--- /dev/null
+/*
+ Palacios VNET Host Bridge
+ (c) Lei Xia, 2010
+ */
+#include <linux/spinlock.h>
+#include <linux/seq_file.h>
+#include <linux/proc_fs.h>
+#include <asm/uaccess.h>
+#include <linux/inet.h>
+#include <linux/kthread.h>
+
+#include <linux/netdevice.h>
+#include <linux/ip.h>
+#include <linux/in.h>
+#include <linux/net.h>
+#include <linux/string.h>
+#include <linux/preempt.h>
+#include <linux/sched.h>
+#include <asm/msr.h>
+
+#include <vnet/vnet.h>
+#include <vnet/vnet_hashtable.h>
+#include "palacios-vnet.h"
+
+#define VNET_SERVER_PORT 9000
+
+struct vnet_link {
+ uint32_t dst_ip;
+ uint16_t dst_port;
+
+ struct socket * sock;
+ struct sockaddr_in sock_addr;
+ vnet_brg_proto_t sock_proto;
+
+ struct nic_statistics stats;
+
+ uint32_t idx;
+
+ struct list_head node;
+};
+
+
+struct vnet_brg_state {
+ uint8_t status;
+
+ uint32_t num_links;
+ uint32_t link_idx;
+ struct list_head link_list;
+ struct hashtable *ip2link;
+
+ spinlock_t lock;
+
+ struct socket * serv_sock;
+ struct sockaddr_in serv_addr;
+ vnet_brg_proto_t serv_proto;
+
+ struct task_struct * serv_thread;
+
+ void * brg_data; /* private data from vnet_core */
+
+ struct vnet_brg_stats stats;
+};
+
+
+static struct vnet_brg_state vnet_brg_s;
+
+
+int vnet_brg_stats(struct vnet_brg_stats * stats){
+ memcpy(stats, &(vnet_brg_s.stats), sizeof(*stats));
+
+ return 0;
+}
+
+static inline struct vnet_link * _link_by_ip(uint32_t ip) {
+ return (struct vnet_link *)vnet_htable_search(vnet_brg_s.ip2link, (addr_t)&ip);
+}
+
+static inline struct vnet_link * _link_by_idx(int idx) {
+ struct vnet_link * link = NULL;
+
+ list_for_each_entry(link, &(vnet_brg_s.link_list), node) {
+
+ if (link->idx == idx) {
+ return link;
+ }
+ }
+ return NULL;
+}
+
+
+static void _delete_link(struct vnet_link * link){
+ unsigned long flags;
+
+ link->sock->ops->release(link->sock);
+
+ spin_lock_irqsave(&(vnet_brg_s.lock), flags);
+ list_del(&(link->node));
+ vnet_htable_remove(vnet_brg_s.ip2link, (addr_t)&(link->dst_ip), 0);
+ vnet_brg_s.num_links --;
+ spin_unlock_irqrestore(&(vnet_brg_s.lock), flags);
+
+ printk("VNET Bridge: Link deleted, ip 0x%x, port: %d, idx: %d\n",
+ link->dst_ip,
+ link->dst_port,
+ link->idx);
+
+ kfree(link);
+ link = NULL;
+}
+
+void vnet_brg_delete_link(uint32_t idx){
+ struct vnet_link * link = _link_by_idx(idx);
+
+ if(link){
+ _delete_link(link);
+ }
+}
+
+static void deinit_links_list(void){
+ struct vnet_link * link;
+
+ list_for_each_entry(link, &(vnet_brg_s.link_list), node) {
+ _delete_link(link);
+ }
+}
+
+static uint32_t _create_link(struct vnet_link * link) {
+ int err;
+ unsigned long flags;
+ int protocol;
+
+ switch(link->sock_proto){
+ case UDP:
+ protocol = IPPROTO_UDP;
+ break;
+ case TCP:
+ protocol = IPPROTO_TCP;
+ break;
+
+ default:
+ printk("Unsupported VNET Server Protocol\n");
+ return -1;
+ }
+
+ if ((err = sock_create(AF_INET, SOCK_DGRAM, protocol, &link->sock)) < 0) {
+ printk("Could not create socket for VNET Link, error %d\n", err);
+ return -1;
+ }
+
+ memset(&link->sock_addr, 0, sizeof(struct sockaddr));
+
+ link->sock_addr.sin_family = AF_INET;
+ link->sock_addr.sin_addr.s_addr = link->dst_ip;
+ link->sock_addr.sin_port = htons(link->dst_port);
+
+ if ((err = link->sock->ops->connect(link->sock, (struct sockaddr *)&(link->sock_addr), sizeof(struct sockaddr), 0) < 0)) {
+ printk("Could not connect to remote VNET Server, error %d\n", err);
+ return -1;
+ }
+
+ spin_lock_irqsave(&(vnet_brg_s.lock), flags);
+ list_add(&(link->node), &(vnet_brg_s.link_list));
+ vnet_brg_s.num_links ++;
+ link->idx = ++ vnet_brg_s.link_idx;
+ vnet_htable_insert(vnet_brg_s.ip2link, (addr_t)&(link->dst_ip), (addr_t)link);
+ spin_unlock_irqrestore(&(vnet_brg_s.lock), flags);
+
+ printk("VNET Bridge: Link created, ip 0x%x, port: %d, idx: %d, link: %p, protocol: %s\n",
+ link->dst_ip,
+ link->dst_port,
+ link->idx,
+ link,
+ ((link->sock_proto==UDP)?"UDP":"TCP"));
+
+ return link->idx;
+}
+
+
+uint32_t vnet_brg_add_link(uint32_t ip, uint16_t port, vnet_brg_proto_t proto){
+ struct vnet_link * new_link = NULL;
+ uint32_t idx;
+
+ new_link = kmalloc(sizeof(struct vnet_link), GFP_KERNEL);
+ if (!new_link) {
+ return -1;
+ }
+ memset(new_link, 0, sizeof(struct vnet_link));
+
+ new_link->dst_ip = ip;
+ new_link->dst_port = port;
+ new_link->sock_proto = proto;
+
+ idx = _create_link(new_link);
+ if (idx < 0) {
+ printk("Could not create link\n");
+ kfree(new_link);
+ return -1;
+ }
+
+ return idx;
+}
+
+
+int vnet_brg_link_stats(uint32_t link_idx, struct nic_statistics * stats){
