* http://www.v3vee.org
*
* Copyright (c) 2009, Lei Xia <lxia@northwestern.edu>
- * Copyright (c) 2009, Yuan Tang <ytang@northwestern.edu>
- * Copyright (c) 2009, Jack Lange <jarusl@cs.northwestern.edu>
- * Copyright (c) 2009, Peter Dinda <pdinda@northwestern.edu>
+ * Copyright (c) 2009, Yuan Tang <ytang@northwestern.edu>
* Copyright (c) 2009, The V3VEE Project <http://www.v3vee.org>
* All rights reserved.
*
* Author: Lei Xia <lxia@northwestern.edu>
* Yuan Tang <ytang@northwestern.edu>
- * Jack Lange <jarusl@cs.northwestern.edu>
- * Peter Dinda <pdinda@northwestern.edu
*
* This is free software. You are permitted to use,
* redistribute, and modify it as specified in the file "V3VEE_LICENSE".
*/
#include <palacios/vmm_vnet.h>
-#include <palacios/vmm_hypercall.h>
#include <palacios/vm_guest_mem.h>
+#include <palacios/vmm_lock.h>
+#include <palacios/vmm_queue.h>
+#include <palacios/vmm_sprintf.h>
#ifndef CONFIG_DEBUG_VNET
#undef PrintDebug
#endif
-struct ethernet_pkt {
- uint32_t size; //size of data
- uint16_t type;
- struct udp_link_header ext_hdr; //possible externel header to applied to data before sent
- char data[ETHERNET_PACKET_LEN];
-};
+struct eth_hdr {
+ uint8_t dst_mac[6];
+ uint8_t src_mac[6];
+ uint16_t type; // indicates layer 3 protocol type
+} __attribute__((packed));
-// 14 (ethernet frame) + 20 bytes
-struct in_pkt_header {
- char ethernetdest[6];
- char ethernetsrc[6];
- unsigned char ethernettype[2]; //layer 3 protocol type
- char ip[20];
-};
-#define VNET_INITAB_HCALL 0xca00 // inital hypercall id
-#define MAX_LINKS 10
-#define MAX_ROUTES 10
-#define HASH_KEY_LEN 16
-#define MIN_CACHE_SIZE 100
-static const uint_t hash_key_size = 16;
-struct link_table {
- struct link_entry * links[MAX_LINKS];
- uint16_t size;
-};
-struct routing_table {
- struct routing_entry * routes[MAX_ROUTES];
- uint16_t size;
-};
+struct vnet_dev {
+
+ uint8_t mac_addr[6];
+ struct v3_vm_info * vm;
+
+ int (*input)(struct v3_vm_info * vm, struct v3_vnet_pkt * pkt, void * private_data);
+ void * private_data;
+
+ int dev_id;
+ struct list_head node;
+} __attribute__((packed));
-static struct link_table g_links;
-static struct routing_table g_routes;
-static struct gen_queue * g_inpkt_q;
-/* Hash key format:
- * 0-5: src_eth_addr
- * 6-11: dest_eth_addr
- * 12: src type
- * 13-16: src index
- */
-typedef char * route_hashkey_t;
+struct vnet_brg_dev {
+ struct v3_vm_info * vm;
+
+ int (*input)(struct v3_vm_info * vm, struct v3_vnet_pkt * pkt, void * private_data);
+ void * private_data;
+} __attribute__((packed));
+
-// This is the hash value, Format: 0: num_matched_routes, 1...n: matches[] -- TY
-struct route_cache_entry {
- int num_matched_routes;
- int * matches;
+
+
+
+struct vnet_route_info {
+ struct v3_vnet_route route_def;
+
+ struct vnet_dev * dst_dev;
+ struct vnet_dev * src_dev;
+
+ struct list_head node;
+ struct list_head match_node; // used for route matching
};
-// the route cache
-static struct hashtable * g_route_cache;
-static void print_packet(uchar_t *pkt, int size) {
- PrintDebug("Vnet: print_data_packet: size: %d\n", size);
- v3_hexdump(pkt, size, NULL, 0);
-}
-#if 0
-static void print_packet_addr(char * pkt) {
- PrintDebug("Vnet: print_packet_destination_addr: ");
- v3_hexdump(pkt + 8, 6, NULL, 0);
-
- PrintDebug("Vnet: print_packet_source_addr: ");
- v3_hexdump(pkt + 14, 6, NULL, 0);
-}
+struct route_list {
+ uint8_t hash_buf[VNET_HASH_SIZE];
-static void print_device_addr(char * ethaddr) {
- PrintDebug("Vnet: print_device_addr: ");
- v3_hexdump(ethaddr, 6, NULL, 0);
-}
-#endif
+ uint32_t num_routes;
+ struct vnet_route_info * routes[0];
+} __attribute__((packed));
-static uint16_t ip_xsum(struct ip_header *ip_hdr, int hdr_len){
- long sum = 0;
- uint16_t *p = (uint16_t*) ip_hdr;
- while(hdr_len > 1){
- sum += *(p++);
- if(sum & 0x80000000)
- sum = (sum & 0xFFFF) + (sum >> 16);
- hdr_len -= 2;
- }
-
- if(hdr_len)
- sum += (uint16_t) *(uchar_t *)p;
-
- while(sum>>16)
- sum = (sum & 0xFFFF) + (sum >> 16);
- return (uint16_t)~sum;
-}
+static struct {
+ struct list_head routes;
+ struct list_head devs;
+
+ int num_routes;
+ int num_devs;
-static inline void ethernet_packet_init(struct ethernet_pkt *pt, uchar_t *data, const size_t size) {
- pt->size = size;
- memcpy(pt->data, data, size);
-}
+ struct vnet_brg_dev *bridge;
-static uint_t hash_from_key_fn(addr_t hashkey) {
- uint8_t * key = (uint8_t *)hashkey;
- return v3_hash_buffer(key, HASH_KEY_LEN);
-}
+ v3_lock_t lock;
-static int hash_key_equal(addr_t key1, addr_t key2) {
- uint8_t * buf1 = (uint8_t *)key1;
- uint8_t * buf2 = (uint8_t *)key2;
- return (memcmp(buf1, buf2, HASH_KEY_LEN) == 0);
-}
+ struct gen_queue * inpkt_q;
+ struct hashtable * route_cache;
-static int init_route_cache() {
- g_route_cache = v3_create_htable(MIN_CACHE_SIZE, &hash_from_key_fn, &hash_key_equal);
+} vnet_state;
