + + char *ifname_peer; +}; + +DEFINE_NETDEV_CAST(VXCAN, VxCan); + +extern const NetDevVTable vxcan_vtable; diff --git a/src/network/netdev/vxlan.c b/src/network/netdev/vxlan.c new file mode 100644 index 0000000..4cb2eca --- /dev/null +++ b/src/network/netdev/vxlan.c @@ -0,0 +1,311 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ + +#include <net/if.h> + +#include "sd-netlink.h" + +#include "conf-parser.h" +#include "alloc-util.h" +#include "extract-word.h" +#include "string-util.h" +#include "strv.h" +#include "parse-util.h" +#include "missing.h" + +#include "networkd-link.h" +#include "netdev/vxlan.h" + +static int netdev_vxlan_fill_message_create(NetDev *netdev, Link *link, sd_netlink_message *m) { + VxLan *v; + int r; + + assert(netdev); + assert(link); + assert(m); + + v = VXLAN(netdev); + + assert(v); + + if (v->id <= VXLAN_VID_MAX) { + r = sd_netlink_message_append_u32(m, IFLA_VXLAN_ID, v->id); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VXLAN_ID attribute: %m"); + } + + if (!in_addr_is_null(v->remote_family, &v->remote)) { + + if (v->remote_family == AF_INET) + r = sd_netlink_message_append_in_addr(m, IFLA_VXLAN_GROUP, &v->remote.in); + else + r = sd_netlink_message_append_in6_addr(m, IFLA_VXLAN_GROUP6, &v->remote.in6); + + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VXLAN_GROUP attribute: %m"); + } + + if (!in_addr_is_null(v->local_family, &v->local)) { + + if (v->local_family == AF_INET) + r = sd_netlink_message_append_in_addr(m, IFLA_VXLAN_LOCAL, &v->local.in); + else + r = sd_netlink_message_append_in6_addr(m, IFLA_VXLAN_LOCAL6, &v->local.in6); + + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VXLAN_LOCAL attribute: %m"); + } + + r = sd_netlink_message_append_u32(m, IFLA_VXLAN_LINK, link->ifindex); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VXLAN_LINK attribute: %m"); + + if (v->ttl != 0) { + r = sd_netlink_message_append_u8(m, IFLA_VXLAN_TTL, v->ttl); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VXLAN_TTL attribute: %m"); + } + + if (v->tos != 0) { + r = sd_netlink_message_append_u8(m, IFLA_VXLAN_TOS, v->tos); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VXLAN_TOS attribute: %m"); + } + + r = sd_netlink_message_append_u8(m, IFLA_VXLAN_LEARNING, v->learning); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VXLAN_LEARNING attribute: %m"); + + r = sd_netlink_message_append_u8(m, IFLA_VXLAN_RSC, v->route_short_circuit); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VXLAN_RSC attribute: %m"); + + r = sd_netlink_message_append_u8(m, IFLA_VXLAN_PROXY, v->arp_proxy); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VXLAN_PROXY attribute: %m"); + + r = sd_netlink_message_append_u8(m, IFLA_VXLAN_L2MISS, v->l2miss); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VXLAN_L2MISS attribute: %m"); + + r = sd_netlink_message_append_u8(m, IFLA_VXLAN_L3MISS, v->l3miss); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VXLAN_L3MISS attribute: %m"); + + if (v->fdb_ageing != 0) { + r = sd_netlink_message_append_u32(m, IFLA_VXLAN_AGEING, v->fdb_ageing / USEC_PER_SEC); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VXLAN_AGEING attribute: %m"); + } + + if (v->max_fdb != 0) { + r = sd_netlink_message_append_u32(m, IFLA_VXLAN_LIMIT, v->max_fdb); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VXLAN_LIMIT attribute: %m"); + } + + r = sd_netlink_message_append_u8(m, IFLA_VXLAN_UDP_CSUM, v->udpcsum); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VXLAN_UDP_CSUM attribute: %m"); + + r = sd_netlink_message_append_u8(m, IFLA_VXLAN_UDP_ZERO_CSUM6_TX, v->udp6zerocsumtx); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VXLAN_UDP_ZERO_CSUM6_TX