+ struct vnet_link * link;
+
+ link = _link_by_idx(link_idx);
+ if(!link){
+ return -1;
+ }
+
+ memcpy(stats, &(link->stats), sizeof(*stats));
+
+ return 0;
+}
+
+
+static int
+_udp_send(struct socket * sock,
+ struct sockaddr_in * addr,
+ unsigned char * buf, int len) {
+ struct msghdr msg;
+ struct iovec iov;
+ mm_segment_t oldfs;
+ int size = 0;
+
+
+ if (sock->sk == NULL) {
+ return 0;
+ }
+
+ iov.iov_base = buf;
+ iov.iov_len = len;
+
+ msg.msg_flags = 0;
+ msg.msg_name = addr;
+ msg.msg_namelen = sizeof(struct sockaddr_in);
+ msg.msg_control = NULL;
+ msg.msg_controllen = 0;
+ msg.msg_iov = &iov;
+ msg.msg_iovlen = 1;
+ msg.msg_control = NULL;
+
+ oldfs = get_fs();
+ set_fs(KERNEL_DS);
+ size = sock_sendmsg(sock, &msg, len);
+ set_fs(oldfs);
+
+ return size;
+}
+
+
+
+static int
+_udp_recv(struct socket * sock,
+ struct sockaddr_in * addr,
+ unsigned char * buf, int len) {
+ struct msghdr msg;
+ struct iovec iov;
+ mm_segment_t oldfs;
+ int size = 0;
+
+ if (sock->sk == NULL) {
+ return 0;
+ }
+
+ iov.iov_base = buf;
+ iov.iov_len = len;
+
+ msg.msg_flags = 0;
+ msg.msg_name = addr;
+ msg.msg_namelen = sizeof(struct sockaddr_in);
+ msg.msg_control = NULL;
+ msg.msg_controllen = 0;
+ msg.msg_iov = &iov;
+ msg.msg_iovlen = 1;
+ msg.msg_control = NULL;
+
+ oldfs = get_fs();
+ set_fs(KERNEL_DS);
+ size = sock_recvmsg(sock, &msg, len, msg.msg_flags);
+
+ set_fs(oldfs);
+
+ return size;
+}
+
+/* send packets to VNET core */
+static int
+send_to_palacios(unsigned char * buf,
+ int len,
+ int link_id){
+ struct v3_vnet_pkt pkt;
+ pkt.size = len;
+ pkt.src_type = LINK_EDGE;
+ pkt.src_id = link_id;
+ memcpy(pkt.header, buf, ETHERNET_HEADER_LEN);
+ pkt.data = buf;
+
+ if(net_debug >= 2){
+ printk("VNET Lnx Bridge: send pkt to VNET core (size: %d, src_id: %d, src_type: %d)\n",
+ pkt.size, pkt.src_id, pkt.src_type);
+ if(net_debug >= 4){
+ print_hex_dump(NULL, "pkt_data: ", 0, 20, 20, pkt.data, pkt.size, 0);
+ }
+ }
+
+ vnet_brg_s.stats.pkt_to_vmm ++;
+
+ return v3_vnet_send_pkt(&pkt, NULL, 1);
+}
+
+
+/* send packet to extern network */
+static int
+bridge_send_pkt(struct v3_vm_info * vm,
+ struct v3_vnet_pkt * pkt,
+ void * private_data) {
+ struct vnet_link * link;
+
+ if(net_debug >= 2){
+ printk("VNET Lnx Host Bridge: packet received from VNET Core ... pkt size: %d, link: %d\n",
+ pkt->size,
+ pkt->dst_id);
+ if(net_debug >= 4){
+ print_hex_dump(NULL, "pkt_data: ", 0, 20, 20, pkt->data, pkt->size, 0);
+ }
+ }
+
+ vnet_brg_s.stats.pkt_from_vmm ++;
+
+ link = _link_by_idx(pkt->dst_id);
+ if (link != NULL) {
+ switch(link->sock_proto){
+ case UDP:
+ _udp_send(link->sock, &(link->sock_addr), pkt->data, pkt->size);
+ vnet_brg_s.stats.pkt_to_phy ++;
+ break;
+ case TCP:
+ vnet_brg_s.stats.pkt_to_phy ++;
+ break;
+
+ default:
+ printk("VNET Server: Invalid Link Protocol\n");
+ vnet_brg_s.stats.pkt_drop_vmm ++;
+ }
+ link->stats.tx_bytes += pkt->size;
+ link->stats.tx_pkts ++;
+ } else {
+ printk("VNET Bridge Linux Host: wrong dst link, idx: %d, discards the packet\n", pkt->dst_id);
+ vnet_brg_s.stats.pkt_drop_vmm ++;
+ }
+
+ return 0;
+}
+
+
+static int init_vnet_serv(void) {
+ int protocol;
+ int err;
+
+ switch(vnet_brg_s.serv_proto){
+ case UDP:
+ protocol = IPPROTO_UDP;
+ break;
+ case TCP:
+ protocol = IPPROTO_TCP;
+ break;
+
+ default:
+ printk("Unsupported VNET Server Protocol\n");
+ return -1;
+ }
+
+ if ((err = sock_create(AF_INET, SOCK_DGRAM, protocol, &vnet_brg_s.serv_sock)) < 0) {
+ printk("Could not create VNET server socket, error: %d\n", err);
+ return -1;
+ }
+
+ memset(&vnet_brg_s.serv_addr, 0, sizeof(struct sockaddr));
+
+ vnet_brg_s.serv_addr.sin_family = AF_INET;
+ vnet_brg_s.serv_addr.sin_addr.s_addr = htonl(INADDR_ANY);
+ vnet_brg_s.serv_addr.sin_port = htons(VNET_SERVER_PORT);
+
+ if ((err = vnet_brg_s.serv_sock->ops->bind(vnet_brg_s.serv_sock, (struct sockaddr *)&(vnet_brg_s.serv_addr), sizeof(struct sockaddr))) < 0) {
+ printk("Could not bind VNET server socket to port %d, error: %d\n", VNET_SERVER_PORT, err);
+ return -1;
+ }
+
+ printk("VNET server bind to port: %d\n", VNET_SERVER_PORT);
+
+ if(vnet_brg_s.serv_proto == TCP){
+ if((err = vnet_brg_s.serv_sock->ops->listen(vnet_brg_s.serv_sock, 32)) < 0){
+ printk("VNET Server error listening on port %d, error %d\n", VNET_SERVER_PORT, err);
+ return -1;
+ }
+ }
+
+ return 0;
+}
+
+static int _udp_server(void * arg) {
+ unsigned char * pkt;
+ struct sockaddr_in pkt_addr;
+ struct vnet_link * link = NULL;
+ int len;
+