- if (g_route_cache == NULL) {
- PrintError("Vnet: Route Cache Initiate Failurely\n");
- return -1;
- }
- return 0;
+
+
+#ifdef CONFIG_DEBUG_VNET
+static inline void mac_to_string(char mac[6], char * buf) {
+ snprintf(buf, 50, "%x:%x:%x:%x:%x:%x",
+ mac[0], mac[1], mac[2],
+ mac[3], mac[4], mac[5]);
}
-static void make_hash_key(route_hashkey_t hashkey,
- char src_addr[6],
- char dest_addr[6],
- char src_type,
- int src_index) {
- int j;
+static void print_route(struct vnet_route_info *route){
+ char str[50];
+
+ mac_to_string(route->route_def.src_mac, str);
+ PrintDebug("Src Mac (%s), src_qual (%d)\n",
+ str, route->route_def.src_mac_qual);
+ mac_to_string(route->route_def.dst_mac, str);
+ PrintDebug("Dst Mac (%s), dst_qual (%d)\n",
+ str, route->route_def.dst_mac_qual);
+ PrintDebug("Src dev id (%d), src type (%d)",
+ route->route_def.src_id,
+ route->route_def.src_type);
+ PrintDebug("Dst dev id (%d), dst type (%d)\n",
+ route->route_def.dst_id,
+ route->route_def.dst_type);
+ if (route->route_def.dst_type == LINK_INTERFACE) {
+ PrintDebug("dst_dev (%p), dst_dev_id (%d), dst_dev_input (%p), dst_dev_data (%p)\n",
+ route->dst_dev,
+ route->dst_dev->dev_id,
+ route->dst_dev->input,
+ route->dst_dev->private_data);
+ }
+}
+
+static void dump_routes(){
+ struct vnet_route_info *route;
+
+ int i = 0;
+ PrintDebug("\n========Dump routes starts ============\n");
+ list_for_each_entry(route, &(vnet_state.routes), node) {
+ PrintDebug("\nroute %d:\n", ++i);
+
+ print_route(route);
+ }
+ PrintDebug("\n========Dump routes end ============\n");
+}
+
+#endif
- for (j = 0; j < 6; j++) {
- hashkey[j] = src_addr[j];
- hashkey[j + 6] = dest_addr[j] + 1;
- }
- hashkey[12] = src_type;
+/*
+ * A VNET packet is a packed struct with the hashed fields grouped together.
+ * This means we can generate the hash from an offset into the pkt struct
+ */
+static inline uint_t hash_fn(addr_t hdr_ptr) {
+ uint8_t * hdr_buf = (uint8_t *)hdr_ptr;
- *(int *)(hashkey + 12) = src_index;
+ return v3_hash_buffer(hdr_buf, VNET_HASH_SIZE);
}
-static int add_route_to_cache(route_hashkey_t hashkey, int num_matched_r, int * matches) {
- struct route_cache_entry * new_entry = NULL;
- int i;
-
- new_entry = (struct route_cache_entry *)V3_Malloc(sizeof(struct route_cache_entry));
- if (new_entry == NULL) {
- PrintError("Vnet: Malloc fails\n");
- return -1;
- }
-
- new_entry->num_matched_routes = num_matched_r;
+static inline int hash_eq(addr_t key1, addr_t key2) {
+ return (memcmp((uint8_t *)key1, (uint8_t *)key2, VNET_HASH_SIZE) == 0);
+}
- new_entry->matches = (int *)V3_Malloc(num_matched_r * sizeof(int));
-
- if (new_entry->matches == NULL) {
- PrintError("Vnet: Malloc fails\n");
+
+static int add_route_to_cache(const struct v3_vnet_pkt * pkt, struct route_list * routes) {
+ memcpy(routes->hash_buf, pkt->hash_buf, VNET_HASH_SIZE);
+
+ if (v3_htable_insert(vnet_state.route_cache, (addr_t)routes->hash_buf, (addr_t)routes) == 0) {
+ PrintError("Vnet: Failed to insert new route entry to the cache\n");
return -1;
}
- for (i = 0; i < num_matched_r; i++) {
- new_entry->matches[i] = matches[i];
- }
-
- // here, when v3_htable_insert return 0, it means insert fails
- if (v3_htable_insert(g_route_cache, (addr_t)hashkey, (addr_t)new_entry) == 0) {
- PrintError("Vnet: Insert new route entry to cache failed\n");
- V3_Free(new_entry->matches);
- V3_Free(new_entry);
- }
-
return 0;
}
static int clear_hash_cache() {
- v3_free_htable(g_route_cache, 1, 1);
-
- g_route_cache = v3_create_htable(MIN_CACHE_SIZE, hash_from_key_fn, hash_key_equal);
-
- if (g_route_cache == NULL) {
- PrintError("Vnet: Route Cache Create Failurely\n");
- return -1;
- }
+
+ v3_free_htable(vnet_state.route_cache, 1, 1);
+ vnet_state.route_cache = v3_create_htable(0, &hash_fn, &hash_eq);
return 0;
}
-static int look_into_cache(route_hashkey_t hashkey, int * matches) {
- struct route_cache_entry * found = NULL;
- int n_matches = -1;
- int i = 0;
+static int look_into_cache(const struct v3_vnet_pkt * pkt, struct route_list ** routes) {
- found = (struct route_cache_entry *)v3_htable_search(g_route_cache, (addr_t)hashkey);
+ *routes = (struct route_list *)v3_htable_search(vnet_state.route_cache, (addr_t)(pkt->hash_buf));
- if (found != NULL) {
- n_matches = found->num_matched_routes;
-
- for (i = 0; i < n_matches; i++) {
- matches[i] = found->matches[i];
- }
- }
-
- return n_matches;
-}
-
-#ifdef CONFIG_DEBUG_VNET
-static inline uint8_t hex_nybble_to_nybble(const uint8_t hexnybble) {
- uint8_t x = toupper(hexnybble);
-
- if (isdigit(x)) {
- return x - '0';
- } else {
- return 10 + (x - 'A');
- }
+ return 0;
}
-static inline uint8_t hex_byte_to_byte(const uint8_t hexbyte[2]) {
- return ((hex_nybble_to_nybble(hexbyte[0]) << 4) +
- (hex_nybble_to_nybble(hexbyte[1]) & 0xf));
-}
-static inline void string_to_mac(const char *str, uint8_t mac[6]) {
- int k;
