X-Git-Url: http://v3vee.org/palacios/gitweb/gitweb.cgi?a=blobdiff_plain;f=palacios%2Fsrc%2Fpalacios%2Fvmm_vnet.c;h=1ec015ee8ab7cc2f100365ba9080da5fd6f7d3f6;hb=24e0221dd2fdcf613cc5e487ec57a30f2fef25e3;hp=d68610cb6300539815dc39db60ce369edb17a84a;hpb=9cabe082ba8acd1b88e6414199b5cc544aadf826;p=palacios.git diff --git a/palacios/src/palacios/vmm_vnet.c b/palacios/src/palacios/vmm_vnet.c index d68610c..1ec015e 100644 --- a/palacios/src/palacios/vmm_vnet.c +++ b/palacios/src/palacios/vmm_vnet.c @@ -7,25 +7,23 @@ * and the University of New Mexico. You can find out more at * http://www.v3vee.org * - * Copyright (c) 2009, Lei Xia - * Copyright (c) 2009, Yuan Tang - * Copyright (c) 2009, Jack Lange - * Copyright (c) 2009, Peter Dinda + * Copyright (c) 2010, Lei Xia + * Copyright (c) 2009, Yuan Tang * Copyright (c) 2009, The V3VEE Project * All rights reserved. * * Author: Lei Xia * Yuan Tang - * Jack Lange - * Peter Dinda -#include #include +#include +#include +#include #ifndef CONFIG_DEBUG_VNET #undef PrintDebug @@ -33,917 +31,718 @@ #endif -struct ethernet_pkt { - uint32_t size; - uint16_t type; - char data[ETHERNET_PACKET_LEN]; -}; +/* for UDP encapuslation */ +struct eth_header { + uchar_t dest[6]; + uchar_t src[6]; + uint16_t type; +}__attribute__((packed)); +struct ip_header { + uint8_t version: 4; + uint8_t hdr_len: 4; + uchar_t tos; + uint16_t total_len; + uint16_t id; + uint8_t flags: 3; + uint16_t offset: 13; + uchar_t ttl; + uchar_t proto; + uint16_t cksum; + uint32_t src_addr; + uint32_t dst_addr; +}__attribute__((packed)); -// 14 (ethernet frame) + 20 bytes -struct in_pkt_header { - char ethernetdest[6]; - char ethernetsrc[6]; - unsigned char ethernettype[2]; // indicates layer 3 protocol type - char ip[20]; -}; +struct udp_header { + uint16_t src_port; + uint16_t dst_port; + uint16_t len; + uint16_t csum;//set to zero, disable the xsum +}__attribute__((packed)); -#define VNET_INITAB_HCALL 0xca00 // inital hypercall id +struct udp_link_header { + struct eth_header eth_hdr; + struct ip_header ip_hdr; + struct udp_header udp_hdr; +}__attribute__((packed)); +/* end with UDP encapuslation structures */ -#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; -}; -static struct link_table g_links; -static struct routing_table g_routes; -static struct gen_queue *g_inpkt_q; +struct eth_hdr { + uint8_t dst_mac[6]; + uint8_t src_mac[6]; + uint16_t type; /* indicates layer 3 protocol type */ +} __attribute__((packed)); + + +struct vnet_dev { + int dev_id; + uint8_t mac_addr[6]; + struct v3_vm_info * vm; + struct v3_vnet_dev_ops dev_ops; + void * private_data; + + int rx_disabled; + + struct list_head node; +} __attribute__((packed)); -/* 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; -// 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_brg_dev { + struct v3_vm_info * vm; + struct v3_vnet_bridge_ops brg_ops; + + int disabled; + void * private_data; +} __attribute__((packed)); + + + +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; +struct route_list { + uint8_t hash_buf[VNET_HASH_SIZE]; -static void print_packet(char *pkt, int size) { - PrintDebug("Vnet: print_data_packet: size: %d\n", size); - v3_hexdump(pkt, size, NULL, 0); -} + uint32_t num_routes; + struct vnet_route_info * routes[0]; +} __attribute__((packed)); -#if 0 -static void print_packet_addr(char *pkt) { - PrintDebug("Vnet: print_packet_destination_addr: "); - v3_hexdump(pkt + 8, 6, NULL, 0); + +#define BUF_SIZE 4096 +struct pkts_buf { + int start, end; + int num; + v3_lock_t lock; + struct v3_vnet_pkt pkts[BUF_SIZE]; +}; + + +static struct { + struct list_head routes; + struct list_head devs; - PrintDebug("Vnet: print_packet_source_addr: "); - v3_hexdump(pkt + 14, 6, NULL, 0); -} + int num_routes; + int num_devs; -static void print_device_addr(char *ethaddr) { - PrintDebug("Vnet: print_device_addr: "); - v3_hexdump(ethaddr, 6, NULL, 0); -} -#endif + struct vnet_brg_dev *bridge; + v3_lock_t