+ printk("Palacios VNET Bridge: UDP receiving server ..... \n");
+
+ pkt = kmalloc(MAX_PACKET_LEN, GFP_KERNEL);
+ while (!kthread_should_stop()) {
+
+ len = _udp_recv(vnet_brg_s.serv_sock, &pkt_addr, pkt, MAX_PACKET_LEN);
+ if(len < 0) {
+ printk("Receive error: Could not get packet, error %d\n", len);
+ continue;
+ }
+
+ link = _link_by_ip(pkt_addr.sin_addr.s_addr);
+ if (link == NULL){
+ printk("VNET Server: No VNET Link match the src IP\n");
+ vnet_brg_s.stats.pkt_drop_phy ++;
+ continue;
+ }
+
+ vnet_brg_s.stats.pkt_from_phy ++;
+ link->stats.rx_bytes += len;
+ link->stats.rx_pkts ++;
+
+ send_to_palacios(pkt, len, link->idx);
+ }
+
+ kfree(pkt);
+
+ return 0;
+}
+
+
+static int _rx_server(void * arg) {
+
+ if(vnet_brg_s.serv_proto == UDP){
+ _udp_server(NULL);
+ }else if(vnet_brg_s.serv_proto == TCP) {
+ //accept new connection
+ //use select to receive pkt from physical network
+ //or create new kthread to handle each connection?
+ }else {
+ printk ("VNET Server: Unsupported Protocol\n");
+ return -1;
+ }
+
+ return 0;
+}
+
+static inline unsigned int hash_fn(addr_t hdr_ptr) {
+ return vnet_hash_buffer((uint8_t *)hdr_ptr, sizeof(uint32_t));
+}
+
+static inline int hash_eq(addr_t key1, addr_t key2) {
+ return (memcmp((uint8_t *)key1, (uint8_t *)key2, sizeof(uint32_t)) == 0);
+}
+
+
+int vnet_bridge_init(void) {
+ struct v3_vnet_bridge_ops bridge_ops;
+
+ if(vnet_brg_s.status != 0) {
+ return -1;
+ }
+ vnet_brg_s.status = 1;
+
+ memset(&vnet_brg_s, 0, sizeof(struct vnet_brg_state));
+
+ INIT_LIST_HEAD(&(vnet_brg_s.link_list));
+ spin_lock_init(&(vnet_brg_s.lock));
+
+ vnet_brg_s.serv_proto = UDP;
+
+ vnet_brg_s.ip2link = vnet_create_htable(10, hash_fn, hash_eq);
+ if(vnet_brg_s.ip2link == NULL){
+ printk("Failure to initiate VNET link hashtable\n");
+ return -1;
+ }
+
+ if(init_vnet_serv() < 0){
+ printk("Failure to initiate VNET server\n");
+ return -1;
+ }
+
+ vnet_brg_s.serv_thread = kthread_run(_rx_server, NULL, "vnet-server");
+
+ bridge_ops.input = bridge_send_pkt;
+ bridge_ops.poll = NULL;
+
+ if( v3_vnet_add_bridge(NULL, &bridge_ops, HOST_LNX_BRIDGE, NULL) < 0){
+ printk("VNET LNX Bridge: Fails to register bridge to VNET core");
+ }
+
+ printk("VNET Linux Bridge initiated\n");
+
+ return 0;
+}
+
+
+void vnet_bridge_deinit(void){
+
+ v3_vnet_del_bridge(HOST_LNX_BRIDGE);
+
+ kthread_stop(vnet_brg_s.serv_thread);
+ vnet_brg_s.serv_sock->ops->release(vnet_brg_s.serv_sock);
+
+ deinit_links_list();
+
+ vnet_free_htable(vnet_brg_s.ip2link, 0, 0);
+
+ vnet_brg_s.status = 0;
+}
+
+
/*
- Palacios VNET Host Bridge
+ Palacios VNET Control Module
(c) Lei Xia, 2010
*/
#include <linux/spinlock.h>
#define VNET_SERVER_PORT 9000
-struct palacios_vnet_route {
+struct vnet_route_iter {
struct v3_vnet_route route;
-
- int route_idx;
+ uint32_t idx;
struct list_head node;
};
-typedef enum {UDP, TCP, RAW, NONE} vnet_bridge_proto_t;
-struct vnet_link {
+struct vnet_link_iter {
uint32_t dst_ip;
uint16_t dst_port;
-
- struct socket * sock;
- struct sockaddr_in sock_addr;
- vnet_bridge_proto_t sock_proto;
-
- unsigned int pkt_rx, pkt_tx;
- uint64_t bytes_rx, bytes_tx;
-
- int link_idx;
+ vnet_brg_proto_t proto;
+ uint32_t idx;
struct list_head node;
};
-struct palacios_vnet_state {
- uint32_t num_routes;
- uint32_t num_links;
-
+struct vnet_ctrl_state {
uint8_t status;
+ uint32_t num_links;
+ uint32_t num_routes;
+
struct list_head route_list;
- struct list_head link_list;
-
- struct hashtable *ip2link; /* map IP to its Link */
+ struct list_head link_iter_list;
spinlock_t lock;
- struct socket * serv_sock;
- struct sockaddr_in serv_addr;
- vnet_bridge_proto_t serv_proto;
-
- /* Socket Receiving Thread */
- struct task_struct * serv_thread;
-
- unsigned int pkt_from_vmm, pkt_to_vmm, pkt_drop_vmm;
- unsigned int pkt_from_phy, pkt_to_phy, pkt_drop_phy;
+ struct proc_dir_entry * vnet_proc_root;
};
-static struct palacios_vnet_state vnet_state;
-
-
-static inline struct vnet_link * link_by_ip(uint32_t ip) {
- return (struct vnet_link *)vnet_htable_search(vnet_state.ip2link, (addr_t)&ip);
-}
-
-static inline struct vnet_link * link_by_idx(int idx) {
- struct vnet_link * link = NULL;
-
- list_for_each_entry(link, &(vnet_state.link_list), node) {
-
- if (link->link_idx == idx) {
- return link;
- }
- }
- return NULL;
-}
-
-static inline struct palacios_vnet_route * route_by_idx(int idx) {
- struct palacios_vnet_route * route = NULL;
-
- list_for_each_entry(route, &(vnet_state.route_list), node) {
-
- if (route->route_idx == idx) {
- return route;
- }
- }
-
- return NULL;
-}