+static struct vnet_dev * find_dev_by_id(int idx) {
+ struct vnet_dev * dev = NULL;
+
+ list_for_each_entry(dev, &(vnet_state.devs), node) {
+ int dev_id = dev->dev_id;
- for (k = 0; k < 6; k++) {
- mac[k] = hex_byte_to_byte(&(str[(2 * k) + k]));
+ if (dev_id == idx)
+ return dev;
}
-}
-static inline void mac_to_string(char mac[6], char * buf) {
- snprintf(buf, 20, "%x:%x:%x:%x:%x:%x",
- mac[0], mac[1], mac[2],
- mac[3], mac[4], mac[5]);
+ return NULL;
}
-#endif
-static int __add_link_entry(struct link_entry * link) {
- int idx;
+static struct vnet_dev * find_dev_by_mac(char mac[6]) {
+ struct vnet_dev * dev = NULL;
- for (idx = 0; idx < MAX_LINKS; idx++) {
- if (g_links.links[idx] == NULL) {
- g_links.links[idx] = link;
- g_links.size++;
-
- return idx;
- }
+ list_for_each_entry(dev, &(vnet_state.devs), node) {
+ if (!memcmp(dev->mac_addr, mac, 6))
+ return dev;
}
- PrintError("No available Link entry\n");
- return -1;
+ return NULL;
}
-static int __add_route_entry(struct routing_entry * route) {
- int idx;
-
- for (idx = 0; idx < MAX_ROUTES; idx++) {
- if (g_routes.routes[idx] == NULL) {
- g_routes.routes[idx] = route;
- g_routes.size++;
-
- return idx;
- }
- }
+int get_device_id_by_mac(char mac[6]){
+
+ struct vnet_dev *dev = find_dev_by_mac(mac);
- PrintError("No available route entry\n");
- return -1;
+ if (dev == NULL)
+ return -1;
+
+ return dev->dev_id;
}
-static int vnet_add_route_entry(char src_mac[6],
- char dest_mac[6],
- int src_mac_qual,
- int dest_mac_qual,
- int link_idx,
- link_type_t link_type,
- int src,
- link_type_t src_type) {
- struct routing_entry * new_route = (struct routing_entry *)V3_Malloc(sizeof(struct routing_entry));
- int idx = -1;
+int v3_vnet_add_route(struct v3_vnet_route route) {
+ struct vnet_route_info * new_route = NULL;
+ unsigned long flags;
- memset(new_route, 0, sizeof(struct routing_entry));
+ new_route = (struct vnet_route_info *)V3_Malloc(sizeof(struct vnet_route_info));
+ memset(new_route, 0, sizeof(struct vnet_route_info));
- if ((src_mac_qual != MAC_ANY) && (src_mac_qual != MAC_NONE)) {
- memcpy(new_route->src_mac, src_mac, 6);
- }
-
- if ((dest_mac_qual != MAC_ANY) && (dest_mac_qual != MAC_NONE)) {
- memcpy(new_route->dest_mac, dest_mac, 6);
- }
-
- new_route->src_mac_qual = src_mac_qual;
- new_route->dest_mac_qual = dest_mac_qual;
- new_route->link_idx= link_idx;
- new_route->link_type = link_type;
- new_route->src_link_idx = src;
- new_route->src_type = src_type;
-
- if ((idx = __add_route_entry(new_route)) == -1) {
- PrintError("Could not add route entry\n");
- return -1;
- }
-
- clear_hash_cache();
+ PrintDebug("Vnet: vnet_add_route_entry: dst_id: %d, dst_type: %d\n",
+ route.dst_id, route.dst_type);
- return idx;
-}
+ memcpy(new_route->route_def.src_mac, route.src_mac, 6);
+ memcpy(new_route->route_def.dst_mac, route.dst_mac, 6);
+ new_route->route_def.src_mac_qual = route.src_mac_qual;
+ new_route->route_def.dst_mac_qual = route.dst_mac_qual;
+ new_route->route_def.dst_id = route.dst_id;
+ new_route->route_def.dst_type = route.dst_type;
+ new_route->route_def.src_id = route.src_id;
+ new_route->route_def.src_type = route.src_type;
-static void * __delete_link_entry(int index) {
- struct link_entry * link = NULL;
- void * ret = NULL;
- link_type_t type;
-
- if ((index >= MAX_LINKS) || (g_links.links[index] == NULL)) {
- return NULL;
+ if (new_route->route_def.dst_type == LINK_INTERFACE) {
+ new_route->dst_dev = find_dev_by_id(new_route->route_def.dst_id);
+ PrintDebug("Vnet: Add route, get device: dev_id %d, input : %p, private_data %p\n",
+ new_route->dst_dev->dev_id, new_route->dst_dev->input, new_route->dst_dev->private_data);
}
- link = g_links.links[index];
- type = g_links.links[index]->type;
-
- if (type == LINK_INTERFACE) {
- ret = (void *)g_links.links[index]->dst_dev;
- } else if (type == LINK_EDGE) {
- ret = (void *)g_links.links[index]->dst_link;
+ if (new_route->route_def.src_type == LINK_INTERFACE) {
+ new_route->src_dev = find_dev_by_id(new_route->route_def.src_id);
}
- g_links.links[index] = NULL;
- g_links.size--;
+ flags = v3_lock_irqsave(vnet_state.lock);
- V3_Free(link);
-
- return ret;
-}
+ list_add(&(new_route->node), &(vnet_state.routes));
+ clear_hash_cache();
-static int find_route_entry(char src_mac[6],
- char dest_mac[6],
- int src_mac_qual,
- int dest_mac_qual,
- int link_idx,
- link_type_t link_type,
- int src,
- link_type_t src_type) {
- int i;
- char temp_src_mac[6];
- char temp_dest_mac[6];
-
- if ((src_mac_qual != MAC_ANY) && (src_mac_qual != MAC_NONE)) {
- memcpy(temp_src_mac, src_mac, 6);
- } else {
- memset(temp_src_mac, 0, 6);
- }
-
- if ((dest_mac_qual != MAC_ANY) && (dest_mac_qual != MAC_NONE)) {
- memcpy(temp_dest_mac, dest_mac, 6);
- } else {
- memset(temp_dest_mac, 0, 6);
- }
-
- for (i = 0; i < MAX_ROUTES; i++) {
- if (g_routes.routes[i] != NULL) {