lock; -//network connection functions -static inline void ethernet_packet_init(struct ethernet_pkt *pt, const char *data, const size_t size) -{ - pt->size = size; - memcpy(pt->data, data, size); -} + uint8_t sidecores; /* 0 -vnet not running on sidecore, > 0, number of extra cores that can be used by VNET */ + uint64_t cores_map; /* bitmaps for which cores can be used by VNET for sidecore, maxium 64 */ -static uint_t hash_from_key_fn(addr_t hashkey) -{ - uint8_t * key = (uint8_t *)hashkey; - return v3_hash_buffer(key, HASH_KEY_LEN); -} + struct hashtable * route_cache; +} vnet_state; -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); -} -static int init_route_cache() -{ - 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 Initiate Failurely\n"); - return -1; - } - return 0; +#ifdef CONFIG_DEBUG_VNET +static inline void mac_to_string(char mac[6], char * buf) { + snprintf(buf, 100, "%d:%d:%d:%d:%d:%d", + 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; - - for (j = 0; j < 6; j++) { - hashkey[j] = src_addr[j]; - hashkey[j + 6] = dest_addr[j] + 1; +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_ops(%p), dst_dev_data (%p)\n", + route->dst_dev, + route->dst_dev->dev_id, + (void *)&(route->dst_dev->dev_ops), + 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 - hashkey[12] = src_type; - *(int *)(hashkey + 12) = src_index; +/* + * 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; + + 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; - } +static int clear_hash_cache() { + + 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) -{ - int n_matches = -1; - int i = 0; - struct route_cache_entry * found = NULL; +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; + return 0; } -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'); - } -} -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 struct vnet_dev * find_dev_by_id(int idx) { + struct vnet_dev * dev = NULL; -static inline void string_to_mac(const char *str, uint8_t mac[6]) -{ - int k; + 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; } -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; } - - return -1; -} -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; - } - } - - return -1; + return NULL; } +int v3_vnet_id_by_mac(char mac[6]){ -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)); + struct vnet_dev *dev = find_dev_by_mac(mac); - memset(new_route, 0, sizeof(struct routing_entry)); + if (dev == NULL) + return -1; - 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; - - int idx = -1; - if ((idx = add_route_entry(new_route)) == -1) - return -1; - - clear_hash_cache(); - - return idx; + return dev->dev_id; } -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; - } - - 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; +int v3_vnet_add_route(struct v3_vnet_route route) { + struct vnet_route_info * new_route = NULL; + unsigned long flags; - g_links.links[index] = NULL; - g_links.size --; + new_route = (struct vnet_route_info *)V3_Malloc(sizeof(struct vnet_route_info)); + memset(new_route, 0, sizeof(struct vnet_route_info)); - V3_Free(link); - - return ret; -} + PrintDebug("Vnet: vnet_add_route_entry: dst_id: %d, dst_type: %d\n", + route.dst_id, route.dst_type); + + 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 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 (new_route->route_def.dst_type == LINK_INTERFACE) { + new_route->dst_dev = find_dev_by_id(new_route->route_def.dst_id); } - - 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); + + if (new_route->route_def.src_type == LINK_INTERFACE) { + new_route->src_dev = find_dev_by_id(new_route->route_def.src_id); } - - for (i = 0; isrc_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; -} -static int delete_route_entry(int index) -{ - struct routing_entry *route; + flags = v3_lock_irqsave(vnet_state.lock); - if (index >= MAX_ROUTES || g_routes.routes[index] == NULL) - return -1; + list_add(&(new_route->node), &(vnet_state.routes)); + clear_hash_cache(); - route = g_routes.routes[index]; - g_routes.routes[index] = NULL; - g_routes.size --; + v3_unlock_irqrestore(vnet_state.lock, flags); + - V3_Free(route); +#ifdef CONFIG_DEBUG_VNET + dump_routes(); +#endif - clear_hash_cache(); - return 0; } -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) { - return -1; - } - - return delete_route_entry(index); -} -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]; - + +/* 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 i; - int max = 0; - int no = 0; - int exact_match = 0; - - for(i = 0; isrc_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; - } + 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; + +#ifdef CONFIG_DEBUG_VNET + { + char dst_str[100]; + char src_str[100]; + + 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); + } +#endif + + 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; + } + + + 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_ANY) && - (g_routes.routes[i]->src_mac_qual == MAC_ANY)) { - matched_routes[num_matches] = i; - values[num_matches] = 3; - 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 (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(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); + } } - - 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_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 ((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 ((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); } + } - 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++; - } - } - } - } - }//end for - - for(i = 0; isrc_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++; - } + // 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; -} -#if 0 -// TODO: To be removed -static int process_udpdata() -{ - struct ethernet_pkt * pt; - - uint32_t dest = 0; - uint16_t remote_port = 0; - SOCK link_sock = g_udp_sockfd; - int length = sizeof(struct ethernet_pkt) - (2 * sizeof(int)); //minus the "size" and "type" - - //run in a loop to get packets from outside network, adding them to the incoming packet queue - while (1) { - pt = (struct ethernet_pkt *)V3_Malloc(sizeof(struct ethernet_pkt)); - - if (pt == NULL){ - PrintError("Vnet: process_udp: Malloc fails\n"); - continue; - } - - PrintDebug("Vnet: route_thread: socket: [%d]. ready to receive from ip [%x], port [%d] or from VMs\n", link_sock, (uint_t)dest, remote_port); - pt->size = V3_RecvFrom_IP( link_sock, dest, remote_port, pt->data, length); - PrintDebug("Vnet: route_thread: socket: [%d] receive from ip [%x], port [%d]\n", link_sock, (uint_t)dest, remote_port); - - if (pt->size <= 0) { - PrintDebug("Vnet: process_udp: receiving packet from UDP fails\n"); - V3_Free(pt); - return -1; + matches = (struct route_list *)V3_Malloc(sizeof(struct route_list) + + (sizeof(struct vnet_route_info *) * num_matches)); + + matches->num_routes = num_matches; + + { + int i = 0; + list_for_each_entry(route, &match_list, match_node) { + matches->routes[i++] = route; } - - PrintDebug("Vnet: process_udp: get packet\n"); - print_packet(pt->data, pt->size); } + + return matches; } -// TODO: To be removed -static int indata_handler( ) -{ - if (!use_tcp) - process_udpdata(); +static int send_to_bridge(struct v3_vnet_pkt * pkt){ + struct vnet_brg_dev *bridge = vnet_state.bridge; - return 0; + if (bridge == NULL) { + PrintError("VNET: No bridge to sent data to links\n"); + return -1; + } + + return bridge->brg_ops.input(bridge->vm, pkt, 1, bridge->private_data); } -// TODO: To be removed -static int start_recv_data() -{ - if (use_tcp){ - - } else { - SOCK udp_data_socket; - - if ((udp_data_socket = V3_Create_UDP_Socket()) < 0){ - PrintError("VNET: Can't setup udp socket\n"); - return -1; - } - PrintDebug("Vnet: vnet_setup_udp: get socket: %d\n", udp_data_socket); - g_udp_sockfd = udp_data_socket; - - store_topologies(udp_data_socket); - - if (V3_Bind_Socket(udp_data_socket, vnet_udp_port) < 0){ - PrintError("VNET: Can't bind socket\n"); - return -1; - } - PrintDebug("VNET: vnet_setup_udp: bind socket successful\n"); - } - V3_CREATE_THREAD(&indata_handler, NULL, "VNET_DATA_HANDLER"); +/* enable a vnet device, notify VNET it can send pkts to it */ +int v3_vnet_enable_device(int dev_id){ + struct vnet_dev *dev = find_dev_by_id(dev_id); + unsigned long flags; + + if(!dev) + return -1; + + if(!dev->rx_disabled) return 0; -} -static void store_test_topologies(SOCK fd) -{ - int i; - int src_mac_qual = MAC_ANY; - int dest_mac_qual = MAC_ANY; - uint_t dest; -#ifndef VNET_SERVER - dest = (0 | 172 << 24 | 23 << 16 | 1 ); - PrintDebug("VNET: store_topologies. NOT VNET_SERVER, dest = %x\n", dest); -#else - dest = (0 | 172 << 24 | 23 << 16 | 2 ); - PrintDebug("VNET: store_topologies. VNET_SERVER, dest = %x\n", dest); -#endif + flags = v3_lock_irqsave(vnet_state.lock); + dev->rx_disabled = 0; + v3_unlock_irqrestore(vnet_state.lock, flags); + + /* TODO: Wake up all other guests who are trying to send pkts */ + dev = NULL; + list_for_each_entry(dev, &(vnet_state.devs), node) { + if (dev->dev_id != dev_id) + dev->dev_ops.start_tx(dev->private_data); + } - int type = UDP_TYPE; - int src = 0; - int src_type= LINK_ANY; //ANY_SRC_TYPE - int data_port = 22; + return 0; } -#endif +/* Notify VNET to stop sending pkts to it */ +int v3_vnet_disable_device(int dev_id){ + struct vnet_dev *dev = find_dev_by_id(dev_id); + unsigned long flags; -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]; + if(!dev) + return -1; - int matches[MAX_ROUTES]; - int num_matched_routes = 0; + flags = v3_lock_irqsave(vnet_state.lock); + dev->rx_disabled = 1; + v3_unlock_irqrestore(vnet_state.lock, flags); - struct in_pkt_header header; - - // 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); -#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); -#endif - - char hash_key[hash_key_size]; - make_hash_key(hash_key, src_mac, dst_mac, LINK_EDGE, src_link_index);//link_edge -> pt->type??? - - num_matched_routes = look_into_cache((route_hashkey_t)hash_key, matches); - - if (num_matched_routes == -1) { //no match - 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++) { - int route_index = -1; - int link_index = -1; - int pkt_len = 0; - - route_index = matches[i]; - - PrintDebug("Vnet: HandleDataOverLink: Forward packet from link according to Route entry %d\n", route_index); - - link_index = g_routes.routes[route_index]->link_idx; - if (link_index < 0 || link_index > MAX_LINKS) - continue; - - struct link_entry *link = g_links.links[link_index]; - if (link == NULL) - continue; - - pkt_len = pkt->size; - if (g_routes.routes[route_index]->link_type == LINK_EDGE) { - - // TODO: apply the header in the beginning of the packet to be sent - if ((link->dst_link->input(pkt->data, pkt_len, NULL)) != pkt_len) - return -1; - } else if (g_routes.routes[route_index]->link_type == LINK_INTERFACE) { - - - if ((link->dst_link->input(pkt->data, pkt_len, NULL)) != pkt_len) - return -1; - } else { - PrintError("Vnet: Wrong Edge type\n"); - return -1; - } + /* TODO: Notify all other guests to stop send pkts */ + dev = NULL; + list_for_each_entry(dev, &(vnet_state.devs), node) { + if (dev->dev_id != dev_id) + dev->dev_ops.stop_tx(dev->private_data); } - return 0; + return 0; } -static int send_ethernet_pkt(char *buf, int length) -{ - struct ethernet_pkt *pt; +int v3_vnet_send_pkt(struct v3_vnet_pkt * pkt, void * private_data) { + struct route_list * matched_routes = NULL; + unsigned long flags; + int i; - pt = (struct ethernet_pkt *)V3_Malloc(sizeof(struct ethernet_pkt)); - ethernet_packet_init(pt, buf, length); //====here we copy sending data once +#ifdef CONFIG_DEBUG_VNET + { + struct eth_hdr * hdr = (struct eth_hdr *)(pkt->header); + char dest_str[100]; + char src_str[100]; + + mac_to_string(hdr->src_mac, src_str); + mac_to_string(hdr->dst_mac, dest_str); + int cpu = V3_Get_CPU(); + PrintDebug("Vnet: on cpu %d, HandleDataOverLink. SRC(%s), DEST(%s), pkt size: %d\n", cpu, src_str, dest_str, pkt->size); + } +#endif + + flags = v3_lock_irqsave(vnet_state.lock); + + look_into_cache(pkt, &matched_routes); - PrintDebug("VNET: vm_send_pkt: transmitting packet: (size:%d)\n", (int)pt->size); - print_packet((char *)buf, length); + if (matched_routes == NULL) { + PrintError("Vnet: send pkt Looking into routing table\n"); - v3_enqueue(g_inpkt_q, (addr_t)pt); - - return 0; -} + 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 0; /* do we return -1 here?