attribute: %m"); + + r = sd_netlink_message_append_u8(m, IFLA_VXLAN_UDP_ZERO_CSUM6_RX, v->udp6zerocsumrx); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VXLAN_UDP_ZERO_CSUM6_RX attribute: %m"); + + r = sd_netlink_message_append_u8(m, IFLA_VXLAN_REMCSUM_TX, v->remote_csum_tx); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VXLAN_REMCSUM_TX attribute: %m"); + + r = sd_netlink_message_append_u8(m, IFLA_VXLAN_REMCSUM_RX, v->remote_csum_rx); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VXLAN_REMCSUM_RX attribute: %m"); + + r = sd_netlink_message_append_u16(m, IFLA_VXLAN_PORT, htobe16(v->dest_port)); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VXLAN_PORT attribute: %m"); + + if (v->port_range.low != 0 || v->port_range.high != 0) { + struct ifla_vxlan_port_range port_range; + + port_range.low = htobe16(v->port_range.low); + port_range.high = htobe16(v->port_range.high); + + r = sd_netlink_message_append_data(m, IFLA_VXLAN_PORT_RANGE, &port_range, sizeof(port_range)); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VXLAN_PORT_RANGE attribute: %m"); + } + + r = sd_netlink_message_append_u32(m, IFLA_VXLAN_LABEL, htobe32(v->flow_label)); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VXLAN_LABEL attribute: %m"); + + if (v->group_policy) { + r = sd_netlink_message_append_flag(m, IFLA_VXLAN_GBP); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append IFLA_VXLAN_GBP attribute: %m"); + } + + return r; +} + +int config_parse_vxlan_address(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + VxLan *v = userdata; + union in_addr_union *addr = data, buffer; + int r, f; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + r = in_addr_from_string_auto(rvalue, &f, &buffer); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "vxlan '%s' address is invalid, ignoring assignment: %s", lvalue, rvalue); + return 0; + } + + r = in_addr_is_multicast(f, &buffer); + + if (streq(lvalue, "Group")) { + if (r <= 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "vxlan %s invalid multicast address, ignoring assignment: %s", lvalue, rvalue); + return 0; + } + + v->remote_family = f; + } else { + if (r > 0) { + log_syntax(unit, LOG_ERR, filename, line, 0, "vxlan %s cannot be a multicast address, ignoring assignment: %s", lvalue, rvalue); + return 0; + } + + if (streq(lvalue, "Remote")) + v->remote_family = f; + else + v->local_family = f; + } + + *addr = buffer; + + return 0; +} + +int config_parse_port_range(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + VxLan *v = userdata; + uint16_t low, high; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + r = parse_ip_port_range(rvalue, &low, &high); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, + "Failed to parse VXLAN port range '%s'. Port should be greater than 0 and less than 65535.", rvalue); + return 0; + } + + v->port_range.low = low; + v->port_range.high = high; + + return 0; +} + +int config_parse_flow_label(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + VxLan *v = userdata; + unsigned f; + int r; + + assert(filename); + assert(lvalue); + assert(rvalue); + assert(data); + + r = safe_atou(rvalue, &f); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to parse VXLAN flow label '%s'.", rvalue); + return 0; + } + + if (f & ~VXLAN_FLOW_LABEL_MAX_MASK) { + log_syntax(unit, LOG_ERR, filename, line, r, + "VXLAN flow label '%s' not valid. Flow label range should be [0-1048575].", rvalue); + return 0; + } + + v->flow_label = f; + + return 0; +} + +static int netdev_vxlan_verify(NetDev *netdev, const char *filename) { + VxLan *v = VXLAN(netdev); + + assert(netdev); + assert(v); + assert(filename); + + if (v->id > VXLAN_VID_MAX) { + log_warning("VXLAN without valid Id configured in %s. Ignoring", filename); + return -EINVAL; + } + + return 0; +} + +static void vxlan_init(NetDev *netdev) { + VxLan *v; + + assert(netdev); + + v = VXLAN(netdev); + + assert(v); + + v->id = VXLAN_VID_MAX + 1; + v->learning = true; + v->udpcsum = false; + v->udp6zerocsumtx = false; + v->udp6zerocsumrx = false; +} + +const NetDevVTable vxlan_vtable = { + .object_size = sizeof(VxLan), + .init = vxlan_init, + .sections = "Match\0NetDev\0VXLAN\0", + .fill_message_create = netdev_vxlan_fill_message_create, + .create_type = NETDEV_CREATE_STACKED, + .config_verify = netdev_vxlan_verify, +}; diff --git a/src/network/netdev/vxlan.h b/src/network/netdev/vxlan.h new file mode 100644 index 0000000..3b273e9 --- /dev/null +++ b/src/network/netdev/vxlan.h @@ -0,0 +1,52 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ +#pragma once + +typedef struct VxLan VxLan; + +#include "in-addr-util.h" +#include "netdev/netdev.h" + +#define VXLAN_VID_MAX (1u << 24) - 1 +#define VXLAN_FLOW_LABEL_MAX_MASK 0xFFFFFU + +struct VxLan { + NetDev meta; + + uint64_t id; + + int remote_family; + int local_family; + + union in_addr_union remote; + union in_addr_union local; + + unsigned tos; + unsigned ttl; + unsigned max_fdb; + unsigned flow_label; + + uint16_t dest_port; + + usec_t fdb_ageing; + + bool learning; + bool arp_proxy; + bool route_short_circuit; + bool l2miss; + bool l3miss; + bool udpcsum; + bool udp6zerocsumtx; + bool udp6zerocsumrx; + bool remote_csum_tx; + bool remote_csum_rx; + bool group_policy; + + struct ifla_vxlan_port_range port_range; +}; + +DEFINE_NETDEV_CAST(VXLAN, VxLan); +extern const NetDevVTable vxlan_vtable; + +CONFIG_PARSER_PROTOTYPE(config_parse_vxlan_address); +CONFIG_PARSER_PROTOTYPE(config_parse_port_range); +CONFIG_PARSER_PROTOTYPE(config_parse_flow_label); diff --git a/src/network/netdev/wireguard.c b/src/network/netdev/wireguard.c new file mode 100644 index 0000000..0c0b16d --- /dev/null +++ b/src/network/netdev/wireguard.c @@ -0,0 +1,778 @@ +/* SPDX-License-Identifier: LGPL-2.1+ */ +/*** + Copyright © 2015-2017 Jason A. Donenfeld <Jason@zx2c4.com>. All Rights Reserved. +***/ + +#include <sys/ioctl.h> +#include <net/if.h> + +#include "sd-resolve.h" + +#include "alloc-util.h" +#include "fd-util.h" +#include "hexdecoct.h" +#include "networkd-link.h" +#include "networkd-manager.h" +#include "networkd-util.h" +#include "parse-util.h" +#include "resolve-private.h" +#include "string-util.h" +#include "strv.h" +#include "wireguard.h" +#include "wireguard-netlink.h" + +static void resolve_endpoints(NetDev *netdev); + +static WireguardPeer *wireguard_peer_new(Wireguard *w, unsigned section) { + WireguardPeer *peer; + + assert(w); + + if (w->last_peer_section == section && w->peers) + return w->peers; + + peer = new(WireguardPeer, 1); + if (!peer) + return NULL; + + *peer = (WireguardPeer) { + .flags = WGPEER_F_REPLACE_ALLOWEDIPS, + }; + + LIST_PREPEND(peers, w->peers, peer); + w->last_peer_section = section; + + return peer; +} + +static int wireguard_set_ipmask_one(NetDev *netdev, sd_netlink_message *message, const WireguardIPmask *mask, uint16_t index) { + int r; + + assert(message); + assert(mask); + assert(index > 0); + + /* This returns 1 on success, 0 on recoverable error, and negative errno on failure. */ + + r = sd_netlink_message_open_array(message, index); + if (r < 0) + return 0; + + r = sd_netlink_message_append_u16(message, WGALLOWEDIP_A_FAMILY, mask->family); + if (r < 0) + goto cancel; + + if (mask->family == AF_INET) + r = sd_netlink_message_append_in_addr(message, WGALLOWEDIP_A_IPADDR, &mask->ip.in); + else if (mask->family == AF_INET6) + r = sd_netlink_message_append_in6_addr(message, WGALLOWEDIP_A_IPADDR, &mask->ip.in6); + if (r < 0) + goto cancel; + + r = sd_netlink_message_append_u8(message, WGALLOWEDIP_A_CIDR_MASK, mask->cidr); + if (r < 0) + goto cancel; + + r = sd_netlink_message_close_container(message); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not add wireguard allowed ip: %m"); + + return 1; + +cancel: + r = sd_netlink_message_cancel_array(message); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not cancel wireguard allowed ip message attribute: %m"); + + return 0; +} + +static int wireguard_set_peer_one(NetDev *netdev, sd_netlink_message *message, const WireguardPeer *peer, uint16_t index, WireguardIPmask **mask_start) { + WireguardIPmask *mask, *start; + uint16_t j = 0; + int r; + + assert(message); + assert(peer); + assert(index > 0); + assert(mask_start); + + /* This returns 1 on success, 0 on recoverable error, and negative errno on failure. */ + + start = *mask_start ?: peer->ipmasks; + + r = sd_netlink_message_open_array(message, index); + if (r < 0) + return 0; + + r = sd_netlink_message_append_data(message, WGPEER_A_PUBLIC_KEY, &peer->public_key, sizeof(peer->public_key)); + if (r < 0) + goto cancel; + + if (!*mask_start) { + r = sd_netlink_message_append_data(message, WGPEER_A_PRESHARED_KEY, &peer->preshared_key, WG_KEY_LEN); + if (r < 0) + goto cancel; + + r = sd_netlink_message_append_u32(message, WGPEER_A_FLAGS, peer->flags); + if (r < 0) + goto cancel; + + r = sd_netlink_message_append_u16(message, WGPEER_A_PERSISTENT_KEEPALIVE_INTERVAL, peer->persistent_keepalive_interval); + if (r < 0) + goto cancel; + + if (peer->endpoint.sa.sa_family == AF_INET) + r = sd_netlink_message_append_sockaddr_in(message, WGPEER_A_ENDPOINT, &peer->endpoint.in); + else if (peer->endpoint.sa.sa_family == AF_INET6) + r = sd_netlink_message_append_sockaddr_in6(message, WGPEER_A_ENDPOINT, &peer->endpoint.in6); + if (r < 0) + goto cancel; + } + + r = sd_netlink_message_open_container(message, WGPEER_A_ALLOWEDIPS); + if (r < 0) + goto cancel; + + LIST_FOREACH(ipmasks, mask, start) { + r = wireguard_set_ipmask_one(netdev, message, mask, ++j); + if (r < 0) + return r; + if (r == 0) + break; + } + + r = sd_netlink_message_close_container(message); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not add wireguard allowed ip: %m"); + + r = sd_netlink_message_close_container(message); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not add wireguard peer: %m"); + + *mask_start = mask; /* Start next cycle from this mask. */ + return !mask; + +cancel: + r = sd_netlink_message_cancel_array(message); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not cancel wireguard peers: %m"); + + return 0; +} + +static int wireguard_set_interface(NetDev *netdev) { + _cleanup_(sd_netlink_message_unrefp) sd_netlink_message *message = NULL; + WireguardIPmask *mask_start = NULL; + WireguardPeer *peer, *peer_start; + uint32_t serial; + Wireguard *w; + int r; + + assert(netdev); + w = WIREGUARD(netdev); + assert(w); + + for (peer_start = w->peers; peer_start; ) { + uint16_t i = 0; + + message = sd_netlink_message_unref(message); + + r = sd_genl_message_new(netdev->manager->genl, SD_GENL_WIREGUARD, WG_CMD_SET_DEVICE, &message); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Failed to allocate generic netlink message: %m"); + + r = sd_netlink_message_append_string(message, WGDEVICE_A_IFNAME, netdev->ifname); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append wireguard interface name: %m"); + + if (peer_start == w->peers) { + r = sd_netlink_message_append_data(message, WGDEVICE_A_PRIVATE_KEY, &w->private_key, WG_KEY_LEN); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append wireguard private key: %m"); + + r = sd_netlink_message_append_u16(message, WGDEVICE_A_LISTEN_PORT, w->port); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append wireguard port: %m"); + + r = sd_netlink_message_append_u32(message, WGDEVICE_A_FWMARK, w->fwmark); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append wireguard fwmark: %m"); + + r = sd_netlink_message_append_u32(message, WGDEVICE_A_FLAGS, w->flags); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append wireguard flags: %m"); + } + + r = sd_netlink_message_open_container(message, WGDEVICE_A_PEERS); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not append wireguard peer attributes: %m"); + + LIST_FOREACH(peers, peer, peer_start) { + r = wireguard_set_peer_one(netdev, message, peer, ++i, &mask_start); + if (r < 0) + return r; + if (r == 0) + break; + } + peer_start = peer; /* Start next cycle from this peer. */ + + r = sd_netlink_message_close_container(message); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not close wireguard container: %m"); + + r = sd_netlink_send(netdev->manager->genl, message, &serial); + if (r < 0) + return log_netdev_error_errno(netdev, r, "Could not set wireguard device: %m"); + } + + return 0; +} + +static WireguardEndpoint* wireguard_endpoint_free(WireguardEndpoint *e) { + if (!e) + return NULL; + e->host = mfree(e->host); + e->port = mfree(e->port); + return mfree(e); +} + +static void wireguard_endpoint_destroy_callback(WireguardEndpoint *e) { + assert(e); + assert(e->netdev); + + netdev_unref(e->netdev); + wireguard_endpoint_free(e); +} + +DEFINE_TRIVIAL_CLEANUP_FUNC(WireguardEndpoint*, wireguard_endpoint_free); + +static int on_resolve_retry(sd_event_source *s, usec_t usec, void *userdata) { + NetDev *netdev = userdata; + Wireguard *w; + + assert(netdev); + w = WIREGUARD(netdev); + assert(w); + + if (!netdev_is_managed(netdev)) + return 0; + + assert(!w->unresolved_endpoints); + w->unresolved_endpoints = TAKE_PTR(w->failed_endpoints); + + resolve_endpoints(netdev); + + return 0; +} + +/* + * Given the number of retries this function will return will an exponential + * increasing time in milliseconds to wait starting at 200ms and capped at 25 seconds. + */ +static int exponential_backoff_milliseconds(unsigned n_retries) { + return (2 << MAX(n_retries, 7U)) * 100 * USEC_PER_MSEC; +} + +static int wireguard_resolve_handler(sd_resolve_query *q, + int ret, + const struct addrinfo *ai, + WireguardEndpoint *e) { + _cleanup_(netdev_unrefp) NetDev *netdev_will_unrefed = NULL; + NetDev *netdev; + Wireguard *w; + int r; + + assert(e); + assert(e->netdev); + + netdev = e->netdev; + w = WIREGUARD(netdev); + assert(w); + + if (!netdev_is_managed(netdev)) + return 0; + + if (ret != 0) { + log_netdev_error(netdev, "Failed to resolve host '%s:%s': %s", e->host, e->port, gai_strerror(ret)); + LIST_PREPEND(endpoints, w->failed_endpoints, e); + (void) sd_resolve_query_set_destroy_callback(q, NULL); /* Avoid freeing endpoint by destroy callback. */ + netdev_will_unrefed = netdev; /* But netdev needs to be unrefed. */ + } else if ((ai->ai_family == AF_INET && ai->ai_addrlen == sizeof(struct sockaddr_in)) || + (ai->ai_family == AF_INET6 && ai->ai_addrlen == sizeof(struct