- if ((memcmp(temp_src_mac, g_routes.routes[i]->src_mac, 6) == 0) &&
- (memcmp(temp_dest_mac, g_routes.routes[i]->dest_mac, 6) == 0) &&
- (g_routes.routes[i]->src_mac_qual == src_mac_qual) &&
- (g_routes.routes[i]->dest_mac_qual == dest_mac_qual) &&
- ( (link_type == LINK_ANY) ||
- ((link_type == g_routes.routes[i]->link_type) && (g_routes.routes[i]->link_idx == link_idx))) &&
- ( (src_type == LINK_ANY) ||
- ((src_type == g_routes.routes[i]->src_type) && (g_routes.routes[i]->src_link_idx == src)))) {
- return i;
- }
- }
- }
-
- return -1;
-}
+ v3_unlock_irqrestore(vnet_state.lock, flags);
+
-static int __delete_route_entry(int index) {
- struct routing_entry * route;
+#ifdef CONFIG_DEBUG_VNET
+ dump_routes();
+#endif
- if ((index >= MAX_ROUTES) || (g_routes.routes[index] == NULL)) {
- PrintDebug("VNET: wrong index in delete route entry %d\n", index);
- return -1;
- }
+ return 0;
+}
- route = g_routes.routes[index];
- g_routes.routes[index] = NULL;
- g_routes.size--;
- V3_Free(route);
- clear_hash_cache();
-
- return 0;
-}
+// At the end allocate a route_list
+// This list will be inserted into the cache so we don't need to free it
+static struct route_list * match_route(const struct v3_vnet_pkt * pkt) {
+ struct vnet_route_info * route = NULL;
+ struct route_list * matches = NULL;
+ int num_matches = 0;
+ int max_rank = 0;
+ struct list_head match_list;
+ struct eth_hdr * hdr = (struct eth_hdr *)(pkt->data);
+ uint8_t src_type = pkt->src_type;
+ uint32_t src_link = pkt->src_id;
-static int vnet_delete_route_entry_by_addr(char src_mac[6],
- char dest_mac[6],
- int src_mac_qual,
- int dest_mac_qual,
- int link_idx,
- link_type_t type,
- int src,
- link_type_t src_type) {
- int index = find_route_entry(src_mac, dest_mac, src_mac_qual,
- dest_mac_qual, link_idx, type, src, src_type);
-
- if (index == -1) {
- PrintDebug("VNET: wrong in delete route entry %d\n", index);
- return -1;
+#ifdef CONFIG_DEBUG_VNET
+ {
+ char dst_str[50];
+ char src_str[50];
+
+ mac_to_string(hdr->src_mac, src_str);
+ mac_to_string(hdr->dst_mac, dst_str);
+ PrintDebug("Vnet: match_route. pkt: SRC(%s), DEST(%s)\n", src_str, dst_str);
}
-
- return __delete_route_entry(index);
-}
+#endif
-static int match_route(uint8_t * src_mac,
- uint8_t * dst_mac,
- link_type_t src_type,
- int src_index,
- int * matches) {
- int values[MAX_ROUTES];
- int matched_routes[MAX_ROUTES];
-
- int num_matches = 0;
- int i;
- int max = 0;
- int no = 0;
- int exact_match = 0;
-
- for(i = 0; i < MAX_ROUTES; i++) {
- if (g_routes.routes[i] != NULL){
- if ( (g_routes.routes[i]->src_type != LINK_ANY) &&
- ( (g_routes.routes[i]->src_type != src_type) ||
- ( (g_routes.routes[i]->src_link_idx != src_index) &&
- (g_routes.routes[i]->src_link_idx != -1)))) {
- PrintDebug("Vnet: MatchRoute: Source route is on and does not match\n");
- continue;
- }
-
- if ( (g_routes.routes[i]->dest_mac_qual == MAC_ANY) &&
- (g_routes.routes[i]->src_mac_qual == MAC_ANY)) {
+ INIT_LIST_HEAD(&match_list);
+
+#define UPDATE_MATCHES(rank) do { \
+ if (max_rank < (rank)) { \
+ max_rank = (rank); \
+ INIT_LIST_HEAD(&match_list); \
+ \
+ list_add(&(route->match_node), &match_list); \
+ num_matches = 1; \
+ } else if (max_rank == (rank)) { \
+ list_add(&(route->match_node), &match_list); \
+ num_matches++; \
+ } \
+ } while (0)
+
+
+ list_for_each_entry(route, &(vnet_state.routes), node) {
+ struct v3_vnet_route * route_def = &(route->route_def);
+
+ // CHECK SOURCE TYPE HERE
+ if ( (route_def->src_type != LINK_ANY) &&
+ ( (route_def->src_type != src_type) ||
+ ( (route_def->src_id != src_link) &&
+ (route_def->src_id != (uint32_t)-1)))) {
+ continue;
+ }
- matched_routes[num_matches] = i;
- values[num_matches] = 3;
- num_matches++;
- }
-
- if (memcmp((void *)&g_routes.routes[i]->src_mac, (void *)src_mac, 6) == 0) {
- if (g_routes.routes[i]->src_mac_qual != MAC_NOT) {
- if (g_routes.routes[i]->dest_mac_qual == MAC_ANY) {
- matched_routes[num_matches] = i;
- values[num_matches] = 6;
-
- num_matches++;
- } else if (memcmp((void *)&g_routes.routes[i]->dest_mac, (void *)dst_mac, 6) == 0) {
- if (g_routes.routes[i]->dest_mac_qual != MAC_NOT) {
- matched_routes[num_matches] = i;
- values[num_matches] = 8;
- exact_match = 1;
- num_matches++;
- }
- }
- }
- }
-
- if (memcmp((void *)&g_routes.routes[i]->dest_mac, (void *)dst_mac, 6) == 0) {
- if (g_routes.routes[i]->dest_mac_qual != MAC_NOT) {
- if (g_routes.routes[i]->src_mac_qual == MAC_ANY) {
- matched_routes[num_matches] = i;
- values[num_matches] = 6;
-
- num_matches++;
- } else if (memcmp((void *)&g_routes.routes[i]->src_mac, (void *)src_mac, 6) == 0) {
- if (g_routes.routes[i]->src_mac_qual != MAC_NOT) {
- if (exact_match == 0) {
- matched_routes[num_matches] = i;
- values[num_matches] = 8;
- num_matches++;
- }
- }
- }
- }
- }
+
+ if ((route_def->dst_mac_qual == MAC_ANY) &&
+ (route_def->src_mac_qual == MAC_ANY)) {