+static struct vnet_ctrl_state vnet_ctrl_s;
static int parse_mac_str(char * str, uint8_t * qual, uint8_t * mac) {
}
-/* Format:
- * add src-MAC dst-MAC dst-TYPE [dst-ID] src-TYPE [src-ID]
- *
- * src-MAC = dst-MAC = not-MAC|any|none|MAC
- * dst-TYPE = edge|interface
- * src-TYPE = edge|interface|any
- * dst-ID = src-ID = IP|MAC
- * MAC=xx:xx:xx:xx:xx:xx
- * IP = xxx.xxx.xxx.xxx
- */
+static inline struct vnet_link_iter * link_by_ip(uint32_t ip) {
+ struct vnet_link_iter * link = NULL;
+
+ list_for_each_entry(link, &(vnet_ctrl_s.link_iter_list), node) {
+
+ if (link->dst_ip == ip) {
+ return link;
+ }
+ }
+
+ return NULL;
+}
+
+static inline struct vnet_link_iter * link_by_idx(int idx) {
+ struct vnet_link_iter * link = NULL;
+
+ list_for_each_entry(link, &(vnet_ctrl_s.link_iter_list), node) {
+ if (link->idx == idx) {
+ return link;
+ }
+ }
+
+ return NULL;
+}
+
+
static int parse_route_str(char * str, struct v3_vnet_route * route) {
char * token = NULL;
- struct vnet_link *link = NULL;
+ struct vnet_link_iter * link = NULL;
// src MAC
token = strsep(&str, " ");
link = link_by_ip(link_ip);
if (link != NULL){
- route->dst_id = link->link_idx;
+ route->dst_id = link->idx;
}else{
printk("can not find dst link %s\n", token);
return -1;
}
- printk("link_ip = %d, link_id = %d\n", link_ip, link->link_idx);
+ printk("link_ip = %d, link_id = %d\n", link_ip, link->idx);
} else if (route->dst_type == LINK_INTERFACE) {
uint8_t mac[ETH_ALEN];
link = link_by_ip(src_ip);
if (link != NULL){
- route->src_id = link->link_idx;
+ route->src_id = link->idx;
}else{
printk("can not find src link %s\n", token);
return -1;
}
-
-
static void * route_seq_start(struct seq_file * s, loff_t * pos) {
- struct palacios_vnet_route * route_iter = NULL;
+ struct vnet_route_iter * route_iter = NULL;
loff_t i = 0;
-
- if (*pos >= vnet_state.num_routes) {
+ if (*pos >= vnet_ctrl_s.num_routes) {
return NULL;
}
- list_for_each_entry(route_iter, &(vnet_state.route_list), node) {
-
+ list_for_each_entry(route_iter, &(vnet_ctrl_s.route_list), node) {
if (i == *pos) {
break;
}
}
-static void * link_seq_start(struct seq_file * s, loff_t * pos) {
- struct vnet_link * link_iter = NULL;
- loff_t i = 0;
-
- if (*pos >= vnet_state.num_links) {
- return NULL;
- }
-
- list_for_each_entry(link_iter, &(vnet_state.link_list), node) {
-
- if (i == *pos) {
- break;
- }
-
- i++;
- }
-
- return link_iter;
-}
-
-
-
static void * route_seq_next(struct seq_file * s, void * v, loff_t * pos) {
- struct palacios_vnet_route * route_iter = NULL;
+ struct vnet_route_iter * route_iter = NULL;
- route_iter = list_entry(((struct palacios_vnet_route *)v)->node.next, struct palacios_vnet_route, node);
+ route_iter = list_entry(((struct vnet_route_iter *)v)->node.next, struct vnet_route_iter, node);
// Check if the list has looped
- if (&(route_iter->node) == &(vnet_state.route_list)) {
+ if (&(route_iter->node) == &(vnet_ctrl_s.route_list)) {
return NULL;
}
return route_iter;
}
+static void route_seq_stop(struct seq_file * s, void * v) {
-static void * link_seq_next(struct seq_file * s, void * v, loff_t * pos) {
- struct vnet_link * link_iter = NULL;
+ return;
+}
-
- link_iter = list_entry(((struct vnet_link *)v)->node.next, struct vnet_link, node);
+static void * link_seq_start(struct seq_file * s, loff_t * pos) {
+ struct vnet_link_iter * link_iter = NULL;
+ loff_t i = 0;
- // Check if the list has looped
- if (&(link_iter->node) == &(vnet_state.link_list)) {
+ if (*pos >= vnet_ctrl_s.num_links) {
return NULL;
}
- *pos += 1;
-
- return link_iter;
-}
-
-
-static void route_seq_stop(struct seq_file * s, void * v) {
-
- return;
-}
-
+ list_for_each_entry(link_iter, &(vnet_ctrl_s.link_iter_list), node) {
+ if (i == *pos) {
+ break;
+ }
-static void link_seq_stop(struct seq_file * s, void * v) {
+ i++;
+ }
- return;
+ return link_iter;
}
static int route_seq_show(struct seq_file * s, void * v) {
- struct palacios_vnet_route * route_iter = v;
+ struct vnet_route_iter * route_iter = v;
struct v3_vnet_route * route = &(route_iter->route);
- seq_printf(s, "%d:\t", route_iter->route_idx);
+ seq_printf(s, "%d:\t", route_iter->idx);
seq_printf(s, "\nSrc:\t");
switch (route->src_mac_qual) {
seq_printf(s, "\nDst-Type:\t");
switch (route->dst_type) {
case LINK_EDGE: {
- struct vnet_link * link = (struct vnet_link *)link_by_idx(route->dst_id);
+ struct vnet_link_iter * link = (struct vnet_link_iter *)link_by_idx(route->dst_id);
seq_printf(s, "EDGE %pI4", &link->dst_ip);
break;
}
seq_printf(s, "\nSrc-Type:\t");
switch (route->src_type) {