sockaddr_in6))) + memcpy(&e->peer->endpoint, ai->ai_addr, ai->ai_addrlen); + else + log_netdev_error(netdev, "Neither IPv4 nor IPv6 address found for peer endpoint: %s:%s", e->host, e->port); + + if (w->unresolved_endpoints) { + resolve_endpoints(netdev); + return 0; + } + + (void) wireguard_set_interface(netdev); + if (w->failed_endpoints) { + _cleanup_(sd_event_source_unrefp) sd_event_source *s = NULL; + + w->n_retries++; + r = sd_event_add_time(netdev->manager->event, + &s, + CLOCK_MONOTONIC, + now(CLOCK_MONOTONIC) + exponential_backoff_milliseconds(w->n_retries), + 0, + on_resolve_retry, + netdev); + if (r < 0) { + log_netdev_warning_errno(netdev, r, "Could not arm resolve retry handler: %m"); + return 0; + } + + r = sd_event_source_set_destroy_callback(s, (sd_event_destroy_t) netdev_destroy_callback); + if (r < 0) { + log_netdev_warning_errno(netdev, r, "Failed to set destroy callback to event source: %m"); + return 0; + } + + (void) sd_event_source_set_floating(s, true); + netdev_ref(netdev); + } + + return 0; +} + +static void resolve_endpoints(NetDev *netdev) { + static const struct addrinfo hints = { + .ai_family = AF_UNSPEC, + .ai_socktype = SOCK_DGRAM, + .ai_protocol = IPPROTO_UDP + }; + WireguardEndpoint *endpoint; + Wireguard *w; + int r = 0; + + assert(netdev); + w = WIREGUARD(netdev); + assert(w); + + LIST_FOREACH(endpoints, endpoint, w->unresolved_endpoints) { + r = resolve_getaddrinfo(netdev->manager->resolve, + NULL, + endpoint->host, + endpoint->port, + &hints, + wireguard_resolve_handler, + wireguard_endpoint_destroy_callback, + endpoint); + + if (r == -ENOBUFS) + break; + if (r < 0) { + log_netdev_error_errno(netdev, r, "Failed to create resolver: %m"); + continue; + } + + /* Avoid freeing netdev. It will be unrefed by the destroy callback. */ + netdev_ref(netdev); + + LIST_REMOVE(endpoints, w->unresolved_endpoints, endpoint); + } +} + +static int netdev_wireguard_post_create(NetDev *netdev, Link *link, sd_netlink_message *m) { + Wireguard *w; + + assert(netdev); + w = WIREGUARD(netdev); + assert(w); + + (void) wireguard_set_interface(netdev); + resolve_endpoints(netdev); + return 0; +} + +int config_parse_wireguard_listen_port(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + uint16_t *s = data; + uint16_t port = 0; + int r; + + assert(rvalue); + assert(data); + + if (!streq(rvalue, "auto")) { + r = parse_ip_port(rvalue, &port); + if (r < 0) + log_syntax(unit, LOG_ERR, filename, line, r, "Invalid port specification, ignoring assignment: %s", rvalue); + } + + *s = port; + + return 0; +} + +static int parse_wireguard_key(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + _cleanup_free_ void *key = NULL; + size_t len; + int r; + + assert(filename); + assert(rvalue); + assert(userdata); + + r = unbase64mem(rvalue, strlen(rvalue), &key, &len); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Could not parse wireguard key \"%s\", ignoring assignment: %m", rvalue); + return 0; + } + if (len != WG_KEY_LEN) { + log_syntax(unit, LOG_ERR, filename, line, EINVAL, + "Wireguard key is too short, ignoring assignment: %s", rvalue); + return 0; + } + + memcpy(userdata, key, WG_KEY_LEN); + return true; +} + +int config_parse_wireguard_private_key(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + Wireguard *w; + + assert(data); + + w = WIREGUARD(data); + + assert(w); + + return parse_wireguard_key(unit, + filename, + line, + section, + section_line, + lvalue, + ltype, + rvalue, + data, + &w->private_key); + +} + +int config_parse_wireguard_preshared_key(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + Wireguard *w; + WireguardPeer *peer; + + assert(data); + + w = WIREGUARD(data); + + assert(w); + + peer = wireguard_peer_new(w, section_line); + if (!peer) + return log_oom(); + + return parse_wireguard_key(unit, + filename, + line, + section, + section_line, + lvalue, + ltype, + rvalue, + data, + peer->preshared_key); +} + +int config_parse_wireguard_public_key(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + Wireguard *w; + WireguardPeer *peer; + + assert(data); + + w = WIREGUARD(data); + + assert(w); + + peer = wireguard_peer_new(w, section_line); + if (!peer) + return log_oom(); + + return parse_wireguard_key(unit, + filename, + line, + section, + section_line, + lvalue, + ltype, + rvalue, + data, + peer->public_key); +} + +int config_parse_wireguard_allowed_ips(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + union in_addr_union addr; + unsigned char prefixlen; + int r, family; + Wireguard *w; + WireguardPeer *peer; + WireguardIPmask *ipmask; + + assert(rvalue); + assert(data); + + w = WIREGUARD(data); + + peer = wireguard_peer_new(w, section_line); + if (!peer) + return log_oom(); + + for (;;) { + _cleanup_free_ char *word = NULL; + + r = extract_first_word(&rvalue, &word, "," WHITESPACE, 0); + if (r == 0) + break; + if (r == -ENOMEM) + return log_oom(); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Failed to split allowed ips \"%s\" option: %m", rvalue); + break; + } + + r = in_addr_prefix_from_string_auto(word, &family, &addr, &prefixlen); + if (r < 0) { + log_syntax(unit, LOG_ERR, filename, line, r, "Network address is invalid, ignoring assignment: %s", word); + return 0; + } + + ipmask = new(WireguardIPmask, 1); + if (!ipmask) + return log_oom(); + + *ipmask = (WireguardIPmask) { + .family = family, + .ip.in6 = addr.in6, + .cidr = prefixlen, + }; + + LIST_PREPEND(ipmasks, peer->ipmasks, ipmask); + } + + return 0; +} + +int config_parse_wireguard_endpoint(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + Wireguard *w; + WireguardPeer *peer; + size_t len; + const char *begin, *end = NULL; + _cleanup_free_ char *host = NULL, *port = NULL; + _cleanup_(wireguard_endpoint_freep) WireguardEndpoint *endpoint = NULL; + + assert(data); + assert(rvalue); + + w = WIREGUARD(data); + + assert(w); + + peer = wireguard_peer_new(w, section_line); + if (!peer) + return log_oom(); + + if (rvalue[0] == '[') { + begin = &rvalue[1]; + end = strchr(rvalue, ']'); + if (!end) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Unable to find matching brace of endpoint, ignoring assignment: %s", rvalue); + return 0; + } + len = end - begin; + ++end; + if (*end != ':' || !*(end + 1)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Unable to find port of endpoint: %s", rvalue); + return 0; + } + ++end; + } else { + begin = rvalue; + end = strrchr(rvalue, ':'); + if (!end || !*(end + 1)) { + log_syntax(unit, LOG_ERR, filename, line, 0, "Unable to find port of endpoint: %s", rvalue); + return 0; + } + len = end - begin; + ++end; + } + + host = strndup(begin, len); + if (!host) + return log_oom(); + + port = strdup(end); + if (!port) + return log_oom(); + + endpoint = new(WireguardEndpoint, 1); + if (!endpoint) + return log_oom(); + + *endpoint = (WireguardEndpoint) { + .peer = TAKE_PTR(peer), + .host = TAKE_PTR(host), + .port = TAKE_PTR(port), + .netdev = data, + }; + LIST_PREPEND(endpoints, w->unresolved_endpoints, TAKE_PTR(endpoint)); + + return 0; +} + +int config_parse_wireguard_keepalive(const char *unit, + const char *filename, + unsigned line, + const char *section, + unsigned section_line, + const char *lvalue, + int ltype, + const