+ UPDATE_MATCHES(3);
+ }
- if ( (g_routes.routes[i]->dest_mac_qual == MAC_NOT) &&
- (memcmp((void *)&g_routes.routes[i]->dest_mac, (void *)dst_mac, 6) != 0)) {
- if (g_routes.routes[i]->src_mac_qual == MAC_ANY) {
- matched_routes[num_matches] = i;
- values[num_matches] = 5;
- num_matches++;
- } else if (memcmp((void *)&g_routes.routes[i]->src_mac, (void *)src_mac, 6) == 0) {
- if (g_routes.routes[i]->src_mac_qual != MAC_NOT) {
- matched_routes[num_matches] = i;
- values[num_matches] = 7;
- num_matches++;
- }
- }
+ if (memcmp(route_def->src_mac, hdr->src_mac, 6) == 0) {
+ if (route_def->src_mac_qual != MAC_NOT) {
+ if (route_def->dst_mac_qual == MAC_ANY) {
+ UPDATE_MATCHES(6);
+ } else if (route_def->dst_mac_qual != MAC_NOT &&
+ memcmp(route_def->dst_mac, hdr->dst_mac, 6) == 0) {
+ UPDATE_MATCHES(8);
+ }
}
+ }
- if ( (g_routes.routes[i]->src_mac_qual == MAC_NOT) &&
- (memcmp((void *)&g_routes.routes[i]->src_mac, (void *)src_mac, 6) != 0)) {
- if (g_routes.routes[i]->dest_mac_qual == MAC_ANY) {
- matched_routes[num_matches] = i;
- values[num_matches] = 5;
- num_matches++;
- } else if (memcmp((void *)&g_routes.routes[i]->dest_mac, (void *)dst_mac, 6) == 0) {
- if (g_routes.routes[i]->dest_mac_qual != MAC_NOT) {
- matched_routes[num_matches] = i;
- values[num_matches] = 7;
- num_matches++;
- }
+ if (memcmp(route_def->dst_mac, hdr->dst_mac, 6) == 0) {
+ if (route_def->dst_mac_qual != MAC_NOT) {
+ if (route_def->src_mac_qual == MAC_ANY) {
+ UPDATE_MATCHES(6);
+ } else if ((route_def->src_mac_qual != MAC_NOT) &&
+ (memcmp(route_def->src_mac, hdr->src_mac, 6) == 0)) {
+ UPDATE_MATCHES(8);
}
}
- }
- } // end for
-
- for (i = 0; i < MAX_ROUTES; i++) {
- if ( (memcmp((void *)&g_routes.routes[i]->src_mac, (void *)src_mac, 6) == 0) &&
- (g_routes.routes[i]->dest_mac_qual == MAC_NONE) &&
- ( (g_routes.routes[i]->src_type == LINK_ANY) ||
- ( (g_routes.routes[i]->src_type == src_type) &&
- ( (g_routes.routes[i]->src_link_idx == src_index) ||
- (g_routes.routes[i]->src_link_idx == -1))))) {
- matched_routes[num_matches] = i;
- values[num_matches] = 4;
- PrintDebug("Vnet: MatchRoute: We matched a default route (%d)\n", i);
- num_matches++;
- }
+ }
+
+ if ((route_def->dst_mac_qual == MAC_NOT) &&
+ (memcmp(route_def->dst_mac, hdr->dst_mac, 6) != 0)) {
+ if (route_def->src_mac_qual == MAC_ANY) {
+ UPDATE_MATCHES(5);
+ } else if ((route_def->src_mac_qual != MAC_NOT) &&
+ (memcmp(route_def->src_mac, hdr->src_mac, 6) == 0)) {
+ UPDATE_MATCHES(7);
+ }
+ }
+
+ if ((route_def->src_mac_qual == MAC_NOT) &&
+ (memcmp(route_def->src_mac, hdr->src_mac, 6) != 0)) {
+ if (route_def->dst_mac_qual == MAC_ANY) {
+ UPDATE_MATCHES(5);
+ } else if ((route_def->dst_mac_qual != MAC_NOT) &&
+ (memcmp(route_def->dst_mac, hdr->dst_mac, 6) == 0)) {
+ UPDATE_MATCHES(7);
+ }
+ }
+
+ // Default route
+ if ( (memcmp(route_def->src_mac, hdr->src_mac, 6) == 0) &&
+ (route_def->dst_mac_qual == MAC_NONE)) {
+ UPDATE_MATCHES(4);
+ }
}
-
- //If many rules have been matched, we choose one which has the highest value rating
+
+ PrintDebug("Vnet: match_route: Matches=%d\n", num_matches);
+
if (num_matches == 0) {
- return 0;
- }
-
- for (i = 0; i < num_matches; i++) {
- if (values[i] > max) {
- no = 0;
- max = values[i];
- matches[no] = matched_routes[i];
- no++;
- } else if (values[i] == max) {
- matches[no] = matched_routes[i];
- no++;
- }
+ return NULL;
}
-
- return no;
-}
-static int handle_one_pkt(struct ethernet_pkt *pkt) {
- int src_link_index = -1; //the value of src_link_index of udp always is 0
- int i;
- char src_mac[6];
- char dst_mac[6];
+ matches = V3_Malloc(sizeof(struct route_list) +
+ (sizeof(struct vnet_route_info *) * num_matches));
- int matches[MAX_ROUTES];
- int num_matched_routes = 0;
+ matches->num_routes = num_matches;
- struct in_pkt_header header;
+ {
+ int i = 0;
+ list_for_each_entry(route, &match_list, match_node) {
+ matches->routes[i++] = route;
+ }
+ }
- char hash_key[hash_key_size];
-
- // get the ethernet and ip headers from the packet
- memcpy((void *)&header, (void *)pkt->data, sizeof(header));
- memcpy(src_mac, header.ethernetsrc, 6);
- memcpy(dst_mac, header.ethernetdest, 6);
+ return matches;
+}
+
+
+int v3_vnet_send_pkt(struct v3_vnet_pkt * pkt, void * private_data) {
+ struct route_list * matched_routes = NULL;
+ unsigned long flags;
+ int i;
#ifdef CONFIG_DEBUG_VNET
- char dest_str[18];
- char src_str[18];
-
- mac_to_string(src_mac, src_str);
- mac_to_string(dst_mac, dest_str);
-
- PrintDebug("Vnet: HandleDataOverLink. SRC(%s), DEST(%s)\n", src_str, dest_str);
+ {
+ struct eth_hdr * hdr = (struct eth_hdr *)(pkt->header);
+ char dest_str[30];
+ char src_str[30];
+
+ mac_to_string(hdr->src_mac, src_str);
+ mac_to_string(hdr->dst_mac, dest_str);
+ PrintDebug("Vnet: HandleDataOverLink. SRC(%s), DEST(%s)\n", src_str, dest_str);
+ }
#endif
- // link_edge -> pt->type???