case LINK_EDGE: {
- struct vnet_link * link = (struct vnet_link *)link_by_idx(route->src_id);
+ struct vnet_link_iter * link = (struct vnet_link_iter *)link_by_idx(route->src_id);
seq_printf(s, "EDGE %pI4", &link->dst_ip);
break;
}
break;
}
-
seq_printf(s, "\n");
return 0;
}
+static void * link_seq_next(struct seq_file * s, void * v, loff_t * pos) {
+ struct vnet_link_iter * link_iter = NULL;
+
+ link_iter = list_entry(((struct vnet_link_iter *)v)->node.next, struct vnet_link_iter, node);
+
+ // Check if the list has looped
+ if (&(link_iter->node) == &(vnet_ctrl_s.link_iter_list)) {
+ return NULL;
+ }
+
+ *pos += 1;
+
+ return link_iter;
+}
+
+static void link_seq_stop(struct seq_file * s, void * v) {
+
+ return;
+}
static int link_seq_show(struct seq_file * s, void * v) {
- struct vnet_link * link_iter = v;
+ struct vnet_link_iter * link_iter = v;
+ struct nic_statistics stats;
+
+ vnet_brg_link_stats(link_iter->idx, &stats);
- seq_printf(s, "%d:\t%pI4\t%d\n\t\tReceived Pkts: %d, Received Bytes %lld\n\t\tSent Pkts: %d, Sent Bytes: %lld\n\n",
- link_iter->link_idx,
+ seq_printf(s, "%d:\t%pI4\t%d\n\t\tReceived Pkts: %lld, Received Bytes %lld\n\t\tSent Pkts: %lld, Sent Bytes: %lld\n\n",
+ link_iter->idx,
&link_iter->dst_ip,
link_iter->dst_port,
- link_iter->pkt_rx,
- link_iter->bytes_rx,
- link_iter->pkt_tx,
- link_iter->bytes_tx);
+ stats.rx_pkts,
+ stats.rx_bytes,
+ stats.tx_pkts,
+ stats.tx_bytes);
return 0;
}
return seq_open(file, &link_seq_ops);
}
-static int inject_route(struct palacios_vnet_route * route) {
- unsigned long flags;
- v3_vnet_add_route(route->route);
- spin_lock_irqsave(&(vnet_state.lock), flags);
- list_add(&(route->node), &(vnet_state.route_list));
- route->route_idx = vnet_state.num_routes++;
- spin_unlock_irqrestore(&(vnet_state.lock), flags);
+static int inject_route(struct vnet_route_iter * route) {
+ unsigned long flags;
+
+ route->idx = v3_vnet_add_route(route->route);
+
+ spin_lock_irqsave(&(vnet_ctrl_s.lock), flags);
+ list_add(&(route->node), &(vnet_ctrl_s.route_list));
+ vnet_ctrl_s.num_routes ++;
+ spin_unlock_irqrestore(&(vnet_ctrl_s.lock), flags);
- printk("Palacios-vnet: One route added to VNET core\n");
+ printk("VNET Control: One route added to VNET core\n");
return 0;
}
+
+static void delete_route(struct vnet_route_iter * route) {
+ unsigned long flags;
+
+ v3_vnet_del_route(route->idx);
+
+ spin_lock_irqsave(&(vnet_ctrl_s.lock), flags);
+ list_del(&(route->node));
+ vnet_ctrl_s.num_routes --;
+ spin_unlock_irqrestore(&(vnet_ctrl_s.lock), flags);
+
+ printk("VNET Control: Route %d deleted from VNET\n", route->idx);
+
+ kfree(route);
+ route = NULL;
+}
+
+
+/* Format:
+ * add src-MAC dst-MAC dst-TYPE [dst-ID] src-TYPE [src-ID]
+ *
+ * src-MAC = dst-MAC = not-MAC|any|none|MAC
+ * dst-TYPE = edge|interface
+ * src-TYPE = edge|interface|any
+ * dst-ID = src-ID = IP|MAC
+ * MAC=xx:xx:xx:xx:xx:xx
+ * IP = xxx.xxx.xxx.xxx
+ *
+ *
+ * del route-idx
+ *
+ */
static ssize_t
route_write(struct file * file,
const char * buf,
}
if (strnicmp("ADD", token, strlen("ADD")) == 0) {
- struct palacios_vnet_route * new_route = NULL;
- new_route = kmalloc(sizeof(struct palacios_vnet_route), GFP_KERNEL);
+ struct vnet_route_iter * new_route = NULL;
+ new_route = kmalloc(sizeof(struct vnet_route_iter), GFP_KERNEL);
if (!new_route) {
return -ENOMEM;
}
- memset(new_route, 0, sizeof(struct palacios_vnet_route));
+ memset(new_route, 0, sizeof(struct vnet_route_iter));
if (parse_route_str(buf_iter, &(new_route->route)) == -1) {
kfree(new_route);
}
if (inject_route(new_route) != 0) {
+ kfree(new_route);
return -EFAULT;
}
} else if (strnicmp("DEL", token, strlen("DEL")) == 0) {
- printk("I should delete the route here\n");
+ char * idx_str = NULL;
+ uint32_t d_idx;
+
+ idx_str = strsep(&buf_iter, " ");
+
+ if (!idx_str) {
+ printk("Missing route idx in DEL Route command\n");
+ return -EFAULT;
+ }
+
+ d_idx = simple_strtoul(idx_str, &idx_str, 10);
+
+ v3_vnet_del_route(d_idx);
+
+ printk("VNET Control: One route deleted\n");
+
} else {
printk("Invalid Route command string\n");
}
return size;
}
-static int create_link(struct vnet_link * link) {
- int err;
+
+static void delete_link(struct vnet_link_iter * link){
unsigned long flags;
- int protocol;
- switch(link->sock_proto){
- case UDP:
- protocol = IPPROTO_UDP;
- break;
- case TCP:
- protocol = IPPROTO_TCP;
- break;
+ vnet_brg_delete_link(link->idx);
- default:
- printk("Unsupported VNET Server Protocol\n");
- return -1;
- }
-