char *rvalue, + void *data, + void *userdata) { + int r; + uint16_t keepalive = 0; + Wireguard *w; + WireguardPeer *peer; + + assert(rvalue); + assert(data); + + w = WIREGUARD(data); + + assert(w); + + peer = wireguard_peer_new(w, section_line); + if (!peer) + return log_oom(); + + if (streq(rvalue, "off")) + keepalive = 0; + else { + r = safe_atou16(rvalue, &keepalive); + if (r < 0) + log_syntax(unit, LOG_ERR, filename, line, r, "The persistent keepalive interval must be 0-65535. Ignore assignment: %s", rvalue); + } + + peer->persistent_keepalive_interval = keepalive; + return 0; +} + +static void wireguard_init(NetDev *netdev) { + Wireguard *w; + + assert(netdev); + + w = WIREGUARD(netdev); + + assert(w); + + w->flags = WGDEVICE_F_REPLACE_PEERS; +} + +static void wireguard_done(NetDev *netdev) { + Wireguard *w; + WireguardPeer *peer; + WireguardIPmask *mask; + WireguardEndpoint *e; + + assert(netdev); + w = WIREGUARD(netdev); + assert(w); + + while ((peer = w->peers)) { + LIST_REMOVE(peers, w->peers, peer); + while ((mask = peer->ipmasks)) { + LIST_REMOVE(ipmasks, peer->ipmasks, mask); + free(mask); + } + free(peer); + } + + while ((e = w->unresolved_endpoints)) { + LIST_REMOVE(endpoints, w->unresolved_endpoints, e); + wireguard_endpoint_free(e); + } + + while ((e = w->failed_endpoints)) { + LIST_REMOVE(endpoints, w->failed_endpoints, e); + wireguard_endpoint_free(e); + } +} + +const NetDevVTable wireguard_vtable = { + .object_size = sizeof(Wireguard), + .sections = "Match\0NetDev\0WireGuard\0WireGuardPeer\0", + .post_create = netdev_wireguard_post_create, + .init = wireguard_init, + .done = wireguard_done, + .create_type = NETDEV_CREATE_INDEPENDENT, +}; diff --git a/src/network/netdev/wireguard.h b/src/network/netdev/wireguard.h new file mode 100644 index 0000000..143e93c --- /dev/null +++ b/src/network/netdev/wireguard.h @@ -0,0 +1,73 @@ +#pragma once + +typedef struct Wireguard Wireguard; + +#include "in-addr-util.h" +#include "netdev.h" +#include "socket-util.h" +#include "wireguard-netlink.h" + +#ifndef IFNAMSIZ +#define IFNAMSIZ 16 +#endif + +typedef struct WireguardIPmask { + uint16_t family; + union in_addr_union ip; + uint8_t cidr; + + LIST_FIELDS(struct WireguardIPmask, ipmasks); +} WireguardIPmask; + +typedef struct WireguardPeer { + uint8_t public_key[WG_KEY_LEN]; + uint8_t preshared_key[WG_KEY_LEN]; + uint32_t flags; + + union sockaddr_union endpoint; + + uint16_t persistent_keepalive_interval; + + LIST_HEAD(WireguardIPmask, ipmasks); + LIST_FIELDS(struct WireguardPeer, peers); +} WireguardPeer; + +typedef struct WireguardEndpoint { + char *host; + char *port; + + NetDev *netdev; + WireguardPeer *peer; + + LIST_FIELDS(struct WireguardEndpoint, endpoints); +} WireguardEndpoint; + +struct Wireguard { + NetDev meta; + unsigned last_peer_section; + + uint32_t flags; + + uint8_t private_key[WG_KEY_LEN]; + uint32_t fwmark; + + uint16_t port; + + LIST_HEAD(WireguardPeer, peers); + + LIST_HEAD(WireguardEndpoint, unresolved_endpoints); + LIST_HEAD(WireguardEndpoint, failed_endpoints); + unsigned n_retries; +}; + +DEFINE_NETDEV_CAST(WIREGUARD, Wireguard); +extern const NetDevVTable wireguard_vtable; + +CONFIG_PARSER_PROTOTYPE(config_parse_wireguard_allowed_ips); +CONFIG_PARSER_PROTOTYPE(config_parse_wireguard_endpoint); +CONFIG_PARSER_PROTOTYPE(config_parse_wireguard_listen_port); + +CONFIG_PARSER_PROTOTYPE(config_parse_wireguard_public_key); +CONFIG_PARSER_PROTOTYPE(config_parse_wireguard_private_key); +CONFIG_PARSER_PROTOTYPE(config_parse_wireguard_preshared_key); +CONFIG_PARSER_PROTOTYPE(config_parse_wireguard_keepalive); |