- make_hash_key(hash_key, src_mac, dst_mac, LINK_EDGE, src_link_index);
-
- num_matched_routes = look_into_cache((route_hashkey_t)hash_key, matches);
-
- if (num_matched_routes == -1) {
- // no match in the cache
- num_matched_routes = match_route(src_mac, dst_mac, LINK_ANY, src_link_index, matches);
-
- if (num_matched_routes > 0) {
- add_route_to_cache(hash_key, num_matched_routes,matches);
- }
- }
-
- PrintDebug("Vnet: HandleDataOverLink: Matches=%d\n", num_matched_routes);
-
- for (i = 0; i < num_matched_routes; i++) {//send packet to all destinations
- int route_index = -1;
- int link_index = -1;
- int pkt_len = 0;
- struct link_entry * link = NULL;
+ flags = v3_lock_irqsave(vnet_state.lock);
- route_index = matches[i];
+ look_into_cache(pkt, &matched_routes);
- PrintDebug("Vnet: HandleDataOverLink: Forward packet from link according to Route entry %d\n", route_index);
+ if (matched_routes == NULL) {
+ PrintDebug("Vnet: send pkt Looking into routing table\n");
- link_index = g_routes.routes[route_index]->link_idx;
+ matched_routes = match_route(pkt);
+
+ if (matched_routes) {
+ add_route_to_cache(pkt, matched_routes);
+ } else {
+ PrintDebug("Could not find route for packet... discards packet\n");
+ v3_unlock_irqrestore(vnet_state.lock, flags);
+ return -1;
+ }
+ }
+
+ v3_unlock_irqrestore(vnet_state.lock, flags);
+
- if ((link_index < 0) ||
- (link_index > MAX_LINKS) ||
- (g_links.links[link_index] == NULL)) {
- continue;
- }
+ PrintDebug("Vnet: send pkt route matches %d\n", matched_routes->num_routes);
+
+ for (i = 0; i < matched_routes->num_routes; i++) {
+ struct vnet_route_info * route = matched_routes->routes[i];
- link = g_links.links[link_index];
- pkt_len = pkt->size;
- if (g_routes.routes[route_index]->link_type == LINK_EDGE) {
-
- //apply the header in the beginning of the packet to be sent
- if (link->dst_link->pro_type == UDP_TYPE) {
- struct udp_link_header *hdr = &(link->dst_link->vnet_header);
- struct ip_header *ip = &hdr->ip_hdr;
- struct udp_header *udp = &hdr->udp_hdr;
- udp->len = pkt_len + sizeof(struct udp_header);
- ip->total_len = pkt_len + sizeof(struct udp_header) + sizeof(struct ip_header);
- ip->cksum = ip_xsum(ip, sizeof(struct ip_header));
-
- int hdr_size = sizeof(struct udp_link_header);
- memcpy(&pkt->ext_hdr, hdr, hdr_size);
-
- pkt_len += hdr_size;
- if ((link->dst_link->input((uchar_t *)&pkt->ext_hdr, pkt_len, link->dst_link->private_data)) != pkt_len) {
- PrintDebug("VNET: Packet not sent properly\n");
- return -1;
- }
- }else {
- PrintDebug("VNET: Link protocol type not support\n");
- return -1;
- }
- } else if (g_routes.routes[route_index]->link_type == LINK_INTERFACE) {
- if ((link->dst_dev->input(pkt->data, pkt_len, link->dst_dev->private_data)) != pkt_len) {
- PrintDebug("VNET: Packet not sent properly\n");
- return -1;
+ if (route->route_def.dst_type == LINK_EDGE) {
+ pkt->dst_type = LINK_EDGE;
+ pkt->dst_id = route->route_def.dst_id;
+
+ if (vnet_state.bridge == NULL) {
+ PrintDebug("VNET: No bridge to sent data to links\n");
+ continue;
+ }
+
+ if (vnet_state.bridge->input(vnet_state.bridge->vm, pkt, vnet_state.bridge->private_data) == -1) {
+ PrintDebug("VNET: Packet not sent properly\n");
+ continue;
+ }
+
+ } else if (route->route_def.dst_type == LINK_INTERFACE) {
+ if (route->dst_dev->input(route->dst_dev->vm, pkt, route->dst_dev->private_data) == -1) {
+ PrintDebug("VNET: Packet not sent properly\n");
+ continue;
}
} else {
PrintDebug("Vnet: Wrong Edge type\n");
- return -1;
+ continue;
}
- }
-
- return 0;
-}
-
-static int send_ethernet_pkt(uchar_t *data, int len) {
- struct ethernet_pkt *pkt;
-
- pkt = (struct ethernet_pkt *)V3_Malloc(sizeof(struct ethernet_pkt));
- memset(pkt, 0, sizeof(struct ethernet_pkt));
- if(pkt == NULL){
- PrintError("VNET: Memory allocate fails\n");
- return -1;
+ PrintDebug("Vnet: HandleOnePacket: Forward packet according to Route\n");
}
-
- ethernet_packet_init(pkt, data, len); //====here we copy sending data once
-
- PrintDebug("VNET: vm_send_pkt: transmitting packet: (size:%d)\n", (int)pkt->size);
- print_packet((char *)data, len);
- v3_enqueue(g_inpkt_q, (addr_t)pkt);
-
return 0;
}
-//send raw ethernet packet
-int v3_vnet_send_rawpkt(uchar_t * buf, int len, void *private_data) {
- PrintDebug("VNET: In V3_Send_pkt: pkt length %d\n", len);
-
- return send_ethernet_pkt(buf, len);
-}
+int v3_vnet_add_dev(struct v3_vm_info *vm, uint8_t mac[6],
+ int (*netif_input)(struct v3_vm_info * vm, struct v3_vnet_pkt * pkt, void * private_data),
+ void * priv_data){
+ struct vnet_dev * new_dev = NULL;
+ unsigned long flags;
-//sending the packet from Dom0, should remove the link header
-int v3_vnet_send_udppkt(uchar_t * buf, int len, void *private_data) {
- PrintDebug("VNET: In V3_Send_pkt: pkt length %d\n", len);
+ new_dev = (struct vnet_dev *)V3_Malloc(sizeof(struct vnet_dev));