- if ((err = sock_create(AF_INET, SOCK_DGRAM, protocol, &link->sock)) < 0) {
- printk("Could not create socket for VNET Link, error %d\n", err);
- return -1;
- }
+ spin_lock_irqsave(&(vnet_ctrl_s.lock), flags);
+ list_del(&(link->node));
+ vnet_ctrl_s.num_links --;
+ spin_unlock_irqrestore(&(vnet_ctrl_s.lock), flags);
- memset(&link->sock_addr, 0, sizeof(struct sockaddr));
+ kfree(link);
+ link = NULL;
+}
- link->sock_addr.sin_family = AF_INET;
- link->sock_addr.sin_addr.s_addr = link->dst_ip;
- link->sock_addr.sin_port = htons(link->dst_port);
- if ((err = link->sock->ops->connect(link->sock, (struct sockaddr *)&(link->sock_addr), sizeof(struct sockaddr), 0) < 0)) {
- printk("Could not connect to remote VNET Server, error %d\n", err);
- return -1;
- }
+static void deinit_links_list(void){
+ struct vnet_link_iter * link;
- // We use the file pointer because we are in the kernel
- // This is only used to assigned File Descriptors for user space, so it is available here
- // link->sock->file = link;
-
- spin_lock_irqsave(&(vnet_state.lock), flags);
- list_add(&(link->node), &(vnet_state.link_list));
- link->link_idx = vnet_state.num_links++;
- vnet_htable_insert(vnet_state.ip2link, (addr_t)&(link->dst_ip), (addr_t)link);
- spin_unlock_irqrestore(&(vnet_state.lock), flags);
+ list_for_each_entry(link, &(vnet_ctrl_s.link_iter_list), node) {
+ delete_link(link);
+ }
+}
- printk("VNET Bridge: Link created, ip 0x%x, port: %d, idx: %d, link: %p, protocol: %s\n",
- link->dst_ip,
- link->dst_port,
- link->link_idx,
- link,
- ((link->sock_proto==UDP)?"UDP":"TCP"));
+static void deinit_routes_list(void){
+ struct vnet_route_iter * route;
- return 0;
+ list_for_each_entry(route, &(vnet_ctrl_s.route_list), node) {
+ delete_route(route);
+ }
}
-
/* ADD dst-ip 9000 [udp|tcp] */
+/* DEL link-idx */
static ssize_t
link_write(struct file * file, const char * buf, size_t size, loff_t * ppos) {
char link_buf[256];
}
if (strnicmp("ADD", token, strlen("ADD")) == 0) {
- struct vnet_link * new_link = NULL;
+ struct vnet_link_iter * link = NULL;
char * ip_str = NULL;
- uint32_t ip;
+ uint32_t d_ip;
+ uint16_t d_port;
+ vnet_brg_proto_t d_proto;
+ int link_idx;
+ unsigned long flags;
ip_str = strsep(&link_iter, " ");
return -EFAULT;
}
- if (in4_pton(ip_str, strlen(ip_str), (uint8_t *)&(ip), '\0', NULL) != 1) {
+ if (in4_pton(ip_str, strlen(ip_str), (uint8_t *)&(d_ip), '\0', NULL) != 1) {
printk("Invalid Dst IP address (%s)\n", ip_str);
return -EFAULT;
}
- new_link = kmalloc(sizeof(struct vnet_link), GFP_KERNEL);
- if (!new_link) {
- return -ENOMEM;
+ d_port = simple_strtol(link_iter, &link_iter, 10);
+ d_proto = UDP;
+
+ link_idx = vnet_brg_add_link(d_ip, d_port, d_proto);
+ if(link_idx < 0){
+ printk("VNET Control: Failed to create link\n");
+ return -EFAULT;
}
-
- memset(new_link, 0, sizeof(struct vnet_link));
- new_link->dst_ip = ip;
- new_link->dst_port = simple_strtol(link_iter, &link_iter, 10);
+ link = kmalloc(sizeof(struct vnet_link_iter), GFP_KERNEL);
+ memset(link, 0, sizeof(struct vnet_link_iter));
- //TODO: Parse UDP|TCP
- new_link->sock_proto = UDP;
+ link->dst_ip = d_ip;
+ link->dst_port = d_port;
+ link->proto = d_proto;
+ link->idx = link_idx;
+
+ spin_lock_irqsave(&(vnet_ctrl_s.lock), flags);
+ list_add(&(link->node), &(vnet_ctrl_s.link_iter_list));
+ vnet_ctrl_s.num_links ++;
+ spin_unlock_irqrestore(&(vnet_ctrl_s.lock), flags);
+ } else if (strnicmp("DEL", token, strlen("DEL")) == 0) {
+ char * idx_str = NULL;
+ uint32_t d_idx;
- if (create_link(new_link) != 0) {
- printk("Could not create link\n");
- kfree(new_link);
+ idx_str = strsep(&link_iter, " ");
+
+ if (!idx_str) {
+ printk("Missing link idx in DEL Link command\n");
return -EFAULT;
}
- } else if (strnicmp("DEL", token, strlen("DEL")) == 0) {
- printk("Link deletion not supported\n");
+ d_idx = simple_strtoul(idx_str, &idx_str, 10);
+
+ vnet_brg_delete_link(d_idx);
+
+ printk("VNET Control: One link deleted\n");
} else {
printk("Invalid Link command string\n");
}
static ssize_t
debug_write(struct file * file, const char * buf, size_t size, loff_t * ppos) {
- char link_buf[32];
- char * link_iter = NULL;
- char * line_str = link_buf;
+ char in_buf[32];
+ char * in_iter = NULL;
+ char * line_str = in_buf;
int level = -1;
if (size >= 32) {
return -EFAULT;
}
- if (copy_from_user(link_buf, buf, size)) {
+ if (copy_from_user(in_buf, buf, size)) {
return -EFAULT;
}
- link_iter = strsep(&line_str, "\r\n");
- level = simple_strtol(link_iter, &link_iter, 10);
+ in_iter = strsep(&line_str, "\r\n");
+ level = simple_strtol(in_iter, &in_iter, 10);