*/ + } + } -int v3_vnet_send_pkt(uchar_t *buf, int length) -{ - PrintDebug("VNET: In V3_Send_pkt: pkt length %d\n", length); - - return send_ethernet_pkt((char *)buf, length); -} + v3_unlock_irqrestore(vnet_state.lock, flags); -static int search_device(char *device_name) -{ - int i; + PrintDebug("Vnet: send pkt route matches %d\n", matched_routes->num_routes); - 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)) { - return i; - } + for (i = 0; i < matched_routes->num_routes; i++) { + struct vnet_route_info * route = matched_routes->routes[i]; + + if (route->route_def.dst_type == LINK_EDGE) { + pkt->dst_type = LINK_EDGE; + pkt->dst_id = route->route_def.dst_id; + + if (send_to_bridge(pkt) == -1) { + PrintDebug("VNET: Packet not sent properly to bridge\n"); + continue; + } + } else if (route->route_def.dst_type == LINK_INTERFACE) { + if (!route->dst_dev->rx_disabled){ + if(route->dst_dev->dev_ops.input(route->dst_dev->vm, pkt, route->dst_dev->private_data) == -1) { + PrintDebug("VNET: Packet not sent properly\n"); + continue; + } + } + } else { + PrintError("VNET: Wrong Edge type\n"); } + + PrintDebug("VNET: Forward one packet according to Route %d\n", i); } - return -1; + return 0; } -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 v3_vnet_add_dev(struct v3_vm_info *vm, uint8_t mac[6], + struct v3_vnet_dev_ops *ops, + void * priv_data){ + struct vnet_dev * new_dev = NULL; + unsigned long flags; - int idx = search_device(dev_name); - if (idx != -1) - return -1; - - if_dev = (struct vnet_if_device *)V3_Malloc(sizeof(struct vnet_if_device)); - - if (if_dev == NULL){ + new_dev = (struct vnet_dev *)V3_Malloc(sizeof(struct vnet_dev)); + + 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 *)V3_Malloc(sizeof(struct link_entry)); - - link->type = LINK_INTERFACE; - link->dst_dev = if_dev; - - idx = add_link_entry(link); - - return idx; -} - -int vnet_unregister_device(char *dev_name) -{ - int idx; - - idx = search_device(dev_name); - - if (idx == -1) { - return -1; + + memcpy(new_dev->mac_addr, mac, 6); + new_dev->dev_ops.input = ops->input; + new_dev->dev_ops.poll = ops->poll; + new_dev->private_data = priv_data; + new_dev->vm = vm; + new_dev->dev_id = 0; + + flags = v3_lock_irqsave(vnet_state.lock); + + 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) { + 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\n", new_dev->dev_id); - return idx; + return new_dev->dev_id; } -int v3_vnet_pkt_process() -{ - struct ethernet_pkt *pt; - PrintDebug("VNET: In vnet_check\n"); - - while ((pt = (struct ethernet_pkt *)v3_dequeue(g_inpkt_q)) != NULL) { - PrintDebug("VNET: In vnet_check: pt length %d, pt type %d\n", (int)pt->size, (int)pt->type); - v3_hexdump(pt->data, pt->size, NULL, 0); - - if(handle_one_pkt(pt)) { - PrintDebug("VNET: vnet_check: handle one packet!\n"); - }else { - PrintError("VNET: vnet_check: fail to forward one packet, discard it!