- uint_t hdr_len = sizeof(struct udp_link_header);
-
- return send_ethernet_pkt((uchar_t *)(buf+hdr_len), len - hdr_len);
-}
-
-static int search_device(char * device_name) {
- int i;
-
- for (i = 0; i < MAX_LINKS; i++) {
- if ((g_links.links[i] != NULL) && (g_links.links[i]->type == LINK_INTERFACE)) {
- if (strcmp(device_name, g_links.links[i]->dst_dev->name) == 0) {
- return i;
- }
- }
- }
-
- return -1;
-}
-
-int vnet_register_device(struct vm_device * vdev,
- char * dev_name,
- uchar_t mac[6],
- int (*netif_input)(uchar_t * pkt, uint_t size, void * private_data),
- void * data) {
- struct vnet_if_device * if_dev;
-
- int idx = search_device(dev_name);
-
- if (idx != -1) {
- PrintDebug("VNET: register device: Already has device with the name %s\n", dev_name);
- return -1;
- }
-
- if_dev = (struct vnet_if_device *)(sizeof(struct vnet_if_device));
-
- if (if_dev == NULL) {
+ if (new_dev == NULL) {
PrintError("VNET: Malloc fails\n");
return -1;
}
-
- strcpy(if_dev->name, dev_name);
- strncpy(if_dev->mac_addr, mac, 6);
- if_dev->dev = vdev;
- if_dev->input = netif_input;
- if_dev->private_data = data;
-
- struct link_entry * link = (struct link_entry *)(sizeof(struct link_entry));
-
- link->type = LINK_INTERFACE;
- link->dst_dev = if_dev;
-
- idx = __add_link_entry(link);
-
- PrintDebug("VNET: Add device %s to link table, idx %d", dev_name, idx);
-
- return idx;
-}
+
+ memcpy(new_dev->mac_addr, mac, 6);
+ new_dev->input = netif_input;
+ new_dev->private_data = priv_data;
+ new_dev->vm = vm;
+ new_dev->dev_id = 0;
-#if 0
-static int vnet_unregister_device(char * dev_name) {
- int idx;
+ flags = v3_lock_irqsave(vnet_state.lock);
- idx = search_device(dev_name);
-
- if (idx == -1) {
- PrintDebug("VNET: No device with name %s found\n", dev_name);
- return -1;
+ if (!find_dev_by_mac(mac)) {
+ list_add(&(new_dev->node), &(vnet_state.devs));
+ new_dev->dev_id = ++vnet_state.num_devs;
}
- struct vnet_if_device * device = (struct vnet_if_device *)__delete_link_entry(idx);
- if (device == NULL) {
- PrintError("VNET: Device %s not in the link table %d, something may be wrong in link table\n", dev_name, idx);
+ v3_unlock_irqrestore(vnet_state.lock, flags);
+
+ // if the device was found previosly the id should still be 0
+ if (new_dev->dev_id == 0) {
+ PrintError("Device Alrady exists\n");
return -1;
}
- V3_Free(device);
+ PrintDebug("Vnet: Add Device: dev_id %d, input : %p, private_data %p\n",
+ new_dev->dev_id, new_dev->input, new_dev->private_data);
- return idx;
+ return new_dev->dev_id;
}
-#endif
-int v3_vnet_pkt_process() {
- struct ethernet_pkt * pkt;
-
- PrintDebug("VNET: In vnet_check\n");
-
- while ((pkt = (struct ethernet_pkt *)v3_dequeue(g_inpkt_q)) != NULL) {
- PrintDebug("VNET: In vnet_check: pt length %d, pt type %d\n", (int)pkt->size, (int)pkt->type);
- v3_hexdump(pkt->data, pkt->size, NULL, 0);
-
- if (handle_one_pkt(pkt)) {
- PrintDebug("VNET: vnet_check: handle one packet!\n");
- } else {
- PrintError("VNET: vnet_check: fail to forward one packet, discard it!\n");
- }
-
- V3_Free(pkt); // be careful here
- }
+int v3_vnet_add_bridge(struct v3_vm_info * vm,
+ int (*input)(struct v3_vm_info * vm, struct v3_vnet_pkt * pkt, void * private_data),
+ void * priv_data) {
+ unsigned long flags;
+ int bridge_free = 0;
+ struct vnet_brg_dev * tmp_bridge = NULL;
- return 0;
-}
-
-
-static void init_empty_link_table() {
- int i;
+
+ flags = v3_lock_irqsave(vnet_state.lock);
- for (i = 0; i < MAX_LINKS; i++) {
- g_links.links[i] = NULL;
+ if (vnet_state.bridge == NULL) {
+ bridge_free = 1;
+ vnet_state.bridge = (void *)1;
}
- g_links.size = 0;
-}
-
-static void init_empty_route_table() {
- int i;
+ v3_unlock_irqrestore(vnet_state.lock, flags);
- for (i = 0; i < MAX_ROUTES; i++) {
- g_routes.routes[i] = NULL;
+ if (bridge_free == 0) {
+ PrintError("Bridge already set\n");
+ return -1;
}
- g_links.size = 0;
-}
+ tmp_bridge = (struct vnet_brg_dev *)V3_Malloc(sizeof(struct vnet_brg_dev));
-static void init_tables() {
- init_empty_link_table();
- init_empty_route_table();
- init_route_cache();
-}
-
-static void init_pkt_queue() {
- PrintDebug("VNET Init package receiving queue\n");
+ if (tmp_bridge == NULL) {
+ PrintError("Malloc Fails\n");
+ return -1;
+ }
+
+ tmp_bridge->vm = vm;
+ tmp_bridge->input = input;
+ tmp_bridge->private_data = priv_data;
- g_inpkt_q = v3_create_queue();
- v3_init_queue(g_inpkt_q);
-}
+ // make this atomic to avoid possible race conditions
+ flags = v3_lock_irqsave(vnet_state.lock);
+ vnet_state.bridge = tmp_bridge;
+ v3_unlock_irqrestore(vnet_state.lock, flags);
-static void free_link_mem(struct link_entry *link){
- V3_Free(link->dst_dev);
- V3_Free(link);
+ return 0;
}
-// TODO:
-static int addto_routing_link_tables(struct routing_entry *route_tab,
- uint16_t num_routes,
- struct link_entry *link_tab,