- printk("Set VNET Debug level to %d\n", level);
+ printk("VNET Control: Set VNET Debug level to %d\n", level);
if(level >= 0){
- vnet_debug = level;
+ net_debug = level;
}
return size;
static int debug_show(struct seq_file * file, void * v){
- seq_printf(file, "Current VNET Debug Level: %d\n", vnet_debug);
+ seq_printf(file, "Current NET Debug Level: %d\n", net_debug);
return 0;
}
static int stat_show(struct seq_file * file, void * v){
struct vnet_stat stats;
+ struct vnet_brg_stats brg_stats;
v3_vnet_stat(&stats);
seq_printf(file, "\tTransmitted Packets: %d\n", stats.tx_pkts);
seq_printf(file, "\tTransmitted Bytes: %lld\n", stats.tx_bytes);
-
+ vnet_brg_stats(&brg_stats);
+
seq_printf(file, "\nVNET Bridge Server\n");
- seq_printf(file, "\tReceived From VMM: %d\n", vnet_state.pkt_from_vmm);
- seq_printf(file, "\tSent To VMM: %d\n", vnet_state.pkt_to_vmm);
- seq_printf(file, "\tDropped From VMM: %d\n", vnet_state.pkt_drop_vmm);
- seq_printf(file, "\tReceived From Extern Network: %d\n", vnet_state.pkt_from_phy);
- seq_printf(file, "\tSent To Extern Network: %d\n", vnet_state.pkt_to_phy);
- seq_printf(file, "\tDropped From Extern Network: %d\n", vnet_state.pkt_drop_phy);
+ seq_printf(file, "\tReceived From VMM: %lld\n", brg_stats.pkt_from_vmm);
+ seq_printf(file, "\tSent To VMM: %lld\n", brg_stats.pkt_to_vmm);
+ seq_printf(file, "\tDropped From VMM: %lld\n", brg_stats.pkt_drop_vmm);
+ seq_printf(file, "\tReceived From Extern Network: %lld\n", brg_stats.pkt_from_phy);
+ seq_printf(file, "\tSent To Extern Network: %lld\n", brg_stats.pkt_to_phy);
+ seq_printf(file, "\tDropped From Extern Network: %lld\n", brg_stats.pkt_drop_phy);
return 0;
}
return -1;
}
debug_entry->proc_fops = &debug_fops;
-
-
- return 0;
-}
-
-
-
-static int
-udp_send(struct socket * sock,
- struct sockaddr_in * addr,
- unsigned char * buf, int len) {
- struct msghdr msg;
- struct iovec iov;
- mm_segment_t oldfs;
- int size = 0;
-
-
- if (sock->sk == NULL) {
- return 0;
- }
-
- iov.iov_base = buf;
- iov.iov_len = len;
-
- msg.msg_flags = 0;
- msg.msg_name = addr;
- msg.msg_namelen = sizeof(struct sockaddr_in);
- msg.msg_control = NULL;
- msg.msg_controllen = 0;
- msg.msg_iov = &iov;
- msg.msg_iovlen = 1;
- msg.msg_control = NULL;
- oldfs = get_fs();
- set_fs(KERNEL_DS);
- size = sock_sendmsg(sock, &msg, len);
- set_fs(oldfs);
-
- return size;
-}
-
-
-
-static int
-udp_recv(struct socket * sock,
- struct sockaddr_in * addr,
- unsigned char * buf, int len) {
- struct msghdr msg;
- struct iovec iov;
- mm_segment_t oldfs;
- int size = 0;
-
- if (sock->sk == NULL) {
- return 0;
- }
-
- iov.iov_base = buf;
- iov.iov_len = len;
-
- msg.msg_flags = 0;
- msg.msg_name = addr;
- msg.msg_namelen = sizeof(struct sockaddr_in);
- msg.msg_control = NULL;
- msg.msg_controllen = 0;
- msg.msg_iov = &iov;
- msg.msg_iovlen = 1;
- msg.msg_control = NULL;
-
- oldfs = get_fs();
- set_fs(KERNEL_DS);
- size = sock_recvmsg(sock, &msg, len, msg.msg_flags);
-
- set_fs(oldfs);
-
- return size;
-}
-
-/* send packets to VNET core */
-static int
-send_to_palacios(unsigned char * buf,
- int len,
- int link_id){
- struct v3_vnet_pkt pkt;
- pkt.size = len;
- pkt.src_type = LINK_EDGE;
- pkt.src_id = link_id;
- memcpy(pkt.header, buf, ETHERNET_HEADER_LEN);
- pkt.data = buf;
-
- if(vnet_debug >= 2){
- printk("VNET Lnx Bridge: send pkt to VNET core (size: %d, src_id: %d, src_type: %d)\n",
- pkt.size, pkt.src_id, pkt.src_type);
- if(vnet_debug >= 4){
- print_hex_dump(NULL, "pkt_data: ", 0, 20, 20, pkt.data, pkt.size, 0);
- }
- }
-
- vnet_state.pkt_to_vmm ++;
-
- return v3_vnet_send_pkt(&pkt, NULL, 1);
-}
-
-
-/* send packet to extern network */
-static int
-bridge_send_pkt(struct v3_vm_info * vm,
- struct v3_vnet_pkt * pkt,
- void * private_data) {
- struct vnet_link * link;
-
- if(vnet_debug >= 2){
- printk("VNET Lnx Host Bridge: packet received from VNET Core ... pkt size: %d, link: %d\n",
- pkt->size,
- pkt->dst_id);
- if(vnet_debug >= 4){
- print_hex_dump(NULL, "pkt_data: ", 0, 20, 20, pkt->data, pkt->size, 0);
- }
- }
-
- vnet_state.pkt_from_vmm ++;
-
- link = link_by_idx(pkt->dst_id);
- if (link != NULL) {
- switch(link->sock_proto){
- case UDP:
- udp_send(link->sock, &(link->sock_addr), pkt->data, pkt->size);
- vnet_state.pkt_to_phy ++;
- break;
- case TCP:
- vnet_state.pkt_to_phy ++;
- break;
-
- default:
- printk("VNET Server: Invalid Link Protocol\n");
- vnet_state.pkt_drop_vmm ++;
- }
- link->bytes_tx += pkt->size;
- link->pkt_tx ++;
- } else {