\n"); - } - - V3_Free(pt); //be careful here +/* TODO: Still need to figure out how to handle this multicore part --Lei + */ +void v3_vnet_poll(struct v3_vm_info *vm){ + struct vnet_dev * dev = NULL; + + switch (vnet_state.sidecores) { + case 0: + list_for_each_entry(dev, &(vnet_state.devs), node) { + if(dev->vm == vm){ + dev->dev_ops.poll(vm, dev->private_data); + } + } + break; + case 1: + break; + case 2: + list_for_each_entry(dev, &(vnet_state.devs), node) { + int cpu_id = vm->cores[0].cpu_id + 2; /* temporary here, should use vnet_state.cores_map */ + struct v3_vnet_dev_xcall_args dev_args; /* could cause problem here -LX */ + dev_args.vm = vm; + dev_args.private_data = dev->private_data; + V3_Call_On_CPU(cpu_id, dev->dev_ops.poll_xcall, (void *)&dev_args); + } + break; + default: + break; } - - return 0; } +int v3_vnet_add_bridge(struct v3_vm_info * vm, + struct v3_vnet_bridge_ops *ops, + void * priv_data) { + unsigned long flags; + int bridge_free = 0; + struct vnet_brg_dev * tmp_bridge = NULL; + + flags = v3_lock_irqsave(vnet_state.lock); -static void init_empty_link_table() -{ - int i; + if (vnet_state.bridge == NULL) { + bridge_free = 1; + vnet_state.bridge = (void *)1; + } - for (i = 0; i < MAX_LINKS; i++) - g_links.links[i] = NULL; + v3_unlock_irqrestore(vnet_state.lock, flags); - - g_links.size = 0; -} - -static void init_empty_route_table() -{ - int i; + if (bridge_free == 0) { + PrintError("Bridge already set\n"); + return -1; + } - for (i = 0; i < MAX_ROUTES; i++) - g_routes.routes[i] = NULL; + tmp_bridge = (struct vnet_brg_dev *)V3_Malloc(sizeof(struct vnet_brg_dev)); - g_links.size = 0; -} + if (tmp_bridge == NULL) { + PrintError("Malloc Fails\n"); + vnet_state.bridge = NULL; + return -1; + } + + tmp_bridge->vm = vm; + tmp_bridge->brg_ops.input = ops->input; + tmp_bridge->brg_ops.xcall_input = ops->xcall_input; + tmp_bridge->brg_ops.polling_pkt = ops->polling_pkt; + tmp_bridge->private_data = priv_data; + tmp_bridge->disabled = 0; + + /* 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 init_tables() { - init_empty_link_table(); - init_empty_route_table(); - init_route_cache(); + return 0; } -static void init_pkt_queue() -{ - PrintDebug("VNET Init package receiving queue\n"); - - g_inpkt_q = v3_create_queue(); - v3_init_queue(g_inpkt_q); -} #if 0 -// TODO: -static int init_routing_tables(struct routing_entry *route_tab, uint16_t size) -{ - //struct routing_entry *route; +int v3_vnet_disable_bridge() { + unsigned long flags; + + flags = v3_lock_irqsave(vnet_state.lock); + if (vnet_state.bridge != NULL) { + vnet_state.bridge->disabled = 1; + } + + v3_unlock_irqrestore(vnet_state.lock, flags); return 0; } -// TODO: -static int init_link_tables(struct link_entry *link_tab, uint16_t size) -{ - //struct link_entry *link; +int v3_vnet_enable_bridge() { + unsigned long flags; + + flags = v3_lock_irqsave(vnet_state.lock); + + if (vnet_state.bridge != NULL) { + vnet_state.bridge->disabled = 0; + } + + v3_unlock_irqrestore(vnet_state.lock, flags); return 0; } - #endif -struct table_init_info { - addr_t routing_table_start; - uint16_t routing_table_size; - addr_t link_table_start; - uint16_t link_table_size; -}; +int v3_init_vnet() { + memset(&vnet_state, 0, sizeof(vnet_state)); + + INIT_LIST_HEAD(&(vnet_state.routes)); + INIT_LIST_HEAD(&(vnet_state.devs)); -static int handle_init_tables_hcall(struct guest_info * info, uint_t hcall_id, void *priv_data) -{ - uint8_t *buf = NULL; - addr_t info_addr = (addr_t)info->vm_regs.rcx; + vnet_state.num_devs = 0; + vnet_state.num_routes = 0; - if (guest_va_to_host_va(info, info_addr, (addr_t *)&(buf)) == -1) { - PrintError("Could not translate buffer address\n"); - return -1; + PrintDebug("VNET: Links and Routes tables initiated\n"); + + if (v3_lock_init(&(vnet_state.lock)) == -1){ + PrintError("VNET: Failure to init lock for routes table\n"); } + PrintDebug("VNET: Locks initiated\n"); - //struct table_init_info *init_info = (struct table_init_info *)buf; - - - return 0; -} + vnet_state.route_cache = v3_create_htable(0, &hash_fn, &hash_eq); + if (vnet_state.route_cache == NULL) { + PrintError("Vnet: Route Cache Init Fails\n"); + return -1; + } -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); + vnet_state.sidecores = 0; + vnet_state.cores_map = 0; - //store_test_topologies(udp_data_socket); + PrintDebug("VNET: initiated\n"); - PrintDebug("VNET Initied\n"); + return 0; } - -