- uint16_t num_links){
- struct routing_entry *route, *new_route;
- struct link_entry *link, *new_link;
- int new_idx;
- int i;
- int link_idxs[MAX_LINKS];
-
- //add all of the links first, record their new indexs
- for (i = 0; i < num_links; i++) {
- link_idxs[i] = -1;
- link = &link_tab[i];
- new_link = (struct link_entry *)V3_Malloc(sizeof(struct link_entry));
-
- if (new_link == NULL){
- PrintError("VNET: Memory allocate error\n");
- return -1;
- }
-
- new_link->type = link->type;
-
- //TODO: how to set the input parameters here
- if (link->type == LINK_EDGE){
- struct vnet_if_device *ifdev = (struct vnet_if_device *)V3_Malloc(sizeof(struct vnet_if_device));
-
- if (ifdev == NULL){
- PrintError("VNET: Memory allocate fails\n");
- return -1;
- }
-
- memcpy(ifdev->name, link->dst_dev->name, DEVICE_NAME_LEN);
-
- // TODO:
- //ifdev->mac_addr
- //ifdev->input
- //ifdev->private_data
-
- new_link->dst_dev = ifdev;
- }else if (link->type == LINK_INTERFACE){
- struct vnet_if_link *iflink = (struct vnet_if_link *)V3_Malloc(sizeof(struct vnet_if_link));
-
- if (iflink == NULL){
- PrintError("VNET: Memory allocate fails\n");
- return -1;
- }
- iflink->pro_type = link->dst_link->pro_type;
- iflink->dest_ip = link->dst_link->dest_ip;
- iflink->dest_port = link->dst_link->dest_port;
- memcpy(&iflink->vnet_header, &link->dst_link->vnet_header, sizeof(struct udp_link_header));
-
- // TODO:
- //iflink->input =
- //iflink->private_data =
-
- new_link->dst_link = iflink;
- }else{
- PrintDebug("VNET: invalid link type\n");
- V3_Free(new_link);
- continue;
- }
+int V3_init_vnet() {
- new_idx = __add_link_entry(new_link);
- if (new_idx < 0) {
- PrintError("VNET: Adding link fails\n");
- free_link_mem(new_link);
- continue;
- }
- link_idxs[i] = new_idx;
- }
+ INIT_LIST_HEAD(&(vnet_state.routes));
+ INIT_LIST_HEAD(&(vnet_state.devs));
- //add all of routes, replace with new link indexs
- for (i = 0; i < num_routes; i++) {
- route = &route_tab[i];
- if (route->link_idx < 0 || route->link_idx >= num_links ||
- ((route->src_link_idx != -1) &&
- (route->src_link_idx < 0 || route->src_link_idx >= num_links))){
- PrintError("VNET: There is error in the intial tables data from guest\n");
- continue;
- }
+ vnet_state.num_devs = 0;
+ vnet_state.num_routes = 0;
- new_route = (struct routing_entry *)V3_Malloc(sizeof(struct routing_entry));
+ PrintDebug("VNET: Links and Routes tables initiated\n");
- if (new_route == NULL){
- PrintError("VNET: Memory allocate fails\n");
- return -1;
- }
- memcpy(new_route, route, sizeof(struct routing_entry));
-
- new_route->link_idx = link_idxs[new_route->link_idx];
-
- if (route->src_link_idx != -1)
- new_route->src_link_idx = link_idxs[new_route->src_link_idx];
-
- if((__add_route_entry(new_route)) == -1){
- PrintDebug("VNET: Adding route fails");
- V3_Free(new_route);
- }
- new_route = NULL;
+ if (v3_lock_init(&(vnet_state.lock)) == -1){
+ PrintError("VNET: Failure to init lock for routes table\n");
}
+ PrintDebug("VNET: Locks initiated\n");
- return 0;
-}
-
-struct table_init_info {
- addr_t routing_table_start;
- uint16_t routing_table_size;
- addr_t link_table_start;
- uint16_t link_table_size;
-};
+ vnet_state.inpkt_q = v3_create_queue();
+ v3_init_queue(vnet_state.inpkt_q);
+ PrintDebug("VNET: Receiving queue initiated\n");
-//add the guest specified routes and links to the tables
-static int handle_init_tables_hcall(struct guest_info * info, uint_t hcall_id, void * priv_data) {
- addr_t guest_addr = (addr_t)info->vm_regs.rcx;
- addr_t info_addr, route_addr, link_addr;
- struct table_init_info *init_info;
- struct link_entry *link_array;
- struct routing_entry *route_array;
+ vnet_state.route_cache = v3_create_htable(0, &hash_fn, &hash_eq);
- if (guest_va_to_host_va(info, guest_addr, &info_addr) == -1) {
- PrintError("VNET: Could not translate guest address\n");
- return -1;
- }
-
- init_info = (struct table_init_info *)info_addr;
-
- if (guest_va_to_host_va(info, init_info->routing_table_start, &route_addr) == -1) {
- PrintError("VNET: Could not translate guest address\n");
- return -1;
+ if (vnet_state.route_cache == NULL) {
+ PrintError("Vnet: Route Cache Init Fails\n");
+ return -1;
}
- route_array = (struct routing_entry *)route_addr;
- if (guest_va_to_host_va(info, init_info->link_table_start, &link_addr) == -1) {
- PrintError("VNET: Could not translate guest address\n");
- return -1;
- }
- link_array = (struct link_entry *)link_addr;
+ PrintDebug("VNET: initiated\n");
- addto_routing_link_tables(route_array, init_info->routing_table_size, link_array, init_info->link_table_size);
-
return 0;
}
-
-
-void v3_vnet_init(struct guest_info * vm) {
- init_tables();
- init_pkt_queue();
-
- v3_register_hypercall(vm, VNET_INITAB_HCALL, handle_init_tables_hcall, NULL);
-
- PrintDebug("VNET Initialized\n");
-}
-
-