- printk("VNET Bridge Linux Host: wrong dst link, idx: %d, discards the packet\n", pkt->dst_id);
- vnet_state.pkt_drop_vmm ++;
- }
+ vnet_ctrl_s.vnet_proc_root = vnet_root;
return 0;
}
-static int init_vnet_serv(void) {
- int protocol;
- int err;
-
- switch(vnet_state.serv_proto){
- case UDP:
- protocol = IPPROTO_UDP;
- break;
- case TCP:
- protocol = IPPROTO_TCP;
- break;
+static void destroy_proc_files(void) {
+ struct proc_dir_entry * vnet_root = vnet_ctrl_s.vnet_proc_root;
- default:
- printk("Unsupported VNET Server Protocol\n");
- return -1;
- }
-
- if ((err = sock_create(AF_INET, SOCK_DGRAM, protocol, &vnet_state.serv_sock)) < 0) {
- printk("Could not create VNET server socket, error: %d\n", err);
- return -1;
- }
-
- memset(&vnet_state.serv_addr, 0, sizeof(struct sockaddr));
-
- vnet_state.serv_addr.sin_family = AF_INET;
- vnet_state.serv_addr.sin_addr.s_addr = htonl(INADDR_ANY);
- vnet_state.serv_addr.sin_port = htons(VNET_SERVER_PORT);
-
- if ((err = vnet_state.serv_sock->ops->bind(vnet_state.serv_sock, (struct sockaddr *)&(vnet_state.serv_addr), sizeof(struct sockaddr))) < 0) {
- printk("Could not bind VNET server socket to port %d, error: %d\n", VNET_SERVER_PORT, err);
- return -1;
- }
-
- printk("VNET server bind to port: %d\n", VNET_SERVER_PORT);
-
- if(vnet_state.serv_proto == TCP){
- if((err = vnet_state.serv_sock->ops->listen(vnet_state.serv_sock, 32)) < 0){
- printk("VNET Server error listening on port %d, error %d\n", VNET_SERVER_PORT, err);
- return -1;
- }
- }
-
- return 0;
+ remove_proc_entry("debug", vnet_root);
+ remove_proc_entry("links", vnet_root);
+ remove_proc_entry("routes", vnet_root);
+ remove_proc_entry("stats", vnet_root);
+ remove_proc_entry("vnet", NULL);
}
-static int vnet_udp_server(void * arg) {
- unsigned char * pkt;
- struct sockaddr_in pkt_addr;
- struct vnet_link * link = NULL;
- int len;
-
- printk("Palacios VNET Bridge: UDP receiving server ..... \n");
- pkt = kmalloc(MAX_PACKET_LEN, GFP_KERNEL);
- while (!kthread_should_stop()) {
-
- len = udp_recv(vnet_state.serv_sock, &pkt_addr, pkt, MAX_PACKET_LEN);
- if(len < 0) {
- printk("Receive error: Could not get packet, error %d\n", len);
- continue;
- }
-
- link = link_by_ip(pkt_addr.sin_addr.s_addr);
- if (link == NULL){
- printk("VNET Server: No VNET Link match the src IP\n");
- vnet_state.pkt_drop_phy ++;
- continue;
- }
-
- vnet_state.pkt_from_phy ++;
- link->bytes_rx += len;
- link->pkt_rx ++;
-
- send_to_palacios(pkt, len, link->link_idx);
- }
-
- kfree(pkt);
-
- return 0;
-}
-
-
-
-static int vnet_server(void * arg) {
-
- if(vnet_state.serv_proto == UDP){
- vnet_udp_server(NULL);
- }else if(vnet_state.serv_proto == TCP) {
- //accept new connection
- //use select to receive pkt from physical network
- //or create new kthread to handle each connection?
- }else {
- printk ("VNET Server: Unsupported Protocol\n");
- return -1;
- }
-
- return 0;
-}
-
-static inline unsigned int hash_fn(addr_t hdr_ptr) {
- return vnet_hash_buffer((uint8_t *)hdr_ptr, sizeof(uint32_t));
-}
-
-static inline int hash_eq(addr_t key1, addr_t key2) {
- return (memcmp((uint8_t *)key1, (uint8_t *)key2, sizeof(uint32_t)) == 0);
-}
-
-
-int palacios_init_vnet(void) {
- struct v3_vnet_bridge_ops bridge_ops;
-
-
- if(vnet_state.status != 0) {
+int vnet_ctrl_init(void) {
+ if(vnet_ctrl_s.status != 0) {
return -1;
}
- vnet_state.status = 1;
+ vnet_ctrl_s.status = 1;
- memset(&vnet_state, 0, sizeof(struct palacios_vnet_state));
+ memset(&vnet_ctrl_s, 0, sizeof(struct vnet_ctrl_state));
- INIT_LIST_HEAD(&(vnet_state.link_list));
- INIT_LIST_HEAD(&(vnet_state.route_list));
- spin_lock_init(&(vnet_state.lock));
+ INIT_LIST_HEAD(&(vnet_ctrl_s.link_iter_list));
+ INIT_LIST_HEAD(&(vnet_ctrl_s.route_list));
+ spin_lock_init(&(vnet_ctrl_s.lock));
init_proc_files();
-
- vnet_state.serv_proto = UDP;
-
- vnet_state.ip2link = vnet_create_htable(10, hash_fn, hash_eq);
- if(vnet_state.ip2link == NULL){
- printk("Failure to initiate VNET link hashtable\n");
- return -1;
- }
- if(init_vnet_serv() < 0){
- printk("Failure to initiate VNET server\n");
- return -1;
- }
+ printk("VNET Linux control module initiated\n");
- vnet_state.serv_thread = kthread_run(vnet_server, NULL, "vnet-server");
+ return 0;
+}
- bridge_ops.input = bridge_send_pkt;
- bridge_ops.poll = NULL;
-
- v3_vnet_add_bridge(NULL, &bridge_ops, HOST_LNX_BRIDGE, NULL);
- printk("Palacios VNET Linux Bridge initiated\n");
+void vnet_ctrl_deinit(void){
+ destroy_proc_files();
- return 0;
+ deinit_links_list();
+ deinit_routes_list();
+
+ vnet_ctrl_s.status = 0;
}
+
